researcharticle bionic vision-based intelligent power line inspection...

Research ArticleBionic Vision-Based Intelligent Power Line Inspection System

Qingwu Li YunpengMa Feijia He Shuya Xi and Jinxin Xu

Key Laboratory of Sensor Networks and Environmental Sensing Hohai University Changzhou 213022 China

Correspondence should be addressed to Qingwu Li li qingwu163com

Received 4 October 2016 Accepted 4 December 2016 Published 19 January 2017

Academic Editor Kaan Yetilmezsoy

Copyright copy 2017 Qingwu Li et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Detecting the threats of the external obstacles to the power lines can ensure the stability of the power system Inspired by theattention mechanism and binocular vision of human visual system an intelligent power line inspection system is presented in thispaper Human visual attention mechanism in this intelligent inspection system is used to detect and track power lines in imagesequences according to the shape information of power lines and the binocular visual model is used to calculate the 3D coordinateinformation of obstacles and power lines In order to improve the real time and accuracy of the system we propose a newmatchingstrategy based on the traditional SURF algorithm The experimental results show that the system is able to accurately locate theposition of the obstacles around power lines automatically and the designed power line inspection system is effective in complexbackgrounds and there are no missing detection instances under different conditions

1 Introduction

The reliability and stability of the power system are essentialfor the social development And the distance between thehigh voltage power lines and the external obstacles mustbe farther away compared to the safety distance The mostcommon external obstacles of the power lines are buildingsand plants Growth of particularly the growing plants suchas trees is irregular and not controlled by the time andenvironment It will lead to the abnormal discharge whenthe plants are close to the power lines and then the powerlines will show the phenomenon of fever and fracture Sothe growth of plants around the power lines is a hiddenthreat to the power lines Therefore complex terrain andbad environment of the growing plants make the powerline inspection tedious expensive time-consuming anddangerous

At present there have been some power line inspectionmethods put into practical work which involves four maindirections foot patrol [1] manned aerial vehicles [2] climb-ing robots [3 4] and unmanned aerial vehicles (UAVs) [5]

(A) Foot Patrol Inspection The components of a foot patrolinspection team were introduced in [1] Usually a team oftwo people walk from pylon to pylon to inspect the power

lines in the foot patrol inspection Visual inspection is carriedout with the help of binoculars and sometimes with infraredradiation and corona detection cameras Such an inspectioncan be highly accurate only when the surfaces of power lineequipment can be well seen from the ground Recently footpatrol inspection requiring the experience of staff has takenthe main responsibility for the detection of fault in the powerlines which will take lots of time and labor However it isnot easy to find hidden faults promptly by this means Manydangerous areas are blind through this kind of inspectionmethod

(B) Manned Aerial Vehicles Helicopter-assisted inspection isalso becoming common for power line inspection The teamis usually made up of three people respectively the pilot theinspector and the recorder In the helicopter over the powerlines the staff will have a wider field of vision and fasterinspection speed and they can get more accurate data withthe help of color infrared radiation and corona detectioncamera There have been several works using this methodfocusing on the sensor fusion as in [6ndash8] However given thehigh costs involved in such inspections their use should onlybe rented in quick inspection of large networks or in placeswhere it is difficult to access land In this way the staff is notnecessary if the aircraft can automatically collect power line

HindawiComputational and Mathematical Methods in MedicineVolume 2017 Article ID 4964287 13 pageshttpsdoiorg10115520174964287

2 Computational and Mathematical Methods in Medicine

data therefore other alternatives have begun to be exploredsuch as climbing robots or UAVs

(C) Climbing Robots Inspection Compared with the footpatrol inspection and the manned aerial vehicles inspectionconducted by climbing robots would be safer less tediousobjective and much faster But this kind of inspectionmethod also has many shortcomings and how to overcomethese shortcomings is the direction of the research Mean-while with the electromagnetic field of the power lines theelectronics and sensors on the robot have to be protected aspointed out in [9 10]

(D) Unmanned Aerial Vehicles (UAVs) Along with the devel-opment of the UAV technology the reconnaissance equip-ment of UAV has been more and more complex developingfrom the early days of the photoelectric camera to opto-electronic platform The UAV provided a new way in powerline inspection which not only reduces the safety risks andcosts but also has a more flexible mode of mobility enablesa wider range of inspections and gains more accurate datainformationMany studies have focused on the integration ofmultiple images collected by UAVs to determine the powerlines status especially in the fusion of infrared and visibleimages as in [10 11] According to the information of varioustypes of images it is easy to find hidden internal faults ofpower lines promptly but it is unable to figure out the threatsof external obstacles to power lines

In this paper we will tackle the mentioned challengesby proposing a new power line inspection system basedon the human visual system model for the detection ofexternal obstacles of the power lines The system can notonly inherit many advantages of traditional UAV inspectionsystem but also precisely calculate distance informationbetween the external obstacles and the power lines Therest of our work is organized as follows Section 2 brieflydescribes the characteristics of human vision Section 3 givesthe description of the UAV inspection system Section 4introduces the processing algorithm in the system basedon the human visual system model Section 5 presents theoverall evaluations and discussions of the system in practiceSection 6 draws conclusions and proposes future work

2 Human Visual Mechanism and InformationProcessing Model

After a long period of evolution biological visual systemsdevelop a strong ability for sensing the world and they helppeoplersquos work and life in a variety of forms [12] In [13]a biological hierarchical model based underwater movingobject detection method was presented In [14] a fly visualperception mechanism based small target detection methodwas proposedThese twomethodsmake use of the advantagesof biological vision in a given environment and have achievedgood results in image processing

Human visual system has some basic properties Firstlyhuman vision can always capture the interesting things in acomplex environment In view of this feature many image

segmentation research works have been carried out as in [1516] Secondly binocular vision enables human to obtain thedepth information of things Many 3D information researchworks have been carried out based on binocular vision as in[17]

