汽车牌照定位系统设计与开发 (文档和资料)
汽车牌照定位系统设计与开发 (文档和资料),汽车牌照定位系统设计与开发,(文档和资料),汽车,牌照,定位,系统,设计,开发,文档,以及,资料
编号无锡太湖学院毕业设计(论文)相关资料题目: 汽车牌照定位系统设计与开发 信机 系 计算机科学与技术专业学 号:学生姓名:指导教师: 2013年5月25日目 录一、毕业设计(论文)开题报告二、毕业设计(论文)外文资料翻译及原文三、学生“毕业论文(论文)计划、进度、检查及落实表”四、实习鉴定表无锡太湖学院毕业设计(论文)开题报告题目: 汽车牌照定位系统设计与开发 信机 系 计算机科学与技术 专业学 号:学生姓名:指导教师: 2012年11月20日 课题来源导师指定科学依据(包括课题的科学意义;国内外研究概况、水平和发展趋势;应用前景等)汽车车牌识别系统是近几年发展起来的计算机视觉和模式识别技术在智能交通领域应用的重要研究课题之一。在车牌自动识别系统中,首先要将车牌从所获取的图像中分割出来实现车牌定位,这是进行车牌字符识别的重要步骤,定位的准确与否直接影响车牌识别率。研究内容车辆牌照定位与识别是计算机视觉与模式识别技术在智能交通领域应用的重要研究课题之一,该技术应用范围非常广泛,其中包括:(1)交通流量检测;(2)交通控制与诱导;(3)机场、港口等出入口车辆管理;(4)小区车辆管理;(5)闯红灯等违章车辆监控;(6)不停车自动收费;(7)道口检查站车辆监控;(8)公共停车场安全防盗管理;(9)计算出行时间;(10)车辆安全防盗、查堵指定车辆等。其潜在在市场应用价值极大,有能力产生巨大的社会效益和经济效益。国外最早提出的是在20世纪80年别系统成功被研制,其主要途径就是对车牌的图像进行分析,自动提取车牌信息,进而确定汽车车牌号码。在现代,以色列Hi-Tech公司的See/Car System系列,新加坡Optasia公司的VLPRS 系列都是比较成熟的产品。其中,VLPRS产品主要适合与新加坡的车牌识别,Hi-Tech公司的See/Car System有多种变形产品来分别适应于某一个国家的车牌识别。See/Car Chinese系统也可以针对中国大陆的车牌进行识别,但是存在着一定的局限性,不能较好的识别车牌中的汉字,另外日本、加拿大、德国、意大利、英国等国家都有适合于本国的车牌识别系统。国内在20世纪90年代已经开始对车牌识别系统进代,该阶段并没有形成完整的理论系统知识,而是对车牌识别的某一种特定环境的应用或某一个问题进行了讨论,采用简单的图像处理技术来解决问题。国外正式开始出现车牌识别系统化的研究是20世纪90年代以后。1990年,国外的第一个车牌识行了相关的研究,其中的北京汉王公司的“汉王眼”、成都西图科技有限公司生产的CIAST2003车牌识别稽查系统、亚洲视觉生产的VECONVIS车辆牌照识别系统以及等产品牌照识别率都达到了95以上。上海交通大学戚飞虎提出了基于彩色分割的拍照识别方法;华中科技大学黄心汉提出了基于模板匹配和神经网络的拍照识别方法。另外,西安交通大学的郑南宁等人提出了多层次纹理分析的牌照识别方法也对车辆牌照识别系统也有极大价值的研究。众多的牌照识别技术的研究促进了适合我国车辆牌照识别产品的问世,国内的牌照识别产品相继问世并且投入使用。拟采取的研究方法、技术路线、实验方案及可行性分析本次毕业设计首先对车牌识别系统的现状和已有的技术进行了深入的研究在此基础上设计并开发了一个基于MATLAB的车牌定位系统通过编写MATLAB文件对各种车辆图像处理方法进行分析、比较,最终确定了车牌预处理、车牌定位的方法。本次设计采取的是基于Sobel的边缘检测先从经过边缘提取后的车辆图像中提取车牌特征,进行分析处理,从而初步定出车牌的区域,再利用车牌的先验知识和分布特征对车牌区域平滑图像进行处理,从而得到车牌的精确区域,并且取得了较好的定位结果。研究计划及预期成果2011年12月4日以前:填写毕业设计开题报告,并按开题报告条款进入毕业设计阶段2011年12月2012年1月:外文资料翻译,系统设计2012年2月:系统设计、编码2012年3月2012年4月:测试、验收,撰写毕业论文2012年5月:上交论文、系统代码、根据导师意见修改毕业论文并完善论文2012年6月2日4日,进行毕业答辩预期成果:每步安排都可以按时完美完成特色或创新之处本次设计采取的是Sobel边缘检测,先从经过边缘提取后的车辆图像中提取车牌特征,进行分析处理,从而初步定出车牌的区域,再利用车牌的先验知识和分布特征对车牌区域进行处理,从而得到车牌的精确区域并且取得了较好的定位结果。已具备的条件和尚需解决的问题已具备的条件:1、硬件方面有一台计算机,一台相机2、软件方面配有MATLAB平台3、网络方面有数条数据线可用指导教师意见 指导教师签名:年 月 日教研室(学科组、研究所)意见 教研室主任签名: 年 月 日系意见 主管领导签名: 年 月 日英文原文1.IntroductionNowadays license plate recognition becomes akey technique to many automated transport systemssuch as road traffic monitoring, automaticpayment of tolls on highways or bridges and parkinglots access control. License plate location is anessential and important stage in this technique,and it has received considerable attention.Researchers have found many diverse methods of license plate location. Rodolfo and Stefano(2000) devised a method based on vector quantization (VQ). VQ image representation is a quadtree representation by the specific coding mechanism,and it can give a system some hints about the contents of image regions, and such information boosts location performance. Park et al. (1999)used neural networks to locate license plate. Neural networks can be used as filters for analyzing small windows of an image and deciding whether each window contains a license plate, and their inputs are HSI values; a post-processor combinesthese filtered images and locates the bounding boxes of license plates in the image. Besides neural networks, other filters have been considered too. For example, some authors used line sensitive filters to extract the plate areas. License plates are identified as image areas with high density of rather thin dark lines or curves. Therefore, localization is handled looking for rectangular regions in the image containing maxima of response to these line filters, which is computed by a cumulative function (Luis et al., 1999). Plate characters can be direct identified by scanning through the input image and looking for portions of the image that were not linked to other parts of the image.If a number of characters are found to be in a straight line, they may make up a license plate (Lim et al., 1998). Fuzzy logic has been applied to the problem of locating