Using heterogeneous computing and edge computing to accelerate anomaly detection in remotely sensed multispectral images

Vol.:(0123456789) The Journal of Supercomputing https://doi.org/10.1007/s11227-024-05918-z 1 3 Using heterogeneous computing andedge computing toaccelerate anomaly detection inremotely sensed multispectral images Javier López‑Fandiño1,2· DoraB.Heras1,2· FranciscoArgüello2 Accepted: 15 January 2024 © The Author(s) 2024 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 multicore 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. Keywords Multispectral· Anomaly detection· Extinction profiles· Heterogeneous computing· Edge computing Javier López-Fandiño, Dora B. Heras and Francisco Argüello contributed equally to this work. * Javier López-Fandiño javier.lopez.f[email protected] Dora B. Heras [email protected] Francisco Argüello [email protected] 1 Centro Singular de Investigación en Tecnoloxías Intelixentes (CiTIUS), Universidade de Santiago de Compostela, SantiagodeCompostela, ACoruña, Spain 2 Departamento de Electrónica e Computación, Universidade de Santiago de Compostela, SantiagodeCompostela, ACoruña, Spain J.López-Fandiño et al. 1 3 1 Introduction Anomaly detection (AD) plays a critical role in numerous fields, including remote sensing, surveillance, and environmental monitoring [1–3]. In the context of multispectral image analysis, AD techniques hold significant potential for identifying and characterizing irregularities or deviations from the norm in the images, providing valuable insights into complex natural systems [4]. The processing of multispectral or hyperspectral images always demands a considerable amount of computational resources due to the high dimensionality and complexity of the data and the need for real-time processing for many applications. This is especially true in the case of AD tasks over very high-resolution images, as AD algorithms require processing the whole spatial and the spectral information available in the images [4–6]. Different computational paradigms, ranging from high-performance computing (HPC) platforms such as clusters, grids, or clouds, to specialized accelerators such as graphics processing units (GPUs), field-programmable gate arrays (FPGAs), or even quantum computing solutions, have been exploited in this context. The choice of the most appropriate computing platform depends on the problem at hand and on the context in which it must be addressed [7]. For instance, in certain scenarios, data may be efficiently offloaded to supercomputers, while in others, it may be more practical to tackle the problem insitu using edge computing solutions or, simply, the commodity hardware available, for instance, a personal computer with a hardware accelerator such as a GPU. Supercomputers play a dominant role in parallel computing applied to remote sensing [8]. A supercomputer is a mixture of shared and distributed memory systems that, in many cases, include heterogeneous nodes. Each node could consist of central processing units (CPUs), GPUs, FPGAs, or any other accelerator. Many models and portable libraries have emerged as possible standards for supercomputer programming, being OpenMP [9] and MPI [10] the most widely used. In fact, MPI is the facto current standard for parallel programming. Accelerators have also played a dominant role in parallel computing and, especially, in image processing including multidimensional images from the remote sensing field [7]. GPUs, renowned for their ability to perform thousands of computations in parallel, excel in tasks involving repetitive mathematical operations. Their highly parallel nature based on multi-threaded, many-core processors and their very high memory bandwidth make them particularly well-suited for processing large-scale datasets and computationally intensive tasks. The development of NVIDIA’s Compute Unified Device Architecture (CUDA) [11] has simplified the programming model for GPUs. Many libraries have been developed to provide the most common operations implemented in CUDA. OpenCL is also a relevant standard for parallel programming across CPUs, GPUs, and other processors [12]. Presently, GPU-based computers integrate the attributes of general-purpose computing, a high degree of parallelism, and high memory bandwidth, at a relatively lower cost compared to other alternatives. This makes them an attractive 1 3 Using heterogeneous computing andedge computing toaccelerate… option compared to a massively parallel system made up of only CPUs [13]. The heterogeneous computing approach that combines the strengths of both CPUs and GPUs can lead to even greater performance gains. By distributing the computational load between CPUs and GPUs based on their respective strengths, a more balanced and efficient workflow can be achieved, thereby maximizing overall processing speed and resource utilization. This kind of architectures has been previously used to tackle other tasks in multispectral images, such as registration [14] or domain adaptation [15], among others. Recently, the use of edge computing architectures has arisen as a promising hardware alternative for remote sensing applications [7]. Edge computing is effective in reducing the delay between the acquisition and the processing of data [16]. Most edge computing devices are heterogeneous computing platforms such as, for example, the NVIDIA Jetson [17] used in this paper. These devices can help to make it possible to perform onboard remote sensing computation. Edge computing architectures are also very effective in reducing power consumption, which, nowadays, is a relevant requirement from both the sustainability and the economic perspectives. This can be especially important in real-time monitoring scenarios when the time needed to transfer the raw data to HPC centers is not an acceptable option [18]. Addressing the challenge of reducing the AD computation times, this paper presents an efficient heterogeneous parallel implementation to perform AD in highresolution multispectral images. The objective is the detection of anomalies corresponding to human constructions within natural fluvial ecosystems. The resulting algorithm is executed on both a supercomputer and an edge computing device. Our methodology allows to automatically identify and characterize anomalies corresponding to human-made constructions by combining the use of extinction profiles (EPs) for extracting spatial information from the images with the well-known Reed–Xiaoli (RX) AD algorithm. The proposed algorithm can contribute to the development of effective conservation and management strategies for fluvial ecosystems, promoting sustainable development while safeguarding the ecological integrity of the ecosystem. The remainder of this paper is structured as follows: Sect.2 reviews the background related to efficient parallel implementations of AD for multispectral and hyperspectral images. In Sect.3, the AD algorithm proposed in this work is introduced, with a particular emphasis on the parallel features of the algorithm. Section4 describes the experimental setup and presents the results obtained on a real-world dataset of multispectral images. Finally, in Sect. 5, the concluding remarks are presented. 2 Related work Processing multispectral images of fluvial ecosystems presents several challenges due to the inherent complexity and variability of natural landscapes [19]. The similarity between the spectral signatures of the various materials in the scene, such as geological features and human-made structures, coupled with the prevalence of vegetation in the landscapes under study, pose difficulties in distinguishing human J.López-Fandiño et al. 1 3 constructions, which seamlessly blend into the surrounding environment [20]. Additionally, multispectral images are vulnerable to atmospheric conditions, sensor noise, and fluctuating illumination, adding complexity to the accurate identification of anomalies in these dynamic environments [21]. Various approaches have been proposed to tackle the challenging problem of AD in multispectral images [19]. These methods can be broadly categorized into unsupervised, supervised, and semi-supervised techniques. Unsupervised methods, such as statistical modeling and clustering algorithms, do not require labeled training data and can automatically identify anomalies based on deviations from the normal data distribution [1, 22, 23]. Supervised methods, on the other hand, rely on labeled training data to learn the characteristics of normal and anomalous samples, enabling them to classify new instances accordingly [2, 24–26]. Semi-supervised methods aim to strike a balance between the two by utilizing a limited amount of labeled data and a larger pool of unlabeled data during the training process. One notable algorithm that has garnered attention in AD is the RX algorithm [27]. The RX algorithm is an unsupervised technique based on the concept of the Mahalanobis distance. It characterizes anomalies by measuring the spectral deviations from the local mean of the surrounding pixels. The effectiveness of the RX algorithm lies in its ability to handle high-dimensional data and identify subtle anomalies, making it a promising candidate for detecting human constructions within fluvial ecosystems in multispectral images. Several variations of the RX algorithm have been applied to improve the detection accuracy of the traditional algorithm [28–31]. In the context of AD in multispectral images within fluvial ecosystems, incorporating spatial information alongside spectral data becomes a crucial aspect of achieving accurate and reliable results, as it has been shown by [32] for AD in hyperspectral images. This has also been explored in the previous classification-oriented works [23, 33–35]. Spectral information alone may not provide sufficient context to differentiate between natural variations and genuine anomalies in the complex and heterogeneous landscapes of fluvial ecosystems. By considering the spatial relationships among neighboring pixels, valuable contextual cues can be extracted, enhancing the discrimination of anomalies from the background. In this work, we incorporate the spatial information together with the spectral one by introducing the filtering technique called EP [36], which is an alternative to the widely recognized attribute