21 Human Visual Attention Mechanism Human beings canrapidly and accurately identify salient regions in their visualfields Human vision has such characteristics in terms ofcolor texture shape and brightness People are always usedto paying attention to bright colors special textures strangeshapes and high brightness areas immediately in a complexenvironment With the improvement of the theory of visualsaliency salience detection is one of the research focuses incomputer vision which simulates human visual attention andvision information processing mechanisms to build saliencycomputation models Simulating such ability in machineaccording to the actual task demand is critical to makemachine handle visual content like humans

22 Binocular Vision Human beings perceive and learn the3D information of the things around them by human visualsystem In this process binocular vision plays an importantrole in obtaining the depth information of things Imagesof the same target are collected in the left and right eyesfrom different perspectives And then our brains determinethe depth of the target through the parallax in the left andright images Nowadays binocular vision technology hasbeenwidely used in computer graphics cognitive psychologyartificial intelligence and many other fields and it has suc-cessfully provided preferable solutions to many challengingproblems in aforesaid fieldsmany times Binocular vision canbe applied to the industrial noncontact detection field for itsoutstanding advantages such as simple structure convenientoperation precise measurement and high speed It is usuallycomposed of twoCCD cameras in structure and can perfectlyaccomplish the task of distance measurement in some specialoccasions

Inspired by the above aspects of visual mechanisms inthe eyes of human beings a power line inspection systembased on the human visual system model is proposed in thispaper In the proposed system the human visual attentionmechanism is used to complete the power line segmentationwork and binocular visual model is used to measure thedistance between the obstacles and the power lines

3 UAV System

The whole UAV system contains many subsystems [5]

(1) Theground control station (GCS) for accommodatingthe system operators and the interfaces between theoperators and the rest of the aerial system

(2) The quadrotor of carrying the payload(3) The system of communication between the quadrotor

and the GCS it transmits controlling inputs to thequadrotor and returns the information from thepayload and other data from the aircraft to the GCS

Computational and Mathematical Methods in Medicine 3

UAV inspection

Radio transmission system

Image sequence

Satellite link

Flight instruction

Operation officeThe ground control station

Wirelessnetwork

(i) Flight control system(ii) A video transmitter(iii) A binocular camera(iv) GPS positioning system(v) Data-communication module

Flight instruction

Figure 1 The architecture of the UAV systems for power line inspection

The hardware system finishes shooting and returns thebinocular vision in image sequences under the control ofthe staff Figure 1 shows the overall deployment of theUAV systems along with their communication and controlsolution In order to collect image information more flexiblywe chose the quadrotor helicopter as the main toolThere aremany other modules in the quadrotor helicopter includingflight control system payload GPS positioning system anddata-communication module

31 Structure of Quadrotor Helicopter The quadrotor heli-copter is a good choice for the power line inspection owingto its good performance in flight Furthermore its originalsmall body can be easily adjusted to the practical needsaccordinglyWhen the quadrotor helicopter is used for powerline inspection it will be equipped with a video transmitter abinocular camera aGPS and a data-communicationmoduleand each subsystem has its own function given as follows

(1) The input video is sampled at the video transmitterand code stream is sent to channel after compression

(2) The binocular camera shoots the power lines throughleft and right eyes at the same time and the imagesequence will be passed to the video transmitter

(3) When the quadrotor helicopter is flying in manualflight it attempts to maintain the current altitude andGPS location

(4) Data-communication module contains two informa-tion transmission channels including radio trans-mission and wireless network Flight instructions

are transmitted by radio transmission and imagesequence is transmitted by wireless network

32 Ground Control Station The GCS is generally movingaround the power line area and serves as an information hubbetween the operation office and the quadrotor helicopterDue to the limited control of the quadrotor helicopter staffsgenerally move with the GCS The GCS includes flight com-mand console information processing system and chargingsystem each subsystem has its own function as follows

(1) Staffs can make flight instructions according to statusof the quadrotor helicopter in the command console

(2) The information processing system records the imagesequence and location of the quadrotor helicopter

(3) The charging system provides charging service for thequadrotor helicopter

The staff can get the flight information and status of thequadrotor helicopter from the command console During theinspection process the staff controls the quadrotor helicopterflight along the direction of the power lines and the consoleshows image processing information and geographic locationsimultaneously

4 Power Line Inspection Method

As given in Figure 2 the visual system includes a binocularcamera and it shoots the image sequence over the power lineThe video transmitter and the data-communication moduletogether send the image sequence to the GCS and the visualprocessing algorithm is implemented in GCS


Left camera

Right cameraRight image

Left image

video transmitter

Communicationmodule

GCS

Figure 2 The architecture of visual system scheme

Left image

Image sequence

Right image

Image preprocessing

Image registration

Calculate the 3D coordinatesof power lines

Power line detection

Calculate the 3D coordinatesof characteristic points

Obstacle detection

Figure 3 The processing chart of the visual system

The visual system contains a number of image processingmodules image preprocessing power line detection imageregistration and obstacle detection as defined in [18 19] andthe processing of visual system is shown in Figure 3 Accord-ing to the actual power line distribution we established thepower line model to obtain a more complex backgroundenvironment so as to check whether our system can sustainbetter applicability in the real work We take the single frameimages respectively selected from the left and right eyes asan example in the systemprocessing In actual processing thesuccessive image frames do not contain the same content andthe system processes all the frame images in turn

41 Image Preprocessing Assume that the left eye image is119879119911 and the right eye image is 119879119910 The image preprocessingcontains image gray transform and edge detection Firstlythe system transforms RGB images 119879119911 and 119879119910 into grayimages119867119911 and119867119910 Secondly the system detects image pointsbelonging to edge There are many common methods ofedge detection each method has its own advantages anddisadvantages However most of them have poor stabilityand lots of background noises when applied to a complexenvironment Therefore to deal with this situation the DoG(Difference of Gaussian) edge detection was used in thispaper Lots of experiments in [20 21] have proven that DoG


