Research On Image Acquisition Control And Processing Technology Of Array Camera

In order to solve the contradiction between the large field of view and high resolution in the traditional imaging system and improve the information capacity of the optical system,under the condition of limited resolution of image sensor,a new system of multi-aperture compound eye imaging emerged.As an important part of the multi-aperture compound eye imaging system,the thesis conducted research work on array camera image acquisition control and processing technology,the system scheme and its working principle were discussed,the array camera image acquisition control system was researched and designed,and the image stitching processing process was discussed.Image stitching was implemented based on SIFT algorithm.Finally,the system experimental verification was performed and results analysis was also given.Aiming at the characteristics of high real-time performance and large data volume of the array camera image acquisition control system,the thesis analyzed the specific working principle based on the system scheme,and a multi-channel camera real-time image acquisition control system based on FPGA was designed and implemented,which focused on the design of real-time image acquisition,storage,display,multi-channel camera control,and host computer software.The logic drive and control module of the array camera image acquisition control system were designed using Verilog hardware description language,which included multi-channel CMOS camera configuration,camera acquisition and image data conversion,read and write FIFO control,read and write arbitration module generation,DDR3 SDRAM read and write control,multiple communication protocol logic driver design,and HDMI interface display.The functions of image I~2C interface acquisition,AXI Bus transmission,double edge dynamic storage,and HDMI color display were effectively implemented.The control mechanism of the multi-channel camera was analyzed and researched.A multi-channel camera working state selection based on multi-channel polling and host computer were designed and implemented.The C#programming language and Visual Studio 2013 development environment were used to implement the design of the host computer,and a test platform was established to verify the two-way communication between it and the main control unit.In order to further improve the effect and real-time performance of image stitching,the thesis focused on the theory and process from three levels of image preprocessing,image registration and image fusion.The possible factors affecting image quality in the image information transmission chain were analyzed.The basic principle and design scheme of image preprocessing were presented,and the specific implementation process of hardware acceleration of image preprocessing algorithm was discussed.Based on FPGA,3×3 image median filtering and used second-order polynomial fitting correction method to deal with non-uniform images were achieved.Two types of image preprocessing improved the real-time performance of image preprocessing and the accuracy of stitching.Then,two image registration methods based on region and feature were described.Image feature extraction and inter-image registration based on SIFT algorithm were studied,and specific feature extraction and image registration results were given.The weighted average method was used to eliminate the stitching seams existing after image stitching and FPGA hardware transplantation was performed.The subjective and objective evaluation of image stitching effect and image quality were given.An array CMOS camera image acquisition control and processing system based on Xilinx Artix-7 FPGA was built,and verified the system functions in the paper.The experimental results show that the system can display at 30fps with a system clock of 50MHz and a single camera image resolution of 640×480,and the picture quality is clear and smooth.The upper computer can control multiple cameras in real time and stably under various baud rate conditions.Compared with other algorithms,the image registration accuracy of SIFT algorithm is high,the stitching effect is good,and the image field of view is increased.Compared with traditional imaging systems,the imaging field of view after stitching is significantly improved.