首页|基于改进YOLOv5算法的无人机巡检图像智能识别方法

基于改进YOLOv5算法的无人机巡检图像智能识别方法

扫码查看
提出一种基于改进YOLOv5算法的无人机巡检图像智能识别方法.该方法构建无人机巡检图像的相邻图像独立坐标系,并利用相对定向法确定图像中共同目标的位置关系.将巡检目标统一转换至同一坐标系下,采用先进的分割技术提取目标纹理特征向量,为后续的图像识别提供了有力支持.在改进YOLOv5算法的过程中,特别注重多尺度网络的选择与融合激活函数及损失函数的优化组合.采用大疆无人机获取建筑裂缝巡检图像进行实验.结果表明,该方法能够在高效率下实现不同类型建筑裂缝的高精度识别,展现出优异的稳定性能.这一研究成果为无人机巡检图像的智能识别提供了新的思路和方法,具有广泛的应用前景和实际价值.
Intelligent Recognition Method for UAV Inspection Images Based on Improved YOLOv5 Algorithm
This paper presents an intelligent recognition method for UAV patrol image based on improved YOLOv5 algorithm.The independent coordinate system of the adjacent image of UAV inspection image is constructed,and the position relation of the common object in the image is determined by the relative orientation method.Then,by converting the inspection target to the same coordinate system,the advanced segmentation technique is used to extract the texture feature vector of the target,which provides strong support for the subsequent image recognition.In the process of improving YOLOv5 algorithm,special attention is paid to the selection of multi-scale network and the optimization combination of fusion activation function and loss function.The DJI UAV is used to acquire the inspection images of building cracks.The results show that the method can iden-tify different types of building cracks with high accuracy and high efficiency,and show excellent stability.This research pro-vides a new idea and method for intelligent recognition of UAV inspection image,which has a wide application prospect and practical value.

UAVinspection imageYOLOv5 algorithmmulti-scale networkintelligent recognition

侯伟、陈雅、宋承继、刘强锋

展开 >

陕西工业职业技术学院,电气学院,陕西,咸阳 712000

广西警察学院,计算机学院,广西,南宁 530023

无人机 巡检图像 YOLOv5算法 多尺度网络 智能识别

陕西工业职业技术学院院级项目陕西省科技厅重点研发项目2023年广西高校中青年教师基础能力提升项目

2022YKYB-0092023-YBNY-21602023KY0910

2024

微型电脑应用
上海市微型电脑应用学会

微型电脑应用

CSTPCD
影响因子:0.359
ISSN:1007-757X
年,卷(期):2024.40(9)