Research On The Technology Of Calibration Of Bionic Compound Eye Visual Systems And Large Field Of View Image Mosaic

Very significant enlightenment can be obtained from animals'sensation organs on information reception and information processing system. Insects'compound eyes, especially the fly,rearhorse and dragonfly's eyes, which have perfect ingenious and extraordinary structure as well as particular function gradually by hundreds of millions of years'evolution, enlighten us greatly on reception and processing of visual information. Currently, to emulate and implement the function of insect compound eye's visual system have become one of the hotspots of bionics.A set of bionic compound eye visual system is designed on the basis of exploring the visual mechanism of insects'compound eye. This system uses several CCD cameras to collect scene images or video information imitating the individual small eyes of insects'compound eye. Classical calibration of camera based on the template was used to get camera's parameters. A new direct image mosaic method was proposed by analyzing the optical principle of the visual system. Direct image mosaic namely does not need algorithm search the same content in two neighbor pictures to determine their relative position, directly synthesize the images obtained from different cameras to form a large field of view quickly. When analysis the errors on the process of image mosaic, the concepts of depth of field and camera auto-focus are introduced. In this paper, a new image processing method is addressed to solve the image dislocation between two adjacent images successfully. This method is used to search optimal suited position of image mosaic edge by calculating the correlation of the mosaic part to obtain the dislocation value between two images. Additionally, improved proportion method is used to eliminate mosaic mark which is engendered by the light intensity difference among the cameras.Besides, the direct mosaic method and traditional mosaic method based on image match is compared. These two methods are both used to mosaic the same images, and the experimental results show that they can both mosaic images well. While, the consumed time of direct image mosaic method is only 6.6% of the traditional method. Therefore, the method put forward in this paper can improve the real-time capability of the visual system effectively. Moreover, the overlapping proportion between the adjacent images can be infinitesimal, and the enlargement proportion of vision field can maximally reach 100% for the two mosaic images.