Data Acquisition And Processing System Of Driverless Vehicular Sensors Based On ARM-Linux

Driverless vehicles are the future development direction of the automobile industry at home and abroad,and have an extremely important strategic position in many fields such as driving safety,social and economic benefits,and scientific and technological development.In order to improve the perception accuracy of the driverless system and ensure driving safety,this paper conducts research on the key technologies in the data acquisition and processing system of drivless vehicular sensors,the main research contents are as follows:(1)Design a computing platform with corresponding interfaces based on the types of driverless vehicular sensors.First,analyze the requirements of the driverless perception system to determine the hardware design.Then,use the "Jetson TX2 core board + carrier board" method to carry out detailed hardware circuit design for the power management module,communication interface module,etc.Finally,perform component layout and wiring work to complete the production and welding of the system prototype.(2)Design the overall software framework of the data acquisition and processing system for driverless vehicular sensors based on the distribute concept of ROS.Firstly,take the vehicle body coordinate system as the reference to complete the calibration of each sensor.Secondly,solve the problems of noise and outliers in the lidar point cloud data,and combine the k neighborhood filtering idea,an improved voxel grid point cloud filtering algorithm is proposed,which effectively solves the real-time problem in the original algorithm.Thirdly,aiming at the problem that the image information of the camera is not clear in a low-light environment,the perception accuracy is reduced,this paper is based on the multi-scale Retinex image enhancement algorithm fusion bilateral filtering,and perform dimensionality reduction on it,effectively solve the loss of detail,noise and lack of real-time performance in low-illuminance images.Fourthly,complete spatio-temporal registration based on the lidar sampling frequency,then,complete the lidar-camera pixel-level fusion,and input sensor data with temporal and spatial consistency.(3)Build an experimental platform and conduct real-vehicle experiments in different driving scenarios,the result shows: The hardware computing platform and the software framework based on ROS are running normally;The improved point cloud and image data processing algorithm and multi-sensor spatio-temporal registration method in this paper effectively improve the target detection rate and improve the perception accuracy;The data processing software method of dviverless vehicular sensors designed in this paper basically meets the real-time needs.