Non-Destructive Internal Defect Detection of In-Shell Walnuts by X-ray Technology Based on Improved Faster R-CNN

The purpose of this study was to achieve non-destructive detection the internal defects in-shell walnuts using X-ray radiography technology based on improved Faster R-CNN network model. First, FPN structure added feature-extraction layer extract richer image information. Then, ROI Align used instead Pooling for eliminating localization bias problem caused by quantization operation. Finally, Softer-NMS module introduced final regression with predicted bounding box improving accuracy candidate boxes. results indicated that proposed model can effectively identify walnuts. Specifically, discrimination accuracies sound, shriveled, and empty-shell were 96.14%, 91.72%, 94.80%, respectively, highest overall 94.22%. Compared original model, achieved an increase 5.86% in mAP 5.65% F1-value. Consequently, method be applied shriveled defects.