SIA OpenIR  > 光电信息技术研究室
基于协方差流形的异常驾驶行为识别方法研究
Alternative TitleAbnormal driving behavior detection based on covariance manifold
李此君1,2
Department光电信息技术研究室
Thesis Advisor刘云鹏
ClassificationTP391.41
Keyword异常驾驶行为识别 图像处理 协方差描述子 黎曼流形 多类LogitBoost分类器
Call NumberTP391.41/L31/2018
Pages55页
Degree Discipline控制工程
Degree Name硕士
2018-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract为了应对因驾驶员因素引发交通事故比例居高不下的现状,辅助安全驾驶技术已经成为当前智能交通领域的研究前沿和研究热点。因此,使用图像处理和模式识别技术,研究一种通过分析驾驶员活动状态对异常驾驶行为正确识别分类的识别方法,对减少由于异常驾驶行为引发的交通事故具有重要的意义。本文研究了一种新的基于协方差流形和多类LogitBoost分类器的异常驾驶行为识别方法。首先对图像进行预处理,通过对图像的预处理以及边缘检测操作,确定方向盘的轮廓及位置,进而得到只包含驾驶员行为活动的图像感兴趣区域,能够提高后续方法识别的精确性,并减少后续算法的计算量。其次提取驾驶行为图像的纹理、颜色和梯度方向等基础特征,以克服基于单一特征识别驾驶行为的不足;并利用协方差流形进行多特征融合,以消除特征冗余,同时降低由于不同特征数值差异过大可能对图像处理及识别带来的影响。最后使用基于二分类器的多类LogitBoost分类器对图片进行分类识别。针对测试数据集,相对传统的直接使用LogitBoost多分类方法,基于二分类思想的多类LogitBoost分类方法较大幅地提高了多分类的正确率。为验证本文所提出的异常驾驶行为识别方法,在MATLAB平台下进行仿真试验,针对测试数据集,平均正确识别率可达81.08%,有效地提高了对多类异常驾驶行为识别的效果。
Other AbstractIn order to cope with the high proportion of traffic accidents caused by driver factors, auxiliary safety driving technology has become the research frontier and research hotspot in the field of intelligent transportation. Therefore, using image processing and pattern recognition technology, this paper studies a recognition method of correct identification of abnormal driving behavior by analyzing driver's active state. It is of great significance to reduce traffic accidents caused by abnormal driving behavior. In this paper, a new anomaly driving behavior recognition method based on covariance manifold and multi class LogitBoost classifier is studied. First, the image is preprocessed. By preprocessing the image and the edge detection operation, the contour and position of the steering wheel are determined, and then the region of interest which contains only the activity of the driver can be obtained, which can improve the accuracy of the follow-up method recognition and reduce the calculation amount of the follow-up method. Secondly, the texture, color and gradient direction of the driving behavior image are extracted to overcome the shortage of driving behavior based on single feature recognition, and the covariance manifold is used for multi feature fusion to eliminate the feature redundancy and reduce the possibility of the image processing and recognition due to the large difference of different features. Influence. Finally, the multi class LogitBoost classifier based on the binary classifiers is used to classify and recognize the pictures. In view of the test data sets, compared with the traditional LogitBoost multi classification method, the multi class LogitBoost classification method based on the binary classifiers has greatly improved the accuracy of multi classification. In order to verify the method of abnormal driving behavior recognition proposed in this paper, the simulation test under the MATLAB platform is carried out. The average correct recognition rate can reach 81.08% for the test data set, which effectively improves the effect of multi class abnormal driving behavior recognition.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/21827
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
李此君. 基于协方差流形的异常驾驶行为识别方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2018.
Files in This Item:
File Name/Size DocType Version Access License
基于协方差流形的异常驾驶行为识别方法研究(2095KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李此君]'s Articles
Baidu academic
Similar articles in Baidu academic
[李此君]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李此君]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.