Research on image segmentation methods of tomato in natural conditions

It is difficult for image segmentation of harvesting objects in natural conditions because of the large range of illuminance variation. In order to improve the performance of adaptation to illuminace variation of image segmentation algorithm, three image segmentation algorithms are researched and comparatively analyzed: segmentation algorithm based on R-G, segmentation algorithm based on normalized R-G, and segmentation algorithm based on band ratio. Experiments showed that segmentation method based on R-G could not meet the need of image segmentation of tomato in natural lighting conditions, but segmentation method based on normalized R-G and band ratio could. Light spot of tomato region couldn't be recognized in front lighting conditions using the second method. In the third method, light spot's recognition could be realized, although error segmentation would happen in some conditions. The segmentation efficiency of method based on normalization chromatic aberration was better than others in weak lighting conditions, the method based on multiband ratio was better than others in front lighting conditions.