(a) (b)

Figure 4 Image preprocessing (a) Original color image (b) Preprocessed image

(a) (b)

Figure 5 Power line detection (a) Morphological processing results (b) Detection results

has good stability and highlights the edge of the power linein the complex background The edge detection results areshown in Figure 4 DoG is the results obtained by subtractingthe Gauss function in different parameters and the imageafter being processed by DoG is defined as

119863 = DoG (119909 119910)= 12120587 [ 112059021 119890

minus(1199092+1199102)212059021 minus 112059022 119890

minus(1199092+1199102)212059022] times 119867

= [119866 (119909 119910 1205901) minus 119866 (119909 119910 1205902)] times 119867(1)

where 1205901 = 06 1205902 = 09 119867 is a gray scale image and asliding filter is defined for the image by the symbol ldquotimesrdquo Thepreprocessed images are depicted in Figure 4 and are definedas119863119911 and11986311991042 Power Line Detection The power lines are detectedbased on the human visual attentionmechanismThe humaneyes will always notice something special in the imageimmediately such as bright colors and unique edges of things[22 23] According to the characteristics of the human eyesa new method is proposed to detect the power lines throughliner edges and the result of the proposed method is shownin Figure 5 The method below is to judge whether there arepower lines

(1) Because the quadrotor helicopterrsquos flight direction isparallel to the direction of the distribution of power linespower lines are vertically distributed in the image Firstly wedilate and erode the image by linear structural factors severaltimes and the dilation and erosion are defined separately asfollows

dilation 119884 = 119864 oplus 119861 = 119910 119861 (119910) cap 119864 = Φ erosion 119883 = 119864 otimes 119861 = 119909 119861 (119909) sub 119864 (2)

where 119861(119909) are structural factors and 119864 is the work space Inthe choice of the direction of the linear structure we first usethe same length linear structure with a plurality of directionsvarying from 0∘ to 180∘ to process the image and then detectthe results and keep the image with the maximum remainingpixels as a target for the next step When the structure factoris relatively short the image will only retain power lines aftera great deal of dilation and erosion

(2) As for detecting line segments in the image linesegments in the same slope range are restored to a straightline Due to the power line throughout the whole imagewe can detect power lines through the distribution of linearconnected domains

(3) Regarding recording coordinates of power lines inleft and right image we will get (1199091198891199111 1199101198891199111) (119909119889119911119899 119910119889119911119899)and (11990910158401198891199111 11991010158401198891199111) (1199091015840119889119911119899 1199101015840119889119911119899) from the left image


(a) (b)

Figure 6 Image registration (a) Original color image (left) (b) Original color image (right)

and (1199091198891199101 1199101198891199101) (119909119889119910119899 119910119889119910119899) and (11990910158401198891199101 11991010158401198891199101) (1199091015840119889119910119899 1199101015840119889119910119899) from the right image

43 Image Registration Traditional binocular image reg-istration methods are slow and inaccurate In the powerline inspection system because the point on the obstacleis connected it is unnecessary to register all the pixelsin the image Based on the traditional SURF algorithm in[24] we make some optimizations in the matching strategyExperiments show that the algorithm can effectively improvethe speed and accuracy of image registration which can bevalidated by the result in Figure 6We can see that the selectedfeature points are randomly distributed in the image Thedetailed procedure for image registration is as follows

(1) Using SURF algorithm detects the feature points of theimage in the scale-space During the image filter processingSURF algorithm selects a window with different sizes andthen the feature points are extracted by using HessianmatrixAnd the point 119863(119909 119910) in the image is defined by Hessianmatrix

119867(119909 120590) = [119871119909119909 (119909 120590) 119871119909119910 (119909 120590)119871119909119910 (119909 120590) 119871119910119910 (119909 120590)] (3)

where120590 varies in different scales and119871119909119909(119909 120590)119871119909119910(119909 120590) and119871119910119910(119909 120590) are defined as

119871119909119909 (119909 120590) = 119863 (119909 119910) lowast 12059721205971199092119892 (120590) 119871119909119910 (119909 120590) = 119863 (119909 119910) lowast 1205972120597119909120597119910119892 (120590) 119871119910119910 (119909 120590) = 119863 (119909 119910) lowast 12059721205971199102119892 (120590)

(4)

(2) Taking the current feature point as the center a blockof size 20120590 is constructed along the main direction of thepoint Then divide the area into 16 small areas and calculatetheHarr wavelet response in each 5120590lowast5120590 region So each sub-area can be represented by V = (sum119889119909sum |119889119909| sum 119889119910sum |119889119910|)

Finally we obtain 64 dimensional descriptors of the pointThe salient points in the image are defined as

Pos1 = (11990910158401 11991010158401) (11990910158402 11991010158402) (1199091015840119898 1199101015840119898) 1 le 119894 le 119898

Pos2 = (1199091 1199101) (1199092 1199102) (119909119899 119910119899) 1 le 119894 le 119899(5)

where Pos1 represents the set of the feature points in the leftimage and Pos2 represents the set of the feature points in theright image

(3) Calculate the Euclidean distance between all points inPos1 and Pos2 and select the point withminimumEuclideandistance as the rough matching point Sort the matchingpoints in ascending order according to the Euclidean dis-tance and delete the abnormal points Select the top119870 pointsin the matching points which are defined as

Pos119870 = (11990910158401 11991010158401) (1199091 1199101) (11990910158402 11991010158402) (1199092 1199102) (1199091015840119870 1199101015840119870) (119909119870 119910119870)

(6)

(4) Next is screen matching points according to the slopebetween the corresponding points in Pos 119870 In detail firstlycalculate the frequency of each slope in all slopes and thenew slope set 119896 new = 1198961 1198962 119896119902 is formed by removingabnormal slopes Secondly a new set of matching points isdefined as

