A MANGO DEFECT DETECTION METHOD BASED ON MULTI-SCALE FEATURE DENSITY ESTIMATION

• Main Authors: Longhao Zhang, Yunfeng Ma, Renhuan Wang, Hongzhen Cui, Shichao Wang
• Format: Patent
• Language: English
• Published: 28.08.2024

A mango defect detection method based on multi-scale feature density estimation belongs to the field of fruit detection technology, and is used to solve the problem of mango defect detection. The appearance image set of mangoes was collected, and the collected data were de-background processed and classified to obtain the image set of mango without background and the image set of mango with defect. The defect location of the image of mango with defect was marked to obtain the defect information of mango with defect. Part of the defect-free mango image set is used as the training data set, and the other defect-free mango image set and defective mango image set are used as the test data set. The mango defect detection model based on multi-scale feature density estimation is trained with the defect-free mango image set, and then the mango appearance image data to be detected after background processing is input. The defect results and defect location markers were obtained, and the multi-scale mango external defects were detected and the location information was introduced to improve the detection efficiency and the accuracy of the defect location markers.