基于深度学习的岩体结构面特征提取研究

Research on Feature Extraction of Rock Mass Discontinuities Based on Deep Learning

爆破是地下金属矿山开采的必要环节,合理的爆破参数能够提升地下矿山整体的安全及生产效率。但在实际生产过程中,若在矿山地质岩体条件发生变化时仍采用原有的参数,则可能会出现因参数固化导致爆破预测效果不准确、爆破效果差的问题。为了实现预期的爆破效果,需要获取爆破的相关参数。基于深度学习技术对岩体结构面的特征提取展开研究,通过对不同算法以及参数的对比选择,最终以MobilenetV3算法作为基础架构并进行优化,将算法的正确率提升了14.07%,达到96.88%。

Blasting is a necessary part of underground metal mining,reasonable blasting parameters can improve the overall safety and production efficiency of underground mines.However,in the actual production process,if the original parameters are still used when the geological conditions of the mine change,there will be the problem of inaccurate blasting prediction and poor blasting effect due to the solidification of parameters.In order to achieve the expected blasting effect,it is necessary to obtain the relevant parameters of blasting.In this paper,the feature extraction of rock structure surface is studied based on deep learning technology,and by comparing and selecting different algorithms and parameters,the MobilenetV3 algorithm is finally used as the infrastructure and optimized to improve the correctness of the algorithm by 14.07%,reaching 96.88%.