license plate by Zimic et al. (1997). The authors made some intuitive rules to describe the license plate, and gave some membership functions for the fuzzy sets bright and dark, bright and dark sequence to getthe horizontal and vertical plate positions. But this method is sensitive to the license plate color and brightness and needs much processing time. Using color features to locate license plate has been studied by Zhu et al. (2002) and Wei et al. (2001), but these methods are not robust enough to the different environments. Edge features of the car image are very important, and edge density can be used to successfully detect a number plate location due to the characteristics of the number plate. Ming et al. (1996) developed a method to improve the edge image by eliminating the highest and lowest portions of the edge density to simplify the whole image. But some of the plate region identity will be lost in this method.This paper further researches the subject of license plate location. The rectangle license platecontains rich edge and texture information, so we consider it in its edge image but very different toMing et al. (1996). We first enhance the original car image to boost up the plate area, then extractthe vertical edge image using Sobel operator, and then remove the background curves and noise inthe edge image, and finally slide a rectangle window to search the plate in the residual image andsegment it out from the original car image. Section2 describes our method of license plate location, and it contains four parts: image enhancement, vertical edge extraction, background curve andnoise removing, plate search and segmentation. Experiments with three sets of car images are performed in Section 3. Section 4 gives the discussion and conclusions.1. The proposed method for license plate locationAll the input car images have 384 288 pixels and 256 gray levels, and an example image is given in Fig. 1. The license plate of the car consists of several characters (such as Latin letters, Arabic numerals, etc.), so the plate area contains rich edge information. But sometimes the background of the car image holds much edge information too. There are two facts that attract our attention: one is that the background areas around the license plate mainly include some horizontal edges; the other is that the edges in the background are mainly long curves and random noises, whereas the edges in the plate area cluster together and produce intense texture feature. If only the vertical edges are extracted from the car image (although the plate will lose a little horizontal edge information, this little loss is to be valuable) and most of the background edges are removed, the plate area will be isolatedout distinctly in the whole edge image. Thus we propose to locate the license plate in its vertical edge image as the following four stages.2.1. Image enhancementIn Fig. 1, the gradients in the license plate area are much lower than those in the contour areas of the car, which is caused by the car shadow in the dazzling sunshine. The car images captured in the gloomy days or dim nights often bring out weak gradients in plate areas too. A few vertical edges will appear in the plate areas, if we extract edge images directly from these car images. Therefore it is important to enhance the car images firstly.The local areas that need to be enhanced in a car image have low variances. Here we use Ii,jto denote the luminance of the pixel Pi,j (row:0 6 i 288, column: 0 6 j 384) in the car image,and use I1i;j to denote the luminance in the enhanced image. We let Ii,j and I1 i;j satisfy Eq. (1), where Wi,j is a window centered on pixel Pi,j, IW i;j and rW i;j are the mean luminance and standard deviation of the pixels in the window Wi,j, I0 andr0 are the expected mean and standard deviation,respectively. (1)In order to represent the local information better, the