仿生复眼视觉系统中全景图拼接算法

模仿昆虫复眼的视觉机理，设计一种仿生球面9复眼视觉系统，并提出一种全景图拼接算法。首先采用抗旋转、抗尺度变化的SIFT图像匹配算法进行特征点的提取与匹配；然后利用稳健的RANSAC算法求出图像间变换矩阵 的初始值，并使用LM非线性迭代算法精炼 ；最后通过改进的加权平滑算法完成多视角图像序列的无缝拼接，得到整个场景的全景图。实验结果表明:全景图的整个拼接过程中复眼视觉系统可以自动地实现对多视角图像序列的无缝拼接，并在多个方向都增大了视场，弥补了传统宽屏幕全景图的不足。

Panorama mosaic in bionic compound eye visual system

A bionic spherical 9 compound eyes visual system was designed through simulating visual mechanism of insect compound eye, and a panoramic image mosaic algorithm was proposed. Firstly, feature points were extracted and matched by using the scale invariant feature transform (SIFT) image-matching algorithm. Secondly, the transformation matrix H was computed by random sample consensus (RANSAC) algorithm and Levenberg-Marquardt (LM) algorithm. Finally, the panoramic image was completed with improved weighted smoothing algorithm. The experimental results show that the whole seamless panorama mosaic process of multi-view image sequence is completed automatically, and the field of view is enlarged in multi-directions, which makes up the defect of the traditional wide screen panorama.