using ultrasonic sensors for on-line weed detection · success) and broadleaf weeds (datura sp....
TRANSCRIPT
Detection systemDetection system
Using ultrasonic sensors for on-line weed detectionUsing ultrasonic sensors for on-line weed detectionD. Andújar 1, A Escolà-Agustí 2, C. Fernández-Quintanilla 1, J. Dorado 1
1 Instituto de Ciencias Agrarias, CSIC, Serrano 115B, 28006 Madrid, Spain2 Universitat de Lleida, Av. Rovira Roure 191, 25195 Lleida, Spain
ObjectiveObjective
ResultsResults
• Two maize fields, assessed at two dates (May 6 and May 25, 2010)
• Fields infested with one grass species (Sorghum halepense), and various broadleaf weeds (Datura ferox, Datura stramonium and Xanthium strumarium)
• In each date, static ultrasonic readings were taken in 180 points with different weed composition of pure and mixed grasses and/orbroadleaves. The system was also tested to acquire dynamic ultrasonic readings
• The validation of the ultrasound readings were performed in all the 180 static points as well as in 87 points for dynamic mapping validation, using plant height and weed biomass
• Two maize fields, assessed at two dates (May 6 and May 25, 2010)
• Fields infested with one grass species (Sorghum halepense), and various broadleaf weeds (Datura ferox, Datura stramonium and Xanthium strumarium)
• In each date, static ultrasonic readings were taken in 180 points with different weed composition of pure and mixed grasses and/orbroadleaves. The system was also tested to acquire dynamic ultrasonic readings
• The validation of the ultrasound readings were performed in all the 180 static points as well as in 87 points for dynamic mapping validation, using plant height and weed biomass
ConclusionsConclusions
The use of ultrasonic sensor readings proved useful to discriminate grasses (Sorghum halepense) (up to 81.1% of success) and broadleaf weeds (Datura sp. and/or Xanthium strumarium) (up to 98.5% of success)The use of ultrasonic sensor readings proved useful to discriminate grasses (Sorghum halepense) (up to 81.1% of success) and broadleaf weeds (Datura sp. and/or Xanthium strumarium) (up to 98.5% of success)
Experimental setupExperimental setup
To assess the use of ultrasonic sensors to detect weeds growing between crop lines, exploring their capabilities and limitations to discriminate weed species of different heights To assess the use of ultrasonic sensors to detect weeds growing between crop lines, exploring their capabilities and limitations to discriminate weed species of different heights
Grasses: dominance of Sorghum halepense(about 40 cm high)
Broad-leaves: dominance of
Datura stramonium (about 20 cm high)
Mixture of grasses and broad-leaves
weeds
GPS antenna
GPS antenna
Wire mesh to avoid crop
interference
Wire mesh to avoid crop
interference
Ultrasonic sensor
Ultrasonic sensor
Canonical Discrimination Analysis of ultrasonic readings, separating three groups depending on the type of weeds
Canonical Discrimination Analysis of ultrasonic readings, separating three groups depending on the type of weeds
5
Func
tion
2
Function 12.50-2.5
2.5
0
-2.5
GrassesMixtureMixtureBroad-leaves
Percentage of correct predictions using ultrasonic sensor to discriminate grasses from broadleaf weeds (2 groups) and considering also a mixture (three groups), on two different dates with static and dynamic readings
Percentage of correct predictions using ultrasonic sensor to discriminate grasses from broadleaf weeds (2 groups) and considering also a mixture (three groups), on two different dates with static and dynamic readings
Two groups Three groups
Dominance of grasses
Dominance of broad-leaves
Dominance of grasses
Mixture Dominance of broad-leaves
Static date1 81.1 98.5 57.5 (17.5*) 38.2 (20.0*) 97.0 (0.0**)
Static date 2 79.6 97.0 71.4 (7.1*) 46.2 (26.9*) 93.9 (0.0**)
Dynamic 77.8 97.0 71.4 (10.7*) 53.8 (23.1*) 87.9 (0.0**)
(*) % of cases misclassified as “broadleaf weeds”; (**) % of cases misclassified as “grasses”