Pos 119870new = (1199091199111 1199101199111) (1199091199101 1199101199101) (1199091199112 1199101199112) (1199091199102 1199101199102) (119909119911119899 119910119911119899) (119909119910119899 119910119910119899)

(7)

44 Obstacle Detection The purpose of using binocularvision camera is to precisely calculate distance informationbetween external obstacles and the power lines In thisprocess we get the 3D information of the obstacles and thepower lines in 2D images according to characteristics ofbinocular vision as in [25] The principle of binocular visionimaging is shown in Figure 7


Focus f

Baseline b

yc

zc

xc

Right imageLeft image

P1(xzn yzn) P2(xyn yyn)

P(xc yc zc)

Figure 7 The principle of binocular vision imaging

We already know the basic parameters of the binocularcamera such as the baseline 119887 and the focus 119891 According topos 119870new the parallax between left image and right image isdefined as

119889 = (119909119911119899 minus 119909119910119899) (8)

The spatial coordinates of a point P in the left cameracoordinate system can be defined as

119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)

By calculating the spatial coordinates of all the featurepoints and the spatial coordinates of the points on the powerline simultaneously we judge whether the current point is abarrier by the minimum Euclidean distance 119869 between thepoint and power lines Finally we deal with all the points andmark the points with less than the threshold value as obstaclesand then enlarge the area of obstacles in a certain scale

5 Evaluations and Discussions

The object of the system in our work is to detect the externalobstacles of power lines based on the human visual systemmodel In the practical application we need to considerboth the accuracy and the stability of the system undervarious conditions In order to test the accuracy of the visualsystem we set up the power line model in the laboratoryIn this power line model we have set up two power lines

Figure 8 Experimental model of the power line inspection (theheight of the obstacle models ranges from 20 to 150mm and theheight of the power line ranges from 150 to 155mm)

and complex background environment including trees andbuildings We can exactly measure the accuracy of thevisual system according to the data we collected from thepower line model In addition the various flight attitudes ofthe quadrotor helicopter are simulated in experiments withexperimental model in Figure 8 In all experiments givenin this paper the time consumptions were measured on astandard PC Windows 81 running at Intel Core i7 25 GHzand 8GRAM the testing software was provided byMATLABR2010a the images are of size 384 lowast 51251 Experiment of Normal Shooting under Forest-Based Envi-ronment The experiment was carried out in a mixed envi-ronment consisting mainly of trees During the normal flightof the quadrotor helicopter power lines are nearly verticallydistributed in images captured by binocular vision camera Inthe process of analysis firstly we mark the external obstacles


(a) (b)

2

3

4

5

1

(c) (d)

Figure 9 Obstacle detection (the marked circles are threats to the power line on the left and the marked squares are threats to the powerline on the left) (a) Original image (left) (b) Original image (right) (c) Standard detection results (left) (d) Actual detection results (left)

Table 1 Experimental analysis

Area 1 Area 2 Area 3 Area 4 Area 5Actual distance 118mm 265mm 51mm 324mm 105mmExperimental distance 13mm 256mm 53mm 298mm 95mmError 12mm 09mm 02 mm 26mm 1mm

Accuracy rate 100 false detection rate 0 image processing time 252 s

through manual measurement and record the measurementresults in the corresponding table And then we analyze theperformance of the system through the comparison of thesystem data and the measured data

In order to get the distance between an area and a powerline we usually first select a set of feature points withinthe area and then calculate the average distance of all theseselected points with the power line The accurate and falsedetection rates are defined as

119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)

where 119878119860 are the actual obstacle connected domains and119878119864 are the experimental obstacle connected domains Inthe experiment a number of connected domains are used

to describe a real obstacle area However if the connecteddomain is not corresponding to the obstacle it will be judgedas a false detection From Example 1 (Figure 9 and Table 1)we can see that in the case of normal shooting the systemcan accurately complete the obstacle detection work and thesystem accuracy rate is 100

In Example 2 (Figure 10 and Table 2) we have trans-formed the background of the experiment and more powerline obstacles appeared in the experiment In this case thesystem is still able to accurately output obstacle areas and theaccuracy rate is 100

The above experiments can validate the fact that ourproposed power line obstacle detection method is feasibleIn a variety of environmental experiments our method hasperformed very well under normal shooting The advantagesof the system are mainly embodied in the less error highaccuracy and low false detection rate

Example 1 See Figure 9 and Table 1


(a) (b)

2

3 4 6 7

51

(c) (d)

Figure 10 Obstacle detection (a) Original image (left) (b) Original image (right) (c) Standard detection results (left) (d) Actual detectionresults (left)


Area 1 Area 2 Area 3 Area 4 Area 5 Area 6 Area 7Actual distance 124mm 139mm 51mm 64mm 82mm 41mm 65mmExperimental distance 135mm 136mm 67mm 53mm 77mm 58mm 63mmError 11mm 03mm 16mm 11mm 05mm 17mm 02mm



52 Experiment of Abnormal Shooting under Building-BasedEnvironment The experiment was carried out in a mixedenvironment consistingmainly of flyovers And the abnormalsituations mainly act in three aspects Firstly the directionof shooting angle may deviate from the power lines distribu-tion Secondly the attitude of the quadrotor helicopter maybe tilted in the horizontal direction Thirdly both above-depicted circumstances showup simultaneously In thework-ing process of the quadrotor helicopter such three factorsas the staff rsquos working state the weather conditions andthe abnormal situations must be taken into considerationHowever these anomalies cannot be avoided completely inthe actual work motivating us to formulate such a systemto deal with a variety of conditions effectively Experimentsshow that the proposed method is effective for the imagesunder abnormal shooting

FromExample 3 (Figure 11 andTable 3) it can be seen thatthe direction of shooting angle is deviated from the powerlines distribution The experimental results show that themeasurement error of the system is very small although thephenomenon of false detection has been detected Overallour proposed method is stable in this condition

