基于吉林一号遥感图像的星载目标快速识别系统

发布时间：2018-04-01 20:09

本文选题：吉林一号卫星　切入点：目标识别　出处：《光学精密工程》2017年01期

【摘要】：针对传统遥感图像地面目标识别系统图像获取周期长,信息实时性差等问题,设计星载目标快速识别系统,用于卫星在轨快速识别,提出改进的基于快速视网膜关键点(FREAK)的特征匹配识别算法,解决遥感图像数据量大、背景复杂的问题。介绍了星载目标快速识别系统的工作原理,提出简化的FREAK特征提取模型,将原有算法的七层模型减少为四层,用于快速提取出遥感图像中目标特征;利用二进制量化空间将高维特征数据量化为二维数据,提高算法的准确度;最后通过匹配,快速识别出遥感目标。实验结果表明,识别算法的准确度平均提高2.3%,识别用时缩短约27.8%,满足遥感卫星在轨目标快速识别的要求。
[Abstract]:Aiming at the problems of long image acquisition period and poor real-time information in traditional remote sensing image ground target recognition system, a fast recognition system for satellite in orbit is designed.An improved feature matching algorithm based on Fast Retinal key Point (FREAK) is proposed to solve the problem of large amount of remote sensing image data and complicated background.This paper introduces the working principle of the fast recognition system for spaceborne targets, and puts forward a simplified FREAK feature extraction model, which reduces the original algorithm's seven-layer model to four layers, and is used to quickly extract the target features from remote sensing images.In order to improve the accuracy of the algorithm, the binary quantization space is used to transform the high-dimensional feature data into two-dimensional data. Finally, the remote sensing target is quickly identified by matching.The experimental results show that the accuracy of the recognition algorithm is improved by 2.3, and the recognition time is shortened by about 27.80.It meets the requirements of fast recognition of objects in orbit of remote sensing satellites.
【作者单位】：中国科学院长春光学精密机械与物理研究所;中国科学院大学;
【基金】：国家863计划基金资助项目(No.2012AA121502)
【分类号】：TP751

【相似文献】