Abstract: The high consequence areas of long-distance oil and gas pipelines are usually provided with technical defense by video surveillance to identify the third-party damage events such as manual excavation, mechanical excavation and heavy vehicles rolling around the pipelines in high consequence areas. However, video surveillance often relies on the screen monitoring of the personnel on duty, which has insufficient effectiveness. For this reason, an intelligent identification method of third-party damage in high consequence areas of the pipeline based on depth learning was proposed. Specifically, the collected images and videos along the line were analyzed, the feature targets were extracted, and an image intelligent identification model based on YOLO v5 was established. Generally, the model improves the optimization speed and target detection accuracy. After 226 iterations, the model training has the process loss function values of training and validation close to 0 and 0.01, respectively, reaching the optimal state. The research method was tested in the video surveillance of a pipeline section in a high consequence area in Tianjin, with a detection accuracy of up to 99.33%, which verified the effectiveness of the method and provided engineering application reference for the subsequent intelligent identification and real-time early warning of the video surveillance system in the high consequence areas.