自适应特征点检测的可见—红外图像配准

发布时间：2018-06-06 01:46

本文选题：自适应特征点检测 + 可见—红外图像配准　；参考：《中国图象图形学报》2017年02期

【摘要】：目的针对可见—红外图像之间配准点的数量不足、分布严重不均匀以及配准点之间的错配率高这3个核心问题,提出一种基于自适应特征点检测的可见—红外图像配准方法。方法本文提出的自适应特征点检测方法,以Harris corner作为基本特征点;以特征点数目与空间分布为检测目标,从而自动地估计合适不同空间位置的特征点的检测阈值。在特征点对匹配中,将梯度方向与互信息相融合有效地添加了相似性函数的空间位置信息。结果自适应Harris corner检测方法能够有效地提供空间分布均匀、数量充足的特征点。而梯度方向与互信息相融合的相似性匹配函数提高特征点的匹配率20%,降低配准误差50%。结论本文提出的多传感器图像配准方法能够快速、准确地实现可见光图像与红外图像之间的配准,在CCD-IR图像融合领域具有很好的实用价值。
[Abstract]:Aim to solve the three core problems such as insufficient number of matching points between visible and infrared images, serious uneven distribution and high mismatch rate between the matching points, a new method for registration of visible infrared images based on adaptive feature point detection is proposed. Methods the adaptive feature point detection method proposed in this paper takes Harris corner as the basic feature point and the number and spatial distribution of the feature point as the detection target, thus automatically estimating the detection threshold of the suitable feature points at different spatial positions. In feature pair matching, the spatial position information of similarity function is effectively added by combining gradient direction with mutual information. Results the adaptive Harris corner detection method can effectively provide a uniform spatial distribution and sufficient number of feature points. The similarity matching function with the fusion of gradient direction and mutual information can improve the matching rate of feature points by 20 and reduce the registration error by 50 percent. Conclusion the multi-sensor image registration method proposed in this paper can realize the registration between visible and infrared images quickly and accurately, and has a good practical value in the field of CCD-IR image fusion.
【作者单位】：南通大学交通学院;
【基金】：国家自然科学基金青年项目(61503201) 江苏省自然科学基金项目(BK20161280) 教育部人文社科项目(16YJCZH086)~~
【分类号】：TP391.41

【相似文献】