From Example 4 (Figure 12 and Table 4) it can be foundthat the attitude of the quadrotor helicopter is tilted in thehorizontal direction based on the distance difference betweenthe power lines The experimental results show that thesystem is stable in this condition without any loss of accuracyrate

In Example 5 (Figure 13 and Table 5) it can be seen thatthe direction of shooting angle is deviated from the powerline distribution and the attitude of the quadrotor helicopteris tilted in the horizontal direction simultaneously In thiscondition the phenomenon of false detection has occurredin our systemThe false detection rate of Example 5 is similarto that of Example 3 Therefore it can be concluded that


(a) (b)

6

2

3

4

1

5

(c) (d)



Area 1 Area 2 Area 3 Area 4 Area 5 Area 6Actual distance 58mm 134mm 162mm 53mm 108mm 156mmExperimental distance 63mm 125mm 157mm 67mm 92mm 171mmError 05mm 09mm 05mm 14mm 16mm 15mm


the direction of shooting angle is deviating from the powerline distribution and constitutes the main reason for falsedetection in the system In order to verify this conclusion wehave done many experiments




Through many experiments the results of Table 6 canprove our conclusion Therefore in the process of imageacquisition we need to try to keep the direction of shootingangle parallel to the direction of power line distributionAlthough the system has the phenomenon of false detectionwe can keep the detection accuracy of the obstacles in theconnected area to be over 85 Under normal circumstancesour proposed method can accurately detect the position

of the power line external obstacles with the error rangeabout 1mm and there is no missing detection From theexperimental results we can see that each frame imageprocessing time is about 25 s which can basically meet therequirements of single frame image processing However theperformance is not good enough in the processing of imagesequences which is to be solved in our future research work

Many methods have been proposed for power line detec-tion In [26] a novel technique based on magnetic fieldsensing is proposed for underground power cable detectionand inspection This innovative algorithm provided goodresults but it is not suitable for the air power line detectionIn [4] the most important achievements in the field ofpower line inspection by mobile robots were presentedAlthough flying robots and climbing robots indeed havespecific benefits both of them are far from practical useMore attention to the control and communication of theUAVwas paid in [27] but no solution was given to the power


(a) (b)

4

6

5

2

3

1

(c) (d)





line detection An unmanned aerial system based on thequadrotor helicopter for high voltage power line inspectionwas proposed in [5] but it just offered a simple solution of thepower line using infrared and visible images In this paperwe present a system for quantitative analysis of power lineobstacles The experimental results show that the proposedsystem can be applied to the actual power line inspectionwithhigh accuracy

6 Conclusion

In this paper we propose a power line inspection systembased on the human visual system model Compared withthe traditional power line inspection systems we introducethe characteristics of human vision into our system Inboth the forest-based environment and the building-basedenvironment our system can not only detect the externalobstacles of the power lines but also accurately measure the

distance between the obstacles and the power linesMoreoverour systemprovides a safer and cheaper power line inspectionmethod The system is of great importance to the electricpower department and can greatly increase the efficiency andaccuracy of power line inspection

In the future the proposed system is to be tested in a realcomplex environment Besides the adaptability of our systemin abnormal situations is expecting further improvementsStill our future work can focus on making the systemmore comprehensive integrating optical cameras infraredcameras and ultraviolet cameras and detecting faults withinthe power lines

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper


(a) (b)

6

5

2

3

4

1

(c) (d)





Table 6 Analysis of false detection

Test times under normal conditionsthenumber of false detection instances

Test times under abnormalconditionsthe number of false detection

instancesThe direction of shooting angle 200 2018The attitude of the quadrotor helicopter 202 201

Acknowledgments

This work was supported by the National Natural ScienceFoundation of China (no 41301448) the Science and Tech-nology Program of Jiangsu Province (no BE2016071) and theScience and Technology Support Program of Changzhou (noCE20150068)

References

[1] D Jones and G Earp ldquoRequirements for aerial inspection ofoverhead electrical power linesrdquo in Proceedings of the 12th RPVsInternational Conference Bristol UK 1996

[2] Shoemaker M Thomas and J E Mack The Linemanrsquos andCablemanrsquos Handbook McGraw-Hill London UK 2007


[3] N Pouliot and S Montambault ldquoGeometric design of thelinescout a teleoperated robot for power line inspection andmaintenancerdquo in Proceedings of the IEEE International Confer-ence on Robotics and Automation (ICRA rsquo08) May 2008

[4] J Katrasnik F Pernus and B Likar ldquoA survey of mobile robotsfor distribution power line inspectionrdquo IEEE Transactions onPower Delivery vol 25 no 1 pp 485ndash493 2010

[5] L F Luque-Vega B Castillo-Toledo A Loukianov and L EGonzalez-Jimenez ldquoPower line inspection via an unmannedaerial system based on the quadrotor helicopterrdquo in Proceedingsof the 17th IEEE Mediterranean Electrotechnical Conference(MELECON rsquo14) April 2014

[6] G Yan J Wang Q Liu et al ldquoAn airborne multi-angle powerline inspection systemrdquo in Proceedings of the IEEE InternationalGeoscience and Remote Sensing Symposium (IGARSS rsquo07) pp2913ndash2915 IEEE Barcelona Spain June 2007

[7] L Zhang H Jiang and L Meng ldquoResearch of the observationsuspended bin for helicopter power line inspectionrdquo in Proceed-ings of the 2009 IEEE International Conference on Mechatronicsand Automation (ICMA rsquo09) pp 1689ndash1694 IEEE ChangchunChina August 2009

[8] L Zhang L Yan LMeng X Li and S Huang ldquoThe applicationstudy of helicopter airborne photoelectric stabilized pod inthe high voltage power line inspectionrdquo in Proceedings of theInternational Conference on Optoelectronics and Microelectron-ics (ICOM rsquo12) Changchun China August 2012

