面向复杂运动场景的目标检测技术研究
Research on Object Detection Technology for Complex Motion Scenes
郑泊文 1李佩阳 1陆华才1
作者信息
- 1. 安徽工程大学 电气工程学院,安徽 芜湖 244061
- 折叠
摘要
针对复杂交通场景下运动目标检测所存在的问题,文章在Yolov5 模型的基础上,提出了二次优化改进聚类Anchor Boxes,引入SE模块算法,以达到提高聚类算法迭代速度的目的,通过Retinex图像增强算法改善图像或视频帧的质量,以提升在夜晚或雾、霾等恶劣天气下对运动目标的识别精度和准确度.实验结果表明,所提出的改进Yolov5模型对运动目标检测平均检测精度有一定的提升,相较于原始Yolov5 模型,改进后的算法在准确率和召回率方面分别提升了 5.8%和2.3%.
Abstract
Aiming at the problems existing in moving object detection in complex traffic scenes,based on the Yolov5 model,this paper proposed a secondary optimization to improve cluster Anchor Boxes and introduced SE module algorithm to improve the iteration speed of clustering algorithms.Retinex image enhancement algorithm was used to improve the quality of image or video frames,so as to improve the precision and accuracy of the moving target recognition in the night or fog,haze,and other bad weather.Experimental results show that the proposed improved Yolov5 model can improve the average detection accuracy of moving target detection to a certain extent.Compared with the original Yolov5 model,the accuracy and recall rate of the improved algorithm are improved by 5.8%and 2.3%respectively.
关键词
运动目标/YOLOv5模型/聚类算法/注意力机制/图像增强Key words
motion target/YOLOv5 model/clustering algorithm/attentional mechanism/image enhancement引用本文复制引用
基金项目
检测技术与节能装置安徽省重点实验室项目(DTEST2022A01)
安徽工程大学研究生教育创新基金(JCKJ2022C01)
出版年
2024
NETL
NSTL
万方数据