汽车里程表读数自动识别系统的研究与实现

Research and Implementation of Automatic Recognition System of Automobile Odometer

为了能够自动、准确的识别出汽车里程表的读数,设计了一种基于机器视觉汽车数显式里程表读数自动识别系统。首先对获取的彩色仪表图像提取R通道图,采用最大类间方差算法获取二值化图像,利用基于最大连通域的方法分割出LCD数显式里程表,然后利用双向投影法分割字符、递归细分法字符特征提取,最后设计一种核函数为RBF的支持向量机(SVM)分类器进行数字识别。系统在VC6.0编程环境下运行,大量的试验表明,本算法耗时低、识别率较高。

In order to be able to automatically and accurately identify a car odometer readings, a kind of automaticidentification system based on machine vision and car digital odometer readings. First to obtain color meter image extractionR channel graph, by means of maximum between class variance algorithm to get a binary image, and use based on the Dalianregion of the segmented LCD digital display odometer. Then the two-way projection character segmentation, recursive subdivisionmethod for character feature extraction, finally design a kernel function for RBF support vector machine (SVM) classifierfor digit recognition. System in the VC6.0 programming environment, a large number of tests show that the algorithm is timeconsuming,low recognition rate.