Abstract
Detection of surface material based on hyperspectral imaging (HSI) analysis is an important and challenging task in remote sensing. It is widely known that spectral-spatial data exploitation performs better than traditional spectral pixel-wise procedures. Nowadays, convolutional neural networks (CNNs) have shown to be a powerful deep learning (DL) technique due their strong feature extraction ability. CNNs not only combine spectral-spatial information in a natural way, but have also shown to be able to learn translation-equivariant representations, i.e. a translation of input features into an equivalent internal CNN feature map. This provides great robustness to spatial feature locations. However, as far as we know, CNNs do not exhibit a natural way to exploit rotation equivariance, i.e. make use of the fact that data patches in a HSI data cube are observed in different orientations due to their orientation or on the varying paths/orbits of the airborne/spaceborne spectrometers. This article presents a rotation-equivariant CNN2D model for HSI analysis, where traditional convolution kernels have been replaced by circular harmonic filters (CHFs). The obtained results over three well-known HSI datasets showcase the potential of the approach.
Highlights
Imaging spectrometry data, in general, and hyperspectral imaging (HSI) in particular [1] are widely used remote sensing technologies for a large variety of applications due their abundant information about earth-surface material, such as: natural resources management [2], for instance in those activities related to forestry [3]–[5], geology/mineralogy [6]–[8] or hydrology [9]–[11]; precision agriculture [12]–[14], soil degradation [15]–[17] and crop stress/disease detection [18]–[20]; urban planning [21]–[23]; risk prevention [24]–[26], and disaster monitoring [27]–[29], among others
We develop a new rotation-equivariant CNN2D architecture for remote sensing HSI data classification to deal with all local rotation information present in HSI data
This article presents a new rotation-equivariant convolutional neural network reCNN2D based on circular harmonic filters (CHFs) for classyfying HSI remote sensing data
Summary
In general, and hyperspectral imaging (HSI) in particular [1] are widely used remote sensing technologies for a large variety of applications due their abundant information about earth-surface material, such as: natural resources management [2], for instance in those activities related to forestry [3]–[5], geology/mineralogy [6]–[8] or hydrology [9]–[11]; precision agriculture [12]–[14], soil degradation [15]–[17] and crop stress/disease detection [18]–[20]; urban planning [21]–[23]; risk prevention [24]–[26], and disaster monitoring [27]–[29], among others. Image segmentation is quite interesting in spatial processing methods, where they split the HSI scene into non-overlapping homogeneous regions according to some partitioning criteria In this sense, Tarabalka et al [74] performed spectral-spatial HSI classification combining pixel-wise SVM classification results and the partitional clustering segmentation map. DEEP LEARNING MODELS FOR SPECTRAL-SPATIAL HSI PROCESSING In contrast to previous ML-based methods, deep learning (DL) models [82], [83] offer a bunch of quite interesting neural architectures to automatically process the spectral-spatial information contained in HSI data, without requiring a priori knowledge about the data distribution/features In this context, deep neural networks (DNNs) can be interpreted as hierarchical stacks of L operational layers, where each layer l applies a transformation function, x(l) = f (l)(x(l−1), P(l)) over its inputs x(l−1) to obtain a certain output data representation x(l), which will depend on the layer parameters P(l) (weights and biases). They are not really effective in exploring some spatial relations among features and their generalization power depends on the transformations present in the training data
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.