基于相关向量机和方向导数的车辆识别
|
摘要
由于图像匹配需预先组建庞大的图片库以及图像识别对灰度梯度的严重依赖性,提出利用相关向量机和灰度的方向导数对车辆进行识别。针对不同车道数将背图像作为图片库的基元,对图片进行编码;对采集到的图像进行车道线提取,确搜索域;并进行相应的网格划分,依据灰度均值确网格的属性标签,从而建立图片的特征向量,将原图与模板特征向量的2范数作为表征指标,并利用相关向量机对车辆进行分类,从而对车辆进行位;对提取的车辆区域灰度进行函数拟合,通过求解函数极值确车辆的中心区域,并建立相应的搜索方向,将满足双阈值的像素点作为最终的轮廓边缘,从而实现对车辆轮廓边缘的细化。用大量真车试验图片对模型进行测试,结果表明:所建模型仅需有限的图片库,搜素量小,识别耗费时间短,准确率高;模型对图像灰度梯度的依赖性小,且通过双阈值有效剔除伪轮廓点,细化车辆边缘。
Because large picture base must be initially established for image match and image recognition is seriously dependent on gray scale gradient,this paper adopts relative vector machine and direction derivative function of gray scale to recognize vehicle.For different lane numbers,background images are taken as the elements of picture base and all imanges are encoded.The search region can be determined by extracting lane line in the collected image.In addition,the image is divided into some grids and the attribute signs of grids can be determined according to the mean of gray scale.Then characteristic vectors are established.The two norm of characteristic vectors between the initial image and the template is to be as index.And then, relative vector machine is adopted to classify vehicles and locate vehicle.The gray scale in vehicle region extracted from the initial image is fitting by functions.The centre of vehicle region can be determined by extreme value solution of function and the corresponding search direction is built.The pixels which meet double thresholds are taken as the last outline edge and the thinning edge of vehicle outline can be obtained from it.A large number of real road vehicle samples are used to test the model and the results show that the presented model has only limited picture base and search capacity is small.Furthmore,the recognition consuming time of this model is short and has higih accuracy.The model is little dependent on gray scale gradient and the fake outline pixels can be removed by double thresholds.In addition,the thinning vehicle outline edge is accomplished.
Because large picture base must be initially established for image match and image recognition is seriously dependent on gray scale gradient,this paper adopts relative vector machine and direction derivative function of gray scale to recognize vehicle.For different lane numbers,background images are taken as the elements of picture base and all imanges are encoded.The search region can be determined by extracting lane line in the collected image.In addition,the image is divided into some grids and the attribute signs of grids can be determined according to the mean of gray scale.Then characteristic vectors are established.The two norm of characteristic vectors between the initial image and the template is to be as index.And then, relative vector machine is adopted to classify vehicles and locate vehicle.The gray scale in vehicle region extracted from the initial image is fitting by functions.The centre of vehicle region can be determined by extreme value solution of function and the corresponding search direction is built.The pixels which meet double thresholds are taken as the last outline edge and the thinning edge of vehicle outline can be obtained from it.A large number of real road vehicle samples are used to test the model and the results show that the presented model has only limited picture base and search capacity is small.Furthmore,the recognition consuming time of this model is short and has higih accuracy.The model is little dependent on gray scale gradient and the fake outline pixels can be removed by double thresholds.In addition,the thinning vehicle outline edge is accomplished.
