Roadway damage detection and safety rating method based on multi-region attention mechanism

The invention discloses a roadway damage detection and safety rating method based on a multi-region attention mechanism, and relates to the technical field of coal mine safety production. The method comprises the steps that firstly, a roadway image is acquired and preprocessed; secondly, classifying roadway guniting damage, constructing a data set, training a YOLOv8 model and a mask image, and dividing the roadway guniting damage into a plurality of confidence frames according to the damage; thirdly, fusing multi-scale information segmentation damage by using an FCN network, and fusing results; then, dividing a result image region, endowing different regions with safety weights, and designing a weighting formula to calculate the proportion of damaged pixels; and finally, four-level safety grading is provided according to coal mine safety regulations, and safety grading is carried out on damage of different degrees. According to the method, the image processing method based on deep learning is applied to coal