An image segmentation method for tomatoes under different lighting conditions

This study analyzes the image segmentation method based on a piecewise idea combined with a recognition algorithm for identifying regions of bright tomato-red color as a viable methodology to decrease the overlapping of color between tomatoes and backgrounds under varying illumination conditions and to realize the recognition of bright tomato-red regions. Decreasing this overlapping of color will help achieve image segmentation under different lighting conditions. Firstly, me concept of illumination prediction using R component in color images was substantiated. Secondly, the two-color features of tomatoes were discussed separately under different lighting conditions. Thirdly, the piecewise segmentation method was designed using these two color features in different R component ranges. However, this method could not successfully separate the bright tomato-red regions from backgrounds, especially under direct lighting conditions. Therefore, we adopted a segmentation method for the bright tomato-red regions. Finally, a recognition algorithm for bright tomato-red regions was employed to separate bright tomato-red regions from bright background regions. Out of the test set of 481 images captured under different lighting conditions, 97 background regions were falsely considered as tomato regions. The average execution time was observed to be 71 ms. Hence, this method can be considered effective in realizing tomato image segmentation under different lighting conditions and recognizing the bright tomato-red regions. Especially under direct illumination conditions, the overall segmentation performance of this method was found to be better than that of the fixed threshold segmentation method, which is based on the normalized color difference. Future studies are necessary to overcome the segmentation problems caused by the color overlapping between tomatoes and complex backgrounds.