[9] LWang F Liu ZWang S Xu S Cheng and J Zhang ldquoDevel-opment of a practical power transmission line inspection robotbased on a novel line walking mechanismrdquo in Proceedings ofthe 23rd IEEERSJ International Conference on Intelligent Robotsand Systems (IROS rsquo10) pp 222ndash227 Taiwan China October2010

[10] J Zhang L Liu B Wang X Chen Q Wang and T ZhengldquoHigh speed automatic power line detection and tracking fora UAV-based inspectionrdquo in Proceedings of the InternationalConference on Industrial Control and Electronics Engineering(ICICEE rsquo12) Xirsquoan China August 2012

[11] X Sun W K Lee Y Hou and P W T Pong ldquoUndergroundpower cable detection and inspection technology based onmagnetic field sensing at ground surface levelrdquo IEEE Transac-tions on Magnetics vol 50 no 7 2014

[12] S Chen Y Zheng C Cattani and W Wang ldquoModeling ofbiological intelligence for SCM systemoptimizationrdquoComputa-tional and Mathematical Methods in Medicine vol 2012 ArticleID 769702 10 pages 2012

[13] J Shen T Fan M Tang Q Zhang Z Sun and F HuangldquoA biological hierarchical model based underwater movingobject detectionrdquo Computational and Mathematical Methods inMedicine vol 2014 Article ID 609801 p 8 2014

[14] F Huang L Xu M Li andM Tang ldquoHigh-resolution remotelysensed small target detection by imitating fly visual perceptionmechanismrdquo Computational and Mathematical Methods inMedicine vol 2012 Article ID 789429 9 pages 2012

[15] H Jing X He Q Han and X Niu ldquoBackground contrast basedsalient region detectionrdquo Neurocomputing vol 124 pp 57ndash622014

[16] S Qi J-G Yu J Ma Y Li and J Tian ldquoSalient object detectionvia contrast information and object vision organization cuesrdquoNeurocomputing vol 167 pp 390ndash405 2015

[17] Q Yang R Yang J Davis and D Nister ldquoSpatial-depthsuper resolution for range imagesrdquo in Proceedings of the IEEE

Computer Society Conference on Computer Vision and PatternRecognition (CVPR rsquo07) Minneapolis Minn USA June 2007

[18] F L Kooi and A Toet ldquoVisual comfort of binocular and 3Ddisplaysrdquo Displays vol 25 no 2-3 pp 99ndash108 2004

[19] G Xu X Li J Su H Pan and G Tian ldquoPrecision evaluationof three-dimensional feature points measurement by binocularvisionrdquo Journal of the Optical Society of Korea vol 15 no 1 pp30ndash37 2011

[20] S Wang W Li Y Wang Y Jiang S Jiang and R ZhaoldquoAn improved difference of gaussian filter in face recognitionrdquoJournal of Multimedia vol 7 no 6 pp 429ndash433 2012

[21] Ph Marthon B Thiesse and A Bruel ldquoEdge detection bydifferences of Gaussiansrdquo in Proceedings of the InternationalTechnical SymposiumEurope International Society for Opticsand Photonics 1986

[22] L Itti C Koch and E Niebur ldquoAmodel of saliency-based visualattention for rapid scene analysisrdquo IEEE Transactions on PatternAnalysis and Machine Intelligence vol 20 no 11 pp 1254ndash12591998

[23] Y Hu X Xie W-Y Ma L-T Chia and D Rajan ldquoSalientregion detection using weighted feature maps based on thehuman visual attentionmodelrdquo in Proceedings of the Pacific-RimConference on Multimedia Springer Tokyo Japan November-December 2004

[24] H Bay A Ess T Tuytelaars and L Van Gool ldquoSpeeded-uprobust features (SURF)rdquo Computer Vision and Image Under-standing vol 110 no 3 pp 346ndash359 2008

[25] P Zhao and N-H Wang ldquoPrecise perimeter measurement for3D object with a binocular stereo vision measurement systemrdquoOptik vol 121 no 10 pp 953ndash957 2010

[26] X Sun W K Lee Y Hou and P W T Pong ldquoUndergroundpower cable detection and inspection technology based onmagnetic field sensing at ground surface levelrdquo IEEE Transac-tions on Magnetics vol 50 no 7 pp 1ndash5 2014

[27] T S Bruggemann J J Ford and R A Walker ldquoControl of air-craft for inspection of linear infrastructurerdquo IEEE Transactionson Control Systems Technology vol 19 no 6 pp 1397ndash1409 2011

Submit your manuscripts athttpswwwhindawicom

Stem CellsInternational

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


MEDIATORSINFLAMMATION

of


Behavioural Neurology

EndocrinologyInternational Journal of



Disease Markers


BioMed Research International

OncologyJournal of



Oxidative Medicine and Cellular Longevity


PPAR Research

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

ObesityJournal of



Computational and Mathematical Methods in Medicine

OphthalmologyJournal of


Diabetes ResearchJournal of



Research and TreatmentAIDS


Gastroenterology Research and Practice


Parkinsonrsquos Disease

Evidence-Based Complementary and Alternative Medicine

Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom


data therefore other alternatives have begun to be exploredsuch as climbing robots or UAVs

(C) Climbing Robots Inspection Compared with the footpatrol inspection and the manned aerial vehicles inspectionconducted by climbing robots would be safer less tediousobjective and much faster But this kind of inspectionmethod also has many shortcomings and how to overcomethese shortcomings is the direction of the research Mean-while with the electromagnetic field of the power lines theelectronics and sensors on the robot have to be protected aspointed out in [9 10]

(D) Unmanned Aerial Vehicles (UAVs) Along with the devel-opment of the UAV technology the reconnaissance equip-ment of UAV has been more and more complex developingfrom the early days of the photoelectric camera to opto-electronic platform The UAV provided a new way in powerline inspection which not only reduces the safety risks andcosts but also has a more flexible mode of mobility enablesa wider range of inspections and gains more accurate datainformationMany studies have focused on the integration ofmultiple images collected by UAVs to determine the powerlines status especially in the fusion of infrared and visibleimages as in [10 11] According to the information of varioustypes of images it is easy to find hidden internal faults ofpower lines promptly but it is unable to figure out the threatsof external obstacles to power lines

