基于改进YOLO的边缘设备实时目标检测算法研究

Research on real-time target detection algorithm for edge devices based on improved YOLO

针对边缘设备上目标检测模型检测精确度低、检测识别效率慢的问题,提出了一种改进YOLO的边缘设备实时目标检测算法。算法设计了多尺度特征融合模块,此模块利用现有方法中各种特征尺度之间缺失路径的组合进行连接来提高检测精确度。实验结果表明,与传统目标检测部署模型相比,改进YOLO算法在边缘设备上识别速度更快检测精确度更高,为目标检测模型向轻量化平台的移植提供了可能。

This paper proposes an improved YOLO algorithm for real-time target detection on edge devices to address the problems of low detection accuracy and slow detection recognition efficiency of target detection models on edge devices.The algorithm introduces a multi-scale feature fusion module.This module uses the combination of missing paths between various feature scales in existing methods to connect to improve the detection accuracy.Experimental results show that the improved YOLO algorithm identifies faster detection accuracy on edge devices compared to traditional target detection deployment models.This offers the possibility of porting the target detection model to lightweight platforms.