遥感云图纹理特征提取算法实时加速设计  

Real-time acceleration design of remote sensing cloud image texture feature extraction algorithm

作　　者：薄振华 管雪元[1] 马训穷 鲁江涛 Bo Zhenhua;Guan Xueyuan;Ma Xunqiong;Lu Jiangtao(National Key Laboratory of Transient Physics,Nanjing University of Science and Technology,Nanjing 210094,China;Huaihai Industry Group,CKangzhi 046012,China)

机构地区：[1]南京理工大学瞬态物理国家重点实验室,南京210094 [2]淮海工业集团,长治046012

出　　处：《电子测量技术》2021年第1期125-129,共5页Electronic Measurement Technology

摘　　要：为了解决灰度共生矩阵对遥感云图特征提取实时图像处理过程中算法复杂度高,运算时间长,数据运算量大等问题,提出了一种Vivado HLS实现卫星遥感云图特征提取算法的硬件加速方法。通过对灰度共生矩阵纹理特征提取算法以及Vivado HLS硬件加速设计进行研究,利用Vivado HLS对灰度共生矩阵纹理特诊提取进行硬件加速,并且封装为可调用的IP核,将PC端遥感云图处理结果与Zynq7020硬件加速后的处理结果进行比较,实验结果表明,该方法能够快速的解算遥感云图纹理特征,加快遥感云图处理速度,同时克服了FPGA设计图像算法难度大的缺点。In order to solve the problems of high algorithm complexity, long operation time and large amount of data calculation in the real-time image processing process of the gray level co-occurrence matrix for remote sensing cloud image feature extraction, a hardware acceleration method of Vivado HLS to realize the satellite remote sensing cloud image feature extraction algorithm is proposed. Studie the gray-level co-occurrence matrix texture feature extraction algorithm and Vivado HLS hardware acceleration design, uses Vivado HLS to hardware-accelerate the gray-level co-occurrence matrix texture special diagnosis extraction, and encapsulates it as a callable IP core to remotely sense cloud images on the PC side. The processing results are compared with the processing results after Zynq7020 hardware acceleration. The experimental results show that this method can quickly resolve the texture characteristics of remote sensing cloud images, speed up the processing speed of remote sensing cloud images, and overcome the disadvantage of FPGA design of image algorithms.

关 键 词：遥感云图 灰度共生矩阵 Vivado HLS 特征提取 硬件加速 

分 类 号：TN911.73[电子电信—通信与信息系统]