Research On The Key Technology Of Steel Plate Surface Defect Detection System Based On Machine Vision

With the rapid development of the visual sensor technology, computer technology andimage processing technology, machine vision technology is increasingly mature, it has becomean indispensable core technology for modern manufacturing industry. Steel plate surface qualityis a very important technology index. How to detect the steel plate surface defects rapidly and toimprove the production levels of steel plate are the problems need to solve for steel plateproducers. Because of the low efficiency, high failure rate and high labor intensity the traditionalmanual detection method can not meet the needs of the development of steel plate production.Compared with the traditional detection techniques, the steel surface defect detection based onmachine vision technology has advantages of speediness, accuracy, reliability and intelligence. Ithas become the main research direction for the majority of scholars at home and abroad and steelproducers.Based on the deep study of the key technology of the steel plate surface defect detectionsystem, this paper focuses on defect image preprocessing algorithm, image segmentationalgorithm, the feature extraction method, feature selection method and classification algorithmbased on support vector machine. In this paper the author provides the overall scheme of thesteel plate surface defect detection system and tests it by experiments. The paper will fulfill themain work as follows:1. Aiming at the characteristics of Steel plate surface defect images, it studies thepreprocessing algorithm and achieves the enhancement of the defect images. The author providesan improved2D maximum between-cluster variance threshold segmentation algorithm, whichcan reduce the time of the defect image segmentation, and get a better segmentation effect.2. It Studies of defect image feature extraction method, extracts and normalize the momentinvariant feature, gray feature, texture feature of defect images. Using principal componentanalysis method to achieve feature selection, so as to produce the new linear independentcharacteristic value, and realize dimension reduction of defect feature.3. IT Studies the defect classification algorithm based on support vector machine, realizesthe design and implementation the muti class classifier based on support vector machine. Doingthe defect sample classification experiment through selection of samples, and get the betterclassification results.4. It puts forward a set of overall scheme of steel plate surface defect detection systembased on the study of image preprocessing, segmentation, feature extraction, feature dimensionreduction and classification algorithm of defects. The author design the system hardwaremodules and introduce the methods of selection. Design the software algorithm flow and validation of the algorithm presented in this paper. And get a better results. experimental resultsshow that this key technology research is effective and feasible.