【作者】 刘雨晴； 李长庚； 庄胤； 陈禾； 龙腾；

【Author】 LIU Yuqing;LI Changgeng;ZHUANG Yin;CHEN He;LONG Teng;Radar Technology Research Institute School, of Information and Electronic, Beijing Institute of Technology;School of Physics and Electronic, Central South University;

【机构】 北京理工大学信息与电子学院雷达技术研究所； 中南大学物理与电子学院；

【摘要】 随着遥感成像技术的快速发展,获取的遥感数据成几何倍数增长,目前面临的问题是获取的海量数据无法有效地下传,形成了巨大的数据资源浪费。因此,急需构建嵌入式实时处理系统有效地处理海量遥感数据,并提取有用的数据进行下传。其中,轻量化智能处理算法是嵌入式实时处理系统的核心,只有构建高精度的轻量化智能处理算法才能满足嵌入式算法映射时的体积、重量、功耗约束。针对这一问题,提出了通过融合多尺度特征优化PCANet结构的轻量化算法,并在UC Merced数据集和SIRI-WHU数据集上进行了详细的实验,实现结果表明提出的算法可以有效的提升PCANet网络性能,在UCMerced数据集和SIRI-WHU数据集上总体分类精度分别提高了4.76%和6.38%。可以应用于对实时性要求较高的场景中。更多还原

【Abstract】 With the rapid development of remote sensing technology, the acquired remote sensing data hasincreased exponentially. The current problem is that the massive amount of data acquired cannot be effectively transmitted, resulting in a huge waste of data resources. There is an urgent need to build an embedded real-time processing system to effectively process massive amounts of remote sensing data. The lightweight and intelligent processing algorithm is the core of the embedded real-time processing system. Only the construction of a high precision lightweight intelligent processing algorithm can satisfy the embedded algorithm mapping volume, weight, power consumption constraints. In response to this problem, a lightweight algorithm that optimizes the PCANet structure by fusing multi-scale features is proposed, and detailed experiments are performed on the UC Merced land-use dataset and the SIRI-WHU dataset. The implementation results show that the proposed algorithm can effectively improve PCANet performance, on the UC Merced dataset and SIRI-WHU dataset, the overall classification accuracy increased by 4.76% and 6.38%, respectively. The proposed algorithm can be applied to scenarios that require high real-time performance.更多还原