使用改进Yolov5的变电站绝缘子串检测方法  

Substation Insulator String Detection Method Using Improved Yolov5

作　　者：冯晗 姜勇[2,3,4] FENG Han;JIANG Yong(School of Information Science and Engineering,Northeastern University,Shenyang 110006,China;Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang 110016,China;Key Laboratory of Networked Control Systems,Chinese Academy of Sciences,Shenyang 110016,China;Institutes for Robotics and Intelligent Manufacturing,Chinese Academy of Sciences,Shenyang 110169,China)

机构地区：[1]东北大学信息科学技术学院,辽宁沈阳110006 [2]中国科学院沈阳自动化研究所,辽宁沈阳110016 [3]中国科学院网络化控制系统重点实验室,辽宁沈阳110016 [4]中国科学院机器人与智能制造创新研究院,辽宁沈阳110169

出　　处：《智能系统学报》2023年第2期325-332,共8页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(52075531)。

摘　　要：针对变电站绝缘子串水冲洗机器人在复杂光照环境下无法准确识别绝缘子的问题,提出了一种基于改进Yolov5的绝缘子检测方法。首先针对逆光环境下图像质量差导致算法失效的问题,提出了一种模拟过曝增强算法,并应用到数据增强过程中;此外,针对变电站绝缘子检测任务,对网络的Neck进行了优化裁剪,使推理速度获得了提升;最后,使用注意力机制改善了裁剪后网络检测精度下降的问题。实验表明,改进后的Yolov5在检测精度基本不变的情况下推理速度提高了25%,并且对于逆光下图像的检测精度获得了大幅提升。An insulator detection method based on improved Yolov5 is proposed for the problem that the substation insulator string water rinsing robot cannot accurately identify insulators in complex lighting environments.Firstly,a simulated overexposure enhancement algorithm is proposed for the problem of algorithm failure due to poor image quality in backlight environment and applied to the data enhancement process;in addition,the network's Neck is optimally cropped for the substation insulator detection task,and the inference speed is improved;finally,the attention mechanism is used to improve the problem of network detection accuracy degradation after the cropping.Experiments show that the improved Yolov5 improves the inference speed by 25%with the same detection accuracy,and achieves a significant improvement in the detection accuracy for images under backlighting.

关 键 词：Yolov5 绝缘子 注意力机制 逆光环境 计算机视觉 深度学习 人工智能 目标检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]