基于多传感器智能汽车环境感知系统研究

发布时间：2018-05-03 22:05

本文选题：智能汽车 + 环境感知系统　；参考：《南昌航空大学》2017年硕士论文

【摘要】：随着汽车产业高速发展过程中,智能汽车逐渐成为汽车领域和各大汽车厂商关注的新方向。智能汽车主要有感知、决策、控制和路径规划等系统构成,其中感知系统是汽车实现智能驾驶的基础。感知系统是通过环境感知传感器收集车辆周围的环境信息,包括车道线、行人和车辆等识别对象。本文的主要研究内容:(1)搭建基于多传感器的智能清扫实验车辆平台,并设计相应的环境感知系统算法,初步实现环境感知系统功能。在智能清扫实验车辆安装相机、长距离毫米波雷达、短距离毫米波雷达和超声波雷达环境感知传感器,实现实验车辆周围360°范围内目标物的检测和对前方车道线识别功能;(2)研究基于通过车辆CAN总线技术获取毫米波雷达检测到障碍物的原始数据,并设计数据预处理的冒泡筛选和时域滤波算法,实现稳定检测和跟踪目标物,其中数据预处理算法也是多传感器数据融合算法的基础;(3)研究基于多传感器数据融合结构的前方目标物检测算法,并设计适用于结构化道路中传感器级和中央级数据融合结构算法,通过设计的数据融合算法提高前方目标物被检测概率和目标物的物理信息可信度;然后对融合后的目标物进行卡尔曼滤波跟踪,实现对车辆前方目标物被稳定检测跟踪的功能;最后,对智能清扫实验车辆进行环境感知系统测试。在结构化道路中,实验车辆前方分别设置行人、车辆和自行车为被检测对象,测试结果表明本文设计的多传感器数据融合算法能够稳定识别跟踪目标物和提高目标物的物理信息可信度;在实验车辆左右二侧和后方分别设置自行车靠近的运动场景,测试结果表明本文设计的数据预处理算法能够稳定识别跟踪自行车目标对象;另外实验车辆前方相机传感器识别车道线;从而初步实现智能清扫实验车的环境感知系统功能,为整车控制器提供车辆全方位的环境信息,保障车辆的安全行驶。
[Abstract]:With the rapid development of automobile industry, intelligent automobile has gradually become a new direction of attention in automobile field and major automobile manufacturers. Intelligent vehicle is composed of perception, decision making, control and path planning, among which the perceptual system is the basis of realizing intelligent driving. The sensing system collects the environmental information around the vehicle through the environment sensing sensor, including lane line, pedestrian and vehicle identification object. The main research content of this paper is to build a multi-sensor intelligent cleaning experimental vehicle platform and design the corresponding environment awareness system algorithm to realize the function of environment sensing system. Installation of cameras, long-range millimeter-wave radar, short-range millimeter-wave radar and ultrasonic radar environment sensing sensors in intelligent cleaning experimental vehicles, The object detection around the vehicle around 360 掳and the recognition function of the front lane are realized. Based on the CAN bus technology of the vehicle, the original data of the obstacle detected by the millimeter wave radar is obtained. The bubbling filtering and time domain filtering algorithms for data preprocessing are designed to realize the stable detection and tracking of the target. Data preprocessing algorithm is also the basis of multi-sensor data fusion algorithm. The algorithm is designed for sensor level and center level data fusion structure in structured road. The data fusion algorithm is designed to improve the detection probability and the reliability of physical information of the object. Then the fusion object is tracked by Kalman filter to realize the function of detecting and tracking the object in front of the vehicle. Finally, the environment sensing system of the intelligent cleaning experimental vehicle is tested. In structured roads, pedestrians, vehicles and bicycles were placed in front of the experimental vehicles. The test results show that the multi-sensor data fusion algorithm designed in this paper can stably identify the tracking object and improve the reliability of the physical information of the object. The test results show that the data preprocessing algorithm designed in this paper can stably identify and track the bicycle target object; in addition, the camera sensor in front of the experimental vehicle can recognize the lane line, thus realizing the environment sensing system function of the intelligent cleaning experimental vehicle. Provide all-round environmental information for the vehicle controller to ensure the safe driving of the vehicle.
【学位授予单位】：南昌航空大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U463.6;TP212

【参考文献】