ps_1.2 -tutorial (bl) - orthophoto, dem (with gcps)

8/16/2019 PS_1.2 -Tutorial (BL) - Orthophoto, DeM (With GCPs)

1/17

Tutorial (Beginner level):

Orthomosaic and DEM Generation with Agisoft Photocan Pro !"#

(with Ground $ontrol Points)

OverviewAgisoft PhotoScan Professional allows to generate georeferenced dense point clouds, textured polygonal

models, digital elevation models and orthomosaics from a set of overlapping images with the correspondingreferencing information. This tutorial describes the main processing steps of DE!"rthomosaic generation

wor#flow for a set of images with the ground control points.

PhotoScan Preferences"pen Photocan Preferences dialog using corresponding command from the Tools menu$

Set the following values for the parameters on the General tab$

Stereo Mode: Anagl%&h %use 'ardware if your graphic card supports &uad 'uffered Stereo(

Stereo Parallax: !"

Write log to file: s&ecif% director% where Agisoft Photocan log will e stored

%in case of contacting the software supportteam it could be re)uired(

Set the parameters in the O&en$* tab as following$

$hec+ on an% O&en$* devices detected % Photocan

in the dialog and reduce the numer of active $P,

cores % one for each O&en$* device enaled"

Set the following values for the parameters on the

Advanced tab$

Project compression level: -

Keep depth maps: enaled

Store absolute image paths: disaled

Chec for updates on program startup: enaled

!nable "#O support: enaled


2/17

$dd PhotosTo add photos select Add Photos""" command from the .or+flow menu or clic# Add Photos button

located on .or+s&ace toolbar.

*n the Add Photos dialog browse the source folder and select files to be processed. +lic# O&en button.

%oad Camera Positions At this step coordinate system for the future model is set using camera positions.

&ote: *f camera positions are un#nown this step could be s#ipped. The align photos procedure,

however, will ta#e more time in this case.

"pen /eference pane using the corresponding command from the 0iew menu.

+lic# 1m&ort button on the /eference pane toolbar and select the file containing camera positions

information in the O&en dialog.

The easiest way is to load simple characterseparated file %-.txt( that contains x and y coordinates

and height for each camera position %camera orientation data, i.e. pitch, roll and yaw values, could also

be imported, but the data is not obligatory to reference the model(.

*n the 1m&ort $0 dialog indicate the delimiter according to the structure of the file and select the rowto start loading from. ote that / character indicates a commented line that is not counted while

numbering the rows. *ndicate for the program what parameter is specified in each column through setting

correct column numbers in the $olumns section of the dialog. Also it is recommended to specify valid

coordinate system in the corresponding field for the values used for camera centers data.

+hec# your settings in the sample data field in 1m&ort $0 dialog.

+lic# O2 button. The data will be loaded into the /eference pane.


3/17

1m&ort E314 button located on the /eference pane can also be used to load camera positions

information if E0*1 metadata is available.

Then clic# on the ettings button in the /eference pane and in the /eference ettings dialog select

corresponding coordinate system from the list, if you have not selected it in the 1m&ort $0 dialog yet.

Set up $amera Accurac% in meters and degrees according to the measurement accuracy$

Ground Altitude should be specified in case of very obli)ue shooting.

+lic# "2 and camera positions will be mar#ed in Model 0iew using their geographic coordinates$

*f you do not see anything in the odel view, even though valid camera coordinates have been

imported, please chec# that Show +ameras button is pressed on the Toolbar. Then clic# 3eset 4iew button

also located on the Toolbar.


4/17

Check Camera Calibration

"pen Tools enu 5 +amera +alibration window.

'y default PhotoScan estimates intrinsic camera parameters during the camera alignment and optimi6ation

steps based on the *nitial values derived from E0*1. *n case &i5el si6e and focal length %both in mm( are

missing in the image E0*1 and therefore in the camera calibration window, they can be input manually prior

to the processing according to the data derived from the camera and lens specifications.

