基于无人机图像的黄河滩区违规牧羊智能检测方法

Intelligent detection method of illegal shepherding in Yellow River beach area based on UAV images

为解决传统的黄河滩区违规牧羊行为人工现场巡检方式效率低、监管难度大的问题,介绍一种针对黄河滩区违规牧羊的智能检测方法。提出一种改进的YOLOv5s目标检测算法,通过减小模型尺寸和计算量,提高算法的实时性能,同时设计一种羊群目标定位算法,通过建立相机中心、二维图像像点、空间地面目标物点三者之间的几何关系模型,判断牧羊行为是否发生在违规区域内。结果表明:改进后的YOLOv5s目标检测算法在CPU上的实时性能提升31.25%,训练自建黄河滩区牧羊数据集后模型大小仅为3.8 MB,目标定位算法的精度可控制在10 m以内,能够更快速精准地识别羊群。基于无人机图像的黄河滩区违规牧羊智能检测方法可实现无人化、智能化的黄河河道巡查,大大提高效率和灵活性,减少人力和物力投入,具有重要的现实意义和应用前景。

In order to solve the problems of low efficiency and difficulty of traditional on-site inspection of illegal shepherding behavior in Yellow River beach area,this study introduces an intelligent detection method for supervising illegal shepherding in Yellow River beach area.Firstly,an improved YOLOv5s target detection algorithm is proposed in this paper,which enhances real-time performance by reducing the model size and computation load.Meanwhile,a sheep herd target location algorithm is designed,which determines whether shepherding is illegal within unauthorized areas by establishing a geometric relationship model between the camera center,two�dimensional image pixel points,and spatial ground target points.The results show that the improved YOLOv5s target detection algorithm achieves a 31.25%increase in real-time CPU performance,and the model size is reduced to just 3.8 MB after training on a custom dataset of shepherding in Yellow River beach area.The accuracy of the target location algorithm can be controlled within 10 meters,enabling the method to identify sheep flocks more quickly and accurately.This intelligent detection method of illegal shepherding in Yellow River beach area based on UAV images allows for unmanned,intelligent monitoring of Yellow River channel,significantly improving efficiency and flexibility while reducing human and material resources,which holds great practical significance and application prospects.