基于FPGA的深度卷积神经网络优化压缩算法研究
    点此下载全文
引用本文:彭泽武 ,蔡 雄,杨秋勇,苏华权.基于FPGA的深度卷积神经网络优化压缩算法研究[J].计算技术与自动化,2021,(4):74-78
摘要点击次数: 75
全文下载次数: 0
作者单位
彭泽武 ,蔡 雄,杨秋勇,苏华权 (广东电网有限责任公司信息中心广东 广州 51000) 
中文摘要:针对现有海量数字图像信息落后,提出了新型的压缩算法,设计出基于FPGA的视频图像采集系统。应用深度卷积神经网络优化视频图像编码算法和聚类算法实现数据特征提取,将图像与距离信息作为深度卷积神经网络的输入与输出,并利用其特征提取能力学习图像特征的距离信息,提取深度卷积神经网络中的全连接层作为编码,通过迭代调整确定图像编码,完成图像压缩。应用测试结果显示,该算法具有较高效率优势,且图像压缩解码后质量较好。
中文关键词:FPGA  深度卷积神经网络  优化压缩  图像采集  编码加速器
 
Research on Optimal Compression Algorithm of Deep Convolution Neural Network Based on FPGA
Abstract:In view of the backwardness of the existing massive digital image information, a new compression algorithm is proposed, a video image acquisition system based on FPGA is designed, and deep convolutional neural network is used to optimize the video image coding algorithm and clustering algorithm to achieve data feature extraction, and image and distance information As the input and output of the deep convolutional neural network, and use its feature extraction ability to learn the distance information of image features, extract the fully connected layer in the deep convolutional neural network as the encoding, determine the image encoding through iterative adjustment, and complete the image compression. The application test results show that the algorithm has the advantages of higher efficiency and better quality after image compression and decoding.
keywords:FPGA  deep convolution neural network  optimized compression  image acquisition  coding accelerator
查看全文   查看/发表评论   下载pdf阅读器