| 摘 要: 随着道路监控系统的数字化和智能化发展，车辆类型识别成为智能交通系统的研究重点之一。针对从道路监控系统中获取的视频图像，考虑如何从图像中提取车标局部区域的显著特征进行分析，提出了联合特征的车标特征点提取和识别方法。基于道路监控系统中的视频图像，对定位的车标图像进行归一化处理，完成了车标的方向梯度直方图特征(HOG)和局部二值模式(LBP)提取，采用支持向量机(SVM)对车标特征矢量进行分类识别。通过从监控视频中分割出来的10 类车标、826 张车标图像对算法效果进行评估，实验结果表明，支持向量机的核函数选择线性核函数，识别率达到95%，优于径向基核函数和多项式核函数。比较了单一特征(HOG或LBP)与联合特征(HOG-LBP)对车标的识别率，联合特征对车标的识别率达到97.27%，识别率最高。基于HOG-LBP联合特征车标区域的智能识别方法，同时利用HOG与LBP的特征优势，提高了车标识别率。 |
| 关键词: 图像识别；车标识别；方向梯度直方图；局部二值模式；支持向量机 |
|
中图分类号: TP311.5
文献标识码: A
|
| 基金项目: 国家自然科学基金面上项目(61671239)；2021年江苏省现代教育技术研究(2021-R-93916). |
|
| Research on Intelligent Recognition Method based on Vehicle Logo Area |
|
QU Aiyan1,2, WU Qiuling1, ZHANG Zheng1, LIANG Yinghong1, HUANG Xiaoting3
|
( 1.Jinling Institute of Technology, Nanjing 211169, China ;
2.Army Engineering University, Nanjing 210001, China ;
3.Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China)
quaiyan@jit.edu.cn; Wuqiuling@jit.edu.cn; zhangzheng@jit.edu.cn; liangyh@jit.edu.cn; 934246903@qq.com
|
| Abstract: With the development of digital and intelligent road monitoring systems, vehicle type recognition has become one of the research focuses of intelligent transportation systems. This paper proposes a method of extracting and identifying feature points of vehicle logo with joint features, after considering how to extract and analyze the significant features of the local part of the vehicle logo from images obtained from road monitoring system. Based on the video image in the road monitoring system, the positioning of vehicle logo is normalized, and the directional gradient histogram feature (HOG) and local binary pattern (LBP) of the vehicle logo are extracted. The support vector machine (SVM) is used to classify the vehicle logo feature vector. 10 types of vehicle logos segmented from surveillance videos and 826 vehicle logo images are used to evaluate the effect of the algorithm. Experimental results show that the recognition rate of linear kernel function is 95%, which is better than radial basis function kernel function and polynomial kernel function. Recognition rate of joint feature (HOG-LBP) for vehicle logo is 97.27%, which is the highest, compared to recognition rate of single feature (HOG or LBP). Intelligent recognition method improves vehicle logo recognition rate based on the HOG-LBP joint feature vehicle logo area and the feature advantages of HOG and LBP. |
| Keywords: image recognition; vehicle logo recognition; directional gradient histogram; local binary mode; support vector machine |
