Preliminary study for weed biomass prediction combining visible images with a plant-growth model

International audienceThis preliminary study was designed to test an image approach (RGB images) to estimate crop and weed aerial biomass (BM) and predict their growths. From an image processing, based on crop/weed discrimination, a calibration was done between photographed leaf area (PLA) and BM deduced from a standard (destructive) approach. Then, a plant growth model component (AZODYN) was used with PLA as input, considering wheat and weeds separately. For wheat, the results confirm that to initialize the model, the image approach may replace the standard one with no impact on the values of predicted BM whereas weed results are more mixed