摘要
随着机动车交通违法行为的增多,民众利用智能手机拍照举报式的监督模式应运而生.针对由手机拍照举报的静态图像的车辆识别问题,提出一种基于局部学习的车辆识别方法.与在整个样本空间里训练一个全局模型的传统方法不同,该方法以局部学习中心选取策略和巴氏距离大小为基础,将样本划分若干子集并在每个子集上训练一个局部分类器.仿真结果表明:与已有形状模型法、超像素级别等图像目标识别方法相比,该方法在静态车辆图像识别的问题上拥有更好的识别率和识别效果.
Along with the increase in vehicle traffic violations, a supervising mode that people use smart phones to take pictures and report the illegal phenomenon comes into being. For the problem that recognizes vehicle from the photoes taken by smart phones, a vehicle recognition method based on local learning is proposed. We divide the sample into several subsets basing on center selection policy for local learning and Bhattacharyya distance, then train a local classifier for each subset, which is different from the traditional training on the whole sample space. The simulation results show that, compared with several existing image target recognition method, this method has better recognition rate and a good recognition effect on the recognition problem of vehicle static image.
作者
赵小敏
孙志刚
夏明
ZHAO Xiaomin SUN Zhigang XIA Ming(College of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310023, China)
出处
《浙江工业大学学报》
CAS
北大核心
2017年第4期439-444,共6页
Journal of Zhejiang University of Technology
基金
国家自然科学基金资助项目(61401397)
浙江省科技厅公益资助项目(2014C33073)
关键词
局部学习
超像素
目标识别
车辆识别
local learning
super pixel
object recognition
vehicle recognition