基于智能手机的自动车标识别系统
发布时间：2018-11-25 12:57
【摘要】：近年来,我国汽车的保有量迅猛增长,交通运输面临着越来越大的压力和挑战,智能交通是现代交通运输业的发展方向。车辆识别是智能交通中的关键技术,具有重要的商业价值和现实意义。将车牌识别技术与车标识别技术结合,能够获取更准确的车辆信息,提高车辆识别系统的识别率和鲁棒性。本文的车标识别系统分为车标定位和车标分类两个阶段。在车标定位阶段,使用基于车牌先验知识的由粗到精的定位方法,先根据车牌在RGB彩色空间和HSV彩色空间的颜色特征,定位车牌位置。然后根据车牌与车标的相对拓扑关系粗略定位车标。最后检测车标粗定位区域图像边缘,将水平边缘图像和垂直边缘图像相与,消除车标周围栅格的干扰,通过数学形态学开运算滤掉了边缘图中的噪声。最后使用边缘投影和图像重心的方法,以重心为起始位置,在水平和垂直投影上分别定位车标上下左右边界。在车标分类的阶段,首先把车标图像归一化为32×32,从中提取出144维的HOG特征,输入SVM线性分类器判断车标类别。利用以上车标识别方法,本文使用Open CV4Android机器视觉库,首次在Android智能手机平台设计开发了车标识别移动应用程序。其中,使用OpenCV Manager管理OpenCV动态链接库,通过CvCameraViewListener接口获取到了摄像头图像。结果表明,本文采用的车标识别算法快速、有效,车标识别移动应用程序的平均处理时间小于0.2秒,满足实时性的要求。
[Abstract]:In recent years, with the rapid growth of automobile ownership in China, traffic and transportation are facing more and more pressure and challenge. Intelligent transportation is the development direction of modern transportation industry. Vehicle identification is a key technology in intelligent transportation, which has important commercial value and practical significance. The combination of license plate recognition technology and vehicle sign recognition technology can obtain more accurate vehicle information and improve the recognition rate and robustness of vehicle recognition system. The identification system is divided into two stages: vehicle sign location and vehicle mark classification. In the phase of vehicle mark location, the location of vehicle license plate is located according to the color features of the license plate in RGB color space and HSV color space based on the priori knowledge of license plate. Then according to the relative topological relationship between the license plate and the vehicle sign, the vehicle logo is roughly located. Finally, the edge of the rough location area image is detected, the horizontal edge image and the vertical edge image are combined with each other, the interference of grid around the vehicle mark is eliminated, and the noise in the edge image is filtered by mathematical morphology. Finally, the edge projection and the image barycenter are used to locate the upper and lower sides of the vehicle mark on the horizontal and vertical projection, taking the center of gravity as the starting position. In the stage of vehicle mark classification, the vehicle mark image is normalized to 32 脳 32, and the 144-dimensional HOG feature is extracted from it, and the SVM linear classifier is input to judge the vehicle mark category. Using the above identification method, this paper designs and develops the mobile application program of vehicle logo recognition on Android smart phone platform using Open CV4Android machine vision library for the first time. Among them, OpenCV Manager is used to manage the OpenCV dynamic link library, and the camera image is obtained through the CvCameraViewListener interface. The results show that the algorithm used in this paper is fast and effective, and the average processing time of the mobile application is less than 0.2 seconds, which meets the requirement of real-time.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495
本文编号：2356187
[Abstract]:In recent years, with the rapid growth of automobile ownership in China, traffic and transportation are facing more and more pressure and challenge. Intelligent transportation is the development direction of modern transportation industry. Vehicle identification is a key technology in intelligent transportation, which has important commercial value and practical significance. The combination of license plate recognition technology and vehicle sign recognition technology can obtain more accurate vehicle information and improve the recognition rate and robustness of vehicle recognition system. The identification system is divided into two stages: vehicle sign location and vehicle mark classification. In the phase of vehicle mark location, the location of vehicle license plate is located according to the color features of the license plate in RGB color space and HSV color space based on the priori knowledge of license plate. Then according to the relative topological relationship between the license plate and the vehicle sign, the vehicle logo is roughly located. Finally, the edge of the rough location area image is detected, the horizontal edge image and the vertical edge image are combined with each other, the interference of grid around the vehicle mark is eliminated, and the noise in the edge image is filtered by mathematical morphology. Finally, the edge projection and the image barycenter are used to locate the upper and lower sides of the vehicle mark on the horizontal and vertical projection, taking the center of gravity as the starting position. In the stage of vehicle mark classification, the vehicle mark image is normalized to 32 脳 32, and the 144-dimensional HOG feature is extracted from it, and the SVM linear classifier is input to judge the vehicle mark category. Using the above identification method, this paper designs and develops the mobile application program of vehicle logo recognition on Android smart phone platform using Open CV4Android machine vision library for the first time. Among them, OpenCV Manager is used to manage the OpenCV dynamic link library, and the camera image is obtained through the CvCameraViewListener interface. The results show that the algorithm used in this paper is fast and effective, and the average processing time of the mobile application is less than 0.2 seconds, which meets the requirement of real-time.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495
【参考文献】
相关期刊论文 前3条
1 钟志伟;徐贵力;王彪;郭瑞鹏;田裕鹏;李开宇;;基于PCA和不变矩的三维物体识别算法研究[J];计算机与现代化;2013年03期
2 李侠;李文举;孙娟红;郑宏亮;;基于模板匹配和边缘检测的车标定位方法[J];计算机工程与设计;2011年10期
3 李文举;孙娟红;韦丽华;李侠;;基于2DPCA-ICA和SVM的车标识别新方法[J];辽宁师范大学学报(自然科学版);2011年02期
，本文编号：2356187
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2356187.html
