Aiming at the problems of poor image quality, susceptibility to interference from the external environment and difficulties in recognition due to high similarity between real defects and fruit stalks in citrus surface defect recognition, we proposed a citrus surface defect recognition method based on a combination of PCSA-2D-Otsu and CGWO-DT-SVM. Firstly, a partial differential equations (PDE) based variational model is used to denoise the captured citrus images, which reduces the blurring of the images while retaining the edge details and impo...
