线缆字符图像的拼接方法研究

摘要

针对线缆字符图像光照不均匀、重叠区域较小、边缘化等问题,现有的图像拼接方法不能准确拼接,提出一种新的图像拼接方法.首先利用HSI彩色空间模型从图像中划分出待配准的线缆区域.然后提出基于线缆光照模型的图像分块算法对线缆区域图像进行分块,并构造相似检测公式对图像块进行分类,对不含字符的图像块直接赋值为白色背景块,对含字符的图像块利用大津法进行处理,得到线缆区域二值图.接着提出以字符宽度为特征的图像配准算法获得重叠区域在两幅线缆区域图像的行对应关系.最后以重叠区域的水平中心直线拼接得到一幅包含完整字符区域的宽图像.实验结果表明,相对于当前主流的FREAK算法,提出的方法能准确地对线缆图像进行拼接,拼接效果良好.%Aiming at the problems such as non-uniform illumination, small overlap area and marginalization, the existing stitching methods can not be accurately stitched, and a new stitching method is proposed.Firstly, the HSI color space model is used to divide the cable region to be registered from the image.Then, an image segmentation algorithm based on the cable illumination model is proposed to segment the image of the cable region, and a similar detection formula is constructed to classify the image blocks.The image blocks without characters are directly assigned to white background blocks, and the image blocks are processed by the Otsu''s algorithm to obtain the cable region binary image.Furthermore, an image registration algorithm based on the character width is proposed to obtain the row-to-line correspondence between the overlapping areas in the two cable region images.Finally, a horizontal center line of the overlapping area is stitched to obtain a wide image containing the entire character region.Experimental results show that the proposed method can stitch the cable image accurately, and the splicing effect is good compared with the current mainstream FREAK algorithm.