profile (AP) [37, 38]. EP is based on the concept of extinction filters (EFs) which are extrema-oriented connected filters that, unlike the AP, preserve the original height of the extrema kept in the image. The parameter tuning is also easier for EPs than for APs due to the fact that they are independent of the kind of attribute being used and only based on the number of extrema to be kept at each level of the EP [36]. Several works have explored the use of parallelization techniques to efficiently tackle the real-time AD problem. [39] presents a GPU implementation of the RX algorithm for AD using CUDA. It remarks the relevance of processing sub-images of the image independently in different hardware to exploit the parallelism minimizing the communications, as they are a common bottleneck in multi-GPU implementations. The paper also analyzes the power consumption of the algorithms, which is especially relevant for real Earth observation missions using onboard computation. 1 3 Using heterogeneous computing andedge computing toaccelerate… The projection of AD to a widely used edge computing device, an NVIDIA Jetson GPU, is explored in [40] over a hyperspectral urban image acquired by the AVIRIS sensor, concluding that there is a promising solution for hyperspectral image processing in low-power consumption scenarios. In turn, [41] proposes the use of FPGAs to perform recursive RX-based AD in hyperspectral images from both urban and natural scenes, also acquired by the AVIRIS sensor. They show how different variants of the traditional RX algorithm can be projected efficiently to FPGA architectures, along with GPU and cloud computing alternatives, concluding that the main limiting factor in FPGAs and GPUs is memory capacity. This limiting factor can be avoided by scaling the number of nodes used. Another FPGA AD algorithm focused on minimizing power consumption is presented in [42], demonstrating that hyperspectral images with thousands of pixels and hundreds of spectral bands can be processed with a power budget of only 1.3 W. Embedded devices such as FPGAs can also be used to perform AD through the use of deep learning techniques. For instance, a deep convolutional neural network is used in [43] for natural anomaly detection in multispectral images. Another possibility to make the algorithms more adequate for edge computing devices is to align the data processing with the data acquisition process. For this purpose, a line-by-line AD technique is presented in [44], which aims to process hyperspectral data in a manner consistent with its collection by push-broom scanners. When assessing a task such as AD is also important to take into account that different architectures may be better suited for each specific processing step. This aspect is explored in [45], where a combination of CPU and GPU is employed to achieve a time-efficient AD technique. The proposed technique is based on the use of multivariate normal mixture models applied to a simulated search and rescue scenario in a real hyperspectral image captured by a HySpex visual and near-infrared (VNIR) hyperspectral camera. 3 Heterogeneous parallel EP‑based AD scheme In this section, the proposed parallel algorithm for detecting anomalies in high-res- olution multispectral images of fluvial ecosystems is presented. A hybrid algorithm using MPI, OpenMP, and CUDA is used to exploit the different levels of parallelism offered by a heterogeneous architecture that, as it will be explained later, includes nodes with a multi-core CPU and a GPU. The outline of the algorithm is shown in Fig.1. It consists of three main stages: First, the spatial information is extracted from the input image by computing an EP over each band of the image separately. The five bands available in the images under study are represented in the figure. As a result, an extended EP is produced by accumulating the results of the individual EPs constructed for each band of the image. Then, a second stage consists of applying the RX anomaly detector over the resulting extended EP. This stage is more efficiently calculated over only one node of the computer platform. A two-dimensional gray-scale intensity image is obtained as a result. Finally, the processing requires calculating and applying a threshold to J.López-Fandiño et al. 1 3 produce the output AD map. This is the objective of the third stage, which applies Otsu’s algorithm, producing a binary map of anomalies as output. More details of the implementation presented in Fig.1 are offered in Algorithm1, where all the computational steps for each stage of the algorithm and the platforms where are computed are annotated on the right side. More in detail, the first stage of the algorithm presented in Fig.1 and detailed in lines 1–18 of Algorithm1, can be considered as a spatial processing where the structures of interest of the image are highlighted through the computation of the EPs. As the anomalies that need to be detected correspond to a set of uniform pixels more than isolated pixels, this spatial processing stage helps in identifying these structures at different levels. This is a very common process followed when changes or anomalies