Study On TV Guidance Target Detection Algorithm Based On FPGA

Nowadays,TV-guided weapons have been more and more widely applied.In the process of the development of TV guidance technology,digital image processing technology is also constantly updated and improved.Especially,the research and application of image target detection algorithm,and at the same time,the depth learning domain has also made breakthrough progress.The convolution neural network architecture can obtain more robush features.In this paper,a new target detection algorithm based on convolutional neural network is proposed,which solves the shortcomings of the traditional manual selection of feature accuracy.In the same time,the recursion method is used instead of the sliding window or the proposed region.The speed of training;at the same time in the training to join the mining difficult to improve the accuracy of the network after training.And three kinds of data sets of training and testing,a detailed analysis of the data.In order to meet the real-time better,the algorithm is mapped to the FPGA to accelerate,to achieve the data conversion module and convolutional neural network design,because a large number of convolutional computing is the concentrated expression of algorithm complexity.According to the advantages of FPGA parallel processing and the characteristics of the parallel structure of the convolution neural network,the parallelism is fully analyzed and utilized.Due to the limited resources of the FPGA,the cache arrangement of the storage system,especially the two-dimensional convolution,is carefully arranged to improve the efficiency of on-chip resources.