In this article, we explore the potential of quantum machine learning for subsurface feature extractions from radar sounder signals. We propose a hybrid quantum-classical learning paradigm that leverages parameterized quantum circuits to generate probability amplitudes based on quantum properties such as superposition and entanglement. These amplitudes are synergistically integrated with the classical deep neural networks that are efficient in learning high-dimension contextual features for downstream prediction tasks. The present research work is structured around two objectives. First, we investigate the role of quantum circuits in the latent space for transferring back-and-forth rich discriminative spatial context from the encoder to the decoder for segmentation. Second, we investigate how the probabilistic amplitudes derived from quantum circuits are significant in integrating into the classical models to provide new insights for radar sounder signals segmentation. The performance of the hybrid architectures has been studied in small-scale settings by simulating the expected behaviour of the quantum circuits on a classical machine. The experimental results have demonstrated the viability of quantum machine learning frameworks on MCoRDS-1 and MCoRDS-3 datasets for radar sounder signal segmentation. Qualitatively, they are capable of delineating the spatial extent of the bedrock from noise. Additionally, we conduct a comparative analysis between theQiskit Aer Simulatorand theIBM FakeBackend Simulatorto highlight the computational trade-offs and validate fidelity of two simulators for scalable experimentation. Therefore, our work opens up new avenues of research for future radar sounder data analysis leading to more precise and efficient subsurface target segmentation.

This article presents a hybrid quantum-classical framework by incorporating quantum feature maps regulated classical Convolutional Neural Network (CNN) architecture in the context of detecting different subsurface targets in the radar sounder signal. The quantum feature maps are generated by quantum circuits to utilize spatially-bound input information from the input training samples. The associated spectral probabilistic amplitudes of the feature maps are further fed as an input to the classical CNN-based network to classify the subsurface targets in the radargram. Experimental results on the MCoRDS and MCoRDS3 dataset demonstrated the capability of contextualizing the classical architecture through quantum feature maps for characterizing the radar sounder data.

Unsupervised semantic segmentation is the method of discovering meaningful semantic contents within the image domain without using any labelled information. The learned semantic contents are then decomposed into distinct semantic segments with known ontology. The core task of an unsupervised feature learning algorithm is to produce dense features for every pixel with rich semantic content to form distinct clusters with compact information for the downstream task. In this work, we extend the previously developed Self-Supervised Transformer with Energy-based Graph Optimization (STEGO) architecture by integrating a convolution-based Expansive Network in the decoder along with the spatial similarity loss function for radar sounder signal segmentation. Experimental results on the Multi-Channel Coherent Radar Depth Sounder (MCoRDS) data confirm the capability of the proposed unsupervised segmentation method.

Radar Sounders (RSs) are space-borne and airborne sensors operating on the nadir-looking geometry to collect sub-surface information by transmitting linearly modulated electro-magnetic (EM) pulses and receiving backscattered (reflected from different subsurface targets) echoes. The echoes are coherently represented to generate radargrams. A radargram is used to characterize subsurface target structures. Interestingly, radargram signals depict sequential structures due to linearly homo-geneous subsurface target features such as ice layers. Several automatic techniques are proposed to characterize the subsurface targets in the radargrams mostly associated with the probabilistic models or CNN-based deep learning models. The CNN-based architectures explicitly model the local spatial high dimensional contexts which are often infeasible for establishing the long-range sequential contextual relationship between local spatial features. Motivated by the aforementioned fact, we propose a hybrid CNN-Transformer-based encoder-decoder architectural framework for addressing the long-range sequential contextual dependencies within the sequential structures of RS signals. We tested the architecture on Multi-channel Coherent Radar Depth Sounder (MCoRDS) dataset. Experimental results confirm the capability of Transformers to characterize the subsurface targets.

Radar Sounders (RSs) are nadir-looking sensors operating in high frequency (HF) or very high frequency (VHF) bands that profile subsurface targets to retrieve miscellaneous scientific information. Due to complex electromagnetic interaction between back-scattered returns, the interpretation of RS data is challenging. The investigations of ice-sheet subsurface structures require automatic techniques to account for both the sequential spatial distribution of subsurface targets and relevant statistical properties embedded in RS signals. Automatic techniques exist for characterizing these targets either related to probabilistic inference models or convolutional neural network (CNN) deep learning methods. Unfortunately, CNN-based methods capture local spatial context and merely model the global spatial context. In contrast to CNN, the Transformer-based models are reliable architectures for capturing long-range sequence-to-sequence global spatial contextual prior. Motivated by the aforementioned fact, we propose a novel Transformer-based semantic segmentation architecture named TransSounder to effectively encode the sequential structures of the RS signals. The TransSounder was constructed on a hybrid TransUNet-TransFuse architectural framework to systematically augment the modules from TransUNet and TransFuse architectures. Experimental results obtained using the Multichannel Coherent Radar Depth Sounder (MCoRDS) dataset confirms the robustness and capability of Transformers to accurately characterize the different subsurface targets.

Coal fire is a serious threat in major coal producing countries across the globe and poses significant constraints in mining operations, often leading to environmental degradation. The applications of thermal and shortwave infrared remote sensing play a substantial role in systematically detecting and monitoring the coal fire. Over the last few decades, researchers have extensively examined the importance of spectral radiance for retrieving reliable pixel-integrated temperature threshold to delineate coal fire from its background. However, such an assumption does not necessarily consider the local information, thereby leading to difficulty in isolating the actual coal fire affected pixels. Therefore, we propose to utilise the channel specific reflectance to retrieve the thermally anomalous pixels in coal fire related applications using Landsat 8 OLI data. This paper explores the practicability of incorporating the active fire detection technique using channel specific reflectances based on both fixed and contextual thresholds in the Jharia coalfield, India.