need to be detected in multi- or hyperspectral images. In this case, the method for extracting spatial information is the use of EPs. The processing of the EP for each image band is independent, making it a good candidate to be computed in parallel. After the tasks corresponding to the processing of each band are distributed through the use of MPI, a hybrid CPU-GPU approach is used to tackle this operation. As each EP calculation consists of some steps where the individual operations are performed over each pixel of the image, it is a perfect fit to be computed in GPU using CUDA. Other steps that are performed over the node-array representation of the image tree [36] offer fewer opportunities for parallelization to be more efficiently computed by the CPU. This is the case for the stages devoted to obtain the parents of each node by means of the union-find algorithm and the computation of the node array (steps 4–5 in pseudocode in Algorithm1). The calculation of the EP of a one-band gray-scale image consists of the application of several opening and closing operations at different granularity levels over the Fig. 1 Parallel RX+EP AD algorithm for heterogeneous computing 1 3 Using heterogeneous computing andedge computing toaccelerate… original images [36]. So, inside each node, different levels of the EP need to be computed. To perform this operation in the most efficient way, the node-array representation of the image is computed first (lines 2–5 in the pseudocode). This representation can be seen as a max-tree [36] computation where the parent of each node is obtained in a union-find process and stored in a structure together with the attribute of interest for each node, that is, the area in this case. Once this information is gathered, it is possible to compute the extinction values of each node as proposed by [36] (line 6). The steps corresponding to the application of the extinction filter for each number of extrema selected (lines 7–9 and 16–18 in Algorithm1) can be computed in parallel through the use of OpenMP once the extinction values of the image have been computed in the previous step, as they are independent computations for each level of the EP. Figure2 illustrates the EP for one band of a multispectral image. The original band is in the middle, and the different components of the profile, corresponding to the result of applying a three-level EP, are represented on both sides. It is worth noting that, for the computation of the closing profiles (lines 10–18), the same code is used but carried out over the negated input image. This is a usual approximation to this task that has been used, for instance, in [38]. The EPs for the different bands of the input multispectral image are concurrently computed in different nodes of the computing platform, through the use of MPI, thus reducing the computation time. The opening and closing profiles individually computed for each band of the image are stored in memory composing an extended EP. This EP will be the input of the second stage of the algorithm. The second stage of the algorithm consists of the application of the anomaly detector, the RX algorithm, over the extended EP. As the second stage of the algorithm (lines 19–22) presents a low computational load, it is efficiently computed in only one node, so the extended EP is computed by gathering the individual EPs, as shown in Fig.1. This part of the algorithm is individually performed Fig. 2 EP computation for one band of a multispectral image (first row) and zoom over a small region (second row) J.López-Fandiño et al. 1 3 over each pixel and, therefore, totally computed in GPU to achieve low execution times. We can identify in the algorithm the steps for the parallel calculation of the RX anomaly detector [27] to obtain a one-band gray-scale image where the higher the intensity of a pixel, the higher the probability of it being an anomaly. It is based on the calculus of the Mahalanobis distance between each pixel of the stacked EP and the average pixel value of the same stacked EP. All the steps are calculated in GPU using CUDA. The third and final stage of the algorithm performs the application of a threshold technique to obtain a binary AD map (lines 23–26 in the algorithm) identifying each pixel as anomaly or non-anomaly. This is tackled through Otsu’s threshold algorithm as it is an automatic threshold technique that has proven to produce the best discrimination with a low computational cost [46, 47]. This algorithm calculates the threshold based on the histogram of the gray levels of the image. This stage of the algorithm is also executed on GPU as both the histogram calculation step and the final binary decision over each pixel of the image can benefit from the highly parallelizable GPU architecture thus achieving lower execution times. Algorithm1 Parallel RX+EP AD algorithm. 