Research On Pipeline Detection Technology Based On Pipeline Robot

Pipeline transportation is widely used in all walks of life.Pipeline transportation has now become one of the most important ways for the transportation of liquid or gaseous materials such as oil,natural gas,domestic water and domestic sewage.In order to ensure smoothness and safety of pipeline transportation,pipeline must be detected.At the same time,a large number of pipeline detection work has created the huge demand of the pipeline robot,but the traditional detection methods of artificial observation image or video collected by pipeline robot can not meet the growing detection task.Therefore,the design and research of pipeline robot is of great significance for the exploration and experiment of pipeline inspection technology based on machine vision.The main contents are as follows:(1)Based on the internal environment of the pipeline and the requirements of the function and design of the pipeline robot,the influence of the overall shape size of the pipe robot on the passage of the bend was analyzed.The geometric constraint conditions and the dynamic constraints of the robot through the bend were proposed,and the overall structure and control system of the pipeline robot was designed.and analyzed.A design scheme of tracked pipe robot with simple structure,high flexibility and high tractive force was put forward.(2)Image processing was done for images collected from pipelines.The main characteristics of the pipeline were studied.The main research targets were the three typical features of the pipe joint,the pipe lateral and the pipe crack,and they were divided into two types according to the morphological characteristics: the irregular shape with obvious morphological features and no obvious morphological features,the algorithm of pipeline image processing and feature extraction was studied.Aiming at the noise in original image,median filtering and gray transformation were applied to de noise and image enhancement of pipeline grayscale image.For the pre-processed images,different two processing algorithms were used in accordance with different categories: mathematical morphology based pipeline joint and pipeline lateral image processing algorithms and statistical characteristics based pipeline crack image processing algorithms.Geometric parameters were extracted for the extracted regions.We used chain code to track each region's edge and extract the edge pixels and the chain code direction.The roundness parameter and flatness parameter were obtained by calculating area perimeter,area and moment of inertia,which provide a basis for subsequent classification and recognition.(3)The identification method of pipeline characteristics was studied.In view of the characteristics of pipeline lateral,pipe joint and pipeline crack defect,a method based on support vector machine algorithm was proposed to identify the type of defect feature.The influence of normalization method,kernel function selection on recognition accuracy and the selection and optimization of the penalty parameter C and kernel parameter g were studied,and the feature parameters of the principal component analysis were applied to the training of support vector machine classifier to improve the classification accuracy.