MS-YOLO: Object Detection Based on YOLOv5 Optimized Fusion Millimeter-Wave Radar and Machine Vision

Millimeter-wave radar and machine vision are both important means for intelligent vehicles to perceive the surrounding environment. Aiming at problem of multi-sensor fusion, this paper proposes object detection method millimeter-wave fusion. Radar camera complement each other, data fusion in network can effectively reduce rate missed under insufficient light conditions, improve accuracy remote small detection. The information is processed by mapping transformation neural obtain mask map, so that visual same scale. A multi-data source deep learning (MS-YOLO) based on was proposed. Homemade datasets were used training testing. This maximized use sensor improved premise ensuring speed. Compared with original YOLOv5 (the fifth version You Only Look Once) network, results show MS-YOLO meets requirements better. Among models, large model has highest an mAP reaching 0.888. good speed, reaches 0.841 while maintaining a high frame 65 fps.