Research on improving YOLOX foreign body identification method under freight train

To address the problem of high workload and low accuracy and speed in manual car bottom object detection, this paper proposes an improved YOLOX cargo train car bottom object detection method. The attention mechanism CBAM (Convolutional Block Attention Module) and adaptive spatial feature fusion module ASFF (Adaptively Spatial Feature Fusion) are introduced into the neck network to enhance the feature extraction and fusion ability of the input feature map, thereby improving the overall detection effect of the model. The high-sensitivity detection network DCBS (Dilated Convolutions Batch normal and Silu) is introduced into the detection network to enhance the ability to utilize the target feature information. The experimental results show that the proposed model achieves an average precision of 80.6% on the car bottom object detection dataset, which is higher than other mainstream comparison models and meets the technical requirements for car bottom object detection.