Stack Workpieces Recognition Model Based on Deep Learning

Abstract

The detection and recognition of stacked workpieces is affected by workpiece occlusion and workpiece overlap, which leads to the problem of difficult detection of workpiece types. This paper proposes a detection method based on the improved Faster R-CNN model, improves the Faster R-CNN feature network, and selects ResNet combined with SENet for feature extraction, which improves the important feature layer and suppresses the non-important feature layer. Introduce the Soft-NMS algorithm to optimize the NMS algorithm to reduce the problem of missed detection and false detection of overlapping or adjacent targets. The test results show that compared with the unimproved Faster R-CNN model, the improved Faster R-CNN model outperforms the traditional algorithm in terms of accuracy, precision, recall and F1 value.