Automatic identification of object of interest in a hyperspectral imagery is promising for remote sensing applications. Spectral knowledge transfer enables autonomous comparison of reference and imagery spectra for expert-independent analysis. Knowledge-transfer based analysis involves comparing image spectra to the reference spectra (spectral libraries) using spectral similarity metrics. However, the reference spectral databases and the imagery acquired by different sensors differ in spectral resolution and bandwidths, limiting the direct comparison of the spectra. Thus, prerequisite process of spectral resampling is required before the analysis. We propose a new method, “Dynamic Spectral Similarity Method (DSSM)” that quantitatively compares spectra from sensors having different spectral resolutions. DSSM geometrically aligns two non-linear spectra and computes an optimal alignment cost through a time-warping process in a dynamic feature space. We demonstrated the potential of DSSM by comparing spectra of diverse landscape elements obtained from various sources (satellites, airborne, spectral libraries) against reference databases. Further, the proposed method is compared with spectral matching methods (SAM, SID, NS3) after a spectral alignment process using a Gaussian Diffusion Model. The results are promising, offering 80% to 90% matching accuracy in all the scenarios. DSSM enables seamless comparison of images with varying spectral characteristics, allowing selective and automatic object identification.

Accurate classification of crops at the patch level based on nutrient status, particularly nitrogen (N) levels, is essential for advancing precision agriculture (PA). While recent advancements in remote sensing, scalable computing, and visualization technologies have enabled high-resolution plant monitoring, the spectral similarity among crops remains a challenge for precise classification using remote sensing data. This study introduces a multisensor fusion approach, integrating terrestrial LiDAR point cloud data and WorldView-III multispectral imagery within a deep learning (DL) framework to classify cabbage, eggplant, and tomato across different N levels. By combining structural and spectral information, this method effectively captures N-induced growth variations, leading to improved crop discrimination. Our results demonstrate that applying a deep convolutional neural network (DCNN) model to the fused dataset enhances classification accuracy by 13%–16% compared to using multispectral data alone. The incorporation of LiDAR data plays a key role in capturing canopy structure, significantly improving classification performance. Additionally, our DL approach outperforms traditional machine-learning methods, including the random forest (RF) classifier, reinforcing the advantages of DL for N-sensitive crop classification. By leveraging multisensor integration and DL, this study presents a robust and scalable approach for enhancing crop classification accuracy, with significant potential for advancing PA and site-specific nutrient management.

A hybrid model for nowcasting of storms has been developed by employing predicted integrated water vapor (IWV) with light gradient boosting machine (GBM) on Global Navigation Satellite System (GNSS) receiver data collected at Gadanki (13.46°N, 79.17°E) and estimated thresholds for three storm predictors. The utilization of predicted IWV allows more lead time for disaster preparedness. The efficacy of light GBM technique in predicting IWV has been tested on 54 stormy days (from the year 2020), identified with a collocated polarimetric weather radar observations. The predicted IWV agrees very well with observed IWV with rms errors0.85) for predictions with a lead time up to 2 h. Among several predictors considered for nowcasting, IWV is found to have a great predictive potential, as the moisture buildup is seen few hours (1–4 h) prior to the occurrence of storm/rainfall. Thresholds for chosen predictors [magnitude of IWV, change in IWV, and change in brightness temperature (Tb)] are finalized using the data from known stormy days (65 days from the years 2018 and 2019). The sensitivity analysis of the predictors independently and in combination in predicting storms reveals that 1 and 2 parameter-based predictions detect storms accurately but produce large false alarm rates. The three-parameter scheme reduced the false alarm rate drastically to 5% and improved the model accuracy to 97%, which is much better than the existing methods.

Crop health is vital in agricultural nations like India. One of India's most important crops, tomato, is farmed and consumed globally. Bacteria, viruses, and fungi may damage tomato plants and other food and vegetable crops. Fusarium wilt is a seed-borne fungal infection that causes severe damage in the growth and yield of tomato crop. The objective of this research is the assessment of spectral discriminability and classification of health and Fusarium wilt diseased-tomato plants using hyperspectral data. In-situ reflectance measurements of healthy and infected plants over a tomato growing region (Tumkur, Karnataka, India) have been collected, processed, and analyzed to spectrally differentiate diseased and healthy plants. Various statistical and machine-learning approaches have been applied for assessing the spectral discrimination of diseased plants with sensitivity to five different levels of disease severity. Results suggest the existence of stable spectral features which differentiate healthy and diseases plants at distinct levels of disease severity. The discrimination has substantial variance with the method used. Amongst the methods used, Linear Discriminant Analysis (LDA), Gradient Boosting, and XGBoost methods offer spectral discrimination of about 80% accuracy.

Cloud detection in satellite imagery is an essential and common preprocessing task in optical remote sensing image analysis. Helping minimize the dominant areal coverage of cloud in high-resolution satellite imagery, various space agencies operating optical remote sensing satellites consciously scout for temporal windows which offer the maximal probability of cloud-free days. Demanded by users across the globe, an acquisition plan for providing cloud-free imagery is critical for all space agencies. Despite many technological advancements, the dynamic nature of weather often poses cloud issues during the day, making the acquired imagery redundant. Thus, having an idea of the probability of an area being cloudy on a given date and time is very helpful in optimizing the resource deployment of space agencies, particularly for optical satellites. Therefore, this requires the acquisition development of long times series cloud image database with a significantly finer spatial resolution of up to 30 meters or better than that.

