Estimating the content of clover and grass in the sward using a consumer camera and image processing

1
Estimating the content of clover and grass in the
sward using a consumer camera and image processing
A. K. Mortensen1, H. Karstoft1, K. Søegaard2 and
R. N. Jørgensen1 1Department of Engineering,
Aarhus University 2Department of Agroecology,
Aarhus University
2
Outline

• Motivation
• Test bed and data acquisition
• Methodology
• Results
• Conclusion and future work

3
Motivation

• Clover and grass is sown as a catch crop and as a
  feed crop
• However, it is unknown how the distribution is in
  the field
• The distribution and total dry matter determines
• the N-uptake in the field
• As feed for dairy cows
• intake
• milk yield
• Current method is destructive analysis
• Labour intensive
• Cumbersome

4
Test bed and data acquisition

• Plot experiment (2013) near Research Center
  Foulum, Denmark
• Two different seed mixtures
• Blanding 35 perennial ryegrass and white
  clover
• Blanding 45 perennial ryegrass, festulolium,
  red clover and white clover
• Photographed and cut on 4 different occasions
• Cuts made within 3 days after photograph
• Photographed from above
• 45 images
• Dry matter analysis in laboratory
• Dry matter 195 kg/ha ? 6111 kg/ha
• Clover 10-72
• Grass 26-90
• Weed 0-3

5
Test bed and data acquisition
6
Test bed and data acquisition
7
Methodology
8
Methodology

• Illumination classification
• Direct and indirect sun light
• Histogram of pixel intensities
• Cross correlation used for classification

9
Methodology

• Coverage estimation
• Remove background
• Soil, dead plant material and deep shadows
• Extract clover leafs
• Inverted edge image erosion
• Grass
• Remaining
• Trained on patches

10
MEthodology

• Transformation of coverage to dry matter
  distribution

11
Results

• Coverage estimation

  Direct light Direct light Direct light Direct light Direct light Direct light Direct light     Indirect light Indirect light Indirect light Indirect light Indirect light Indirect light Indirect light
  Sh So G C F W ?     Sh So G C F W ?
Sh 66 1 25 8 0 0 2955   Sh 52 11 29 9 0 0 3934
So 0 0 0 0 0 0 0   So 11 87 2 1 0 0 2422
G 9 0 73 17 0 0 4745   G 11 1 65 23 0 0 6757
C 13 0 39 48 0 0 6565   C 3 0 29 68 0 0 13491
F 30 20 46 4 0 0 254   F 18 0 77 5 0 0 79
W 24 0 71 4 0 0 181   W 7 0 52 41 0 0 617
? 3361 98 6986 4255 0 0 -   ? 3541 2654 9795 11310 0 0 -
12
Results

• Dry matter distribution
• No correlation between error and mixture, clover
  dry matter or total dry matter.

  Error (-points) Error (-points) Error (-points) Error (-points)   Absolute error (-points) Absolute error (-points) Absolute error (-points) Absolute error (-points)
  Mean Std. Max Min   Mean Std. Max Min
Test set Test set Test set Test set Test set          
Clover -2.1 9.8 19.1 -15.6   7.9 5.8 19.1 0.0
Grass 2.6 10.4 17.4 -18.7   8.8 5.7 18.7 1.3
Training set Training set Training set Training set Training set          
Clover 0.0 12.9 18.9 -50.3   8.6 9.4 50.3 0.2
Grass 0.0 13.5 51.0 -20.7   9.2 9.7 51.0 0.2
13
Conclusion and future work

• It is possible with a reasonable accuracy
  (8-9-points)
• Greatest source of error
• Coverage estimation
• Room for improvement
• Better estimation of coverage
• Texture analysis
• Illumination invariant model
• Include growth models
• Time since last harvest
• Temperature sum
• Available water

14
(No Transcript)