Determination of crop soil cover by means of digital colour photos and automatic computer image analysis
How quickly are plants growing? The soil cover is an important parameter in plant growing experiments. But visual assessment of the soil cover by experts is inaccurate and subjective.
The soil cover is an important plant growth parameter. The more the plants cover the soil, the more leaf area may effectively used for photosynthesis and the better the soil is protected against evaporation and erosion. Visual assessment by experts is possible but inaccurate and subjective. In the project, a standardized robust procedure for the objective accurate soil cover determination will be developed.
In the project we first want do develop a practical method for people with plant growth experiments. Today in most experiments soil cover is still estimated visually by experts. The project will give further insight into problems and possible solutions for the automatic segmentation of colour photos into plant and soil for various other automation tasks.
See "Data and methods".
In the project many photos of various agricultural crops were already taken under partially standardized illumination conditions. A knowledge based classification algorithm will be implemented and trained that will be able to evaluate new photos in the future without new training.
Preliminary results show a high segmentation quality under optimal conditions. Under certain adverse conditions, considerable errors are possible that cannot be prevented with the selected approach. The results shall be improved by further training, and the performance of the method under various conditions will be evaluated and documented.
1.2004 - 12.2016