Segmentation Algorithm of Armor Plate Surface Images Based on Improved Visual Attention Mechanism

ABSTRACT

In the sight of difficulty in segmentation of armor plate surface defects, visual attention mechanism is applied to segment defects images of steel strip surface, the key step of visual attention mechanism-fusion method of saliency images is improved, thinking of the contribution of salient area size, number and distribution to the comprehensive salient image, gray level consistency and GLCM’s entropy are used to calculate weight coefficients for feature images fusion, and comprehensive saliency image is obtained, then maximum entropy method is used to segment the comprehensive saliency image. The segmented results are compared to the results of clustering and region-growing methods, the improved visual attention mechanism has better effects.

Keywords:

Feature fusion, salient feature, segmentation of steel strip surface defect images, visual attention mechanism.