引文
[1]You Feng Wang Rongben&Zhang Ronghui.Algorithm on Lane Changing and Tracking Control Technology for Intelligent Vehicle[C].Proceedings of the 2007 IEEE International Conference on Robotics and Biomimetics,December 15-18,2007,Sanya,China,2007:1888-1893
[2]Eui Yoon Chung,Ho Choul Jung,Eugene Chang,eatl.Vision Based for Lane Change Decision Aid System[C].IFOST2006,Oct.18-Oct.20,2006:10-13
[3]Wei Liu,XueZhi Wen,Bobo Duan,etal.Rear Vehicle Detection and Tracking for Lane Change Assist[C].Proceedings of the 2007IEEE Intelligent Vehicles Symposium Istanbul,Turkey,June 13-15,2007:252-257
[4]Bing-Fei Wu,Wei-Hsin Chen,Chai-Wei Chang,etal.A New Vehicle Detection with Distance Estimation for Lane ChangeWarning Systems[C].Proceedings of the 2007 IEEE Intelligent Vehicles Symposium Istanbul,Turkey,June 13-15,2007:698-703
[5]邓天明,于勇,邵毅明.一种新的车辆图像识别分类算法[J].重庆交通大学学报:自然科学版,2008,27(06):1142-1145
[6]ViolaP,JonesM.Rapid Object Detection Using a Boosted Cascade of Simple Feature[C].IEEE Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2001:511-518
[7]FriedmanJ,HastieT,TibshiraniR.Additive Logistic Regression:a Statistical View of Boosting[J].The Annals of Statistics,2000,28(2):337-407
[8]张亮修,王玉林,张鲁邹.一种树形结构的实时道路车辆识别方法[J].青岛大学学报:工程技术版,2008,23(03):49-54
[9]连静,王珂,杨兆升.多尺度图像边缘检测技术在车辆识别中的应用[J].公路交通科技,2007,24(9):7-11
[10]杨树仁,沈洪远.基于相关向量机的机器学习算法研究与应用[J].计算技术与自动化,2010,29(1):43-47
[11]TZIKAS D,LIKAS A,GALATsANOs N.Large scale multikernel RVM for object detection[J].Lecture notes in cornputer science,2006,3955:389-395
[12]CAMPS VALLS G,MARTNEZ RAMN M,ROJO LVAREZ J L,etal Nonlinear System Identification With Composite Relevance Vector Machines[J].IEEE signal processing letters,2007(14):279-298
[13]李娜.贝叶斯分类器的应用[J].北京工业职业技术学院学报,2008,7(02):7-10
[2]Eui Yoon Chung,Ho Choul Jung,Eugene Chang,eatl.Vision Based for Lane Change Decision Aid System[C].IFOST2006,Oct.18-Oct.20,2006:10-13
[3]Wei Liu,XueZhi Wen,Bobo Duan,etal.Rear Vehicle Detection and Tracking for Lane Change Assist[C].Proceedings of the 2007IEEE Intelligent Vehicles Symposium Istanbul,Turkey,June 13-15,2007:252-257
[4]Bing-Fei Wu,Wei-Hsin Chen,Chai-Wei Chang,etal.A New Vehicle Detection with Distance Estimation for Lane ChangeWarning Systems[C].Proceedings of the 2007 IEEE Intelligent Vehicles Symposium Istanbul,Turkey,June 13-15,2007:698-703
[5]邓天明,于勇,邵毅明.一种新的车辆图像识别分类算法[J].重庆交通大学学报:自然科学版,2008,27(06):1142-1145
[6]ViolaP,JonesM.Rapid Object Detection Using a Boosted Cascade of Simple Feature[C].IEEE Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2001:511-518
[7]FriedmanJ,HastieT,TibshiraniR.Additive Logistic Regression:a Statistical View of Boosting[J].The Annals of Statistics,2000,28(2):337-407
[8]张亮修,王玉林,张鲁邹.一种树形结构的实时道路车辆识别方法[J].青岛大学学报:工程技术版,2008,23(03):49-54
[9]连静,王珂,杨兆升.多尺度图像边缘检测技术在车辆识别中的应用[J].公路交通科技,2007,24(9):7-11
[10]杨树仁,沈洪远.基于相关向量机的机器学习算法研究与应用[J].计算技术与自动化,2010,29(1):43-47
[11]TZIKAS D,LIKAS A,GALATsANOs N.Large scale multikernel RVM for object detection[J].Lecture notes in cornputer science,2006,3955:389-395
[12]CAMPS VALLS G,MARTNEZ RAMN M,ROJO LVAREZ J L,etal Nonlinear System Identification With Composite Relevance Vector Machines[J].IEEE signal processing letters,2007(14):279-298
[13]李娜.贝叶斯分类器的应用[J].北京工业职业技术学院学报,2008,7(02):7-10
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |