基于改进DeeplabV3+的矿区植被覆盖度变化检测

Detection on Vegetation Coverage Changes in Mining Areas Based on Improved DeepLabV3+

针对传统的露天矿区植被覆盖度变化检测方法存在的监测信息不完整、目标区域边缘模糊等问题,提出一种将改进DeepLabV3+和注意力机制相结合的矿区植被覆盖度变化检测方法,即M-CS-DeepLabV3+模型检测方法。该方法引入带有扩张卷积的MobileNetV2降低DeepLabV3+模型的网络参数量和计算量,从而提高植被区域目标检测的精度和效率,同时运用改进的DeepLabV3+优化模型,实现对矿区植被状况的遥感监测与分析。结果表明,该方法可对矿区植被受损状况进行时序检测,且分割效果完整准确,检测准确率达到86.16%,速度可达68帧/s,可为露天矿区植被变化检测和生态环境治理提供有效依据。

Aiming at the problems of incomplete monitoring information and blurred edge of the target area in the traditional detection method of vegetation coverage change in open-pit mine,a detection method combining improved DeepLabV3+and attention mechanism was proposed on vegetation coverage change in mining area,namely M-CS-DeepLabV3+model detection method.This method introduced MobileNetV2with dilated convolution to reduce the number of network parameters and computation of DeepLabV3+model,so as to improve the accuracy and efficiency of target detection in vegetation area.At the same time,the improved DeepLabV3+optimization model was used to realize the remote sensing monitoring and analysis of the vegetation status in the mining area.The results show that the method can detect the time series of vegetation damage in the mining area,and the segmentation effect is complete and accurate.The detection accuracy is 86.16%,and the speed can reach 68frames/s.The study can provide an effective basis for vegetation change detection and ecological environment management in open-pit mine areas.