Method of Peanut Pod Quality Detection Based on Improved ResNet

Peanuts are prone to insect damage, breakage, germination, mildew, and other defects, which makes the quality of peanuts uneven. The difference in peanut pod price economic benefit also have a big difference. classification pods according is an important part improving product grade market competitiveness. Real-time, accurate, non-destructive detection can effectively improve utilization commercial value peanuts. strong subjectivity manual low efficiency accuracy mechanical caused considerable wastage. Therefore, present study proposed new convolutional neural network for algorithm (PQDA) based on improved ResNet. Compared previous models, this model more practical with high accuracy, lightweight, easy nesting. Firstly, effects ResNet18, AlexNet, VGG16 compared, ResNet18 was determined be best backbone feature extraction training. Secondly, three models were designed optimize algorithm. KRSNet module added make lightweight. CSPNet learning each layer. Convolutional Block Attention Module (CBAM) its ability capture information about pods. experimental ablation results show that precision PQDA reaches 98.1%, size parameters only 32.63 M. Finally, optimized applied varieties generalization experiments, reached 89.6% 90.0%, indicating effectiveness model. Furthermore, suitable deployment embedded resource-limited devices, such as mobile terminals, achieve real-time accurate quality.