1 3 Using heterogeneous computing andedge computing toaccelerate… 3.1 Comparison withother parallel AD implementations The algorithm proposed in this paper aims to exploit the capabilities of the available hardware on a heterogeneous computing platform at different levels. This has been the methodology used by other parallel AD implementations in the literature, such as [45], where a dual Quad-Core Intel Xeon CPU and a NVidia GeForce 8800 Ultra GPU are used to reduce the execution times of an AD method for hyperspectral images based on the use of multivariate normal mixture models. Similarly in [41], the authors conclude that embedded, GPU, and cloud architectures should be combined to achieve efficient processing of remote sensing data. In particular, in our paper, the first stage of the algorithm, as we described in Algorithm1, exploits a multi-node implementation for extracting parallelism. The reason is that the EP calculation computed by this stage of the algorithm can be individually processed for each band. The implementation combines the use of MPI, OpenMP, and CUDA to efficiently distribute the computational load of this stage among the available nodes of a supercomputing device. The same approach is applied, for instance, by [41], which introduces a multi-node cloud implementation of an AD algorithm based on the use of RX. The second stage, the RX detector, is performed entirely in a single GPU by using CUDA in our algorithm. Given that the amount of data to be processed is small in this stage, it would be difficult to compensate for the communication times needed to use a distributed architecture. This task has been tackled similarly in other parallel AD implementations such as [39, 41]. The same remarks apply to the Otsu’s thresholding needed to obtain the final AD map. Finally, when prioritizing the minimization of power consumption, the use of embedded devices has been established in the literature as the optimal alternative [40, 42]. Different parallel AD algorithms in the literature, for instance, [41] or [43], have been projected to FPGAs. In our paper, an NVIDIA Jetson platform [17] is used. This platform presents, in comparison with a FPGA, the advantage that the implementation required is more similar to that of general-purpose architectures. 4 Experimental results This section is devoted to summarize the experiments carried out for the validation of the algorithm presented in this work for the AD of human-made constructions in fluvial ecosystems. First, the dataset and experimental setup selected for the experimentation will be presented in Sects.4.1 and 4.2. Then, the achieved accuracy and performance results will be analyzed in Sect.4.3. 4.1 Dataset The experiments were conducted utilizing data captured by the MicaSense RedEdge multispectral sensor, onboard a specialized unmanned aerial vehicle (UAV). This J.López-Fandiño et al. 1 3 Table 5 Execution times, in seconds, and speedups for z1 and z2 images for the RX+EP(64,8,4,2) and RX+EP(32,16,2,1) configurations, respectively, on FinisTerrae III. Times include CPU processing, network communication costs, and GPU computations Step z1 image z2 image Single-node Multi-node Single-node Multi-node CPU CPU-GPU CPU-GPU CPU CPU-GPU CPU-GPU EP opening Order image 0.6539 0.0809 8.1 × 0.0163 40.2 × 0.0988 0.0232 4.3 × 0.0047 21.1 × Get parents 2.6112 2.5340 1.0 × 0.5052 5.2 × 0.6729 0.6462 1.0 × 0.1297 5.2 × Node array 0.5631 0.5933 0.9 × 0.1181 4.8 × 0.1484 0.1531 1.0 × 0.0308 4.8 × Extinction values 0.1207 0.1021 1.2 × 0.0199 6.1 × 0.0465 0.0403 1.2 × 0.0082 5.7 × EP application 0.2264 0.1619 1.4 × 0.0334 6.8 × 0.0586 0.0460 1.3 × 0.0090 6.5 × EP closing Negate image 0.0623 0.0012 52.4 × 0.0002 264.8 × 0.0390 0.0004 96.0 × 0.0001 469.3 × Order image 0.6604 0.0776 8.5 × 0.0142 46.4 × 0.0990 0.0238 4.2 × 0.0050 19.7 × Get parents 2.5567 2.4906 1.0 × 0.5006 5.1 × 0.6572 0.6335 1.0 × 0.1275 5.2 × Node array 0.5153 0.5128 1.0 × 0.1025 5.0 × 0.1387 0.1387 1.0 × 0.0279 5.0 × Extinction values 0.0964 0.0704 1.4 × 0.0140 6.9 × 0.0423 0.0342 1.2 × 0.0069 6.1 × EP application 0.2114 0.1320 1.6 × 0.0273 7.7 × 0.0512 0.0359 1.4 × 0.0071 7.2 × EP total 8.2778 6.7568 1.2 × 1.3519 6.1 × 2.0526 1.7752 1.2 × 0.3569 5.8 × RX: mean 0.5363 0.0064 83.3 × 0.0064 83.5 × 0.1293 0.0012 108.2 × 0.0012 108.4 × RX: covariance 24.5064 0.8390 29.2 × 0.8390 29.2 × 5.8359 0.1996 29.3 × 0.1996 29.2 × RX: inverse cov 0.0001 0.0007 0.1 × 0.0007 0.1 × 0.0001 0.0007 0.1 × 0.0007 0.1 × RX: Mahalanobis 20.0732 0.1191 168.5 × 0.1191 168.6 × 4.8048 0.0292 164.6 × 0.0292 164.6 × Otsu’s threshold 0.0106 0.0038 2.8 × 0.0038 2.8 × 0.0019 0.0008 2.2 × 0.0008 2.2 × RX+Otsu total 45.1266 0.9682 46.5 × 0.9682 46.6 × 10.7719 1.2316 8.8 × 1.2316 8.7 × Total 53.4044 7.7251 6.9 × 2.3201 23.0 × 12.8245 3.0068 4.3 × 1.5885 8.1 × 1 3 Using heterogeneous computing andedge computing toaccelerate… spatial dimension. The time is shown separately for the EP stage and for the RX and Otsu stages. As can be seen, the time needed for the execution of the algorithm for these images scales nearly linearly with the image size. Nevertheless, in the 4 × size image, it starts to be noticeable that the spatial part of the processing, the EP, loses some performance as it needs 4.2 × the time needed for the original image. On the other hand, the RX detector achieves better performance (it only needs 3.69 × the time needed for the original image). This can be explained because the spectral part is more suitable for pixel-level