付利军,张福泉,杨金劳.基于信息特征耦合夹角一致性规则的图像匹配算法[J].包装工程,2018,39(9):190-198.
FU Li-jun,ZHANG Fu-quan,YANG Jin-lao.Image Matching Algorithm Based on Information Characteristics and Coupling Angle Consistency Rule[J].Packaging Engineering,2018,39(9):190-198. |
| 基于信息特征耦合夹角一致性规则的图像匹配算法 |
| Image Matching Algorithm Based on Information Characteristics and Coupling Angle Consistency Rule |
| 投稿时间：2017-11-23 修订日期：2018-05-10 |
| DOI：10.19554/j.cnki.1001-3563.2018.09.033 |
| 中文关键词: 图像匹配 Forstner算法 Hessian矩阵 信息特征 距离模型 夹角一致性规则 |
| 英文关键词: image matching Forstner algorithm Hessian matrix information characteristics distance function angle consistency rule |
| 基金项目:国家教育部博士点基金（20121101110037）；山西省自然科学基金（2013011121-1） |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 针对当前基于灰度特征的图像匹配算法在遇到匹配图像存在较大的光照变换时，会引起较多的误匹配和漏匹配等问题，提出一种基于信息特征耦合夹角一致性规则的图像匹配算法。方法 首先，利用Forstner算子来检测图像的特征点，接着用Hessian矩阵最大特征值与其最小的特征值做比值计算，优化Forstner算子的检测特征点。然后，以特征点为原点，构建极坐标系，将特征点的邻域进行分割。再利用信息熵模型求取每个分割块中的信息特征，以生成特征描述子。最后，利用特征描述子构造距离模型，搜索指定特征点的最近邻特征点和次近邻特征点，并通过距离比值方法完成特征点的匹配。通过匹配特征点之间形成的夹角，建立夹角一致性规则，对匹配特征点的可靠性进行度量，剔除错误匹配特征点，从而完成图像匹配。结果 实验结果显示，与当前图像匹配算法相比，所提图像匹配算法图像在旋转角度10~100范围内，识别率为94.6%~88%，平均识别时间为5.48 s，具有更高的匹配精度与鲁棒性。结论 所提算法具有较高的检测精度，在印刷防伪与信息安全等领域具有较好的应用价值。 |
| 英文摘要: |
| The work aims to propose an image matching algorithm based on information characteristics and coupling angle consistency rule regarding the problems of many mismatches and missing matches caused by current image matching algorithm based on grayscale feature when there is a large illumination transformation in matching images. First of all, the Forstner operator was used to detect the feature points of the image, and then the ratio of the maximum eigenvalue to the minimum eigenvalue of the Hessian matrix was calculated to optimize the feature points detected by the Forstner descriptor. Then, the polar coordinate system was constructed with the feature points as the origin, and the neighborhood of the feature points was segmented. The information feature of each segmentation block was obtained with the information entropy model to generate the feature descriptor. Finally, the distance model was constructed by feature descriptor, and the nearest neighbor feature points and the next nearest neighbor feature points of the specified feature points were searched, and the feature points were matched by distance ratio method. By matching the angle formed between the feature points, an angle consistency rule was established to measure the reliability of the matched feature points, and the wrongly matched feature points were eliminated to complete the image matching. The experimental results showed that, compared with the current image matching algorithm, when the rotation angle was 10~100, the recognition rate was 94.6%~88% and the average recognition time was 5.48 s, the proposed image matching algorithm had higher matching accuracy and stronger robustness. The proposed algorithm has high detection accuracy and good application value in the field of printing anti-counterfeiting and information security. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
渝公网安备 50010702501716号