基于图像拼接的并行车牌识别系统

发布时间：2019-03-20 17:06

【摘要】：智能交通系统是当前智能城市建设的重要组成部分。智能交通系统同时利用了多种先进技术,如信息通信技术,传感器技术,电子控制技术以及计算机技术等[1],其目的在于全方位,多层次,大范围的确保综合运输和管理系统的实时性,准确性和高效性。车牌识别在智能交通系统中是十分重要的一个部分。传统的,如今广泛应用的车牌识别系统(LPR)包括图像预处理、车牌定位与分割、字符的切割与识别、输出显示等四个环节。该识别系统在需要单张依次输出时工作出色,当今已普遍应用于智能收费,智能停车等领域。然而,当车牌数量增多,大量摄像头下的车辆截图都需要在短时间内被及时识别时,传统的车牌识别系统就会效率低下,急需新的算法体系来解决效率问题。基于此种大量车牌图像需要被识别的场景,本文研究并提出了一种基于图像拼接的并行车牌识别算法,并通过大量的数据验证了其可行性。具体步骤如下:(1)依次定位并提取每张图片的车牌。针对城市摄像头拍下的车牌图像,利用基于边缘检测以及跳变点的车牌粗定位、细定位算法,准确定位出每张待识别车牌的位置并提取。(2)车牌拼接。设每张图像的宽度为w,高度为h,那么对于N张车牌图像,利用图像拼接算法将其横向拼接成一幅大的图像。新图像的宽度为W*N,是原图像的N倍,而高度仍然为h,保持不变。(3)字符切分与识别。主要利用投影法,实现对长度为w*N,高度为h的图像的车牌字符切分,采用支持向量机实现字符样本的训练和识别。(4)系统输出。识别过后,每7个字符输出一次,分N次共输出7*N个字符。最后,本文对以上新提出基于图像拼接的并行车牌识别系统进行了模拟仿真,并对结果进行综合分析。
[Abstract]:Intelligent transportation system is an important part of the current intelligent city construction. The intelligent transportation system utilizes a variety of advanced technologies, such as information and communication technology, sensor technology, electronic control technology and computer technology, etc. Its aim is to be all-round and multi-level. Ensure the real-time, accuracy and efficiency of the integrated transportation and management system on a large scale. License plate recognition is a very important part of intelligent transportation system. The traditional and widely used license plate recognition system (LPR) includes image preprocessing, license plate location and segmentation, character cutting and recognition, output display and so on. The recognition system has been widely used in the fields of intelligent charging, intelligent parking and so on. However, when the number of license plates increases and a large number of vehicle screenshots under the camera need to be recognized in a short time, the traditional license plate recognition system will be inefficient, and a new algorithm system is urgently needed to solve the efficiency problem. Based on the scene in which a large number of license plate images need to be recognized, a parallel license plate recognition algorithm based on image splicing is studied and proposed in this paper, and its feasibility is verified by a large number of data. The concrete steps are as follows: (1) locate and extract the license plate of each picture in turn. Aiming at the license plate image captured by the city camera, the location of each license plate to be recognized is accurately located and extracted by the rough location and fine location algorithm based on edge detection and jump point. (2) license plate stitching. If the width of each image is w and the height of each image is h, then for N license plate images, the image splicing algorithm is used to concatenate them horizontally into a large image. The width of the new image is N times that of the original image, and the height of the new image is still h, keeping the same. (3) character segmentation and recognition. In this paper, the projection method is used to segment the license plate character of the image with length and height h, and support vector machine is used to train and recognize the character samples. (4) the output of the system is given. After recognition, every 7 characters output once, N times output a total of 7 N characters. Finally, a new parallel license plate recognition system based on image splicing is simulated and the results are analyzed.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U495;TP391.41

【相似文献】