Study On Nighttime Vehicle Detection Technology In Intelligent Assisted Driving

Smart city is a advanced form of urban informatization,which makes full use of the new generation of information technology in all walks of life in the city based on knowledge society.Its purpose is to realize the deep integration of informatization,industrialization and urbanization,and intelligent transportation is a very important application scenario.Intelligent transportation system(ITS)is an advanced method to improve transportation problems.As the main body of transportation,intelligent driving is an important part of its research.At this stage,there is still a certain distance from fully automatic driving,but many vehicles have begun to be equipped with Advanced Driver-Assistance Systems(ADAS),and vehicle detection technology is an important part of ADAS.Compared with the daytime vehicle detection,the nighttime vehicle detection has the problems of low field brightness and low contrast between the background and the object.It is not easy to detect the vehicle,which makes the nighttime vehicle detection more challenging.Moreover,the actual traffic environment is complex and changeable,and traffic accidents occur frequently at night.Nighttime vehicle detection at night is very important for ADAS.In recent years,with the rapid development of deep network,many studies show that deep network can also achieve good results in the field of object detection.However,some deep networks suitable for daytime vehicle detection can not achieve good performance at night.For driving at night,the vehicle light is an essential part,so the vehicle light information can be used to assist vehicle detection.In this paper,we propose a deep network scheme assisted by light information with good generalization to detect vehicle at night.Our approach is divided into two branches,the object stream and the pixel stream.The object stream generates a batch of bounding boxes,and the pixel stream utilizes the vehicle light information to calibrate the bounding boxes of the object stream.In the object stream,we propose a new structure,Direction Attention Pooling(DAP),to improve the accuracy of the prior boxes.DAP leads into attention mechanism.The feature maps obtained from backbone network is divided into two branches.One branch obtains direction perception information through IRNN layer,and the other branch learns attention weights.The weights are multiplied with the direction perception features in an element-wise manner.In the pixel stream,we propose a corner localization algorithm based on Bayes to get more accurate corners with the vehicle light pixels.The locations of the corners are considered as a discrete random variable.When the mask of the object is known,solving the probability distribution of the corner of the object.The corners with the highest probability is the correct corner.Experiments show that the proposed algorithm can achieve good performance in nighttime vehicle detection,and can achieve good performance in different datasets,which shows that it has generalization ability.