针对红外图像与可见光图像的自动配准问题，提出了一种基于图像角点特征和 透视变换模型的方法。首先采用自适应阈值对红外与可见光图像进行分割，然后利用Harris因子分别在 分割后的红外和可见光图像上检测角点。通过分析角点邻域在原始图像上的相关性实现角点的粗匹配。接着 通过RANSAC算法对角点进行细匹配，删除outliers，再由Hu矩不变量进一步提纯角点。最后利用最终匹配的角点 作为透视变换的控制点，得到透视变换模型。用该模型对待配准图像进行透视变换，可实现图像配准。实验 结果表明，该方法的配准精度高，可以很好地完成红外与可见光图像的自动配准。

According to the automatic registration of visible and infrared images, a new method based on Harris Corner and Hu invariant is proposed. In this method, the adaptive thresholds are used to segment visible and infrared images and the Harris factors are used to detect the corners in the segmented visible and infrared images respectively. After analyzing the correlation of the corner neighborhoods in the original images, the coarse corner matching is realized. Then, the RANSAC algorithm is used to remove the outliers in the coarse matched corners and the Hu invariant is used to further filter the wrong matching. Finally, by using the matched corner as the control point of projection transformation, a projection transformation model is obtained. The model can be used to conduct the projection transformation of the images to be matched and make the images registered. The experimental result shows that the method has a high accuracy in registration and can automatically register the visible and infrared images.