GPU parallelism and the bigger the number of pixels to process, the bigger the opportunities for exploiting the high number of processing cores available in the GPU. Regarding the edge computing Jetson platform introduced in Sect.4.2, experiments have been carried out with three different power budgets (15 W, 30 W, and MAXN (60 W)). The findings previously commented in the FinisTerrae III platform remain valid for this platform: The Get parents and Node array steps of the EP computation are still the most time-consuming steps, and the speedups achieved when the available GPU is exploited also remain similar. Nevertheless, as was expected, the more limiting the power budget selected, the higher the execution times achieved for the same computations. The most parallelizable steps are those that benefit the most from the higher power budgets, as they can exploit both, the larger number of cores available and the increased frequency of the CPU and GPU. In this way, the RX: mean and the RX: Mahalanobis steps are the ones with a higher increase in speedup when the CPU version is compared with the CPU-GPU one. The speedup increasing on these steps ranges from 4 × to 10 × for the 15 W power budget up to 9.7 × to 17.7 × for the 60 W power budget. As it is shown in Table6, the reduction in the available power and the disabling of some hardware components, resulting in fewer parallel computing opportunities for the algorithm, greatly increase the time needed for the computation up to 7.1 × . This makes it clear that the power requirements must be carefully chosen in an Fig. 7 Execution time scaling when the size of the z1 image is increased. The times are expressed in multiples of the execution time over the times for the original z1 image. Times include CPU processing, network communication costs, and GPU computations J.López-Fandiño et al. 1 3 edge computing environment to achieve the right balance between autonomy and performance. 5 Conclusions This paper introduces a computationally efficient parallel algorithm for AD. The algorithm is specifically designed to run on heterogeneous computing platforms, comprising nodes with multi-core CPUs and GPUs. AD is accomplished through a combination of an extended extinction profile for spatial information extraction and a detector known as the RX algorithm. The resulting parallel hybrid MPI+OpenMP+CUDA algorithm outperforms a traditional RX approach, detecting up to 27% more anomalies in the images presented in this paper. Experiments were conducted using the FinisTerrae III multinode supercomputer, analyzing high-resolution multispectral images of fluvial ecosystems. Speedups of up to 23 × were achieved. Furthermore, the same algorithm was executed on a mobile embedded system, specifically a NVIDIA Jetson. This aimed to assess the feasibility of running the algorithm under various power consumption limitations. Experiments have shown an increase of up to 7.1 × in execution time when the power consumption is limited to 15 W, compared to the situation with a limit of 60 W. A challenge for the future would be addressing workload imbalances when processing images, such as hyperspectral ones, with a higher number of spectral bands. In such cases, it may not be practical to allocate as many nodes as there are bands in the image. Additionally, adapting the implementation to work with series of images covering more extensive spatial areas would allow the application of the algorithm to a wide variety of remote sensing applications. Table 6 Summary of execution times, in seconds, and speedups, with respect to the 15 W version, for different versions of the algorithm on the Jetson AGX Orin. Times include CPU and GPU processing Version z1 image z2 image CPU CPU-GPU CPU CPU-GPU 15 W EP 140.2101 105.9384 6.4297 5.0576 RX+Otsu 87.7020 29.9635 21.7822 7.9783 Total 227.9121 135.9019 28.2119 13.0360 30 W EP 23.8902 20.9899 5.7982 5.7447 RX+Otsu 60.4277 16.6504 14.9905 3.7830 Total 84.3178 37.6403 20.7887 9.5278 60 W EP 14.4470 13.1580 3.5670 3.6841 RX+Otsu 45.1647 6.0055 11.1657 1.2641 Total 59.6117 19.1635 14.7327 4.9481 Speedup 30 W 2.7 × 3.6 × 1.4 × 1.4 × Speedup 60 W 3.8 × 7.1 × 1.9 × 2.6 × 1 3 Using heterogeneous computing andedge computing toaccelerate… Acknowledgements The authors want to acknowledge Centro de Supercomputación de Galicia (CESGA) for providing access to the supercomputer FinisTerrae III. We also would like to thank Augas de Galicia from Xunta de Galicia for the support in the construction of the reference information for the multispectral images. Funding Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work was supported in part by the Civil Program UAVs Initiative, promoted by the Xunta de Galicia and developed in partnership with the Babcock company to promote the use of unmanned technologies in civil services. It was also supported by grants PID2019–104834GB–I00, PID2022–141623NB-I00, and TED2021–130367B–I00 funded by MCIN/AEI/10.13039/501100011033 and by “European Union Next- GenerationEU/PRTR.” We also have to acknowledge the support by Xunta de Galicia—Consellería de Cultura, Educación, Formación Profesional e Universidades [Centro de investigación de Galicia accreditation 2019–2022 ED431G-2019/04 and Reference Competitive Group accreditation, ED431C-2022/16], by Junta de Castilla y León [Project VA226P20 (PROPHET–II)], and by European Regional Development Fund (ERDF). Data Availability Statement Restrictions apply to the availability of the data used in this study. Datasets are the property of Babcock International and are used with permission. Declarations Conflicts of interest The authors have no competing interests as defined by Springer, or other interests that might be perceived to influence the results and/or discussion reported in this paper. Ethical approval Not applicable. