基于SIFT的鞋印图像特征提取及配准研究

发布时间：2018-05-11 00:41

本文选题：图像配准 + SIFT　；参考：《河南理工大学》2015年硕士论文

【摘要】：随着科学技术的飞速发展,对案发现场证据采集及检测工作有了更高的要求。鞋印图像自动识别技术是利用计算机对犯罪现场遗留下的鞋印进行处理、分析和识别,从而协助侦破案件。本文将对基于SIFT(Scale Invariant Feature Transform)的鞋印图像特征提取及图像匹配开展深入研究,主要工作如下:针对鞋印图像的定位,提出一种二次定位法。首先对鞋印图像进行预处理,根据鞋印图像特征,运用顶点角度偏转法进行第一次偏转定位;在此基础上,采用中垂线法对其进行二次定位,最终使鞋印基本处于同一竖直位置。针对SIFT算法在描述特征点时出现的数据量大、计算复杂等问题,本文提出一种改进的SIFT算法来对鞋印图像进行特征匹配:一方面,通过二次定位法对鞋印图像进行提前旋转,从而省去SIFT算法中为保证旋转不变性所产生的大量运算;另一方面,在特征向量匹配部分,以最小欧氏距离作为判断特征点是否匹配的标准,再通过最小欧氏距离与次小欧氏距离之比来剔除无效匹配点,SIFT特征匹配效率得到提高。通过实验验证,本文提出的鞋印二次定位算法与改进的SIFT算法,在保证图像配准率和算法鲁棒性的前提下,既缩短了配准时间,也提高了图像的配准精度。
[Abstract]:With the rapid development of science and technology, there are higher requirements for the collection and detection of evidence in crime scene. The automatic recognition technology of shoeprint image is to use computer to process, analyze and identify the shoeprints left by the crime scene, so as to assist in the detection of cases. In this paper, the feature extraction and image matching of shoeprint image based on SIFT(Scale Invariant Feature Transform) are deeply studied. The main work is as follows: aiming at the location of shoeprint image, a secondary location method is proposed. First of all, the shoeprint image is preprocessed, and the vertex angle deflection method is used to locate the first time according to the characteristics of the shoeprint image, and on this basis, the secondary location is carried out by using the method of mid-vertical line. Finally, the shoe printing is basically in the same vertical position. Aiming at the problems of large amount of data and complicated calculation in describing feature points in SIFT algorithm, this paper proposes an improved SIFT algorithm to match the feature of shoeprint image: on the one hand, the shoeprint image is rotated ahead of time by the second location method. In addition, in the feature vector matching part, the minimum Euclidean distance is taken as the criterion to judge whether the feature points match or not. Then the sift feature matching efficiency is improved by the ratio of the minimum Euclidean distance to the sub-Euclidean distance to eliminate the invalid matching points. The experimental results show that the proposed secondary location algorithm and the improved SIFT algorithm not only shorten the registration time but also improve the registration accuracy under the premise of ensuring the image registration rate and robustness of the algorithm.
【学位授予单位】：河南理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41

【相似文献】