Green Plums Surface Defect Detection Based on Deep Learning Methods

Green plums are a characteristic fruit resource in China, with long history of cultivation. Many surface defects will appear the growth, transportation and preservation green which seriously affect processing quality by-products. The existing manual sorting method is limited by experience workers. It difficult to ensure speed detection. Therefore, formation automatic detection great significance development plum industry. According plums, this paper divides into five categories: rot, cracks, scars, spots normal. A total 1235 images were obtained self-built image acquisition device. WideResNet50-AdamW-Wce model based on WideResNet was built classify plums. Accuracy, recall F1-measure selected as indexes evaluate accuracy classification. classification reached 98.95 %, rain spots, normal, rot crack 100 99.56 98.59 98.25 % 96.10 respectively. Comparing performance ResNet50-SGD, WideResNet50-SGD, WideResNet50-SGD-Wce WideResNet50-AdamW network models, F1-Measure highest each defect, more greengage defect features can be learned. results meet production needs deep enterprises – evaluating 1800 per hour assembly line.