基于Adaboost的安全带检测方法

发布时间：2018-06-15 05:25

本文选题：安全带检测 + Adaboost算法　；参考：《合肥工业大学》2015年硕士论文

【摘要】：安全带检测是智能交通系统中的一个重要研究课题,该技术的实现可以在很大程度上对那些安全意识淡薄和无视交通法规的驾乘人员起到警告或提醒作用,提高了驾驶员遵守交通法规的意识,从而达到降低交通事故伤亡率的效果。本文主要研究基于Adaboost的安全带检测方法,其主要研究工作和成果如下：1.提出了一种基于Adaboost的安全带检测方法,实现了一个完整的安全带检测系统。在道路监控摄像机采集到道路车辆图像后,图像首先进入车窗区域粗定位模块,找到图像上若干个车窗候选区域,然后进入驾驶员区域粗定位模块,找到图像上若干个驾驶员候选区域,接着将结果送入安全带检测模块,找到若干个安全带检测区域。2.采用SVM分类器对安全带候选区域进行后处理。将利用正面车辆自身几何关系并结合车窗、驾驶员、和安全带候选区域在检测时的置信度来构建基于下文信息的特征向量,然后利用SVM算法进行后处理,最终剔除掉误检区域。3.文中详述了经典的安全带检测方法,从最基本的理论及其算法假设出发,对比分析了各自的优缺点。并在实验环节中,在不同的应用场景、不同的应用难度上进行测试,完成了实验结果对比分析,基于Adaboost的安全带检测方法的优越之处。
[Abstract]:Seat belt detection is an important research topic in intelligent transportation system. The realization of this technology can to a large extent serve as a warning or reminder to drivers who are weak in safety awareness and who ignore traffic laws and regulations. The driver's consciousness of obeying traffic regulations is improved, and the casualty rate of traffic accident is reduced. This paper mainly studies the seatbelt detection method based on Adaboost, the main research work and results are as follows: 1. A seatbelt detection method based on Adaboost is proposed, and a complete seatbelt detection system is implemented. After the road vehicle image is captured by the road surveillance camera, the image first enters the coarse location module of the window area, finds several window candidate areas on the image, and then enters the coarse location module of the driver area. Several candidate driver regions were found in the image, and then the results were put into the seat belt detection module to find a number of seat belt detection areas. 2. SVM classifier is used for post-processing of seat belt candidate area. The feature vector based on the information below will be constructed by using the geometric relationship of the frontal vehicle and the confidence degree of the window, driver and seat belt candidate area in the detection, and then the SVM algorithm will be used for post-processing, finally eliminating the false detection region. In this paper, the classical seatbelt detection methods are described in detail, and their advantages and disadvantages are compared and analyzed based on the most basic theory and algorithm hypothesis. And in the experiment link, in the different application scene, the different application difficulty carries on the test, has completed the experiment result contrast analysis, based on the Adaboost seatbelt detection method superiority.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP391.41

【相似文献】