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. References 1. Guo Q, Pu R, Cheng J (2016) Anomaly detection from hyperspectral remote sensing imagery. Geosciences 6(4):56 2. Hu X, Xie C, Fan Z, Duan Q, Zhang D, Jiang L, Wei X, Hong D, Li G, Zeng X etal (2022) Hyperspectral anomaly detection using deep learning: a review. Remote Sens 14(9):1973 3. Racetin I, Krtalić A (2021) Systematic review of anomaly detection in hyperspectral remote sensing applications. Appl Sci 11(11):4878 4. Su H, Wu Z, Zhang H, Du Q (2021) Hyperspectral anomaly detection: a survey. IEEE Geosci Remote Sens Mag 10(1):64–90 5. Han W, Zhang X, Wang Y, Wang L, Huang X, Li J, Wang S, Chen W, Li X, Feng R etal (2023) A survey of machine learning and deep learning in remote sensing of geological environment: challenges, advances, and opportunities. ISPRS J Photogramm Remote Sens 202:87–113 6. Di, L., Yu, E.: Challenges and opportunities in the remote sensing big data. Remote Sens Big Data, pp 281–291 (2023) 7. Cavallaro G, Heras DB, Wu Z, Maskey M, Lopez S, Gawron P, Coca M, Datcu M (2022) Highperformance and disruptive computing in remote sensing: Hdcrs-a new working group of the grss J.López-Fandiño et al. 1 3 earth science informatics technical committee [technical committees]. IEEE Geosci Remote Sens Mag 10(2):329–345 8. Plaza A, Du Q, Chang Y-L, King RL (2011) Foreword to the special issue on high performance computing in earth observation and remote sensing. IEEE J Select Top Appl Earth Observ Remote Sens 4(3):503–507 9. Dagum L, Menon R (1998) Openmp: an industry standard api for shared-memory programming. IEEE Comput Sci Eng 5(1):46–55 10. Dongarra J, Walker D, Lusk E, Knighten B, Snir M, Geist A, Otto S, Hempel R, Lusk E, Gropp W etal (1994) Mpi-a message-passing interface standard. Int J Supercomput Appl High Perform Comput 8(3–4):165 11. NVIDIA, Vingelmann P, Fitzek, FHP (2020) CUDA, release: 10.2.89 https:// devel oper. nvidia. com/ cudatoolk it 12. Munshi A (2009) The opencl specification. In: 2009 IEEE Hot Chips 21 Symposium (HCS), pp 1–314 IEEE 13. Christophe E, Michel J, Inglada J (2011) Remote sensing processing: from multicore to gpu. IEEE J Select Top Appl Earth Observ Remote Sens 4(3):643–652 14. Ordóñez, Á., Heras, D.B., Argüello, F.: Multi-gpu registration of high-resolution multispectral images using hsi-kaze in a cluster system. In: IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium, pp. 5527–5530 (2022). IEEE 15. Garea AS, Heras DB, Argüello F, Demir B (2023) A hybrid cuda, openmp, and mpi parallel tcabased domain adaptation for classification of very high-resolution remote sensing images. J Supercomput 79(7):7513–7532 16. Haut JM, Bernabé S, Paoletti ME, Fernandez-Beltran R, Plaza A, Plaza J (2018) Low-high-power consumption architectures for deep-learning models applied to hyperspectral image classification. IEEE Geosci Remote Sens Lett 16(5):776–780 17. NVIDIA (2022) Nvidia jetson agx orin series technical brief v1.2. Technical report, NVIDIA https:// www. nvidia. com/ conte nt/ dam/ enzz/ Solut ions/ gtcf21/ jetson- or in/ nvidiajetson- agxorin- techn icalbrief. pdf 18. Liu J, Xiang J, Jin Y, Liu R, Yan J, Wang L (2021) Boost precision agriculture with unmanned aerial vehicle remote sensing and edge intelligence: A survey. Remote Sens 13(21):4387 19. Rhee DS, Kim YD, Kang B, Kim D (2018) Applications of unmanned aerial vehicles in fluvial remote sensing: an overview of recent achievements. KSCE J Civil Eng 22:588–602 20. Argüello F, Heras DB, Garea AS, Quesada-Barriuso P (2021) Watershed monitoring in galicia from uav multispectral imagery using advanced texture methods. Remote Sens 13(14):2687 21. Gxokwe S, Dube T, Mazvimavi D (2020) Multispectral remote sensing of wetlands in semi-arid and arid areas: a review on applications, challenges and possible future research directions. Remote Sens 12(24):4190 22. Ma N, Peng Y, Wang S, Leong PH (2018) An unsupervised deep hyperspectral anomaly detector. Sensors 18(3):693 23. Truax DD (2004) Comparing spectral and object based approaches for classification and transportation feature extraction from high resolution multispectral imagery 24. Leon-Lopez KM, Mouret F, Arguello H, Tourneret J-Y (2021) Anomaly detection and classification in multispectral time series based on hidden markov models. IEEE Trans Geosci Remote Sens 60:1–11 25. Mouret F, Albughdadi M, Duthoit S, Kouamé D, Rieu G, Tourneret J-Y (2021) Outlier detection at the parcel-level in wheat and rapeseed crops using multispectral and sar time series. Remote