基于颜色特征及投影变换的集装箱标识符分割

Container Code Segmentation Based on Color Features and Projective Transformation

集装箱标识符的自动识别对实现电子通关、无人值守有重要作用。然而集装箱图像受环境、拍摄、运动等影响,其标识符的准确定位及分割有一定难度。本文在分析集装箱图像的基础上,设计了一套标识符定位分割方法。本方法首先对集装箱图像进行基于颜色特征的预处理,完成图像的二值化;利用水平投影及字符区域面积、字符个数的限定等,实现噪声区域的剔除和标识符区域的准确定位;采用Radon变换进行倾斜矫正;最后结合水平与垂直投影实现单个标识符的分割。实验结果表明,本文的方法能够很好地实现标识符的快速精确定位和分割,有效解决了复杂背景下准确定位难的问题,为字符的识别打下了良好基础。

The automatic identification of container identifiers plays an important role in the realization of electronic customs clearance and unattended checkpoint. However, the container image is easy to be affected by the environment, shooting and motion, and it is difficult to locate and segment the identifiers accurately. In this paper, based on the analysis of the container image, a set of identifier location and segmentation method is designed. Firstly, preprocessing based on color feature is used to process the image to get the binary image. Secondly, the horizontal projection combining the limitation of the identifier area and the identifiers number and so on, is utilized to eliminate the noise areas and locate the identifier area accurately. And then Radon transformation is used to correct the image tilt. At last, level and vertical projections are adopted to segment the single identifier. Experimental results show that this method can achieve fast, precise location and segmentation of identifiers, effectively solve the problem of locating accurately under complex background, and establish a good foundation for characters recognition.