基于Mask R-CNN的煤矿巷道掘进迎头裂隙检测与定位算法

Research on crack detection and localization algorithm for advancing face in coalmine roadways based on Mask R-CNN

针对当前煤矿巷道掘进迎头裂隙图像识别精度不高、井下环境难以批量化快速识别的难题,提出基于Mask R-CNN的煤矿巷道掘进迎头裂隙检测与定位算法。首先,建立一个包含1000张图像的煤矿巷道掘进迎头裂隙图像数据库。然后,基于Mask R-CNN深度学习网络构建一个裂隙检测与定位框架,并选择该数据库对神经网络模型进行训练和测试,进而开展煤矿井下干扰环境下的鲁棒性和适应性评价,并与传统图像处理算法进行对比。结果表明,在相同样本条件下,基于Mask R-CNN的深度学习算法能够高效实现煤矿巷道迎头裂隙的检测与定位,该算法能有效避免低照度、多尺度边缘、截割刻痕、非均匀光照等干扰因素的影响,具有更高的分割准确度和计算速度,可满足煤矿井下裂隙批量化快速识别的要求,为煤矿巷道裂隙迹线的检测与定位提供了新路径。

This paper proposes a deep learning algorithm based on Mask R-CNN for crack detection in advancing faces of coal mine roadways,addressing the issues of low recognition accuracy and difficulties in achieving batch and rapid identification of cracks in underground coal mines.Firstly,a database with 1000 images of coal mine roadway advancing face crack images is established,which is used to train and test for the neural network model.Based on the deep learning network named Mask R-CNN,a framework for crack detection and localization is constructed.Finally,the robustness and adaptability of the model are evaluated under common interference environments in coal mines,which is compared with traditional image processing algorithms.Experimental results demonstrate that the deep learning algorithm based on Mask R-CNN efficiently detects and locates cracks in coal mine roadway advancing faces under the same sample conditions.Furthermore,the algorithm reduce the influence of factors such as low illu-mination,multi-scale edges,cutting marks,and non-uniform lighting.Compared to traditional image recognition algorithms,the proposed model exhibits higher segmentation accuracy and computational speed,meeting the requirements of batch and rapid crack identification in underground coal mines.This research provides a new approach to crack detection and localization under similar conditions.