Abstract Accurate registration of high-resolution multispectral UAV orthomosaics acquired on different dates and sensors is essential for a wide range of remote sensing applications. However, this task remains challenging due to variations in acquisition conditions, including seasonal changes, differences in illumination, weather, and sensor characteristics. This article presents a parallel multilevel registration method that combines and improves two existing algorithms: HSI-KAZE, a feature-based approach, and HYFM, an area-based method. The three-level approach first applies an optimized HSI-KAZE (OHSI-KAZE) for coarse estimation of scale, rotation, and translation, followed by HYFM for fine correction. The multi-node multi-GPU proposed implementation, leveraging MPI, OpenMP, and CUDA, enables efficient processing on GPU-accelerated HPC clusters. Experiments on six real multispectral orthomosaic pairs from river environments, compared against state-of-the-art classical and deep learning methods, achieve high registration accuracy with RMSE values below 1.57 pixels and a 30 $$\times $$ × speedup, confirming both the accuracy and scalability of the proposed method.

The monitoring of semi-natural ecosystems has become increasingly critical due to the rising impact of ecological disturbances, including natural disasters and unauthorized human-made constructions. Anomaly detection (AD) in multispectral imagery serves as a fundamental tool in this context. Deep learning-based techniques are particularly effective at capturing the intricate spectral and spatial patterns of anomalies. This paper proposes a new AD technique called ACNN, designed to enhance AD performance in multispectral images of high spatial resolution for the detection of human-made constructions. The model integrates attention mechanisms to prioritize informative features while suppressing irrelevant background information, thereby improving sensitivity to subtle and rare anomalies. Experimental results on multispectral datasets from semi-natural ecosystems show that the proposed approach outperforms existing deep learning (DL) techniques in terms of detection accuracy. These findings highlight the potential of attention-based models as a robust framework for environmental monitoring and AD in complex remote sensing scenarios.

Feature-based methods are widely used for the registration of remote sensing images because of their robustness to viewpoint, scale, and light changes. However, they are computationally demanding, especially when dealing with multi or hyperspectral images. Hyperspectral Maximally Stable Extremal Regions (HSI-MSER) is a hyperspectral remote sensing image registration method based on MSER for feature detection and Scale Invariant Feature Transform (SIFT) for feature description. This article presents a first approach to a parallel implementation of the HSI-MSER algorithm for the registration of multispectral images on a heterogeneous computing platform. The results of the registration capabilities under extreme scaling and rotating conditions show that the proposed parallel implementation obtains a speedup of 3.33× compared to the sequential implementation making it suitable for applications with execution time constraints.

Abstract This paper proposes a parallel algorithm exploiting heterogeneous computing and edge computing for anomaly detection (AD) in remotely sensed multispectral images. These images present high spatial resolution and are captured onboard unmanned aerial vehicles. AD is applied to identify patterns within an image that do not conform to the expected behavior. In this paper, the anomalies correspond to human-made constructions that trigger alarms related to the integrity of fluvial ecosystems. An algorithm based on extracting spatial information by using extinction profiles (EPs) and detecting anomalies by using the Reed–Xiaoli (RX) technique is proposed. The parallel algorithm presented in this paper is designed to be executed on multi-node heterogeneous computing platforms that include nodes with multi-core central processing units (CPUs) and graphics processing units (GPUs) and on a mobile embedded system consisting of a multi-core CPU and a GPU. The experiments are carried out on nodes of the FinisTerrae III supercomputer and, with the objective of analyzing its efficiency under different energy consumption scenarios, on a Jetson AGX Orin.

Image registration is a crucial process that involves determining the geometric transformation required to align multiple images. It plays a vital role in various remote sensing image processing tasks that involve analyzing changes among images. To enable real-time response, it is essential to have computationally efficient registration algorithms, especially when dealing with large datasets as is the case of hyperspectral images. This article presents a comparative analysis of two descriptors used to characterize local features of images prior to their matching and registration. The objective is to analyze whether the LATCH binary keypoint descriptor, which produces compact descriptors, provides similar results to the gradient-based SURF descriptor in terms of execution time and registration precision. To obtain the best computational performance, multithreaded implementations using OpenMP have been proposed. LATCH has proven to be 7× faster and as reliable as SURF in terms of accuracy on scale differences of up to 1.2×.

