基于卷积神经网络的铁道工程勘探地质岩芯图像识别系统的开发和实践

Development and Practice of Image Recognition System of Geological Core in Railway Engineering Exploration Based on Convolutional Neural Network

在铁道工程地质勘探中,钻探是最常用且有效的一种勘探手段,其成果是进行工程地质评价和设计、施工的基础。通过搭建卷积神经网络模型,实现对勘探现场岩芯原始图像的岩芯类型识别。开发移动端App图像识别系统对岩芯图像进行获取,通过调用服务器端可自主学习的识别模型进行岩芯类型鉴别,并将结果反馈给勘探现场数据采集人员。该系统在多个轨道交通项目勘探工作中进行了测试,准确率可达90%以上。这项工作有效推动了轨道交通勘探业务的数字化、标准化建设,可显著提升勘探项目生产工作效率。

In railway engineering geological exploration,drilling is the most commonly used and effective exploration method,and its results are the basis for engineering geological evaluation,design,and construction.By constructing a convolutional neural network model,we can achieve the recognition of core types from the original images of cores at the exploration site.Therefore,a mobile App image recognition system is developed to capture images of the cores,which then calls a server-side self-learning recognition model to identify the type of core,and feeds back the results to the data collection personnel at the exploration site.This system is tested in multiple rail transit project explorations,and the accuracy rate is over 90%.The work effectively promotes the digitization and standardization of rail transit exploration services,and can significantly improve the production efficiency of exploration project.