基于视频分析技术的车距测量及预警系统设计  

Design of a Vehicle Distance Measurement and Early Warning System Based on Video Analysis Techniques

作　　者：王传钦[1] 曹江涛[1] 姬晓飞[2] 

机构地区：[1]辽宁石油化工大学信息与控制工程学院,辽宁抚顺113001 [2]沈阳航空航天大学自动化学院,辽宁沈阳110136

出　　处：《计算机技术与发展》2016年第9期87-90,共4页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61103123;61203021)

摘　　要：车距测量及预警是汽车主动安全技术中的一个重要组成部分，而基于视觉的车距测量及预警系统一直是智能车系统和辅助安全系统中研究的热点。为了提高车距测量的精确度和实时性，以Visualc＋＋6．0集成开发环境和OpenCV开源计算机视觉库为实验平台，设计并实现了一种基于视频分析技术的车距测量及预警系统。该系统具有车辆检测、车辆跟踪、距离测量及预警等功能。以Haar-like特征作为图像描述，结合Adaboost算法训练分类器实现道路中车辆的检测；采用CamShift和Kalman相结合的方法实现目标车辆的跟踪及预测；提出一种基于RBF神经网络的车距测量及预测方法。实验结果表明，该系统能较准确地实现1～15m范围内的车辆检测及车距测量，且具有良好的实时性。Vehicle distance measuring and early warning is an important component in vehicle active safety technology. And the vision- based vehicle distance measuring and early warning system has been a research hotspot of the intelligent vehicle system and secondary safety system. In order to improve the accuracy and processing speed of vehicle distance measurement,vehicle distance and early warning system is designed and implemented based on video analysis techniques in Visual C＋＋6.0 and OpenCV software environment,which has functions like vehicle detection, vehicle tracking, distance measurement and early warning and so on. Haar-like features is chosen as im- age descriptions, and Adaboost algorithm is combined to train classifiers to achieve vehicles detection. A combination method of CamShift and Kalman is used to track the target vehicle, and the measurement and prediction of distance is achieved by using RBF neural network. Experiments show that the system can accurately realize the range of 1 ~ 15 m for vehicle detection and distance measurement,and it has good real-time performance.

关 键 词：车辆检测 HAAR-LIKE特征 ADABOOST算法 跟踪 RBF神经网络 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]