一种无人机图像识别技术体系研究与应用

The Study and Application of a UAV Image Recognition Technology System

为了实现水利工程巡检、河湖岸线、河湖环境监测等海量无人机图像异常特征物的智能化检测识别,提高管理效率,满足智慧水利建设的需求,基于YOLO v3算法框架,通过引入注意力模块SE,构建了高精度的YOLO v3-SE目标检测算法,形成了无人机图像识别技术体系,并将其成功应用于多个水利工程的海量无人机图像的异常特征物检测识别中。结果表明:通过自建66000张图片数据的训练集和35514张图片数据的测试集,本算法与原始YOLO v3算法、改进的SKSet-YOLO v3算法和CBAM-YOLO v3算法相比,在积水、塌方、运输船、滑坡、聚集型垃圾和分散型垃圾等6类目标物的检测精度AP均有较大幅度的提升;平均检测精度mAP也分别从59.83%提升至90.17%、从79%提升至90.17%、从72%提升至90.17%,精度得到明显提升,满足水利工程智慧化监控的需求。

In order to make the intelligent detection and identification of abnormal features of massive UAV images,such as water conservancy project inspection,river and lake shoreline,and the realization of river and lake environmental monitoring,the enhancement of management efficiency,the satisfaction of demands for intelligent water conservancy construction based on YOLO v3 algorithm framework,a highprecision YOLO v3-SE target detection algorithm is constructed by the introduction of the SE,the UAV image recognition technology system is formed and it is also successfully applied to the detection and identification of massive UAV images in multiple water conservancy projects.It is shown based on the results that are compared with the original YOLO v3.SKSet-YOLO v3 and CBAM-YOLO v3 algorithm are improved by a self-built 66000 picture data training set and a test set of 35514 picture data,for the algorithm,the detection accuracy of 6 types of targets such as water,landslide,carrier,landslide,aggregate garbage and decentralized garbage has been greatly improved.With regard to the average detection accuracy mAP,it has also increased from 59.83%to 90.17%,from 79%to 90.17%,and from 72%to 90.17%.