| With the increasing number of cars,the traffic situation are becoming more and more complex,people’s demand for driving assistance is growing.The advent of automatic driving assistance can help the driver to deal with specific driving environment under different conditions.But the frequent accidents of automatic driving make people worry about the safety of automatic driving.In many automatic driving accidents,most of the accidents are caused by the wrong recognition of the target,resulting in the wrong judgment of driving performance,and finally lead to the accident happens.Therefore,for the two mainstream target recognition hardware,radar and camera,it is very necessary and to make use of their respective advantages to make up for the lack of single hardware performance,carry out target fusion and screening,and improve the target recognition algorithm of the two hardware.This thesis analyzes the current situation of target recognition in the field of automatic driving in domestic and abroad.I introduce the system architecture of Baidu Apollo automatic driving platform and the selection of target recognition hardware by Tesla and Google.Regarding hardware and target recognition algorithms used by different automatic driving companies,I decide to use middle range radar and micro camera as object detection hardware according my working experience.I develop a whole radar and camera system software and design the target object fusion module.The existing target recognition algorithm is improved.According to the target detected by the sensor,the calculation formula of the possibility of target existence is designed.According to the detected target trajectory,the expected driving trajectory of the vehicle is planned.According to the self driving trajectory and the possibility of the target,the targets related to the self driving trajectory are finally selected to complete the process of target recognition.After the software algorithm development process is completed,flash the whole target recognition system software into the radar and camera hardware,and install it on the real vehicle for public road test.During the road test,we record the real target recognition data and analyze the scene of target loss or target false recognition,and adjust the logic of target screening to make the whole target recognition algorithm more accurate in practical application.Finally,the whole sensor system is more accurate and effective,which is more helpful and safer for the vehicle to realize the automatic driving assistance process. |