*f precalibrated camera is used, it is possible to load calibration data in one of the supported formats using

7oad button in the window. To prevent the precalibrated values from being ad8usted by PhotoScan during

processing, it is necessary to chec# on 4i5 $aliration flag.

PhotoScan can process the images ta#en by different cameras in the same pro8ect. *n this case in the left

frame of the $amera $aliration window multiple camera groups will appear, split by default according to

the image resolution, focal length and pixel si6e. +alibration groups may also be split manually if it is

necessary.

*n case ultrawide or fisheye angle lens is used, it is recommended to switch camera type from 4rame

%default( to 4ishe%e value prior to processing.


5/17

$lign PhotosAt this stage PhotoScan finds matching points between overlapping images, estimates camera

position for each photo and builds sparse point cloud model.

Select Align Photos command from the .or+flow menu.

Set the following recommended values for the parameters in the Align Photos dialog$

$ccurac': 'igh %lower accuracy setting can be used to get rough camera positions in a

shorter time(

Pair preselection: /eference %in case camera positions are un#nown Generic preselection

mode should be used(

Constrain features b' mas: Disaled % Enaled in case any areas have been mas#ed prior to

processing(

Ke' point limit: 7 (ie point limit: !

+lic# O2 button to start photo alignment. *n a short period of time %depends on the number of

images in the pro8ect and their resolution( you will get sparse point cloud model shown in the

odel view. +amera positions and orientations are indicated by blue rectangles in the view

window$


6/17

Place Marers ar#ers are used to optimi6e camera positions and orientation data, which allows for better model

referencing results.

To generate accurately georeferenced orthomosaic at least 9: ; 9< ground control points %=+Ps(

should be distributed evenly within the area of interest.

To be able to follow guided mar#er placement approach %which would be faster and easier(, you need

to reconstruct geometry first.Select Build Mesh command from the .or+flow menu and specify following parameters in the Build

Mesh dialog$

+lic# "2 button.

Then, when geometry is built %it usually ta#es a few seconds to reconstruct mesh based on the sparse point cloud(, open a photo where a =+P is visible in Photo 4iew by doubleclic#ing on its icon on the

Photos pane. >oom in to locate the =+P on the photo and place a mar#er in the corresponding point of

the image using $reate Mar+er command from the photo context menu available on rightclic# on

the opened photo in the corresponding position$


7/17

Select the mar#er on the /eference pane. Then filter images in Photos pane using 4ilter %

Mar+ers option in the context menu available by rightclic#ing on the mar#ers label in the .or+s&ace

pane.

ow you need to chec# the mar#er location on every related photo and refine its position if necessary

to provide maximum accuracy. "pen each photo where the created mar#er is visible. >oom in and dragthe mar#er to the correct location while holding left mouse button. 3epeat the described step for every

=+P.

)nput Marer Coordinates


8/17

1inally, import mar#er coordinates from a file. +lic# 1m&ort button on the /eference pane toolbar

and select file containing =+P coordinates data in the O&en dialog. The easiest way is to load simple

characterseparated file %-.txt( that contain mar#ers name, x, y coordinates and height.

*n 1m&ort $0 dialog indicate the delimiter according to the structure of the file and select the row to

start loading from. ote that / character indicates a commented line that is not counted while numbering

the rows. *ndicate for the program what parameter is specified in each column through setting correct

column numbers in the $olumns section of the dialog.

Also it is recommended to specify valid coordinate system in the corresponding field for the values used

for camera center data.

+hec# your settings in the sample data field in 1m&ort $0 dialog$

+lic# O2 button. The data will be loaded into the /eference pane.

Optimi*e Camera $lignment To achieve higher accuracy in calculating camera external and internal parameters and to correct

possible distortion %e.g. ?bowl effect@ and etc.(, optimi6ation procedure should be run. This step is

especially recommended if the ground control point coordinates are #nown almost precisely ; within

several centimeters accuracy %mar#er based optimi6ation procedure(.

+lic# the ettings button in the /eference pane and in the /eference ettings dialog select

corresponding coordinate system from the list according to the =+P coordinates data.

