A Study On Vehicle Detection And Localization Based On The Fusion Of Multi-sensors

Nowadays,autonomous vehicle technology is at the forefront and host spot in the field of transportation.It monitors the surrounding environment of the vehicle and the state of the vehi cle through advanced in-vehicle sensor system,information processing systems and execution systems.It can replace the driver to complete various driving operational functions.As one of the most important modules of the autonomous vehicle,vehicle naviga tion system can guide the vehicle how to arrive at the destination effectively from the starting point.Based on the analysis and summary of the existing vehicle detection and localization technology,this paper mainly focuses on the improvement of previous work from the perspective of multi-sensor information fusion.The main works completed are as follows:(1)Front vehicle detection based on the fusion of the millimeter-wave radar and visionIn this system,millimeter-wave radar and vision sensor are calibrated jointly in advance to determine conversion relationship between radar coordinate system and.camera coordinate system.In front objects ' identification processing,the region of interest is firstly formed in figures based on the filtered radar data.Then the shadow feature in the region of interest is extracted to recognize and analyze.Finally,the recognition width is obtained according to the inverse perspective transformation.Moreove r,the results show that the algorithm has high environmental adaptability and recognition accuracy,and would offset the shortcomings of the single sensor.(2)Cooperative vehicle localization method based on the fusion of GPS and VANETIn this paper,a cooperative vehicle localization method is proposed based on the Bayesian framework.The data from GPS sensor can be denoised by the fusion the GPS data and the inter-vehicle distance and bearing angle information,and finally the vehicle position and its m otion trajectory can be obtained combined with the vehicle kinematics equation via kalman filter.In addition to this,the appropriate number of nearby vehicles for cooperative localization is determined while balancing the accuracy and computing burden.At last,for verifying the feasibility and robustness of the all system,then this paper carries out simulation analysis under various multiple experimental scenarios.And the results show that the proposed method provides a well-robustness performance as well as localization accuracy in various experimental scenarios,which can meet the positioning requirements of intelligent vehicles.