Rock image segmentation based on salient target detection method

Abstract

∗In the process of rock image classification, the existence of rock image background affects the accuracy of rock recognition. In this paper, a saliency target detection network based on the attention mechanism is used to generate the saliency map of the rock image, and the local and global information of the image is extracted through a grid structure of multiple resolutions, and they are merged into prediction features; The saliency map uses an adaptive threshold method to segment the background area of the rock image, and only retains the rock area image, which further improves the accuracy of rock sample classification. By comparing the traditional histogram-based threshold segmentation, the edge detection method based on the Canny algorithm and the salient target detection method based on the attention mechanism, the results show that the result of using the salient target detection method to achieve rock image segmentation is more accurate. It proves its effectiveness for rock image segmentation.