Design And Implementation Of Distributed Aperture Vehicle Panoramic Infrared Image Generation System

In both military and civil aspects,safety accidents caused by driver’s visual blind zone occur from time to time.Therefore,with the continuous development of intelligent driving technology,there is an increasing demand for vehicle panoramic all-weather situational awareness system.At present,the vehicle mounted panoramic situation awareness system is mainly based on visible 360° look around imaging or infrared imaging system with small field of view,which is difficult to meet the needs of panoramic all-weather military and civil dual-use.In contrast,the ultra wide-angle infrared staring imaging system has the advantages of large airspace coverage,low time-domain alarm leakage rate,simple structure and can meet all-weather work,It has become the focus of research all over the world.Based on the distributed aperture mechanism,this thesis uses six infrared cameras to build a panoramic infrared situation awareness system that can cover hemispherical airspace,and completes the design of software,hardware and related algorithms of the system.The specific research contents are as follows:1.A set of software and hardware system for hemispherical airspace infrared panoramic image generation is designed.The system is based on Hisilicon hi3559av100 processor,which realizes the real-time simultaneous acquisition,compression and coding of six channels of video,and outputs it through CVBS interface.The hardware design part of the system includes main control module,video acquisition module,compressed video output module,power supply module,external interface and mechanical structure;The software functions include: uniform light enhancement of images synchronously collected by multi-channel cameras,panoramic stitching,pixel fusion in transition area,panoramic display,etc.2.Aiming at the problem of image exposure difference caused by inconsistent imaging conditions of multiple infrared detectors,a uniform light enhancement algorithm based on reference image is proposed.In order to solve the optimal gain value of each image corresponding to the function,a loss function to minimize the sum of the gray differences of each image pixel in each overlapping area in the panoramic image is constructed by using the pixel mean of the overlapping area between the images,and the gradient descent method is used to iteratively optimize the loss function.Based on the optimal gain value corresponding to each image and the multiplication operation function,the pixel gray value of each image is processed,and each image is adjusted to varying degrees,so that the exposure degree of the processed image is as close as possible,so as to obtain a higher quality panoramic effect.3.A generation model of hemispherical spatial panoramic image is constructed.Firstly,the uniform light enhancement algorithm based on the reference image designed in this paper is used to preprocess the six images,and then the plane images taken by each camera are spliced according to the relative position relationship between each camera,Then,by constructing the plane spherical projection model,the pixels of each plane image are mapped to a hemispherical surface,and the pixel fusion is carried out in the overlapping area of each image to eliminate the seam.Finally,the generated hemispherical spatial panoramic image is projected into its corresponding two-dimensional plane circular area for display.