Research On The Key Technology Of Vision System For Pipeline Patrol Robot

Pipeline inspection robot is one of the key equipment in pipeline operation.It is an important technical equipment to realize intelligent pipeline inspection,reduce serious pipeline accidents and improve pipeline labor conditions.At present,the focus of the technology research of pipeline inspection robot is unmanned,automation and intelligence,and the vision system of pipeline inspection robot has become the key to realize robot automation control,and based on this,the vision system of pipeline inspection robot has become the key to realize automatic control of robot.In this paper,the vision system of robot in pipeline inspection process can not effectively detect the siltation inside the pipeline,and can not accurately locate the location information of the siltation inside the pipeline and so on.Based on the laboratory single-drive bi-directional peristaltic pipeline inspection robot,according to the working process and working environment characteristics of pipeline inspection,the visual system of pipeline inspection robot is built according to the research object and content of this subject.Mainly includes the visual system hardware selection and software design.Based on the theory of deep learning,this paper simulates the siltation in pipeline environment,establishes the image database of pipeline siltation,and uses the deep learning convolution network to study and train the siltation image.Aiming at the problems existing in the detection of silted objects,an optimized YOLO detection model is proposed.Compared with the traditional YOLO model,the training time of the network is greatly reduced under the condition that the number of iterations is equal to that of the traditional model.Secondly,it improves the ability of identifying the details of silts,which greatly improves the problem that small-size silts can not be detected in real-time by large-scale silts.Finally,the problem of low detection accuracy of small-scale silts is solved effectively.Through The experimental results show that the detection accuracy of the new model can reach 94.3%,and it can be effectively applied to the vision system of pipeline inspection robot.On the basis of the vision system can effectively detect the pipeline siltation,this paper designs the monocular vision ranging and positioning system of pipeline inspection robot.Based on monocular vision theory,the pipeline siltation is established.Coordinate transformation relationship between industrial camera and pipeline inspection robot.Based on Zhang Zhengyou calibration method,the selected industrial camera is calibrated by the calibration toolbox of MATLAB,and the initial parameters of monocular camera are obtained.The initial parameters of monocular camera are optimized by simulated annealing algorithm.Finally,a distance location model based on ground plane constraint is proposed.The experimental results show that the model is effective for pipeline deposition.The location error is less than 3 cm,which provides a new scheme for vision system of pipeline inspection robot.