海面舰船红外与可见光图像配准

The Registration of Warship Infrared and Visible Images

在利用特征点集的红外与可见光舰船图像配准过程中,经常会存在点集的一致性差而无法配准的情况,本文针对可见光图像中背景干扰大,纹理丰富容易出现较多非舰船目标轮廓特征点的情况,利用全局广义直方图均衡化和显著性增强对可见光图像进行增强,然后进行Canny轮廓提取并在轮廓的基础上提取舰船目标边缘角点作为匹配点集;对于红外舰船图像,海面背景与舰船温度差异较大,其舰船目标成像效果较好,存在的干扰较少,本文直接进行Canny算子运算并提取角点作为特征点集。实验效果显示,本算法实用性较强,误差小于3个像素,能够满足工程使用要求。

In the registration process which uses feature point set of infrared and visible light images, the consistency of the point sets is poor and can＇t always be used in registration. Focusing on the fact that the visible images are always with a lot of background and rich textures, we use the global generalized histogram equalization and significant enhancements to visible light image enhancement, make contour extraction by ＇Canny＇, and then detect the comer points as feature sets from the ＇Canny＇ map. For infrared ship image, the temperature difference of background and ship targets is bigger, the ship target imaging better, and less interference. This paper directly uses the ＇Canny＇ operator to operate and extract the comers as feature point sets. Experimental results show that the algorithm is practical, the error is less than 3 pixels, which can satisfy the engineering demand.