Threatening marine life, coastal erosion, and human health, environmental pollution caused by plastic-litter is a global problem. The process of identifying and mapping plastic-litter is arduous, time-consuming, and expensive. Environmental parameters may be mapped and monitored with the use of hyperspectral remote sensing. Theoretical advances in recent years have shown that hyperspectral imagery may be used to detect and foresee plastics in a variety of geographical settings. Except a few efforts which have used simulated datasets for assessing spectral signatures, there are no studies which have attempted to assess plastic-litter abundance using remote sensing imagery. Plastic-litter are pieces of plastic that are less than five millimetres in size. Sub-pixel methods are required for processing and analysis. In this study, we explore the use of hyperspectral remote sensing to identify and map microplastic pollution. Since plastic-litter are uncommon in pixels, spectral unmixing is employed in conjunction with known endmembers. Mapping and detecting micro-plastic species have yielded varying results. Advancing crucial environmental monitoring application, results suggest that hyperspectral imaging is a potential data source for mapping the plastic litter.

Algal bloom events distress coastal fisheries, tourism, and recreational activities and affect a nation’s economy. We have studied the large-scale spatio-temporal dynamics of algal bloom events in Indian waters. We have assessed the algal bloom trend over 20 years using a non-parametric Mann-Kendall test. The cross-correlation test is used to estimate the influence of oceanographic processes on bloom. Besides, temporal prediction of bloom events is carried out using the multivariate autoregression model. Study reveals that algal bloom coverage is declining over the coastal waters in all seasons except post-monsoon. The cross-correlation results indicate significant changes in the regional bloom patterns and are due to the substantial changes in the regional oceanographic processes owing to climate change. We observe that the autoregression models with a variable time lag up to 2 months are more suitable for predicting bloom occurrences. Furthermore, this study is useful in making appropriate adaptation measures against climate change.

Remote sensing-based crop mapping has emerged as one of the crucial elements of precision agriculture. Crops can be differentiated from one another based on their distinctive spectral signatures. Drone-based hyperspectral imagery can provide plant-level spectral signatures for further discriminant analysis. The high dimensionality and spatial auto-correlation common to agricultural landscapes often make the processing and analysis of hyperspectral imagery challenging for application in an operational environment. Although dimensionality reduction and band selection approaches have been explored in various application domains, including agriculture, there are no studies that have addressed dimensionality reduction at the crop level, referring to plant-level objects where the differences in spectral signature across crops are small. The goal of this work is to explore the potential of evolutionary optimisation algorithms (GWO and ALO) to optimise the band selection of high-resolution drone-based hyperspectral imaging of agricultural areas. The findings highlight how band reduction can be accomplished without impacting the spectral integrity of the data or classification accuracy.

Hyperspectral remote sensing has gained prominence in the past two decades. The high resolution imagery of Unmanned Aerial Vehicle (UAV), Terrestrial Hyperspectral spectroradiometer (THS) have gained popularity. In this study, we develop a Convolutional Neural Network (CNN) based architecture to accurately classify UAV, THS dataset for the region of Bangalore, India. To carry out feature extraction, classification, prediction and analysis a CNN Conv-4 and CNN Conv-6 model is designed with varying kernel sizes thus, extracting multiple features. The results show a visual and quantitative measure of how CNN Conv-6 has better capability and higher accuracy. Also, a graphical plot of training, validation accuracy and loss as a function of number of iterations is generated as an output at the end of prediction. This is to clarify the threshold limit for number of epochs required to run the model for classification and prediction.

Long-term measurements of GNSS receiver at Gadanki, India, have been used to develop a machine learning technique – lightGBM for the prediction of integrated water vapor (IWV) with different lead times. A variety of data sets related to IWV (representing source, sink and transport) and short-scale features of IWV (gradients, sinusoidal pattern) have been used to train the model. Model performance is validated in different seasons and also on storm days. The predicted IWV at different lead times (30–120 min) perfectly captures the temporal variability of measured IWV with a correlation coefficient >0.99. The RMSE of predicted IWV with 30 minutes lead time is less than 1 mm in all seasons. Nevertheless, the RMSE for predicted IWV with longer lead times increases with lead time but always remains <3 mm. The bias is slightly larger during the monsoon, mainly due to the higher occurrence of longer duration rainy events. Even in those days, the model is able to accurately predict the enhanced IWVbefore the rain occurrence. Sensitivity analysis and feature importance analysis on different predictors used in the model reveal that the IWV features are more important for short-scale prediction, like 30 min, whereas the importance of other predictors is high for longer lead time prediction (1-2 hours) and on storm days.

