旋转扫描序列图像的全景图拼接方法

Panoramic Image Mosaic Method For Rotary Scanning Serial Image

针对火炮身管内表面观测的特殊需求,考虑到传统的图像拼接方法效率低或精度低的缺陷,提出了一种硬件和软件相结合的基于旋转扫描序列内表图像的全景拼接方案和图像拼接算法。该方案用步进电机驱动45°平面反射扫描镜和CCD相机同步旋转来实现对等距内表面序列图像的快速获取,利用获取的序列图像根据步进电机的步进角度完成内表面全景图像的粗级拼接,即在拼接中首先对步进旋转角度θ和图像素距离P进行标定,然后根据标定结果选取待配准的子图像区域,在子图像区域中采用基于相位相关的图像配准方法对图像进行像素级配准,从而完成对序列图像的精级拼接。为了实现图像的平滑过渡避免出现拼接痕迹,最后提出了基于人眼视觉特性的高斯函数权值加权图像融合过渡的方法。试验和仿真结果表明:该方案能够实现内表图像的快速高精度全景拼接,拼接的精度达到像素级,拼接时间小于500 ms,能够满足内腔表面观测的实时性需求,非常适合工程实际应用。

In the light of the special requirement of inner surface observation of cannon barrel and considering the shortage of traditional image mosaic, a kind of panoramic image mosaic method is presented combining hardware and software, which is based on rotated scanning serial inner surface image. In this method, a reflector with 45° angle and synchronous ratation of a CCD camera driven by a step motor are used to get the equidistant serial inner surface images quickly, then the inner surface full image are coarse mosaic according to the step angle based on the serial images. During the mosaic the rotator step angle θ and the image pixel distance P calibrated first, then the sub-district to be matched are chosen according to calibrating, after that the precision matching on sub-pixel degree are implemented based on phase correlation, so mosaic of the serial images is realized in high precision. In order to obtain a smooth transition in full image seam Gaussian functions weighted amalgamation algorithm based on human visual system（HVS） character is used to blend images. Simulation and test shows that the algorithm can generate the seamless full image of the inner surface quickly and in high precision, the mosaic precision can reach sub-pixel and the time consumption is less than 500 ms,which can meet the real-time observation requirement, and suit to engineering application.