Research On Defects Detection Technology Of Pipe Internal Surface Based On Active Stereo Omnidirectional Vision

Pipeline transportation is widely used in many fields,such as chemical industry,energy supply,and plumbing supply.However,due to the erosion of external force and transport material and other reasons,the inner surface of the pipeline often appear in a variety of diseases,such as cracks,corrosion,exfoliation.To ensure the safety and the serviceability of pipeline infrastructure it is essential to inspect and assess its physical and functional condition.The two key problems to be solved in pipeline defects detection are the fast acquisition of the inner wall texture images and the feature extraction and recognition.Manual inspection has been the primary method of pipeline defects detection,which is highly dependent on the operator’s interpretation and assessment.In order to solve the issues mentioned above,a defect detection system based on active stereo omnidirectional vision is designed and implemented in this thesis.The main research work and achievements of this thesis are as follows:1.An active stereo omnidirectional vision detection device is designed to acquire a 360 degree panoramic wall texture information,avoiding the traditional camera multi angle shooting and image splicing problem.2.In order to eliminate the image distortion caused by panoramic imaging,the expansion algorithm is used for panoramic images,and the image preprocessing algorithms are used for image noise reduction,brightness and contrast adjustment as well as the suspected defect segmentation and extraction.3.Four of the most common defects such as cracks,corrosion,root,and branch crossing the pipe are detected,and two kinds of defect recognition algorithms were used.Firstly,this thesis uses the recognition algorithm based on the geometric features to calculate the area,perimeter,roundness,convexity,angle and other geometric parameters of the defect area,then determine the categories of defects.In order to solve the difficulty of the above methods,this thesis uses the CNN to extract and recognize the pipeline defects,and improve the recognition accuracy and speed.Finally,in order to achieve one-time position detection and species identification of defects,Faster R-CNN method is adopted.Panoramic images are directly used as the input of network,the detection accuracy is improved,while highlighting the advantages of panoramic imaging system.In this thesis,the pipeline defect detection system is designed and realized,and the function modules of the system are described in detail.The experimental results show that the pipeline defects detection system designed in this thesis can realize the collection,processing,defect detection and identification automatically,achieving good results in recognition accuracy and inspection speed.