Double-mode concealed pipe detection robot based on machine vision and pressure sensors

Abstract Soil salinization is a global issue, and one of the primary methods for improving saline-alkali soil is concealed pipe salt drainage technology. To address the concealed pipe system inspection challenges, this paper proposes a novel double-mode concealed pipe detection robot based on machine vision and pressure sensors. Initially, the 3D mechanical structures are constructed and printed using Rhino7 and subsequently assembled to form a complete pipe robot. The camera integrated in the front of the robot is utilized to capture real-time video of the interior of the pipeline, which are subsequently subjected to image classification utilizing a pipeline recognition network based on knowledge distillation(KD). This paper uses the Resent101 teacher model to guide ResNet18 student model training, which can accurately classify pipeline conditions while reducing computational requirements.Additionally, the paper presents a pipe-diameter deformation detecting device based on an adaptive flexible linkage (PDDD-AFL) to detect the pressure value of the pipe wall and provide the pipe deformation feedback. Experimental tests demonstrate that the robot can use the visual monitoring and deformation feedback to realize the double-mode intelligent analysis of concealed pipe. Such an approach holds significant potential in achieving fine performance, quantitative description, and qualitative analysis of pipeline systems, and realizing the automation and intelligence of concealed pipe salt drainage.