size of the window should be smaller than the estimated size of the plate. In this paper, we select a 48 36 rectangle as the window Wi,j and thus 8 8 windows can cover over the whole 384 288 car image. Let I0 be equal to IW i;j and r0 be a constant independent of pixel Pi,j. Now we need to know the values IW i;j and rW i;j at each pixel. Computing out all the values is not advisable, and we can use the bilinear interpolation algorithm to get them. First we cut the car image into 8 8 blocks equably; and then compute out the IW i;j and rW i;j values at the vertexes of blocks, where i = 36m, j = 48n,m,n = 0,1,2, . . . ,8; finally compute out every IW i;j and rW i;j by the bilinear interpolation Eqs. (2) and (3) (Fig. 2), where 36m 6 i 36(m + 1), 48n 6 j 0) Mi,j = maxMi1,j1,Mi1,j,Mi1,j+1, Mi,j1 + 1; else Mi,j = maxMi2,j1,Mi2,j,Mi2,j+1, Mi1,j2,Mi1,j+2,Mi,j2 + 1; end end end end3. for each row i from bottom-to-top do for each column j from right-to-left do if (Ei,j= =1) if (Ei+1,j1 + Ei+1,j + Ei+1,j+1+Ei,j+1 0) Ni,j = maxNi+1,j1,Ni+1,j,Ni+1,j+1, Ni,j+1 + 1; else Ni,j = maxNi+2,j1,Ni+2,j,Ni+2,j+1, Ni+1,j2,Ni+1,j+2,Ni,j+2 + 1; end end endend4. for each row i from top-to-bottom do for each column j from left-to-right do if (Ei,j= =1) if (Mi,j + Ni,j TlongkMi,j + Ni,j Tshort) Ei,j = 0; end end endendIn the above algorithm, we accumulate the edge lengths through observing the concerned neighborhood pixels (CNP) of the current pixel Pi,j. Fig. 7 shows the CNP in shadow grids.The result (Fig. 8) shows that most of the background and noise edges have been eliminated, but the license plate edges are almost fully saved.2.4. License plate search and segmentationAfter non-plate edges have been heavily removed, license plate location becomes much easier.We can shift a rectangle window whose size is just bigger than that of the license plate (forexample 80 32) from left-to-right and top-to-bottom in the edge image. Count the total number of the edge points in the window. If the number is above a certain percentage of the area of the window, there may be a license plate in the corresponding window.In order to expedite the search process, we let the window shifted by some Xstep and Ystep (for example Xstep = 8,Ystep = 8) instead of by pixel.And the following four steps tell how to locate the license plate: Cut the 384 288 edge image into many 8 8 blocks equably, and count the number of edge points in each little block, and form a 48 36 block image B. Use a (80/8) (32/8) = 10 4 matrix W (each element is equal to 1) to denote the window.Convolve the image B with the window W,and export the image B0. If is above the threshold and is the local maximum, then record position as one of the plate candidates. Search all the candidates in their local regions,sort them by their B values and segment them out from the original car image.The convolution result is shown in Fig. 9(a). The response in the plate center is far higher than the other areas, so only one candidate will be searched in general. The plate segmented from the car image in Fig. 1 is shown in Fig.9(b). The rest tasks are plate distortion correction, characters cutting and characters recognition, but they are not discussed in this paper.3. Experimental resultsIn this section we compare the performance ofthe proposed method against some other used methods: line sensitive filters (Luis et al.,1999), row-wise and column-wise DFTs (Parisi et al., 1998), and edge image improvement(Ming et al., 1996). The vector quantization(Rodolfo and Stefano, 2000) is mainly used for imagecoding; and color feature (Wei et al., 2001; Zhu et al., 2002) is not robust enough for weather conditions, extra lights or dirty plates; and fuzzy logic approach (Zimic et al., 1997) works well under the assumption that the majority of plates are white with black characters, while most of Chinese license plates are blue with white