The Research of Imitation Compound Eye Visual System’s Dynamic Characteristics and Image Mosaic Technology
|School||Xi'an University of Technology|
|Course||Signal and Information Processing|
|Keywords||Compound Eye Since the focusing lens Motion detection Image mosaic|
With the development of digital image processing technology, the actual project requirements for target detection technology is getting higher and higher, target detection technology for existing single perspective also demonstrated the shortcomings of the limited detection range, multi-angle image sensor developed target detection technology has become a research hotspot. The unique skills of the insect compound eyes on the receiving and processing of visual information for researchers provides reliable clues to solve the problems of these two areas. Thesis insect compound eyes physiological structure as the reference model, the application of self-focusing lens array and planar array CCD device composed imitation compound eye visual system prototype structure, 2 * 3 parallel-type compound eyes of visual function. The method to solve the integration issues to achieve a large number of small compound eyes eye imaging problems and low-cost implementation of the main challenges of the three major problems. Processing motion information function of the nervous system of the fly compound eye, the paper presents the background suppression of the two-layer model of target motion detection system, the original image data stream first passes through the spatial filter to weaken the interference of background information, reflecting the target sequence. The thesis study against the static background suppression gradation variation gradient using the still images and moving images, a series of frame image three-dimensional matrix (x, y, t) is converted to a matrix (x, t, y), along the y-direction the (x, t) Series image for high-pass filtering process, filtering the background information, the projection target information, obtained highlights the image of the sequence of information of the target feature. Fly compound eye of the second layer of the model design the papers fly compound eye the Reichardt related primary movement detector (EMD) as the basis algorithm proposed use class of video sequences the Reichardt related sequence model algorithm to get a little eye image sequence delayed by different time image sequence with all of the other small eyes, search out the largest correlation coefficient of the image, according to the laws of the delay time is inversely proportional to the target motion, and finally the velocity of the target motion and the image gray value combination, intuitive target motion state diagram. By the the repeated experiments demonstrate papers target detection algorithm can accurately estimate the state of motion of the target object, consistent with the actual state of motion. And technical research on the image mosaic, the static characteristics of the image for the four different orientations of small eyes collected by the overlapping area between the rough estimation image, greatly reduced the range of the image registration; the case in order to avoid a single VALID the gray image registration algorithm searches for the best combined ratio method and template matching inlay position; obvious suture traces directly mosaic gray difference, and finally smoothing factor of image mosaic of overlapping area of ??data fusion processing to obtain a complete smoothing of the large field of view image. Experimental results show that the paper imitation of the total field of view of the structure of the fly eye increase more small eye image mosaic aspect when the algorithm is applied to obtain a visual field will increase more than the original single small eye visual field . Compound eye image mosaic deal with both a strong theoretical significance, more wide application value.