Font Size: a A A

Research On Digital Image Processing Based Seam Tracking Algorithms

Posted on:2010-07-02Degree:MasterType:Thesis
Country:ChinaCandidate:Z X XuFull Text:PDF
GTID:2178360275453564Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Pipeline transport,as a high-effective and special means of transportation,plays a more and more important role in gas or oil transportation.At present,domestic large diameter pipelines are made of spiral welded tubes.To ensure the quality of spiral tubes, especially the easily leak placeā€”the weld seam,the unbiased welding is an important precondition during the welding process,and non-destructive testing is an important means of improving the quality after being welded.No matter whether the former or the latter,both have to ensure the precise position of weld seams and weld tracks.The automatic tracking system of weld seams can ensure the precise position for the welding and testing systems.The system studied in this thesis is used to track and locate weld seams during the process of non-destructive testing.According to the requirements of the terminal users,a new version of the weld seam tracking system developed by Signal and Image Processing Institute of Dalian Maritime University is released.The main works are as follows:To weaken the effect of light intensity,an software color filter is proposed.The processes are:firstly removing the physical filter from the original system,secondly obtaining the image of structured-light weld seam through the CCD color camera, thirdly extracting the red structured-light region,and finaly this region is passed to the current tracking algorithm to track the seam.Experimental results show that the proposed algorithm is robust to uneven ilumination and water reflections.To make the system more robust and independent to the structure light,two other block matching based algorithms are proposed.The rough positions are firstly predicted by Kalman filter or three layers' BP neural network respectively,and then block matching is used to refine the position.The experimental results have proven that the two algorithms can improve the efficiency of the system and meet the users' requirements.The proposed algorithms have been programed by VC++6.0,and the system adopting the proposed algorithms is to be used in some petrol-pipe plants.
Keywords/Search Tags:Weld Seam Tracking, Block Matching, Neural Network, Kalman Filter
PDF Full Text Request
Related items