Study On Underwater Weld Detection Technology Based On Visual Sensing

With the acceleration of the construction of water and hydropower projects and marine projects,the corresponding underwater welding technology has also received more and more attention.There are some problems in traditional diving welding,such as low efficiency and difficult to guarantee welding quality.The combination of vision sensing technology and underwater vehicle technology to realize underwater automation and intelligent welding is one of the effective ways to solve the above problems.Because the underwater imaging environment is different from that on the land,the traditional visual detection method of the welding seam on the land can not meet the practical application requirements,so it is very important to study the underwater welding seam detection technology based on visual sensing.In order to realize the automatic detection and recognition of underwater weld seam,this paper discusses the imaging of underwater camera and the calibration of underwater camera,and focuses on the processing methods of each link in the process of underwater weld image processing algorithm.The main research work is as follows:According to the research requirement of underwater weld detection,an underwater visual inspection system was constructed,the components of the system were selected,and the phase secret encapsulation device was designed.According to the principle of camera aperture imaging,the transformation relationship between imaging coordinates is established,and the underwater imaging characteristics of the camera are analyzed,and the imaging model of underwater non-linear camera is established.At the same time,aiming at the problem of underwater camera calibration,a focal length change compensation method is proposed to deal with the influence of water and lens medium.Zhang Zhengyou calibration algorithm is used to calibrate the camera.The results show that the calibration error is less than 0.06 pixels and the accuracy is high.Aiming at the low contrast of underwater weld image and the unclear feature information,an underwater weld image enhancement algorithm based on edge information fusion is proposed.The algorithm enhances the original image and the high frequency component of the image respectively.Aiming at the problem that underwater weld image contains a lot of mixed noise,this paper analyses the denoising effect of different filtering algorithms,and proposes a mixed filtering method,which is bilateral first and then median,to ensure the denoising effect while retaining the image edge information to the greatest extent.Otsu algorithm is used to separate the region of interest of the image.To overcome the shortcomings of the algorithm,an improved Otsu algorithm is proposed to improve the segmentation effect of the histogram with single peak distribution.By comparing and analyzing the effects of various edge detection algorithms on seam edge feature extraction,it is found that the traditional Canny edge detection algorithm has the best extraction effect.At the same time,an improved Canny algorithm based on hybrid filtering and fuzzy clustering is proposed to improve the self-adaptability of the algorithm.Aiming at the problem of seam centerline extraction,the effects of different thinning algorithms are analyzed,and Z-S thinning algorithm is adopted to extract seam centerline.Aiming at the problem of local discontinuity in the extracted weld centerline,the least square method is used to fit the centerline,and the intersection method is used to get the feature points of the weld.On the basis of the above key technologies,the visual software for seam detection is developed by VS2015,QT and OpenCV,and various algorithms involved in the paper are programmed and implemented.The implementation results show that the proposed method can not only accurately extract the weld feature information,but also the processing time is short and meets the design requirements.