Contraband Detection Based on YOLOv7-Tiny X-Ray Image

In order to solve the problems of high stacking of items and difficult to locate in X-ray contraband detection, a contraband target detection algorithm based on YOLOv7-tiny was proposed. Firstly, the SPPMFCSPC module was designed to perform multi-scale spatial pyramid pooling on the input feature maps to obtain receptive fields of different scales. In addition, an efficient multi-scale attention mechanism for cross-spatial learning is introduced into the Neck part of the network to strengthen the correlation between each channel. Secondly, the R-ELAN structure is designed in the backbone network to enhance the network’s ability to fuse features. Finally, the Neck part of the network adopts a progressive pyramid network to avoid a large semantic gap in the network. Experimental results show that the improved algorithm achieves 90.3%, 83.1% and 84.0% detection accuracy on SIXray, PIDray and CLCXray contraband datasets, respectively, which effectively improves the accuracy of contraband detection.