Target detection (TD) in spectral imagery is an evolving analytical perspective with broader application potential. The perceived distinctness of the spectral signatures of the materials of interest is exploited for detecting targets in hyperspectral imagery. Space-time varying atmospheric perturbances on the radiation reaching a remote sensor are major limitations for designing a successful TD framework. Incorporating atmospheric components into a target detection framework is vital for practical applicability. Considered a general approach for flexibility, scalability, and optimal prediction, deep learning (DL) methods are increasingly used in various remote sensing applications. However, their potential for TD is relatively unexplored. Especially, the ability to provide training data sufficient for DL models and maintaining the functional relevance of the sparsely distributed targets in hyperspectral imagery are crucial for TD frameworks. This letter presents a novel method for training of DL architecture, called Deep Spectral Target Detector (DSTD). The proposed method includes a semi-supervised multi-scenario forward radiative transfer modelling (RTM) for the simulation of spectral signatures of various targets as training data suitable for the functional requirements of a typical DL architecture. We implemented the DSTD on a TD application-specific benchmark AVIRIS-NG airborne hyperspectral imagery acquired over a study site near Ooty, India. Compared to state-of-art statistical target detectors, the detection performance of the DSTD is superior to equivalent. Further, RTM-based training yields a robust model, impervious to the atmospheric mismatches between target collection and TD environments, indicating the potential for a similar approach to developing efficient DL-based methods for TD in the future.

Classification is one of the widely used techniques to quantify and mine the rich information present in hyperspectral imagery. However, the realization of trustworthy classification still continues to be a challenging task. This is due to the presence of various uncertainties, such as incomplete a priori knowledge on the actual number of classes and noise. Especially, mismatching between the number of spectral classes and the number of information classes leads to substantial omission or commission errors in the classification. In this letter, we present a new multiclass classification framework, named supervised cascaded classifier system (SC2S), that addresses the abovementioned problem by providing reliable results. The SC2S method is a two-stage cascaded classification procedure that involves quantifying the uncertainty and then classifying the samples. In the first stage, pixels for which no reliable training samples can be found in the training stage are detected. In the second stage, pixels that are matched with the respective training distribution are classified using a supervised learning algorithm, otherwise labeled as unknown. The proposed SC2S framework has been implemented on eight different classification scenarios with a varying number of unknown classes (UCs) using hyperspectral and multispectral imageries. The performance of the proposed SC2S method is compared with other widely used classifiers with and without reject-option. The experimental results indicate that our method offers superior classification results even in the case of data sets with a large number of unknown spectral classes.

In this paper, we introduce an active learning (AL)-based object training library generation for a multiclassifier object-oriented image analysis (OOIA) system. While several AL approaches do exist for pixel-based training library generation and for hyperspectral image classification, there is no standard training library generation strategy for OOIA of very high spatial resolution images. Given a sufficient number of training samples, supervised classification is the method of choice for image classification. However, this strategy becomes computationally expensive with the increase in the number of classes or the number of images to be classified. The above-mentioned issue is solved in this proposed method, where an optimized training library of objects (superpixels) is generated based on a batch mode AL approach. A softmax classifier is used as a detector in this method, which helps in determining the right samples to be chosen for library updation. To this end, we construct a multiclassifier system with max-voting decision to classify an image at pixel level. This algorithm was applied on three different very high-resolution airborne data sets, each with varying complexity in terms of variations in geographical context, sensors, illumination, and view angles. Our method has empirically outperformed the traditional OOIA by producing equivalent accuracy with a training library that is orders of magnitude smaller. In addition, the most distinctive ability of the algorithm is experienced in the most heterogeneous data set, where its performance in terms of accuracy is around twice the performance of the traditional method in the same situation. The generality of this classification strategy is proved through its performance on multispectral images and for cross-domain application. Finally, the robustness of this method is identified by comparing its performance with an alternative AL approach-self-learning-based semisupervised SVM. The capability of the proposed method to handle highly heterogeneous data is identified as the primary reason for its robustness.

Multiview registration is an important stage in three‐dimensional modelling pipeline. Motion averaging is an efficient approach for multiview registration which utilises the redundancy in overlap among the scans. The averaging of the underlying relative motions is performed in the corresponding Lie‐algebra elements of the SE (3) transformation matrices. However, this method is non‐robust and affected by the presence of outliers in the set of relative motions. The authors present a graph‐based approach to filter out the outliers before performing averaging of motions. The relative motions are assigned weights based on their agreement with global motions and other relative motions. The results indicate that the authors’ approach can efficiently filter out the outliers and can thus introduce robustness to multiview registration using motion averaging.

Multiple classifier system (MCS) is one of the effective strategies for hyperspectral image classification. Deploying different dimensionality reduction methods as the input data source to the MCS creates diversity among the base classifiers. The performance of the MCS is guaranteed when the base classifiers are accurate and diverse. Moreover the presence of the bad classifiers may negatively influence the performance of the MCS. In order to form a strong MCS, which are accurate as well as diverse, in this work the dynamic classifier system is developed. The dynamic classifier system selects the adaptive classifier from a pool of classifier for each dimensionality reduction method. The selected classifier relative to each dimensionality reduction method is further combined by different combination functions. Our experimental results on five multi-site hyperspectral images show the potential of dynamic classifier system to increase the classification accuracy significantly.