The Research On Steel Defects Detection Technology Based On Visual Attention Mechanism

Strip steel is one of the important steel enterprise products, whose quality directly decides the quality of related terminal products; The traditional quality detection of strip surface mainly depends on artificial sampling, which inevitably cause residual detection, can hardly meet the needs of modern production. With the development and application of machine vision and image processing technology, the steel defect detection based on image processing algorithms has become a hot study among strip steel enterprises at home and abroad.Currently, the classical strip defect detection based on image processing methods can not meet the needs of segmentation integrity and the accuracy of detect strip image. On the basis of human visual perception theory, this paper proposed a strip defect detection method based on visual attention mechanism. Firstly, The bottom-up visual attention mechanism detection model is formed through the operations of the feature extraction、the wavelet multi-resolution filtering 、 the center-surround difference operation 、 feature integration and so on; On this basis, a new strip defect detection algorithm is formed, to solve the problem of detection accuracy, segmentation integrity and robust; Finally, the article verified the effectiveness of the proposed method through comparing this algorithm with the traditional detection algorithm experiments.The experiment results show that the present algorithm can realize accurate detection a variety of usual defects and solve the problem of incomplete segmentation and low detecting accuracy in the traditional defect detection process, thus a new efficient detection method, which has broad application prospects, is provided for strip steel production enterprises to enhance their competitive ability in the market.