Sens 13(5):956 26. Coca M, Datcu M (2021) Anomaly detection in post fire assessment. In: 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, pp 8620–8623 IEEE 27. Reed IS, Yu X (1990) Adaptive multiple-band cfar detection of an optical pattern with unknown spectral distribution. IEEE Trans Acoust Speech Signal Process 38(10):1760–1770 28. Zhou J, Kwan C, Ayhan B, Eismann MT (2016) A novel cluster kernel rx algorithm for anomaly and change detection using hyperspectral images. IEEE Trans Geosci Remote Sens 54(11):6497–6504 29. Imani M (2017) Rx anomaly detector with rectified background. IEEE Geosci Remote Sens Lett 14(8):1313–1317 30. Yang X, Huang X, Zhu M, Xu S, Liu Y (2022) Ensemble and random rx with multiple features anomaly detector for hyperspectral image. IEEE Geosci Remote Sens Lett 19:1–5 1 3 Using heterogeneous computing andedge computing toaccelerate… 31. Molero JM, Garzón EM, García I, Plaza A (2012) Anomaly detection based on a parallel kernel rx algorithm for multicore platforms. J Appl Remote Sens 6(1):061503–061503 32. Imani M (2018) Anomaly detection using morphology-based collaborative representation in hyperspectral imagery. Eur J Remote Sens 51(1):457–471 33. Dalla Mura M, Villa A, Benediktsson JA, Chanussot J, Bruzzone L (2010) Classification of hyperspectral images by using extended morphological attribute profiles and independent component analysis. IEEE Geosci Remote Sens Lett 8(3):542–546 34. Liu C, Tao R, Li W, Zhang M, Sun W, Du Q (2020) Joint classification of hyperspectral and multispectral images for mapping coastal wetlands. IEEE J Select Top Appl Earth Observ Remote Sens 14:982–996 35. Gao Y, Li W, Zhang M, Wang J, Sun W, Tao R, Du Q (2021) Hyperspectral and multispectral classification for coastal wetland using depthwise feature interaction network. IEEE Trans Geosci Remote Sens 60:1–15 36. Ghamisi P, Souza R, Benediktsson JA, Zhu XX, Rittner L, Lotufo RA (2016) Extinction profiles for the classification of remote sensing data. IEEE Trans Geosci Remote Sens 54(10):5631–5645 37. Dalla Mura M, Benediktsson JA, Waske B, Bruzzone L (2010) Morphological attribute profiles for the analysis of very high resolution images. IEEE Trans Geosci Remote Sens 48(10):3747–3762 38. Bascoy PG, Quesada-Barriuso P, Heras DB, Argüello F, Demir B, Bruzzone L (2019) Extended attribute profiles on gpu applied to hyperspectral image classification. J Supercomput 75:1565–1579 39. Molero JM, Garzón EM, García I, Quintana-Ortí ES, Plaza A (2014) Efficient implementation of hyperspectral anomaly detection techniques on gpus and multicore processors. IEEE J Select Top Appl Earth Observ Remote Sens 7(6):2256–2266 40. Wu Y, Gao L, Zhang B, Yang B, Chen Z (2015) Embedded gpu implementation of anomaly detection for hyperspectral images. In: High-Performance Computing in Remote Sensing 9646: 66–71 SPIE 41. Du Q, Tang B, Xie W, Li W (2021) Parallel and distributed computing for anomaly detection from hyperspectral remote sensing imagery. Proc IEEE 109(8):1306–1319 42. Caba J, Díaz M, Barba J, Guerra R, Escolar S, López S (2022) Low-power hyperspectral anomaly detector implementation in cost-optimized fpga devices. IEEE J Select Top Appl Earth Observ Remote Sens 15:2379–2393 43. Coca M, Datcu M (2023) Fpga accelerator for meta-recognition anomaly detection: Case of burned area detection. IEEE J Select Top Appl Earth Observ Remote Sens 44. Díaz M, Guerra R, Horstrand P, López S, Sarmiento R (2019) A line-by-line fast anomaly detector for hyperspectral imagery. IEEE Trans Geosci Remote Sens 57(11):8968–8982 45. Tarabalka Y, Haavardsholm TV, Kåsen I, Skauli T (2009) Real-time anomaly detection in hyperspectral images using multivariate normal mixture models and gpu processing. J Real-Time Image Proc 4:287–300 46. Otsu N (1979) A threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybernet 9(1):62–66 47. Chang C-I, Lin C-Y, Chung P-C, Hu PF (2023) Iterative spectral-spatial hyperspectral anomaly detection. IEEE Trans Geosci Remote Sens 61:1–30 48. CESGA: FinisTerrae-II supercomputer. Accessed: 03 Jan 2022. https:// www. cesga. es/ en/ infra struc tures/ compu ting/ 49. NVIDIA CORPORATION ‘I &’ AFFILIATES: Jetson Orin Nano Series, Jetson Orin NX Series and Jetson AGX Orin Series: Supported Modes and Power Efficiency. https:// docs. nvidia. com/ jetson/ archi ves/ r35.3. 1/ Devel operG uide/ text/ SD/ Platf ormPo werAn dPer f orman ce/ Jetso nOrin NanoS eries Jetso nOrin NxSer iesAn dJets onAgx OrinS eries. html Accessed 2023-09-15 50. Han C, Rundo L, Murao K, Noguchi T, Shimahara Y, Milacski ZÁ, Koshino S, Sala E, Nakayama H, Satoh S (2021) Madgan: unsupervised medical anomaly detection gan using multiple adjacent brain mri slice reconstruction. BMC Bioinformat 22(2):1–20 51. Han Y, Li W, Liu M, Wu Z, Zhang F, Liu X, Tao L, Li X, Guo X (2021) Application of an anomaly detection model to screen for ocular diseases using color retinal fundus images: design and evaluation study. J Med Int Res 23(7):27822 52. Huang C, Xu Q, Wang Y, Wang Y, Zhang Y (2022) Self-supervised masking for unsupervised anomaly detection and localization. IEEE Trans Multimed Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.