One of the major challenges in the classification of remotely sensed hyperspectral data is the limited number of labelled pixels within an image. This means that the labelled and unlabelled pixels can have different distributions, leading to poor performance by semi-supervised machine learning models. Our work addresses this problem of domain shift, and shows how a domain aware model can be used to overcome this problem. We employ our model on the Pavia University and Pavia Centre dataset, and compare the performance of our model with a traditional deep neural network model. We observed an improvement of up to 10 percent in the classification accuracy. We also analysed the effect of spatial and spectral features on our domain aware model, and came to a basic conclusion that spectral features provide a more global information about a pixel’s class, whereas spatial features provide a more local information, which means that domain shifts are likely to be more pronounced in spatial data as compared to spectral data.

Hyperspectral image (HSI) classification is widely used for many critical remote sensing applications that require high accuracy. Their classification is a challenging task due to the large size of HSI datasets. One approach to reduce the size of HSI datasets is to exploit patch similarity. Pixels with the same class tend to exhibit similar features in space, so similar patches are unlikely to provide any new information for classification. (Distributed Architecture for Fusion Transformer training Acceleration) DAFTA is a multi-modal data fusion transformer that integrates information from multiple modalities and sources, such as HSI and LiDAR, to improve classifi-cation accuracy. This work proposes a method for the DAFTA framework to reduce the size of HSI datasets with less clustering time than the existing approach. We use a clustering approach to identify and eliminate similar patches that belong to the same class with the least amount of time. The experimental results show that the proposed method can significantly reduce the size of the training dataset in minimum time without sacrificing classification accuracy, which enhances the efficiency of distributed training.

In today’s world, environmental pollution is worsening and having detrimental effects on the environment. The current focus of human endeavors is aimed at reducing this pollution to ensure a sustainable future for our planet. In this pursuit, we are striving to create a framework that can estimate the carbon fluxes within a city. The accurate estimation of carbon fluxes from cities is crucial for understanding and mitigating their impact on global greenhouse gas emissions. This paper presents a novel approach for estimating carbon fluxes at a spatial resolution of 1 km x 1 km grids using remotely sensed data. Our research relies on indicators such as population and impervious surface area as proxies for estimating carbon dioxide (CO2) levels with a linear regression model. To test the generality of the method and models, we have validated our grid-model coefficients on point data and vice versa.

Classification of remotely sensed data is the mainstay of analysis methods for generating actionable insights. Classification of remotely sensed data is inherently a semi-supervised classification problem. Often, the labeled pixels and the unlabeled pixels in the image may have different distribution. Hence classification accuracy of such images is affected. We propose a umbrella framework for semi-supervised learning that considers the domains shifts in labeled and unlabeled pixels (called Domain Aware Semi-supervised learning- DASSL). The method learns the deep features in a such a that they are invariant of the pixel source, i.e, labeled or unlabeled. We employed DASSL for classification of hypersepctral image of Pavia University. We compared DASSL with self-training iterations performed using SVM and Convolutional Neural Network. We used spectral features, spatial features, and fused spectral-spatial features. The results are encouraging. We observed the reasonable improvement in classification by DASSL over self-training iterations.

Calculating CO2emissions from a city using Global Protocol for Communities (GPC) guidelines is a data intensive exercise. Moreover, the outcome is difficult to verify because of the data quality issues. We have developed a quick and approximate method for calculating scope 1 and scope 2 CO2emitted by a city, using Landsat 8 data. We used Vegetation-Impervious surfaces-Soil (VIS) classes extracted from the satellite image, and city population as independent variables and related it with CO2emissions. The model results match well with the reported total annual carbon emissions by the world carbon budget report.

Selection of appropriate features is important for classification of urban materials using hyperspectral data. Urban materials lack dominant diagnostic absorption and hence features representing complete spectrum are likely to provide better classification performance. Furthermore, selection of appropriate features for a given data requires empirical assessment. In the present work, we introduce artificial color features that take into account complete spectrum. In addition to the color features, we use reflectance values of all the noise free wavelengths of EO-l Hyperion (set A), and a wavelength set H={445,576,638,759,1100, 1316, 1989} reported in literature. We classify EO-l Hyperion image of Pune city using multiple classifiers and compare their outcome. The color values, set A, and H provide similar results. Set H and color features results in minor drop of accuracies in urban classes such as industrial roofs and residential concrete roofs.

Dictionary learning (DL) is one of the popular sparse coding machine learning techniques. In image processing literature, every input image is represented as the sparse linear combination of basis vectors. DL has been shown to have wide applications for image restoration as well as pattern recognition problems. In DL, the input image is factorized into dictionary and sparse codes. This factorization always leaves a residual or approximation error. Very few works in the literature had focused on to leverage the information present in this residual. In this paper, we use residuals within our framework and show that the restoration performance or accurate prediction of missing label in multi-spectral images can be significantly improved over conventional DL based techniques. We initially show that the higher order frequencies are propagated through residuals. Then we show that incorporating this residual in the image restoration methodology can significantly improve the outcomes. Finally, we propose a technique to solve the problem of missing label prediction by using a restoration based deep residual dictionary learning framework.