A study for measuring the sunlit areas of onions and weeds in the field by image processing

摘要

Measurement and calculation of sunlit areas~(1) of leaves allows evaluating the developments of onions and weeds in the field and studying the co-concurrence between crop and weeds,. This is traditionally done by hand, in the field, with a rule. The operation is very laborious. This paper introduces another method to achieve this work using image processing. Color pictures are taken in the field under sun illumination, at short distance from the ground and are digitalized. Color index of onions, weeds and soil are tested using diffeent color spaces. HSI space allows the best distinction. Two algorithms were developed using this color space. The first algorithm is based on shape recognition after image segmentation. The second algorithm is based on classification between onions and weeds using color indexes and adjacent region criteria. A combination between the criterion of minimum distance and adjacent region allows determining whether a pixel of vegetation belongs to onions or weeds. After classification, the percentages of sunlit areas occupied by onions and by weeds are calculated. The performances of these two algorithms and their limits are discussed.