Research On Strip Steel Surface Defects Inspection Algorithm Based On Machine Vision

With the demand of social development, strip steel products are more and more widely used in industry production and social life. Strip surface quality has a direct impact on its follow-up products quality, and the real-time and recognition rate of existing detection methods cannot match high-speed steel strip lines, which makes it a bottleneck that constricts the strip industrial development, so strip surface defect detection method has become one of the hot research.Defect detection technology base on machine vision is nondestructive, intelligent, precision and fast, which makes it a trend in defect detecting technology. Based on machine vision technology, the study focuses on the core strip surface defect detection algorithms and key software technology, in order to achieve a high speed and accuracy of defect detection, details are as follows:1. Defect-object detection algorithm:a comparative analysis has been undertaken, which reveals the limitations of defect-object detection algorithm based on various edge-detection operators, so the background difference algorithm is used in strip surface defect-object detection, the experimental results show that the algorithm has a strong anti-interference ability and a good self adaptability, and it also can locate defects accurately, which is in favor of follow-up defect segmentation.2. Defect-region extraction algorithm:fast-labeling algorithm is used for marking defect objects, then the marked defect objects is located, and defect bitmap which contains only a single defect object is cut out. which is in favor of follow-up defect feature extraction.3. Defect-feature optimization algorithm:morphological features, gray features and texture features of individual defects are extracted, and principal component analysis is used for feature optimal selection of these three categories original features, then the Optimized features as BP neural network inputs are used for defect classification.4. Defects recognition algorithm:improved BP neural network is adopted to classify strip surface defects, and it has a high recognition rate. 5. Multi-threading and database programming technology:the programming of whole algorithm is carried out in VC++environment, and multi-threading technology is aplied to achieve rapidity of algorithm, moreover defect information is stored into the database in real time, and also the defect information management operation in the database is achieved.