In this paper we will tackle the mentioned challengesby proposing a new power line inspection system basedon the human visual system model for the detection ofexternal obstacles of the power lines The system can notonly inherit many advantages of traditional UAV inspectionsystem but also precisely calculate distance informationbetween the external obstacles and the power lines Therest of our work is organized as follows Section 2 brieflydescribes the characteristics of human vision Section 3 givesthe description of the UAV inspection system Section 4introduces the processing algorithm in the system basedon the human visual system model Section 5 presents theoverall evaluations and discussions of the system in practiceSection 6 draws conclusions and proposes future work

2 Human Visual Mechanism and InformationProcessing Model

After a long period of evolution biological visual systemsdevelop a strong ability for sensing the world and they helppeoplersquos work and life in a variety of forms [12] In [13]a biological hierarchical model based underwater movingobject detection method was presented In [14] a fly visualperception mechanism based small target detection methodwas proposedThese twomethodsmake use of the advantagesof biological vision in a given environment and have achievedgood results in image processing

Human visual system has some basic properties Firstlyhuman vision can always capture the interesting things in acomplex environment In view of this feature many image

segmentation research works have been carried out as in [1516] Secondly binocular vision enables human to obtain thedepth information of things Many 3D information researchworks have been carried out based on binocular vision as in[17]

21 Human Visual Attention Mechanism Human beings canrapidly and accurately identify salient regions in their visualfields Human vision has such characteristics in terms ofcolor texture shape and brightness People are always usedto paying attention to bright colors special textures strangeshapes and high brightness areas immediately in a complexenvironment With the improvement of the theory of visualsaliency salience detection is one of the research focuses incomputer vision which simulates human visual attention andvision information processing mechanisms to build saliencycomputation models Simulating such ability in machineaccording to the actual task demand is critical to makemachine handle visual content like humans

22 Binocular Vision Human beings perceive and learn the3D information of the things around them by human visualsystem In this process binocular vision plays an importantrole in obtaining the depth information of things Imagesof the same target are collected in the left and right eyesfrom different perspectives And then our brains determinethe depth of the target through the parallax in the left andright images Nowadays binocular vision technology hasbeenwidely used in computer graphics cognitive psychologyartificial intelligence and many other fields and it has suc-cessfully provided preferable solutions to many challengingproblems in aforesaid fieldsmany times Binocular vision canbe applied to the industrial noncontact detection field for itsoutstanding advantages such as simple structure convenientoperation precise measurement and high speed It is usuallycomposed of twoCCD cameras in structure and can perfectlyaccomplish the task of distance measurement in some specialoccasions

Inspired by the above aspects of visual mechanisms inthe eyes of human beings a power line inspection systembased on the human visual system model is proposed in thispaper In the proposed system the human visual attentionmechanism is used to complete the power line segmentationwork and binocular visual model is used to measure thedistance between the obstacles and the power lines

3 UAV System

The whole UAV system contains many subsystems [5]

(1) Theground control station (GCS) for accommodatingthe system operators and the interfaces between theoperators and the rest of the aerial system

(2) The quadrotor of carrying the payload(3) The system of communication between the quadrotor

and the GCS it transmits controlling inputs to thequadrotor and returns the information from thepayload and other data from the aircraft to the GCS


UAV inspection


Image sequence

Satellite link

Flight instruction


Wirelessnetwork


Flight instruction

















Left camera


Left image

video transmitter

Communicationmodule

GCS


Left image

Image sequence

Right image

Image preprocessing

Image registration




Obstacle detection





(a) (b)


(a) (b)



119863 = DoG (119909 119910)= 12120587 [ 112059021 119890

minus(1199092+1199102)212059021 minus 112059022 119890

minus(1199092+1199102)212059022] times 119867

= [119866 (119909 119910 1205901) minus 119866 (119909 119910 1205902)] times 119867(1)








(a) (b)





119867(119909 120590) = [119871119909119909 (119909 120590) 119871119909119910 (119909 120590)119871119909119910 (119909 120590) 119871119910119910 (119909 120590)] (3)



(4)



Pos1 = (11990910158401 11991010158401) (11990910158402 11991010158402) (1199091015840119898 1199101015840119898) 1 le 119894 le 119898

Pos2 = (1199091 1199101) (1199092 1199102) (119909119899 119910119899) 1 le 119894 le 119899(5)



Pos119870 = (11990910158401 11991010158401) (1199091 1199101) (11990910158402 11991010158402) (1199092 1199102) (1199091015840119870 1199101015840119870) (119909119870 119910119870)

(6)


Pos 119870new = (1199091199111 1199101199111) (1199091199101 1199101199101) (1199091199112 1199101199112) (1199091199102 1199101199102) (119909119911119899 119910119911119899) (119909119910119899 119910119910119899)

(7)



Focus f

Baseline b

yc

zc

xc



P(xc yc zc)



119889 = (119909119911119899 minus 119909119910119899) (8)


119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)







(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















UAV inspection


Image sequence

Satellite link

Flight instruction


Wirelessnetwork


Flight instruction

















Left camera


Left image

video transmitter

Communicationmodule

GCS


Left image

Image sequence

Right image

Image preprocessing

Image registration




Obstacle detection





(a) (b)


(a) (b)



119863 = DoG (119909 119910)= 12120587 [ 112059021 119890

minus(1199092+1199102)212059021 minus 112059022 119890

minus(1199092+1199102)212059022] times 119867

= [119866 (119909 119910 1205901) minus 119866 (119909 119910 1205902)] times 119867(1)








(a) (b)





119867(119909 120590) = [119871119909119909 (119909 120590) 119871119909119910 (119909 120590)119871119909119910 (119909 120590) 119871119910119910 (119909 120590)] (3)



