Research On Product Defect Detection Technology Based On X-ray Image Feature Extraction

X-ray inspection technology is one of the common methods for non-destructive inspection of internal defects of products,it is widely used in the field of defect detection for related products in industries such as aerospace,military explosive,civil automobile,high-speed rail and power equipment.Product quality requirements are increasing day by day with the rapid development of industrial production technology,especially in some industries that produce precision electromechanical products in large quantities,the demand for fast and accurate full inspection of products has been raised.These are still the current difficult problems in the industry about how to improve the detection efficiency and reduce the detection cost.In the process of product testing,it is difficult to avoid the tilt of the product and the testing table,these will cause mismatched recognition and affect the recognition accuracy.And it will affects the real-time performance of recognition due to imaging and recognition in multiple orientations during actual detection.In the view of the problem that the tilt of the product and the testing table,the Hough transform is used to extract the characteristic straight lines in the image,and then the measurement and correction of the horizontal tilt and vertical tilt of the product in the current state are completed based on the analysis of the geometric mapping characteristics of the imaging system.The proposed method was simulated and verified by visible light simulation experiment.In the view of the problem of excessive detection data leading to increased detection time,this paper proposes a fast detection algorithm based on assembly detection area for the characteristics of structure products.Based on the prior knowledge,the structure with obvious characteristics and no error in the structure of the product is taken as the circumferential position feature recognition area after product imaging,the feature vector of the area is extracted,and the current imaging azimuth is searched from the circumferential position feature library.The images of all the sub-target regions to be detected under this azimuth are segmented after the product imaging azimuth is determined.The feature vectors of each sub-target region and the standard feature vectors in the sample library are extracted for similarity detection to complete the product defect detection.The proposed method has been simulated and verified by simulation experiments.Experimental results show that this paper can accurately detect defects and has a faster detection speed than SIFT algorithm.Based on the above tilt correction and feature-based fast recognition algorithm,this paper carried out defect detection and verification of an industrial product,the average detection speed is 0.0477 s under the premise of meeting the identification requirements.The research work in this paper has laid a good foundation for the rapid defect detection of products,and is of great significance for the research of rapid automated X-ray inspection.