Identification of apple diseases in digital images by using the Gaining-sharing knowledge-based algorithm for multilevel thresholding

Identifying the defects in apples is commonly done with visual examination techniques. However, it is a slow and laborious process. Image processing techniques have begun to be used to help and make the diagnosis of fruit diseases more efficient. In image processing systems, the segmentation of regions in the scenes is a crucial step. Specifically for images from apples, disease segmentation is a complicated task due to the different elements that affect the acquisition of the images. In addition, apple diseases also have features that need to be segmented. In this work, an efficient approach that uses the Gaining-sharing Knowledge-based (GSK) algorithm is proposed to optimize the minimum cross-entropy thresholding (MCET) for the segmentation of apple images highlighting the diseases defects. The proposed MCET-GSK has been tested for experimental purposes over different images and compared with various metaheuristics. The experiments were conducted to provide evidence of the GSK’s optimization capabilities by performing the Wilcoxon test and applying a set of metrics to verify the quality of the segmented images. The experimental results validate the performance of the MCET-GSK in the segmentation of apple images by adequately separating the regions with damage produced by a disease. The quality of the segmentation is superior compared with other similar approaches.