基于深度学习的智能爆破矿岩块度自动分析系统

在露天开采领域，自动、准确地获取爆堆矿岩块度信息是优化爆破设计的关键。针对目前块度分析系统存在无法自动采集和自动批量处理图像的问题，提出一种基于深度学习的矿岩块度自动分析系统，该系统主要由基于MobileNet分类模型的自动采集子系统和基于U-Net语义分割模型的自动分析子系统组成。系统自动连续采集旋回破碎站的电动轮卸矿图像，通过4G网络上传云平台进行块度信息自动分析，分别对分类模型和分割模型进行定量、定性评估，其中分类模型在测试集上的精度达到98.08%，矿石分割模型的矿石类别IoU达到78.43%。将系统部署到某矿露天采区旋回站，通过一年多的工业生产实践，结果表明，本系统达到了设计要求，实现从采集到分析、信息展示全流程的自动化、无人化、智能化，可以进一步为智能爆破提供数据支持。

An intelligent automatic analysis system of ore fragmentation in blasting based on deep learning

In the field of open-pit mining, the key to ingoptimizing the blasting design is to automatically and accurately obtain the ore fragmentation information of the blasting pile. because of the problem that the current fragmentation analysis system cannot automatically acquire and automatically process images in batches, this paper proposes an automatic fragmentation analysis system of ore based on deep learning. The system is mainly composed of an automatic acquisition subsystem based on the MobileNet classification model and an automatic analysis subsystem based on U-Net semantic segmentation model. The system takes pictures of an electric dump truck unloading at the crushing station automatically and continuously and uploads them to the cloud platform by 4G network for automatic analysis of fragmentation information. Evaluating the classification model and segmentation model qualitatively and quantitatively, the accuracy of the classification model on the test set reaches 98.08%, and the ore category IOU of the ore segmentation model reaches 78.43%. The system has been running at the cycle station in the open-pit mining area of a mine for more than a year. The results show that the system meets the design requirements, and realizes the automation and intellectualization of the whole process from collection to analysis and information display, which can further provide data support for intelligent blasting.