Research On Vehicle-Road Collaborative Perception Fusion

In today’s society,cars are no longer a luxury for people,but a daily travel item.The popularity of automobiles has greatly improved the travel efficiency,but the accompanying problems such as traffic congestion and safety accidents have also occurred frequently,and have become a worldwide problem.Although the transportation infrastructure has been continuously improved,it has never been able to meet the needs of human beings.Since the 1970 s,researches on traffic intelligence and vehicle automation have been carried out at home and abroad.Therefore,the birth of intelligent traffic system is inevitable.Intelligent traffic system aims to provide more and more accurate information to traffic participants and traffic managers through technologies such as information technology and network communication.With the continuous development of sensor technology,computer technology and communication technology,the internet of vehicles and autonomous driving have also become a hot research field in the world today.In the process of realizing autonomous driving,one of the most crucial research module is the cooperative vehicle-Infrastructure system,connecting the vehicle end and the road end through the wireless network to realize the exchange and sharing of the vehicle end information and the road end information.Especially for the car end,the traffic environment information perceived by the car end may have visual blind spots and visual limitations,etc.,and it is impossible to fully perceive the surrounding environment.The information received from the roadside through wireless communication enables the vehicle to obtain a full range of traffic environment information,thereby providing more accurate traffic environment information for the vehicle’s next driving behavior.For the vehicle-road coordination system,the speed of obtaining them and the accuracy of the perception results are the indicators that need to be focused on in the research.Therefore,this paper conducts research experiments on the speed of perceptual fusion and the accuracy of perceptual fusion results.The specific contents are as follows:This paper proposes a perceptual fusion algorithm based on hash partitioning in the vehicle-road coordination system,and performs spatial compensation on the fusion results.The perceptual fusion algorithm used in this paper is Non-Maximum Suppression.The most accurate perceptual target result is obtained by solving the values of the two target candidate frames in turn for all candidate frames in this frame.Since the candidate boxes with large geographic ranges cannot be the same target,it is meaningless to compare them.Therefore,this paper proposes to use the idea of hashing to perform hash partitioning according to the position of the center point of the target candidate frame.Through the fusion of the first partition and the second partition twice,the fusion result can reduce the fusion time while ensuring the basic fusion effect.It is verified by comparative experiments that when the number of fusions is 253,the fusion rate of secondary partition fusion is 92.00%,and the fusion time is 34.18 times lower than that of the direct fusion method without partition.And as the order of magnitude of the sensing target increases,the fusion speed increases more,and the efficiency of sensing fusion under vehicle-road coordination is greatly improved.In addition,there is a certain delay from the time when the vehicle and road sensors perceive the traffic environment information to the completion of the fusion,which reduces the timeliness of the perception data.Therefore,in order to make the fusion result more accurate,this paper proposes to perform spatial compensation for this time delay,and according to the motion state of the target,the spatial compensation of the target in the x and y directions is performed respectively.It is verified by comparative experimen ts that when the speed reaches 50km/h,the accuracy of the fusion result after spatial compensation is increased by 4.05 times.In summary,the perceptual fusion algorithm based on hash partitioning and spatial compensation in the vehicle-road coordination system proposed in this paper can provide vehicles with faster and more accurate traffic environment information.