The bionic compound eye has the advantages of small volume,light weight,large field of view and high sensitivity.It has a broad application prospect in radar system,micro aircraft,naval search and tracking system and other fields.A bionic compound eye imaging system has been set up in our laboratory.The system is based on a multi-micro-surface fiber faceplate with a diameter of 25 millimeters.The work flow is: 9 sub-eye lenses' images formed the back end of the fiber faceplate,and then take a SLR camera to get the initial image of the bionic compound eye.This article has done the following three tasks on the basis of the original experimental system:1.By redesigning the aperture of the center sub-eye len,the previously-exposed central sub-eye image has a certain improvement in overexposure;2.Some research on image registration algorithm was carried out.Previously Yang Chao selected the SIFT algorithm to extract the feature points,and adopted the globally projective method.This method has the problem of insufficient homography matrix binding.If there is a depth change in the overlapping region of the sub-eyes,the stitching image will appear blurred and ghosted.In order to solve this problem,this paper attempts to use As Projective As Possible Warps.In order to improve the registration speed,ORB feature extraction is preferred,and the SURF feature point detection is adopted when ORB feature extraction fails;3.Using NVIDIA Jetson TX2 as the hardware of video processing,the power of the only credit card size platform is less than 7.5 watts,which provides the necessary basis and conditions for the miniaturization and portability of the follow-up bionic compound eye imaging system.The video stitching algorithm is based on the hybrid compilation of CUDA and Open CV,and realizes the stitching of bionic compound eye video. |