characters. Therefore, these three methods were not employed in our comparative experiments.The line sensitive filters method consists of three steps: subsampling image, applying line sensitive filters, and looking for rectangle regions containing maxima of response. The row-wise and column-wise DFTs method involves four steps: decomposing expected harmonics by using horizontal DFT on the image, averaging the harmonics in the spatial frequency domain, finding the horizontal stripe of the image containing the plate by maximizing the energy, and finding the vertical position of the plate in the same way by using vertical DFT on the candidate stripes. And the edge image improvement method contains five steps: extracting the edge image using Sobel operators, computing the horizontal projections of the edge image, calculating the medium range of the edge density level, eliminating the highest and lowestportion of the horizontal projections to simplify the whole image, and finding the candidates of license plates.Three sets of Chinese vehicle images were used in our experiments. The first set has 163 images, and they were captured on a gate of our campus.The second set has 218 images, and they were captured in the shadow of strong sunlight near a road. The third set has 784 images, in which there are many complex backgrounds such as trees, parked bicycles and so on, and these images were taken from morning till night.We performed all the four methods on the three sets. The Table 1 shows the experimental results. For each image, we located 1 3 plate candidates, and the numbers of license plates hit by the first, the second and the third candidates are listed in Table 1, respectively. From this table we can see that the proposed method outperforms much over the other three methods: most of the license plates are found by the first candidates, and the location rates are almost all 100% (The two missing plates in the third set are too small). And the high location rates on the three sets reveal the robustness and efficiency of our method in license plate location.For the other three methods, most mislocated plates happened with the images containing some special objects (brands, radiators, bumpers) or complex backgrounds (trees, bicycles), and the images captured against the strong sunlight or under the gloomy light.The computational times of the four methods are shown in Table 2, when they run on a Pentium-4 2.4 GHz, 256 MB RAM PC. The proposed method is the slowest one among the four methods.The average processing times for the four stages of the proposed method are listed in Table 3. A lot of the time is consumed on the first stage image enhancement. The total time of processing one 384 288 image is 47.9 ms, and it meets the requirement of real time processing.4. ConclusionThe proposed method of license plate location makes use of the rich edge information in the plate area. In Section 2.1, we enhance the local areas in the original car image, but it is an alternative to enhance the gradient image to intensify the texture of the plate region. To avoid the interference factors around the plate, only vertical edges are extracted in Section 2.2. If the vertical edges, the left diagonal edges and the right diagonal edges are all extracted, a better continuity in edge curves can be attained at the expense of more computation time. The survival isolated short edges in Fig. 8 remain to be eliminated but not necessary.The method has still some drawbacks. In Section 2.1, the IW i;j and rW i;j calculated by bilinear interpolation algorithm are not the actual values at the point Pij. The integral image algorithm can solve this problem, but it takes much computation time too. Some numbers in this paper are relative to the estimated size of the license plate.So if all the license plates in the images have the same size, the method will work better.A great effec
收藏