基于超宽选票的图像拼接方法

Image mosaic method for ultra-wide ballot

针对机器视觉的电子选举系统中出现的超宽选票无法被现在主流的电子选举系统进行加载识别的问题,提出一种改进的图像拼接方法。首先,在特征提取部分简化了尺度不变特征变换(SIFT)算法中金字塔的层数以提升特征提取的效率;然后,在图像拼接阶段,先使用近邻(NN)算法进行特征粗匹配,再使用渐进一致采样(PROSAC)算法提高匹配精度,实现图像配准;最后,对重合区域使用渐入渐出融合法进行融合处理,实现了电子选票图像的拼接。由实验结果可以看出,该算法在保证了图像拼接质量的基础上,拼接平均耗时大概为30 ms,拼接效率高,满足了电子选举系统对超宽选票机器采集的需求。

Aiming at the problem that the ultra-wide ballots in the electronic voting system of machine vision can not be loaded and recognized by the current mainstream electronic voting systems,an improved image stitching method was proposed. Firstly,in the feature extraction part,the number of pyramid layers in the Scale-Invariant Feature Transform(SIFT)algorithm was simplified to improve the efficiency of feature extraction. Then,in the image stitching stage,the Nearest Neighbor(NN)algorithm was used for rough feature matching,and the PROgressive SAmple Consensus(PROSAC)algorithm was used to improve matching accuracy and realize image registration. Finally,the overlapped area was merged by using the gradual-in and gradual-out fusion method to achieve the splicing of electronic ballot images. From the experimental results,it shows that the average stitching time of the algorithm is about 30 ms while ensuring the quality of image stitching,which greatly improves the stitching efficiency and meets the needs of the electronic voting system for ultra-wide ballot machine collection.