Radiographic Testing Using The Digital Image Processing Technology Based On Mathematical Morphology

With the fast development of manufacturing and its more important role in national economy,it means that radiographic inspection become more and more important and urgent in the welding defect detection.Radiographic inspection is an effective method to detect welding defect and the product's quality is dominated and improved by it. But for the lower SNR of radiographic inspection in the image, seeking the effective method is imperative .This paper is based on X-ray detection of welding defect, and the key of automatic detection and recognition of welding defect is thoroughly studied, finally the corresponding software of automatic detection and recognition of welding defect in X-ray image is developed. The main is listed below.(1) Under the multi-scale mathematical morphology filtering framework, the complex welding defect image pretreatment is discussed.The new pretreatment of image welding defects based on multi-scale mathematical morphology filter and the details of the multi-scale filter operator is proposed. The filter can deal with the filtering of complicated welding defect image, and has obvious effect on the welding defect image with complex background, or especially the small-objects image. The effect of multi-scale and multi-structure structure element on the mathematical morphology filtering is discussed,and on the basis, a multi-scale opening and closeing filter is proposed and good filtering effects is obtained by the filter with compounding structural elements.(2)A adaptive matching features segmentation algorithm under the multi-scale mathematical morphology filtering framework is proposed. Under the multi-scale mathematical morphology framework,the algorithm transforms traditional thresholding segmentation algorithm into seeking for a suitable segmentation scale and has no sensitivity to the threshold selection. A perfect segmentation of welding image can be obtained by a constant threshold under multi-scale mathematical morphology framework. The feature histogram, matching characteristics histogram, matching characteristics histogram area, relative ratio of matching features is calcaulated by the multi-scale morphological characteristics sampling and the optimal scale is obtained adaptively. All the computationis is based on the local polynomial nonlinear methods and good result is get.(3) Post-treatment of the segmented image is discussed in this paper. The morphological reconstruction and welding defects edge reconstruction and region growing based on the Hausdoff distance is proposed. Firstly the morphological reconstruction method is constructed and the image is handled by morphological reconstruction to clean the noise points in the binary image, and then the methods of welding defect edge reconstruction and region growing based the Hausdorff distance is proposed and the original shape of the welding defect is kept and corrected farthest.(4) After analysis of the welding defect features, five typical features are successfully extracted from the welding defect features. Binary decision tree is used to make a greatest distinction between welding defects. On the former basis, support vector machines based on binary decision tree is proposed, which owns good performance in the classfication and the noticeable achievements are made in the classification of welding defects by the method .(5) Integrated image preprocessing, weling defect image segmentation and weling defect recognition research results proposed in the paper, a automatic inspection and recognition system of welding defect is developed to satisfy the actual demand and performance of the welding defect detection, and the software system Integrates with the monitoring, analysis and management of welding defect, and has the function of image denoising, image enhancement, the welding defects edge or feature extraction, the welding defects analysis and recognition.