使用时打开此例题目录下pic中的图片，然后依次单击按钮“转”、“1”、“2”、“3”、“4”和“5”，就可以实现精确的车牌定位。





具体步骤



1.24位真彩色->256色灰度图。





2.预处理:中值滤波。





3.二值化:用一个初始阈值T对图像A进行二值化得到二值化图像B。



初始阈值T的确定方法是:选择阈值T=Gmax-(Gmax-Gmin)/3,Gmax和Gmin分别是最高、最低灰度值。



该阈值对不同牌照有一定的适应性,能够保证背景基本被置为0,以突出牌照区域。





4.削弱背景干扰。对图像B做简单的相邻像素灰度值相减,得到新的图像G,即Gi,j=|Pi,j-Pi,j-1|i=0,1,…,439 j=0,1,…,639Gi,0=Pi,0,左边缘直接赋值,不会影响整体效果。





5.用自定义模板进行中值滤波



区域灰度基本被赋值为0。考虑到文字是由许多短竖线组成,而背景噪声有一大部分是孤立噪声,用模板(1,1,1,1,1)T对G进行中值滤波,能够得到除掉了大部分干扰的图像C。





6.牌照搜索:利用水平投影法检测车牌水平位置,利用垂直投影法检测车牌垂直位置。





7.区域裁剪,截取车牌图像。-Use the open directory examples the image in the PIC, which in turn click button "turn", "1", "2", "3" and "4" and "5", can achieve precise license plate location.





Specific steps



1.24 a true color-> 256 color gray image.





2. Pretreatment: median filter.





3. Binary: with an initial threshold value of the image on A T binary get binary image B.



The initial threshold value method to determine the T is: select threshold T = Gmax-(Gmax-Gmin)/3, Gmax and Gmin respectively is the maximum and minimum gray values.



The threshold for different licence have certain adaptability, can guarantee that background basic is set to 0, to highlight licence area.





4. Weaken the background interference. Image B do simple adjacent pixels XiangJian grey value, get the new image G, namely Gi, j = | Pi, j-Pi, j-1 | I = 0, 1,... , 439 J = 0, 1,..., 639 Gi, 0 = Pi, 0, the left edge direct assignment, and will not affect the overall effect.





5. Use custom templates by median filter