Image saliency detection method for safe operation of substation

Abstract

Aiming at the problem that methods based on background templates in existing substation scenes cannot completely detect salient targets with similar characteristics in some areas and the background, a salient target integrity detection method based on contour detection is proposed. First, extract the initial contours of the input image, use the proposed virtual connection-based contour processing scheme to merge adjacent contours and remove isolated contours, and use the designed shortest path-based closed-loop search scheme to merge the contours with longer distances to obtain Saliency map for contour detection. Then, an adaptive threshold segmentation algorithm is used to process the saliency map based on background template suppression to obtain the binarized saliency map and the salient pixels. By removing the complete area of the contour with the proportion of salient pixels smaller than the specified threshold, an optimized saliency map based on contour detection is obtained. Finally, merge it with the binarized saliency map to obtain a complete saliency map. The results prove that the method in this paper can obtain a better saliency map for images where the saliency target is located at any position.