Design of intelligent rain-polluted pipeline inspection robot based on embedded vision

The structure and size of urban underground sewage pipes are varied and the environment inside them is complex and changeable. The local damage and blockage will threaten people's lives and property safety and cause disastrous consequences. Traditional artificial pipeline inspection is time-consuming, laborious, and prone to misdetection and missing detection. Therefore, it is of great significance to design an intelligent pipeline robot based on deep learning to carry out inspection of pipeline visual defects and effectively prevent pipeline lesions. In this paper, the hardware structure of the wheeled robot with amphibious crawling ability is designed first, and the robot's motion control under complex conditions in the tube is realized. Then, the detection algorithm based on deep learning is carried out to complete the detection and type identification of the pipeline. Deploy the trained deep learning model to the Jetson Nano embedded device while accelerating the model's reasoning process using TensorRT. The test results show that the inspection robot can effectively identify the pipeline defects and the recognition rate of the type of pipeline defects reaches 88.2%, which confirms the feasibility and effectiveness of the proposed technology.