The Research On The Extraction Of The Weld Flaw Images

Ship welding is the key factor to impact the quality of the ship in ship-construction, the quality of which can directly affect the ship's durability and safety, and it is the important prerequisite of the ship's quality and safety navigation, as well as safety working onboard. So the detection of the welding flaw in ship-construction is an important step. Currently, the most common method in welding detection is the X Radiography method, which can produce plenty of welding film. However, people usually take the manual assessment method to check the flaws in the film, and because of the different technical level between different people, the manual assessment method has the feature of heavy workload, lower efficiency and poor accuracy.The computer Assistant Assess System has greatly reduced the workload of the worker, and improve the working efficiency is some extent, thus, the process of the film assessment is more scientific and more standard. But due to the specification of the film image in the formation of the welding film image, there are many difficulties in the exact extraction of the flaw information and the accurate feature calculation as well as the following classification. The classic features of the film image are too many noises, blur flaw edge and low contrast.Against the problems in the present process of the image-processing with computer assistant assess system, the thesis made in-depth research and analysis to each step of the image-processing, the mainly tasks are as follows:Firstly, against the specificity of the ship's film image acquisition, we focus on the image's digital feature, and get the properly method of the image processing through the comparative study. Begin with the formation of the digital image, we analyzed the imaging features and the digital process, by analyzing the gray features of the digital image, the features of the welding image were obtained, which include complex noises such as gauss noise, the weld edge and the flaw edge are blur, the contrast between the weld area and the flaw is low; Next, based on those features, different denoising and enhancement algorithm was used for the comparison, in that way, we get the best approach of the image process is gauss smoothing and the fuzzy enhancement.Secondly, based on the image preprocessing, we use the flaw marking method and the boundary tracking algorithm to mark and count the flaws. For each flaw, we use the eight-connected marking algorithm and the boundary tracking algorithm to mark each flaw and record the flaw's boundary information in order for the flaw's feature calculation.Thirdly, calculate the flaw's geometry feature. After the extraction and the marking of the flaw in the welding image, calculate the pixel's feature and made the calibration conversion. In this thesis, each flaw's area, Circumference and the minor axis was calculated and proved to have good effect, so we have made a good foundation for the later classification and rating.