基于改进Mask R-CNN的露天矿边坡裂隙智能检测算法

为了预防因露天矿边坡表面恶化而产生节理、裂隙或断裂等破坏边坡完整性所引发的安全事故，同时解决传统图像处理算法以及经典的深度学习模型直接应用于露天矿边坡裂隙检测效果不甚理想的问题，提出了一种基于改进的Mask R-CNN的露天矿边坡裂隙智能检测算法，运用了Mask R-CNN在目标检测、语义分割以及目标定位方面的集成性特点，改进了其在掩膜分支的边缘不清晰以及误检等缺点，构建了一种针对露天矿边坡裂隙图像的检测分割框架。该方法在掩膜分割分支引入了空洞卷积神经网络以及分类分割迭代上采样操作，能够解决边坡裂隙分割边缘粗糙的问题，实验结果表明，与传统的裂隙分割算法相比，该算法具有更高的识别精度以及更好的分割效果。

An intelligent detection method for open-pit slope fracture based on the improved Mask R-CNN

To prevent the unexpected accidents caused by the failure of slope integrity, we propose an intelligent fracture detection algorithm based on improved Mask R-CNN which can address the limitations of traditional image processing algorithm and the poor performance of the application of the classical deep learning model directly to the open-pit mine slope crack detection. In this paper, we use the integrated features of Mask R-CNN in target detection, segmentation and location to improve the shortcomings of Mask branch, such as unclear edges and false detections, and construct a detection and segmentation framework for slope fracture images of the open-pit mine. This method introduces dilated convolution neural network and a classify segmentation iterative up-sampling operation into the mask branch, solving the problem of slope fracture mask''s rough edge. Experimental results show that compared with the traditional crack segmentation algorithm, this method has higher recognition accuracy and better segmentation effect.