md_Destructive_LAI

Code for destructive LAI measurement

0. install necessary libraries

pip install -r requirements.txt

1. Print the calibration file and scan before each leaf scanning You can modify the markers_list.json to customise the cultivars scanned.

cd markers
python generate_markers.py -o <output dir> -j <json filepath>

2. Prepare a directory with the leaf scans, and a separate directory for the calibration scans. See ./data/leaves and ./data/calibration for an example.
3. Set the area to crop out in data/calibration.yaml

python crop_and_threshold.py -d ./data/leaves -o ./data

3. Run calibration with the calibration dir to create calib.txt

python calibration_board_one_dir.py \
       -c ./data/calibration \
       -o ./data \
       --calib_fp data/calibration.yaml

4. (optional) Run vegetation masks. You can manually correct this mask if required.

mkdir ./data/vis_vm
python generate_veg_mask.py \
      --img_dir ./data/leaves \
      --calib_fp ./data/calibration.yaml \
      --out_dir ./data/vis_vm

5. Process leaves images to obtain leaf areas

python process_images.py -d data/leaves --is_vis --vm_dir ./data/vis_vm/labeller/semantics

You may need to adjust the tracker configuration in ./data/tracker_hyps.cfg to match your setup. By adding the --is_vis flag, you can check the visualisation of the tracker in ./data/test_dbg and ./data/vis_frame.

6. Check the leaf area csv (la.csv) We encode further information in the csv as follows:

• stddev=-1 means long leaf, so no repeated measurements for stddev calculations
• stddev=-2 means the leaf could not be fully seen (e.g. was placed too high or low in camera FOV)