Research On Intelligent Recognition Mechanism Of X-ray Images For Welding Defects

Welding is acknowledged as the vital processing technology to complete material connection.The quality of welding will influence the structure property directly.With the general improvement of the intelligent degree of manufacturing industry,the manual evaluation method of X-ray image is gradually weeded out on account of subjective factors,which was used for the quality of welding defects detection.In recent years,in accordance with the development direction of efficiency and intelligence,automatic detection and recognition of welding defects has become an urgent need of many industries,which has theoretical significance and engineering value.Focusing on the automatic recognition method of welding defects based on vision,this paper aims to solve the problems of image denoising,image segmentation,defects feature selection and welding defects classification,which provides significant technologies for promoting intelligent automatic detection in the welding industry.In this regard,the major work of this article is as follows:Aiming at the characteristics of uneven illumination and low contrast in welding ray images,a quadratic polynomial transformation algorithm is used to complete the grey level transformation of the ray image.Based on the problem that the welding images have large noise interference,a secondary filtering algorithm which combines adaptive median filtering and multi-scale morphology is proposed to solve the detailed problem of suppressing noise while keep welding ray image as much as possible.Experiments show the effectiveness of the gray-scale transform and quadratic filtering algorithm in the preprocessing of welding defects images.The welding ray images often appear the problems such as target edge blur,low contrast,and diversity of defects,which are the main influence factor of the segmentation of welding ray images.In this paper,the average sliding histogram is introduced to estimate the non-parametric probability density in the traditional active contour model,so that the model evolves according to the gradient force and the statistical pressure value of the active contour models based on the average sliding histogram.The statistical information solves the problem that the welding defects ray images are greatly affected by noise.An immune optimized active contour model with strong global search ability and parallel processing by immune algorithm is proposed.The immune cloning algorithm is applied to optimize the weights of the control factor,gradient force and constraining force in the active contour model that affects the running speed and feature extraction of target.In order to get the optimal value of the three parameters in the active contour model,the cloning selection operator,the cloning amplification operator and the cloning mutation operator are used to complete the immune optimization algorithm.The active contour model combining the immune algorithm solves the problems of limited initialization and difficulty in accurately segmenting the boundary of the welding defects.The results of applying these two algorithms to X-ray image segmentation of welding defects indicate that the algorithm can be used to complete the weak edge target segmentation very well.The combined performance is better than the traditional active contour model obviously.Aiming at the feature description and selection of defects in the ray image of welding,the kernel principal component analysis and the locally linear representation manifolds margin based on global and local information fusion is applied in the welding defects selection in the paper.Firstly,the kernel principal component analysis algorithm is used to retain the global nonlinear information of the original data to the greatest extent;then,locally linear representation manifolds margin is used to mine the local manifold structure information of the data.The experimental results show that in the method the local and global information can be maintained comprehensively,the selection of welding features can be completed effectively,and the redundant and invalid features can be removed.Finally,aiming at the classification of welding defects as unbalanced sample classification,the improved chaotic immune algorithm is proposed to solve optimization problem and the ergodicity of the chaotic algorithm is used to optimize the support vector machine based on the good learning performance and generalization ability of support vector machine.In the chaotic immune system,an improved Logistic map is constructed,which can effectively overcome the attractor problem in Logistic map and improve the ergodicity of chaos optimization.Furthermore,chaotic perturbation parameter is improved to ensure better optimization results.The experiment proves that the algorithm is superior to the traditional classification method and significantly improves the classification accuracy of welding defects.