Research On Intelligent Vehicle Environment Perception Algorithm Based On The Data Fusion By Vision And Radar

Due to issues such as safety,congestion,energy consumption,and environmental pollution,intelligent vehicle and automatic driving technology have attracted much attention.Environmental sensing technology integrates a variety of sensor information to accurately sense and deeply understand vehicle environment,and provides information for vehicle intelligent decision-making and control.It is one of the key common technologies for realizing high-level automatic driving,and it is also a bottleneck technology to be solved urgently.This paper focuses on the technology of environmental awareness in vehicle safety driving assistance system.We studied the detection methods of pedestrians and vehicles based on vision and millimeter-wave radar from the perspective of sensor multi-source information fusion.To improve the accuracy of target recognition,detection algorithms and information fusion methods were established based on deep learning.The main content of the paper is as follows:The method of visual-based lane detection was studied.The image preprocessing algorithm was used to extract the lane line information.The combination of gradient and color filter was used to obtain lane line pixels.An efficient method of lane detection was developed based on region of interest using sliding window and polynomial fitting.The results showed that this method can quickly and accurately extract lane lines in high-speed road.Based on directional gradient histogram combined with support vector machine algorithm,an intelligent vehicle detection classifier was designed using a multi-view vehicle sample database,and the performance of the detection classifier was tested under actual working conditions.The results showed that the algorithm basically meets the requirements of vehicle detection under day high speed road.In order to realize the real-time detection of intelligent vehicles' environment perception,multi-objective detection and recognition under multiple operating conditions were considered.A vehicle environment target detection model was built under the framework of Keras and Tensorflow,and an environmental perception target classification detector based on YOLO-v2 depth learning algorithm was designed.The validity of the algorithm was tested by the collected data of vehicle.It improves the accuracy and real-time performance of target recognition in intelligent vehicles' environment perception.Single sensor has its limitations.In order to improve the reliability and real-time performance of target detection,a method of target detection and recognition in front of vehicle based on sensor fusion was proposed.Firstly,a time/space joint calibration model of camera and millimeter-wave radar was established.Then,we studied the millimeter-wave radar filtering algorithm,analyzed the data characteristics of the ESR radar target output.The region of interest was established on the visual image by using the distance,speed and angle information obtained by the radar.Finally,the detection and recognition of target based on millimeter-wave radar and visual information fusion were achieved using vehicle ROI area target consistency detection algorithm.