In recent decades, lithological mapping techniques using hyperspectral remotely sensed imagery have developed rapidly. The processing chains using visible-near infrared (VNIR) and shortwave infrared (SWIR) hyperspectral data are proven to be available in practice. The thermal infrared (TIR) portion of the electromagnetic spectrum has considerable potential for mineral and lithology mapping. In particular, the abovementioned rocks at wavelengths of 8–12 μm were found to be discriminative, which can be seen as a characteristic to apply to lithology classification. Moreover, it was found that most of the lithology mapping and classification for hyperspectral thermal infrared data are still carried out by traditional spectral matching methods, which are not very reliable due to the complex diversity of geological lithology. In recent years, deep learning has made great achievements in hyperspectral imagery classification feature extraction. It usually captures abstract features through a multilayer network, especially convolutional neural networks (CNNs), which have received more attention due to their unique advantages. Hence, in this paper, lithology classification with CNNs was tested on thermal infrared hyperspectral data using a Thermal Airborne Spectrographic Imager (TASI) at three small sites in Liuyuan, Gansu Province, China. Three different CNN algorithms, including one-dimensional CNN (1-D CNN), two-dimensional CNN (2-D CNN) and three-dimensional CNN (3-D CNN), were implemented and compared to the six relevant state-of-the-art methods. At the three sites, the maximum overall accuracy (OA) based on CNNs was 94.70%, 96.47% and 98.56%, representing improvements of 22.58%, 25.93% and 16.88% over the worst OA. Meanwhile, the average accuracy of all classes (AA) and kappa coefficient (kappa) value were consistent with the OA, which confirmed that the focal method effectively improved accuracy and outperformed other methods.

Due to the relatively low temporal resolutions of high spatial resolution (HR) remotely sensed images, land-cover change detection (LCCD) may have to use multi-temporal images with different resolutions. The low spatial resolution (LR) images often have high temporal repetition rates, but they contain a large number of mixed pixels, which may seriously limit their capability in change detection. Soft classification (SC) can produce the proportional fractions of land-covers, on which sub-pixel mapping (SPM) can construct fine resolution land-cover maps to reduce the low-spatial-resolution-problem to some extent. Thus, in this paper, sub-pixel land-cover change detection with the use of different resolution images (SLCCD_DR) is addressed based on SC and SPM. Previously, endmember combinations within pixels are ignored in the LR image, which may result in flawed fractional differences. Meanwhile, the information of a known HR land-cover map is insignificantly treated in the SPM models, which leads to a reluctant SLCCD_DR result. In order to overcome these issues, a novel approach based on a back propagation neural network (BPNN) with different resolution images (BPNN_DR) is proposed in this paper. Firstly, endmember variability per pixel is considered during the SC process to ensure the high accuracy of the derived proportional fractional difference image. After that, the BPNN-based SPM model is constructed by a complete supervised framework. It takes full advantage of the prior known HR image, whether it predates or postdates the LR image, to train the BPNN, so that a sub-pixel change detection map is generated effectively. The proposed BPNN_DR is compared with four state-of-the-art methods at different scale factors. The experimental results using both synthetic data and real images demonstrated that it can outperform with a more detailed change detection map being produced.

Hyperspectral images contain abundant spectral information, which provide great potential for target detection. However, it also introduces a critical spectral variability problem for hyperspectral target detection, which makes the hyperspectral target detection much difficult than the classical spectral match issue. Many traditional detection methods have been proposed to deal with the spectral variability. However, these algorithms are still highly susceptible to the target spectral variability. The single input restriction and the inherent spectral characteristics mining problem are the main issues with these methods. The multitask learning (MTL) technique may have the potential to solve the above hyperspectral target detection issues since it can extract the inherent similarity and difference within multiple priori target spectra to learn a robust target spectral signature. This paper proposed a MTL-based reliability analysis method for hyperspectral target detection (MultiRely). This approach: 1) utilizes multiple priori target spectra to better represent the target spectral characteristics and construct multiple related detection tasks; 2) takes full advantage of the multitask learning technique to explore the spectral similarity and difference between multiple priori target spectra; 3) and applies the reliability analysis to obtain a reliable target spectrum in order to alleviate the target spectral variability. Experiments on two real hyperspectral datasets and one synthetic hyperspectral dataset illustrated the effectiveness of the proposed algorithm compared to the state-of-the-art detectors.

Extracting subpixel land-cover change detection (SLCCD) information is important when multitemporal remotely sensed images with different resolutions are available. The general steps are as follows. First, soft classification is applied to a low-resolution (LR) image to generate the proportion of each class. Second, the proportion differences are produced by the use of another high-resolution (HR) image and used as the input of subpixel mapping. Finally, a subpixel sharpened difference map can be generated. However, the prior HR land-cover map is only used to compare with the enhanced map of LR image for change detection, which leads to a nonideal SLCCD result. In this letter, we present a new approach based on a back-propagation neural network (BPNN) with a HR map (BPNN_HRM), in which a supervised model is introduced into SLCCD for the first time. The known information of the HR land-cover map is adequately employed to train the BPNN, whether it predates or postdates the LR image, so that a subpixel change detection map can be effectively generated. In order to evaluate the performance of the proposed algorithm, it was compared with four state-of-the-art methods. The experimental results confirm that the BPNN_HRM method outperforms the other traditional methods in providing a more detailed map for change detection.

Deep learning has achieved good performance in hyperspectral image classification (HSIC). Many methods based on deep learning use deep and complex network structures to extract rich spectral and spatial features of hyperspectral images (HSIs) with high accuracy. During the process, how to accurately extract the features and information from pixel blocks in HSIs is important. All of the spectral features are treated equally in classification, and the input of the network often contains much useless pixel information, leading to a low classification result. To solve this problem, an enhanced spectral-spatial residual attention network (ESSRAN) is proposed for HSIC in this article. In the proposed network, the spectral-spatial attention network (SSAN), residual network (ResNet) and long-short term memory (LSTM) are combined to extract more discriminative spectral and spatial features. More specifically, SSAN is first applied to extract image features by using the spectral attention module to emphasize useful bands and suppress useless bands. The spatial attention module is used to emphasize pixels that have same category with the central pixel. Then, these obtained features are fed into an improved ResNet, which adopts LSTM to learn representative high-level semantic features of the spectral sequences, since the use of ResNet can prevent gradient disappearance and explosion. The proposed ESSRAN model is implemented on three commonly used HSI datasets and compared to some state-of-the-art methods. The results confirm that ESSRAN effectively improves accuracy.