Apple image segmentation using teacher learner based optimization based minimum cross entropy thresholding

Fruit image segmentation is the primary phase during fruit image analysis to develop an artificial intelligence system for fruit classification. In this paper, apple images are considered for segmentation using the concept of teaching learning strategy. In the proposed approach, firstly cross entropy based objective function is designed and then teacher leaner based optimization algorithm is applied to minimize the objective function for finding optimal threshold values at the different levels. Selected threshold values by the proposed approach are used to segment red, green and golden apple images. The proposed approach is called TLBO-MCET. The proposed approach is inspired by teaching learning philosophy, where students learn from teacher in the classroom and from each other mutually. For performance evaluation, PSNR and uniformity measures are used. The results of proposed approach are compared with GA-MCET and HBMO-MCET. From simulation and experimental works, it has been observed that the performance of proposed approach is quite promising. In future, the proposed work will be used for automatic grading of different varieties of apple.