On behalf of the Organizing Committee, we welcome you to the 31st Euromicro International Conference on Parallel, Distributed, and Network-Based Processing (PDP2023), organized by the Department of Science and Technology of the University of Naples “Parthenope”. The conference was hosted in Naples in the prestigious Villa Doria d'Angri from the 1st to the 3rd of March 2023.

Welcome to the 31st Euromicro International Conference on Parallel, Distributed, and Network-Based Processing (PDP2023).

Abstract Domain Adaptation (DA) is a technique that aims at extracting information from a labeled remote sensing image to allow classifying a different image obtained by the same sensor but at a different geographical location. This is a very complex problem from the computational point of view, specially due to the very high-resolution of multispectral images. TCANet is a deep learning neural network for DA classification problems that has been proven as very accurate for solving them. TCANet consists of several stages based on the application of convolutional filters obtained through Transfer Component Analysis (TCA) computed over the input images. It does not require backpropagation training, in contrast to the usual CNN-based networks, as the convolutional filters are directly computed based on the TCA transform applied over the training samples. In this paper, a hybrid parallel TCA-based domain adaptation technique for solving the classification of very high-resolution multispectral images is presented. It is designed for efficient execution on a multi-node computer by using Message Passing Interface (MPI), exploiting the available Graphical Processing Units (GPUs), and making efficient use of each multicore node by using Open Multi-Processing (OpenMP). As a result, an accurate DA technique from the point of view of classification and with high speedup values over the sequential version is obtained, increasing the applicability of the technique to real problems.

Feature-based registration methods have been demonstrated to be very effective to register multispectral images with large distortions or transformations despite the higher execution time that they require.In this paper, a first approach to a multi-node, multi-GPU implementation of the Hyperspectral KAZE (HSI-KAZE) method for co-registering bands and multispectral images is presented.Different multispectral datasets are distributed among the available nodes of a cluster using MPI and exploiting the parallel stream-based capabilities of the GPUs inside each node using CUDA.

Registration is required as a previous step for processing multispectral images. The different bands captured by each sensor for each image, as well as the different images corresponding to the same area, need to be aligned. In this paper, a 2-level registration scheme comparing the results obtained by the hyperspectral Fourier-Mellin (HYFM) and hyperspectral KAZE (HSI-KAZE) registration methods is proposed. It is designed for efficient implementation in a multi-GPU system in which different scenes are registered in parallel on different GPUs.

The alignment of images, also known as registration, is a relevant task in the processing of hyperspectral images. Among the feature-based registration methods, Speeded Up Robust Features (SURF) has been proposed as a computationally efficient approach. In this paper HSI-SURF is proposed. This is a method to register hyperspectral remote sensing images based on SURF that takes advantage of the full spectral information of the images. In this sense, the proposed method selects specific bands of the images and adapts the keypoint descriptor and the matching stages to benefit from the spectral information, thus increasing the effectiveness of the registration.

Change detection (CD) in multitemporal datasets is a key task in remote sensing. In this paper, a scheme to perform multiclass CD for remote sensing hyperspectral datasets extracting features by means of Stacked Autoencoders (SAEs) is introduced. The scheme combines multiclass and binary CD to obtain an accurate multiclass change map. The multiclass CD begins with the fusion of the multitemporal data followed by Feature Extraction (FE) by SAEs. The binary CD is based on the spectral information by calculating pixel-wise distances and thresholding, and it also incorporates spatial information through watershed segmentation. The processed image is filtered by using the binary CD map and later classified by a Support Vector Machine or an Extreme Learning Machine algorithm. The scheme was evaluated over a multitemporal hyperspectral dataset obtained from the Hyperion sensor. Experimental results show the effectiveness of the proposed scheme using a SAE for extracting the relevant features of the fused information when compared to other published FE methods.

