Font Size: a A A

Research On Weld Pool Visual Image Geometrical Characteristics Extraction Based-on Improved Threshold Segmentation And CV Model

Posted on:2014-01-25Degree:MasterType:Thesis
Country:ChinaCandidate:Y ZhangFull Text:PDF
GTID:2248330395982982Subject:Materials Processing Engineering
Abstract/Summary:PDF Full Text Request
A new weld pool image segmentation technique based on threshold segmentation is proposed in this paper, and single wide and continuous weld pool contour is gained by the technique, which includes three image processing techniques:improved threshold segmentation method, median filter, contour extraction. The convolution sum of Prewitt operator of all points on the weld pool contour is selected as the judge function to find the optimal threshold for the threshold segmentation. The threshold according to the max Prewitt convolution sum is chosen as the optimal threshold for weld pool image segmentation.In the paper, the high-proficiency median filter based on sliding histogram increases the efficiency of the whole weld pool image processing algorithm in a large extent. In the hardware environment of Intel I Core i5-2450M CPU2.5GHZ,4G memory, and the software environment of VC6.0, the above algorithm takes the shortest time of0.046s to extract the weld pool region contour of a400×400MAG weld pool image.The extraction of the weld pool region with CV model algorithm is studied in the paper. For a preprocessed weld pool image, Gauss filter is used to remove the noise of the image. Given an initial contour line of a circle shape, the weld pool region contours with different weld shaping can be automatically extracted by the CV model. The optimal parameters of CV model for the contour extraction of the weld image pool region and its influence on the detection efficiency is studied mainly in the paper. The experiments show that:the increase of parameters s and At can lead to a faster speed of detecting the edge points far from the original contour line.In this paper, the regularized Dirac function of CV model is replaced by the gradient norm of the level function. The improved model has a dull sensibility of the model parameters, a good adaptability in weld pool images segmentation, and a better ability of detecting the edge far from initial contour line which resulted with a shorter detecting time. A weld pool image processing program based on CV model is developed in the MFC software environment. The program can realize the contour extraction of MAG weld pool images with different weld configuration, and achieve a continuous contour line of the weld pool region. The experiments show that:The algorithm can detect the weld pool edge in a short time, and has a good extraction precision at the same time.
Keywords/Search Tags:threshold segmentation, Prewitt operator, high-proficiency median filter, CVmodel, Level Set
PDF Full Text Request
Related items