当前位置: X-MOL 学术Infrared Phys. Technol. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Hyperspectral Image Classification Using CNN with Spectral and Spatial Features Integration
Infrared Physics & Technology ( IF 3.3 ) Pub Date : 2020-06-01 , DOI: 10.1016/j.infrared.2020.103296
Radhesyam Vaddi , Prabukumar Manoharan

Abstract Hyperspectral image (HSI) classification is very important task having numerous applications in the remote sensing field. Many methods have been proposed in the recent years. Among them Convolutional Neural Network (CNN) based algorithms have shown higher performance. But these algorithms need high computational power and storage capacity. This Paper presents an approach for remote sensing hyper spectral image classification based on data normalization and CNN. HSI data is first normalized by reducing its scalar values by retaining complete information. Then, spectral and spatial information is extracted using Probabilistic Principal Component Analysis (PPCA) and Gabor filtering respectively. Further, the spectral and spatial information is integrated to form fused features. Finally classification task is done using simply designed CNN framework. Experiments are performed on three benchmark hyperspectral datasets (Indian Pines, Pavia University and Salinas). The proposed approach has achieved significant performance over the state-of-art methods. This can be useful in real world applications like agriculture, forestry and food processing.

中文翻译:

使用具有光谱和空间特征集成的 CNN 进行高光谱图像分类

摘要 高光谱图像(HSI)分类是一项非常重要的任务,在遥感领域有着广泛的应用。近年来提出了许多方法。其中基于卷积神经网络(CNN)的算法表现出更高的性能。但是这些算法需要很高的计算能力和存储能力。本文提出了一种基于数据归一化和CNN的遥感高光谱图像分类方法。HSI 数据首先通过保留完整信息来减少其标量值来标准化。然后,分别使用概率主成分分析 (PPCA) 和 Gabor 滤波提取光谱和空间信息。此外,光谱和空间信息被整合以形成融合特征。最后使用简单设计的 CNN 框架完成分类任务。在三个基准高光谱数据集(印度松树、帕维亚大学和萨利纳斯)上进行了实验。与最先进的方法相比,所提出的方法取得了显着的性能。这在农业、林业和食品加工等实际应用中非常有用。
更新日期:2020-06-01
down
wechat
bug