Research And Design On FPGA-based Synchronously Binocular Vision System For Image Acquisition

With the rapid development of modern video image processing technology and machine vision technology,people's pursuit of related technologies in the visual field has also been continuously improved,especially with regard to the portability and real-time aspects of image processing platforms.In today's intelligent applications,binocular vision technology has played an important role in aerospace,smart robots,security,intelligent transportation,industrial inspection and other industries and has a wealth of application scenarios.However,with the increase in the complexity of binocular vision applications,system performance based on traditional embedded development is gradually unable to meet people's needs.In order to meet the actual needs of people in the application,first of all,this paper design a binocular vision synchronization acquisition system,the system uses Altera's Cyclone V series FPGA as the core processing chip in the system,in order to complete the two OV7725 COMS camera The drive and configuration,image acquisition,dual video synchronization cache design optimization and display output to the VGA interface solve the high synchronization and high real-time requirements for input dual-channel images in binocular vision systems for further research and Binocular stereo vision matching algorithm and image splicing technology provide a reliable image acquisition platform.Secondly,in order to cater for the application development trend of binocular vision synchronization acquisition system,this paper further studies the fisheye image correction algorithm based on FPGA.Through the research of fisheye correction algorithm,the correction algorithm is implemented by Matlab.According to the corrected coordinate information extracted from the Matlab experimental results,the second-order differential operation is performed.A compression method based on self-correlation is proposed.The compression of pixel coordinate points is corrected and the encoding is performed by combining the distribution properties of the second-order differential values.In this way,the complicated model equation solution that the FPGA is not good at can be converted into the good parallel addition and subtraction calculation,that is,the decompression process of the pixel coordinates of the corrected image.The complete restoration of the image is finally corrected by the bilinear difference.The fisheye correction scheme proposed in this paper not only saves the hardware resources needed for fisheye correction,but also improves the processing speed of fisheye correction.This article has completed the design of the binocular vision synchronization acquisition system,and achieved a 1280 * 720 @ 60 Hz dual real-time synchronous video acquisition and display.On this basis,in order to improve the function of binocular vision synchronization acquisition,the fisheye image correction solution based on FPGA is further completed,and the compression rate involving the compression algorithm is up to 4.3%.And on the premise that the hardware resource consumption is greatly reduced,it can meet the real-time correction of fisheye images.For binocular vision applications have a high reference value.