Prior to optimi6ation it is also possible to remove the points with the highest repro8ection error values

using corresponding criterion in Edit enu 5 =radual Selection dialog.


9/17

Set the following values for the parameters in Measurement accurac% section and chec# that valid

coordinate system is selected that corresponds to the system that was used to survey =+Ps$

Marer accurac': "8 %specify value according to the measurement accuracy(.

Scale bar accurac': "!

Projection accurac': "!

(ie point accurac': 7

+lic# "2 button.

"n the /eference pane unchec all photos and chec# on the mar#ers to be used in optimi6ation

procedure. The rest of the mar#ers that are not ta#en into account can serve as validation points to

evaluate the optimi6ation results. *t is recommended since camera coordinates are usually measured with

considerably lower accuracy than =+Ps, also it allows to exclude any possible outliers for camera

positions caused by the onboard =PS device failures.

+lic# O&timi6e button on the /eference pane toolbar.

Select camera parameters you would li#e to optimi6e. +lic# O2 button to start optimi6ation process.


10/17

Set #ounding #ox 'ounding 'ox is used to define the reconstruction area.

'ounding box is resi6able and rotatable with the help of 3esi6e 3egion and 3otate 3egion

tools from the Toolbar.

)mportant: The redcolored side of the bounding box indicates the plane that would be treated as

ground plane and has to be set under the model and parallel to the 0 plane. This is important if mesh

is to be built in Beight 1ield mode, which is reasonable for aerial data processing wor#flow.

#uild +ense Point Cloud'ased on the estimated camera positions the program calculates depth information for each camera to

be combined into a single dense point cloud.

Select Build Dense $loud command from the .or+flow menu.

Set the following recommended values for the parameters in the Build Dense $loud dialog$

,ualit': Medium %higher )uality ta#es )uite a long time and demands morecomputational resources, lower )uality can be used for fast processing(

+epth filtering: Aggressive %if the geometry of the scene to be reconstructed is complex with

numerous small details or untextured surfaces, li#e roofs, it is recommended to set

Mild depth filtering mode, for important features not to be sorted out(


11/17

Points from the dense cloud can be removed with the help of selection tools and Delete9$ro&

instruments located on the Toolbar.

#uild Mesh -optional: can be sipped if pol'gonal model is not re.uired as a finalresult/

After dense point cloud has been reconstructed it is possible to generate polygonal mesh model based

on the dense cloud data.

Select 'uild esh command from the Cor#flow menu.

Set the following recommended values for the parameters in the Build Mesh dialog$

Surface t'pe: 'eight 4ield

Source data: Dense cloud

Pol'gon count: Medium %maximum number of faces in the resulting model. The valuesindicated next to 'igh9Medium9*ow preset labels are based on the number of points in

the dense cloud. +ustom values could be used for more detailed surface

reconstruction(.

)nterpolation: Enaled


12/17

+lic# O2 button to start mesh reconstruction.

!dit 0eometr'Sometimes it is necessary to edit geometry before building texture atlas and exporting the model.

nwanted faces could be removed from the model. 1irstly, you need to indicate the faces to be

deleted using selection tools from the toolbar. Selected areas are highlighted with red color in the odel

4iew. Then, to remove the selection use Delete Selection button on the Toolbar %or Del #ey( or use +rop

Selection button on the Toolbar to remove all but selected faces.

*f the overlap of the original images was not sufficient, it may be re)uired to use $lose 'oles

command from the Tools menu at geometry editing stage to produced holeless model. *n $lose 'oles

dialog select the si6e of the largest hole to be closed %in percentage of the total model si6e(.

PhotoScan tends to produce D models with excessive geometry resolution. ThatFs why it is

recommended to decimate mesh before exporting it to a different editing tool to avoid performance


13/17

decrease of the external program.

To decimate D model select Decimate Mesh""" command from the Tools menu. *n the Decimate

Mesh dialog specify the target number of faces that should remain in the final model. 1or PD1 export

tas# or webviewer upload it is recommended to downsi6e the number of faces to 9:::::G:::::.