This paper proposes an intrinsically distributed cellular automata (CA) based approach to address the perennial problem of real time segmentation and classification of high dimensional images, such as remote sensing hyperspectral images. This approach is efficiently implemented on GPUs providing results that improve on the state of the art algorithms presented in the literature. It is based on the evolutionary generation of the CA rule sets under two basic premises: During the segmentation process, the CAs must work over the whole dimensionality of the images without any projection onto lower dimensionalities, and the rule sets that are generated must be adapted to the segmentation level required by the user. The performance of the approach is tested over a benchmark set of well-known hyperspectral images and the results compared to the state of the art in the literature for two implementations, one using a SVM based classification stage and another that considers an ELM based classification stage.

SUMMARY One of the main factors that affect the performance of the sparse matrix–vector product (SpMV) is the low data reuse caused by the irregular and indirect memory access patterns. Different strategies to deal with this problem such as data reordering techniques have been proposed. The computational cost of these techniques is typically high because they consider all the nonzeros of the sparse matrix in order to find an appropriate permutation of rows and columns that improves the SpMV performance. In this paper, we analyze the possibility of increasing the locality of the SpMV using incomplete information in the reordering process. This partial information comes as a consequence of considering only a subset of the nonzero elements of the matrix. These nonzeros are obtained from the original matrix through a sampling process. In particular, two different sampling methods have been considered: a random sampling and an event‐based sampling using hardware counters. We have detected that a small number of samples is enough to obtain quality reorderings. As a consequence, using sampling‐based reorderings leads to noticeable performance improvements with respect to the non‐reordered matrices, reaching speedup values up to 2.1 × . In addition, an important reduction in the computational time required by the reordering technique has been observed. Copyright © 2012 John Wiley & Sons, Ltd.

Nowadays GPUs are commodity hardware containing hundreds of cores and supporting thousands of threads that can be used to accelerate a wide range of applications. From a programmer's perspective, GPUs offer a stream processing model which requires the application of new techniques to exploit their capabilities. In this paper we present the application of the split-and-merge technique to the following parallel tridiagonal system solvers on the GPU: cyclic reduction and recursive doubling. The split-and-merge technique naturally splits the algorithm flow in parallel paths that can be solved in shared memory, and later merged in global memory. In this way, we can solve large systems of equations efficiently exploiting the memory hierarchy of the GPU. The results obtained show a significant acceleration compared with the direct implementation of the algorithms on the GPU.

The watershed transform is a widely used method for non-supervised image segmentation, especially suitable for low-contrast images. In this paper we show that an algorithm calculating the watershed transform based on a cellular automaton is a good choice for the most recent GPU architectures, especially when the synchronization rules are relaxed. In particular we compare a synchronous and an asynchronous implementation of the algorithm. The results show high speedups for both implementations, especially for the asynchronous one, indicating the potential of this kind of algorithms for new architectures based on hundreds of cores.

The techniques for segmentation and classification of hyperspectral images are very costly, which makes them good candidates for parallel and, in particular, GPU processing. In this paper we present a GPU implementation of a segmentation strategy for hyperspectral images consisting in the calculation of a morphological gradient operator that reduces the dimensionality of the hyperspectral image followed by the calculation of a watershed transform over the resulting 2D image. We have studied the main issues for the efficient implementation of the algorithms in GPU: the exploitation of thousands of threads available in this architecture and the adequate use of the device bandwidth. The tests show the efficiency of the GPU implementation indicating that the processing of hyperspectral images can be performed in real-time even on commodity GPUs like the one used in the experiments.

Abstract In this paper the problem of the locality of sparse algebra codes on simultaneous multithreading (SMT) architectures is studied. In these kind of architectures many hardware structures are dynamically shared among the running threads. This puts a lot of stress on the memory hierarchy, and a poor locality, both inter‐thread and intra‐thread, may become a major bottleneck in the performance of a code. This behavior is even more pronounced when the code is irregular, which is the case of sparse matrix ones. Therefore, techniques that increase the locality of irregular codes on SMT architectures are important to achieve high performance. This paper proposes a data reordering technique specially tuned for these kind of architectures and codes. It is based on a locality model developed by the authors in previous works. The technique has been tested, first, using a simulator of a SMT architecture, and subsequently, on a real architecture as Intel's Hyper‐Threading. Important reductions in the number of cache misses have been achieved, even when the number of running threads grows. When applying the locality improvement technique, we also decrease the total execution time and improve the scalability of the code. Copyright © 2009 John Wiley & Sons, Ltd.