Third-party construction intrusion detection of natural gas pipelines based on improved YOLOv5

With the continuous development of unmanned aerial vehicle (UAV) detection technology, UAV line detection has been well applied in the oil and gas storage and transportation industry. Regularly patrol the high-risk areas through UAV, and monitor whether there are third-party intrusion events near oil and gas pipelines, including construction machineries such as excavators and bulldozers. When the third-party intrusion event occurs near the pipeline, an alarm will be issued, and the line patrol personnel will investigate in time. In this paper, the single-stage target detection algorithm based on deep learning is combined with the UAV patrol line of natural gas pipelines, and an algorithm model based on improved YOLOv5 is proposed. The Recurrent Criss-Cross Attention module is added to the backbone network to improve the feature extraction capability. The feature fusion module uses the TridentNet network to improve the performance of the multi-scale feature fusion of the model. WBF post-processing is used to replace the original NMS, improving the prediction algorithm's regression accuracy. Experiments show that the improvement has effectively improved the detection effect of third-party intrusion of natural gas pipelines.