+lic# O2 button to start mesh decimation procedure.

#uild (exture -optional1 applicable onl' to pol'gonal models/

This step is not really needed in the orthomosaic export wor#flow, but it might be necessary to inspecta textured model before exporting it or it might be helpful for precise mar#er placement.

Select Build Te5ture command from the .or+flow menu.

Set the following recommended values for the parameters in the Build Te5ture dialog$

Mapping mode: "rthophoto

#lending mode: Mosaic(exture si*e2count: H9IG %width J height of the texture atlas in pixels(

!nable color correction: disaled %the feature is useful for processing of data sets with

extreme brightness variation, but for general case it could be left unchec#ed to save the

processing time(

+lic# O2 button to start texture generation.


14/17

#uild +!M

Digital elevation model can be generated based on the dense cloud or mesh model. sually first

option is preferred, as it provides more accurate results %lowpoly model, being used as a source

data, may result in inaccurate DE( and allows for faster processing, since mesh generation step

can be s#ipped.

Select Build DEM command from the Cor#flow menu$

+oordinate system should be specified in accordance with the system used for the model referencing.

At the export stage it will be possible to pro8ect the results to a different geographical coordinate system.

After DE generation process is finished, it is possible to open the reconstructed model in "rtho view

by doubleclic#ing on the DE label in the Cor#space pane$


15/17

#uild Orthomosaic

Select Build Orthomosaic command from the Cor#flow menu$

Select desired surface for orthomosaic generation process$ mesh or DE, and blending mode.

Pixel si6e will be suggested according to the average ground sampling resolution of the original

images. According to the surface si6e and the input pixel si6e the total si6e of the orthomosaic %in pixels(

will be calculated and shown in the bottom of the dialog box.

=enerated orthomosaic can be reviewed in "rtho mode similar to the digital elevation model. *t can

be opened in this view mode by doubleclic#ing on the orthomosaic label in the Cor#space pane.


16/17

!xport Orthomosaic

Select E5&ort Orthomosaic 5 E5&ort PEG9T1449P;G command from 4ile menu.

Set the following recommended values for the parameters in the E5&ort Orthomosaic dialog$

Projection: Desired coordinate s%stem

Pixel si*e: desired export resolution %please note that for C=SHK coordinate system units

should be specified in degrees. se Metres button to specify the resolution in metres(.

Split in blocs: ! x ! %if the exported area is large it is recommended to enable

&lit in Bloc+s feature, since the memory consumption is rather high at exportingstage(

3egion: set the boundaries of the modelFs part that should be pro8ected and presented as

orthomosaic. Also polygonal shapes drawn in the "rtho view and mar#ed as

boundaries will be ta#en into account for the orthomosaic export.

T*11 compression and LPE= )uality should be specified according to the 8ob re)uirements.

'igT*11 format allows to overcome the T*11 file si6e limit for the large orthomosaics, but it

is not supported by some applications.

+lic# E5&ort""" button and then specify target file name and select type of the exported file

%e.g. =eoT*11(. +lic# ave button to start orthomosaic generation.


17/17

!xport +!M

Select E5&ort DEM 5 Export =eoT*11!'*7!0> command from 4ile menu.

Set the following recommended values for the parameters in the E5&ort DEM dialog$

Projection: Desired coordinate s%stem

&o4data value: value for not visible pointsM should be specified according to the re)uirements

of the post processing application.

Pixel si*e: desired export resolution

Split in blocs: ! x ! %if the exported area is large, it is recommended to enable

&lit in loc+s feature, since the memory consumption is rather high at exporting

stage(

3egion: set the boundaries of the modelFs part that should be pro8ected and presented asDE. Also polygonal shapes drawn in the "rtho view and mar#ed as boundaries will

be ta#en into account for the DE export.

+lic# E5&ort""" button and then specify target file name and select type of the exported file

%e.g. =eoT*11(. +lic# ave button to start DE generation.

ps_1.2 -tutorial (bl) - orthophoto, dem (with gcps)

Documents