面向Bayer真彩色航空影像特征匹配的彩色描述符适用性分析

Performance analysis of color descriptors for aerial Bayer real-color image

针对Bayer真彩色遥感影像进行特征匹配时,彩色描述符的效果与适用性问题一直研究较少。结合Bayer真彩色遥感影像成像变化规律和彩色描述符算法特点,从理论上分析彩色描述符的不变性。提出模拟数据评价、不同类别影像评价、任务总体适用性评价等三种实验方法对彩色描述符适用性进行实验验证和分析。通过理论分析和实验评估,总体上彩色描述符中RGBSIFT算法效果最优,对Bayer真彩色航空影像特征匹配有较好的适用性,并且不同地物属性影像在特征匹配时有不同的最优彩色描述符。

It is unclear that what color descriptors best fits feature matching of aerial remote sensing Bayer real-color image. To analyze variation of imaging of the aerial image and algorithm characteristics of color descriptors, a theoretical analysis of invariant properties of color descriptors was presented. The effect and applicability of descriptors were verified and analyzed by three experimental methods which are evaluations of simulation data, different categories images and general applicability of task. The theoretical and experimental results show that RGBSIFT has best applicability for the feature matching of aerial Bayer real-color image. The usefulness of color descriptor is category-specific.