Intelligent veins recognition method for slope rock mass geological images in complex background noise

Abstract

The veins formed by magma intrusion often exist in the rock mass of slopes in the form of weak structural planes, thereby triggering landslide engineering disasters. Therefore, accurately identifying veins information from complex slope geological images is crucial. In light of this, this paper tackles the challenges encountered by traditional edge detection and deep learning algorithms in effectively detecting veins amidst complex background noise. These challenges include poor detection performance, low-quality manually annotated data, and the significant workload associated with manual annotation. To address these issues, we propose a vein intelligent recognition algorithm that combines OpenCV, AnyLabeling, and Mask R-CNN. By employing OpenCV, this study utilizes denoising and cropping techniques on the original images characterized by complex background noise. By preprocessing the raw dataset, it removes image noise and optimizes feature details within the images, thereby enhancing the quality of the dataset; Utilizing AnyLabeling for objective and automated annotation of veins information, this process eliminates the subjectivity associated with manual annotations, resulting in the creation of a high-quality library of veins image recognition samples; Building upon this foundation, we develop a Mask R-CNN model for veins contour segmentation. This model accurately and efficiently identifies veins information in complex geological images with background noise. Its successful application has been rigorously validated on a specific open-pit mine slope. The results indicate that this method successfully addresses the challenge of accurately identifying veins contours under complex background noise; The preprocessing technique, which combines OpenCV and AnyLabeling, demonstrates a significant enhancement in both the accuracy and efficiency of dataset annotation. This improvement contributes to heightened precision in the recognition results. Effectively discerned intricate veins details within the specific open-pit mine slope, resulting in an impressive 93.6% increase in MIoU value. This notable enhancement substantially improves the precision of slope stability calculation outcomes. The research findings are of significant importance for the analysis of veins-type geological hazards.