(4)



Pos1 = (11990910158401 11991010158401) (11990910158402 11991010158402) (1199091015840119898 1199101015840119898) 1 le 119894 le 119898

Pos2 = (1199091 1199101) (1199092 1199102) (119909119899 119910119899) 1 le 119894 le 119899(5)



Pos119870 = (11990910158401 11991010158401) (1199091 1199101) (11990910158402 11991010158402) (1199092 1199102) (1199091015840119870 1199101015840119870) (119909119870 119910119870)

(6)


Pos 119870new = (1199091199111 1199101199111) (1199091199101 1199101199101) (1199091199112 1199101199112) (1199091199102 1199101199102) (119909119911119899 119910119911119899) (119909119910119899 119910119910119899)

(7)



Focus f

Baseline b

yc

zc

xc



P(xc yc zc)



119889 = (119909119911119899 minus 119909119910119899) (8)


119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)







(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















Left camera


Left image

video transmitter

Communicationmodule

GCS


Left image

Image sequence

Right image

Image preprocessing

Image registration




Obstacle detection





(a) (b)


(a) (b)



119863 = DoG (119909 119910)= 12120587 [ 112059021 119890

minus(1199092+1199102)212059021 minus 112059022 119890

minus(1199092+1199102)212059022] times 119867

= [119866 (119909 119910 1205901) minus 119866 (119909 119910 1205902)] times 119867(1)








(a) (b)





119867(119909 120590) = [119871119909119909 (119909 120590) 119871119909119910 (119909 120590)119871119909119910 (119909 120590) 119871119910119910 (119909 120590)] (3)



(4)



Pos1 = (11990910158401 11991010158401) (11990910158402 11991010158402) (1199091015840119898 1199101015840119898) 1 le 119894 le 119898

Pos2 = (1199091 1199101) (1199092 1199102) (119909119899 119910119899) 1 le 119894 le 119899(5)



Pos119870 = (11990910158401 11991010158401) (1199091 1199101) (11990910158402 11991010158402) (1199092 1199102) (1199091015840119870 1199101015840119870) (119909119870 119910119870)

(6)


Pos 119870new = (1199091199111 1199101199111) (1199091199101 1199101199101) (1199091199112 1199101199112) (1199091199102 1199101199102) (119909119911119899 119910119911119899) (119909119910119899 119910119910119899)

(7)



Focus f

Baseline b

yc

zc

xc



P(xc yc zc)



119889 = (119909119911119899 minus 119909119910119899) (8)


119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)







(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)


(a) (b)



119863 = DoG (119909 119910)= 12120587 [ 112059021 119890

minus(1199092+1199102)212059021 minus 112059022 119890

minus(1199092+1199102)212059022] times 119867

= [119866 (119909 119910 1205901) minus 119866 (119909 119910 1205902)] times 119867(1)








(a) (b)





119867(119909 120590) = [119871119909119909 (119909 120590) 119871119909119910 (119909 120590)119871119909119910 (119909 120590) 119871119910119910 (119909 120590)] (3)



(4)



Pos1 = (11990910158401 11991010158401) (11990910158402 11991010158402) (1199091015840119898 1199101015840119898) 1 le 119894 le 119898

Pos2 = (1199091 1199101) (1199092 1199102) (119909119899 119910119899) 1 le 119894 le 119899(5)



Pos119870 = (11990910158401 11991010158401) (1199091 1199101) (11990910158402 11991010158402) (1199092 1199102) (1199091015840119870 1199101015840119870) (119909119870 119910119870)

(6)


Pos 119870new = (1199091199111 1199101199111) (1199091199101 1199101199101) (1199091199112 1199101199112) (1199091199102 1199101199102) (119909119911119899 119910119911119899) (119909119910119899 119910119910119899)

(7)



Focus f

Baseline b

yc

zc

xc



P(xc yc zc)



119889 = (119909119911119899 minus 119909119910119899) (8)


119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)







(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)





119867(119909 120590) = [119871119909119909 (119909 120590) 119871119909119910 (119909 120590)119871119909119910 (119909 120590) 119871119910119910 (119909 120590)] (3)



(4)



Pos1 = (11990910158401 11991010158401) (11990910158402 11991010158402) (1199091015840119898 1199101015840119898) 1 le 119894 le 119898

Pos2 = (1199091 1199101) (1199092 1199102) (119909119899 119910119899) 1 le 119894 le 119899(5)



Pos119870 = (11990910158401 11991010158401) (1199091 1199101) (11990910158402 11991010158402) (1199092 1199102) (1199091015840119870 1199101015840119870) (119909119870 119910119870)

(6)


Pos 119870new = (1199091199111 1199101199111) (1199091199101 1199101199101) (1199091199112 1199101199112) (1199091199102 1199101199102) (119909119911119899 119910119911119899) (119909119910119899 119910119910119899)

(7)



Focus f

Baseline b

yc

zc

xc



P(xc yc zc)



119889 = (119909119911119899 minus 119909119910119899) (8)


119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)







(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















Focus f

Baseline b

yc

zc

xc



P(xc yc zc)



119889 = (119909119911119899 minus 119909119910119899) (8)


119909119888 = 119887 sdot 119909119911119899119889 119910119888 = 119887 sdot 119910119911119899119889 119911119888 = 119887 sdot 119891119889

(9)







(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)

2

3

4

5

1

(c) (d)







119860 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119860 ) times 100

119864 = (sum119878119864 sub [(119878119860 cap 119878119864) = 0]119878119864 ) times 100

(10)







(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)

2

3 4 6 7

51

(c) (d)











(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)

6

2

3

4

1

5

(c) (d)













(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)

4

6

5

2

3

1

(c) (d)






6 Conclusion




Competing Interests



(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(a) (b)

6

5

2

3

4

1

(c) (d)









Acknowledgments


References



































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















researcharticle bionic vision-based intelligent power line inspection...

Documents