A Spiral-Driven In-Pipe Detection Robot with RGB-D Recognition and Quantification Algorithm for Thermal Bulge Inspection

Abstract

To address the limitations of conventional pipeline hot-spot detection methods-such as low efficiency, poor accuracy, and high costs-this paper proposes an autonomous mobile pipeline inspection robot system based on the Convolutional Block Attention Module (CBAM). The system integrates a stereo camera, an inertial measurement unit (IMU), and brushless motors, and employs an improved wheeled drive mechanism combined with a helical transmission structure to achieve autonomous navigation inside pipelines. By developing an RGB-D fusion-based image recognition algorithm and leveraging disparity calculation principles, the system accurately quantifies the proportion of hot spots relative to the total pipeline length, thereby evaluating their impact on subsequent dismantling operations. Experimental results demonstrate that the proposed system achieves a hot-spot detection accuracy of 96.7% and a localization error of less than 3 cm, significantly enhancing both the automation level and precision of pipeline inspection. This work provides an advanced technical solution to longstanding challenges in pipeline maintenance.