Research And Implementation Of Target Location Technology For Smart Lamp-post

With the continuous popularization of 5G communication technology,the Internet of things industry represented by Smart City has ushered in a new chapter of development.As an important part of Smart City,Smart lamp-post plays an irreplaceable role.The multi-sensor information fusion of millimeter wave radar,network camera and other sensors based on Smart lamp-post can not only overcome the performance limitation brought by the use of single sensor,but also carry out real-time monitoring of road traffic information to achieve the purpose of accurate perception and positioning of abnormal events,which is of great significance to improve the level of road traffic safety supervision.In this paper,the Smart lamp-post is used as the application scene,and the millimeter wave radar and camera sensor are fused to detect and locate the target in video surveillance.A target detection and location method based on the fusion of millimeter wave radar and visual sensor is proposed,and a Smart lamp-post monitoring system based on 5G network communication technology is designed and implemented.The main contents of this paper are as follows:(1)Aiming at the limited memory size and computing power of many embedded hardware platforms,this paper designs and implements a real-time target detection and tracking method based on YOLOv3 model compression.Based on the network model of YOLOv3,the model is thinned and pruned,and the lightweight network model of YOLOv3 is obtained.Then the model is combined with the Deep SORT target tracking algorithm,and different scenes are selected to test the tracking effect.Experimental results show that after pruning the model channel,the memory consumption is greatly reduced,and the dependence on the hardware environment is greatly reduced.The algorithm can track and locate the target stably and continuously,and can overcome the occlusion and overlap of the target under certain conditions,so it has good robustness.(2)Aiming at the performance defects of using a single sensor,it is difficult to achieve ideal environmental perception effect.This paper proposes a multi-sensor information fusion target location method based on millimeter wave radar and camera.The data fusion model of millimeter wave radar and camera is built,and the fusion of radar signal and visual signal in space domain and time domain is completed.Finally,based on the proposed target fusion output strategy,the data association matching between radar and camera and the positioning results of radar vision fusion are analyzed.The experimental results show that the target sequence detected by radar and the result detected by camera can be correlated and matched correctly.The new target output strategy combines the advantages of the two sensors and realizes the target detection,location and tracking.(3)Using 5G network communication technology,a kind of Smart lamp-post monitoring system based on 5G network transmission is designed and implemented.This monitoring system takes NVIDIA Jetson Xavier NX as the basic development platform,and completes the construction of RTSP streaming media server through this platform.The system selects RM500Q-GL 5G module of Yi Yuan communication as the intermediate data transmission platform.Finally,combined with target detection tracking algorithm and radar vision fusion algorithm,the whole monitoring system is tested and verified,which proves the effectiveness and stability of RTSP streaming media server and monitoring system.