Research Of Real-time Image Texture Retrieval Technology Based On FPGA

With the development of Internet Technology and the increase of video and images data,transmission,storage and retrieval of video and image information have an increasingly important impact on the Big Data.How to quickly and accurately retrieve their interests in a lot of video and images has aroused widespread concern of scholars.The traditional image retrieval system has the problem of low retrieval speed and accuracy.This dissertation uses neural network algorithm to build an image retrieval system on embedded FPGA platform to improve retrieval speed and accuracy.At present,the requirements for the speed and accuracy of image retrieval in the field of image retrieval are getting higher.This dissertation improves the speed of the image retrieval by setting up an image retrieval system on the embedded FPGA platform,and in order to improve the accuracy of the image retrieval,an improved scheme of Spiking Cortical Model is proposed.Because the video and images contain a large amount of feature data,it is not conducive to computing and caching operations,this dissertation signs images based on neural network algorithm to reduce the data of the images.Next,the digital images signatures are extracted for feature extraction to reduce the image data,and then the image similarity measure and retrieval identification are further performed.In order to improve the accuracy of image retrieval,this dissertation presents a scheme of self-optimizing the digital image signature algorithm of image retrieval based on improved Genetic Algorithm and building an adaptive Spiking Cortical Model.In the process of implementing the image retrieval system based on FPGA,aiming at the problem of large memory consumption,large amount of images data to be processed and low image processing speed,this dissertation proposes several optimization schemes,such as the algorithms of Spiking Cortical Model,feature extraction and image similarity measurement,storage of sample eigenvectors and serial-parallel conversion.The dissertation set up the real-time image texture retrieval system based on FPGA.The experimental results evaluate the speed and accuracy of the retrieval and the utilization of FPGA on-chip resources.Compared with the existing image retrieval system,it has superior performance.The contrast experiments verifies the superiority of the image signature algorithm and feature extraction algorithm and shows the effects of the parameters self-optimization of Spiking Cortical Model.