基于多尺度卷积神经网络的矿山监控图像识别研究

Research on Mine Monitoring Image Recognition Based on Multi-scale Convolutional Neural Network

现有矿山复垦监测识别方法主要有遥感卫星监测、土壤环境指标判断以及无人机监测等,但这些方法存在精度较低、时效性差等问题。为此,文章提出了一种基于多尺度卷积神经网络的矿山监控图像识别方法,通过采集高清摄像头传送的视频流数据,对关键帧图像进行分析处理,然后通过对比任意时刻图像的差异来判断矿山生态修复治理质量状况。该方法在安徽省境内露天矿山进行了实地应用,结果表明此方法具有高稳定性的生态修复识别能力,可为露天矿山的环境治理提供有效的技术支持。

The present methods for monitoring and recognizing mine reclamation mainly include remote sensing satellite monitoring,soil environmental indicator judgment,and drone monitoring.However,these methods suffer from low accuracy and poor timeliness.Therefore,this paper proposes a mine monitoring image recognition method based on the multi-scale convolutional neural network,which analyzes and processes key frame images captured from the video stream data transmitted by a high-definition camera,and judges the quality of mine ecological restoration and management by comparing the differences between images at any given time.The method has been applied in open-pit mines in Anhui Province,demonstrating a high stability in identifying ecological restoration,thus providing effective technical support for environmental management in open-pit mines.