论文部分内容阅读
研究基于图像识别的CTCS-3级列控系统车载设备人机界面信息的识别方法。首先,根据人机界面信息具有位置固定、字体和大小确定、信息量有限等特点,将人机界面信息划分为图标、数字和字母、汉字3种类型。根据各类信息的特点,分别进行灰度、二值化、切分和归一化等预处理,其中采用改进的基于最大宽度回溯的字切分法提高切分单个汉字的准确度。然后,对图标采用改进的6主色法提取颜色特征,采用统计法提取面积特征;对数字和字母提取欧拉数和8等分面积特征;对汉字提取欧拉数、细化的面积特征和笔画复杂性指数。最后,针对3种类型信息,分别构造决策树对提取的特征进行分类,实现车载设备人机界面信息的识别。以图标和汉字为例的实验结果表明,本文的方法能够准确地实现DMI界面信息的识别。
This paper studies the recognition method of human-machine interface information of vehicle-mounted equipment of CTCS-3 train control system based on image recognition. First of all, based on the characteristics of human-machine interface information such as fixed position, fonts and size determination, and limited amount of information, the human-machine interface information is divided into three types of icons, numbers, letters and Chinese characters. According to the characteristics of all kinds of information, preprocessing such as grayscale, binarization, segmentation and normalization are respectively carried out. The improved word segmentation based on maximum width backtracking enhances the accuracy of segmentation of single Chinese characters. Then, using the improved 6-primary color method to extract the color features, using the statistical method to extract the area features; extracting the Euler number and the 8-point area features from the numbers and letters; extracting the Euler number, refining the area features and Stroke complexity index. Finally, for the three types of information, the decision tree is respectively constructed to classify the extracted features to realize the recognition of the man-machine interface information of the vehicle-mounted equipment. Experimental results using icons and Chinese characters show that this method can accurately identify DMI interface information.