车牌识别 [英] car plate recognition

问题描述

任何想法如何从jpeg图像中提取车牌号码？

我试着写这段代码但我需要一些帮助.....
介意与我分享......源代码会很好.... [叹气]

any idea how to extract car plate number from jpeg image??

i trying to write this code but i need some assistance.....
mind to share with me...a source code would be nice....[sigh]

推荐答案

不幸的是，你不会得到这样做的代码，因为这是一个非常复杂的区域，商业系统需要巨额资金。通常，ANPR系统使用几个阶段来识别车牌，这些阶段是：



1.车牌位置。此步骤负责识别图片中的印版位置。

2.印版翻译。定位和调整图像大小，补偿偏斜。

3.图像标准化。调整图像的亮度/对比度。

4.分割。查找字符本身（即删除标记等项目）。

5.光学字符识别。



现在，ANPR要求摄像头拍摄图像速度非常快，并且要求处理速度非常快，这意味着.NET可能不是处理图像的最合适选择（通常需要亚秒级处理）。通常这些系统是用C ++或C编写的，经过高度优化，并利用图像处理技术，如边缘检测和锐化滤波器。

Unfortunately, you aren't going to be given the code to do this as this is a hugely complicated area, with commercial systems going for huge amounts of money. Typically ANPR systems make use of several stages to identify the number plate, these stages being:

1. Plate location. This step is responsible for identifying where the plate is in a picture.
2. Plate translation. Orienting and sizing the image, compensating for skew.
3. Image normalisation. Adjusting the brightness/contrast of the image.
4. Segmentation. Finding the characters themselves (i.e. removing items such as flags).
5. Optical Character Recognition.

Now, ANPR requires cameras to take pictures at very high speeds, and requires the processing to take place very quickly, which means that .NET is probably not going to be the most appropriate choice for processing the image (which typically requires sub-second processing). Typically these systems are written in C++ or C, are highly optimised and make use of image processing techniques such as edge detection and sharpening filters.

嗯，大多数车牌都是用标准字体完成的

如果你可以创建一个图像/向量数组来显示每个可能的角色的框架

（即F的骨架只是基本的线条），然后让它搜索这个形状在车牌的每个字母内。

您必须考虑的是照明，拍摄照片的角度，可能的不同尺寸，汽车速度等。

寻找框架将让您获得非常准确的测量结果，你可以针对具体问题制定对策（即车牌上的图像等）

问候
James

Well, most car plates are done in a standard font

If you could create an array of images / vectors showing the framework for each possible character

(ie the skeleton of an F would be just the basic lines), and then had it search for this shape within each letter in the license plate.

What you have to allow for are lighting, the angle at which the photo was taken, possible different sizes, speeds of car etc.

Looking for a framework will allow you a pretty accurate measurement, and you can make countermeasures against specific problems (ie images on license plates etc)

Regards
James

嗨jianloong

关注这个

1-找出这个盘子有多少颜色，假设它是红色和白色的2，然后在黑色背景上重绘你的图像

2-最喜欢的盘子不平衡我的意思是它应该以一个小角度倾斜或旋转，所以你必须将它旋转回来，直到你得到宽度与水平线对齐

你可能还需要将它倾斜

3-你应事先知道盘子的尺寸，使用这些数据你可以将盘子缩放到它的正常尺寸，你可以知道你的相机有多远，除非你有aradar

4- supp ming盘子有3个数字和3个字符，你应该将宽度分成6（3digit +3 char）段我将参考这个作为6个sprits

5-每个sprit存储器中的数据2D更多likley char数组

6-记住盘子的内部颜色是红色所以我们应该搜索每个有红色的像素

7知道颜色的颜色像素我们可以很容易地计算它的位置x，y

8你应该为所有数字存储数据和存储所有alphabtic的数据

9 - 你需要比较每个sprite与那些存储的数据

10-你应该知道光线不是同等浓度的alawys

所以这会累积你的数据错误，忘记夜班工作

11-确保在慢速计算之前汽车不会ESCAPE

所以你应该选择选择怀特语言和最好的街道我不建议高速公路这个摄像机应该低速度街

12-最多重要的是错误结果的可能性和获得错误数量的其他汽车。我相信它应该与你的计划相切

你需要大量的研究

khalid sabtan

saudia arabia





11 -

Hi jianloong
follow this
1- find out how many color the plate has ,assuming it is cosist of 2 red and white,then redraw your image on black backcolor
2- most likley the plate will not be balanced i mean it should be tilted or rotated at some small angle so you have to rotatet it back untill you get the width aligne with horizontal line
you may need to skew it too
3- you should know in advance the dimension of the plate ,using this data you can scale the plate to its normal dimension you can realize how far is the car from your camera unless if you have aradar
4- assuming the plate has 3 digits and 3 character,u should divide the width to 6 (3digit +3 char ) segment i will refere to this as 6 sprits
5- for each sprit store the data in 2D more likley char array
6- remember the inner color of the plate is red so we should be searching for each pixel that has a color red
7 knowing the color of the pixel we can easily calculate its location x,y
8 you should have data stored for all digits and data which store all alphabtic
9 - you need to compare each sprite with those stored data
10- you should be carfull that light is not alawys at same concentration
so this will accumulate errors to your data ,forget about night work
11- make sure the car will not ESCAPE before your slow calculation
so you should be selective in chosing the wright language and the best street i do not recomend highway this camera should be in low speed street
12- the most important thing is the likleyhood of wrong result and obtainning wrong number of other cars . which i beleive it should be tangent to your program
you need alot of research
khalid sabtan
saudia arabia


11-

这篇关于车牌识别的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！