郁　浩,张荣福,程金光,郭世平

(上海理工大学 光电信息与计算机工程学院,上海　200093)

基于图像处理的钢包标牌识别方法

郁浩,张荣福,程金光,郭世平

(上海理工大学 光电信息与计算机工程学院,上海200093)

摘要由于钢厂车间环境的复杂性对钢包标牌的实时检测和识别提出了较高的要求。为了实现对钢包的实时和动态跟踪,提出了一种基于HSV色彩空间信息的图像分割方法,根据HSV色彩空间的颜色和亮度无关性,将彩色图像从RGB空间变换到HSV空间。基于颜色和形状特征选取原则对标牌区域进行粗定位,再使用最小二乘直线拟合的方法检测标牌边框的倾斜角度,对标牌进行倾斜校正。实验表明,该方法能准确地进行标牌定位与倾斜校正,具有较高的可靠性与稳定性。

关键词HSV色彩空间;彩色图像处理;倾斜校正

Ladle License Plate Recognition Based on Image Processing

YU Hao,ZHANG Rongfu,CHENG Jinguang,GUO Shiping

(School of Optical-Electrical and Computer Engineering,University of Shanghai for

Science and Technology,Shanghai 200093,China)

AbstractThe complexity of the steel plant environment leads to high requirements for real-time detection and identification of ladle license plate.An image segmentation method based on theHSVcolor space information is proposed for real-time detection and dynamic tracking.According to theHSVcolor space,regardless of the relation of brightness and color,the color image is converted fromRGBspace toHSVspace.We can locate the position of the ladle license plate according to feature selection principles based on color and shape of the numeral cards,and the lean angle of image is found out according to the method of linear least squares fitting.Then the original oblique image can be adjusted by using the method of rotate transformation.Experiments show that the method can locate the Ladle license plate and correct tilt with high reliability and stability.

KeywordsHSVcolor space;color image process;adjustment

在现代炼钢工艺流程中,钢厂车间环境的复杂性对于钢包标牌的识别是一个具大的挑战。为了能够更准确地识别钢包标牌,钢包标牌统一做成蓝底白字的形式。

标牌识别系统包括标牌定位、倾斜校正和数字识别3部分,标牌定位是识别系统的关键技术,对识别起着重要作用。由于标牌图像都是采集于背景比较复杂的车间环境,这给标牌的定位带来困难。目前车牌定位主要有基于车牌底色[1]、基于车牌纹理特征[2]以及基于特征颜色边缘[3]的定位方法。

本文考虑到钢包标牌的特点,在借鉴车牌定位算法的基础上,综合考虑车间环境和各种方法的复杂度及定位效果,提出了一种基于HSV颜色空间的标牌定位方法。……