Research On Multi-sensor Data Fusion Algorithm And Deployment Strategy For Pedestrian Detection

With the rapid development of economy and automobile industry,there has been a significant increase in vehicle ownership,which has caused serious traffic problems,including inefficient operation of traffic system,worrisome traffic safety,serious environmental pollution,intensified energy consumption,severely hindering the continuous and rapid development of economy and society.Key to build a comprehensive sensing system of pedestrian,vehicles,roads,and the Internet,smart roads will resolve the aforementioned problems by greatly promoting the construction of intelligent transportation.Meanwhile,with the gradual development of 5G technology,higher transmission rates,lower latency and larger communication capacity can effectively support the development and popularization of smart roads,and meet the needs of the scenarios.As an important component of smart roads,pedestrian detection system is considered an effective measure to improve road safety,and widely applied in intelligent driving assistance,intelligent monitoring and big data analysis.Different from the traditional machine vision solution which is complex and expensive,this thesis proposes a roadside pedestrian detection syetem with lower cost and wider coverage based on the Internet of Things(Io T)technology.In this syetem,multi-sensor data fusion,large-scale sensor network are used to realize accurate ientification and early warning of pedestrians’ abnormal behavior of crossing the road.In addition,the optimal deployment of large-scale wireless sensing networks will directly improve the sensing and detection of the entire networks.The deployment of detection nodes can optimize network resource and ensure accurate environmental awareness and information acquisition,which is the basis of the entire system to achieve efficient and accurate detection.Therefore,in view of the analysis and requirements above,this study consists of the following two aspects.On the one hand,the urban traffic scenario fatures ever-changing road environment,complex traffic condition,dense building layout and ever-changing density of traffic flow,among others.This greatly increases the difficulty of designing pedestrian detection systems both in real-time and accurate pedestrian identification and accident avoidance.Starting from the reliability and adaptability of pedestrian detection systems,this thesis comprehensively utilizes Doppler microwave radar sensors,passive infrared sensors,geomagnetic sensors and other Io T sensing devices,and uses multi-sensor data fusion technology to enhance the real-time performance of the system and information utilization rate to extend the survival time and space coverage rate of the entire system.The pedestrian detection system based on multisensor data fusion algorithm enables each sensor to unleash its own advantages,enhances the system’s detection accuracy and improves its adaptability to complex road environment.On the other hand,in large-scale sensing networks,equipment’s deployment strategy influences the networks’ sensing ability in the entire deployed region,and ultimately affects the detection performance of the whole sensing networks.Therefore,based on the theoretical research and system construction of the above-mentioned pedestrian detection system,this research sets the deployment scenario as a rectangular monitoring region.Given the initial deployment of the sensing nodes and the sensing model of module nodes in this region,the problem of node redeployment is solved using the minimum exposure path algorithm.By designing the pedestrian detection module based on the minimum exposure path algorithm to optimize the deployment strategy,the coverage quality of the entire monitoring region on moving pedestrian targets is improved.Simulation results show that the optimized deployment strategy can improve the deployment effect of pedestrian detection module in the entire monitoring region and better the coverage quality of the entire sensing network.