基于改进YOLOv7的高分二号遥感影像滑坡识别算法研究

RESEARCH ON LANDSLIDE RECOGNITION ALGORITHM OF GF-2 REMOTE SENSING IMAGE BASED ON IMPROVED YOLOV7

为快速准确地识别滑坡地质灾害,本文提出基于YOLOv7的轻量滑坡检测模型。实验结果表明,改进后YOLOv7网络模型整体mAP达到94.6%,与原始YOLOv7网络模型相比,参数量减少了20.8M,计算量减少47.3G,整体mAP0.5提高3.8%,检测速度FPS提高14.3f/s,对滑坡灾害具有出色的检测效果。

In order to quickly and accurately identify landslide geological hazards,we propose a lightweight landslide detection model based on YOLOv7.The experimental results show that the overall mAP of the improved YOLOv7 network model reaches 94.6%.Compared with the original YOLOv7 network model,the parameter and calculation amount have been reduced by 20.8M and 47.3G,respectively,while the overall mAP0.5 and the detection speed FPS have been increased by 3.8%and 14.3 f/s,respectively,indicating that the model has excellent detection performancefor landslide disasters.