Surface defect detection method based on visual saliency map and support vector machine

Abstract

The invention discloses a surface defect detection method based on a visual saliency map and a support vector machine. The method comprises the following steps: firstly, calculating the visual saliency map of a surface image sample of a product to be detected by using a GBVS (Graph-based Visual Saliency) model, carrying out adaptive threshold segmentation on the visual saliency map through a method of maximum classes square error, extracting a visual saliency region, calculating the average gray-values of the visual saliency map and the average gray-values of the visual saliency region in the visual saliency map, and respectively carrying out normalization processing to form two-dimensional features, then taking the obtained two-dimensional features of the visual saliency map as training samples of the support vector machine, selecting two dimensions to classify the optimal classification line, based on the optimal classification line, classifying the two-dimensional features, thereby distinguishing whether the product in the map has defects or not. The surface defect detection method disclosed by the invention can effectively save labor, lower the labor intensity and improve the work efficiency and has high identification accuracy.

Description

technical field [0001] The invention relates to a surface defect detection method, in particular to a surface defect detection method based on a visual saliency map and a support vector machine. Background technique [0002] In textile, metal and other production fields, how to detect and solve product surface defects in time is a widely concerned issue. At present, most domestic manufacturers use manual visual inspection to complete the inspection work. This method is labor-intensive, has low work efficiency, and is easily affected by the subjective factors of inspectors. Therefore, the research on automatic surface defect detection technology is of great significance. [0003] Visual saliency model is one of the research hotspots of domestic and foreign scholars in recent years, and it is widely used in remote sensing, metallurgy, textile, agricultural production and other fields. This model imitates the human visual attention mechanism and integrates image features thro...

AI Technical Summary

Problems solved by technology

At present, most domestic manufacturers use manual visual inspection to complete the inspection work. This method is labor-intensive, has low work efficiency, and is easily affected by the subjective factors of inspectors.

Images