To a certain extent, foreign objects on road surfaces can cause damage to vehicles and increase the risk of accidents, which is even more immeasurably harmful in such special scenarios as airports. Effective detection...
详细信息
Since its birth, football has become a world-class sport with countless fans. With the continuous development of football, fans are increasingly looking forward to seeing more exciting football games. The article is b...
详细信息
In this paper, an algorithm based on local binary pattern (LBP) is proposed to obtain clear remotesensingimages under the premise of unknown causes of blurring. We find that LBP can completely record the texture fea...
详细信息
In this paper, an algorithm based on local binary pattern (LBP) is proposed to obtain clear remotesensingimages under the premise of unknown causes of blurring. We find that LBP can completely record the texture features of the images, which will not change widely due to the generation of blur. Therefore, LBP prior is proposed, which can filter out the pixels containing important textures in the blurry image through the mapping relationship. The corresponding processing methods are adopted for different types of pixels to cope with the challenges brought by the rich texture and details of remotesensingimages and prevent over-sharpening. However, the existence of LBP prior increases the difficulty of solving the model. To solve the model, we construct the projected alternating minimization (PAM) algorithm that involves the construction of the mapping matrix, the fast iterative shrinkage-thresholding algorithm (FISTA) and the half-quadratic splitting method. Experiments with the AID dataset show that the proposed method can achieve highly competitive processing results for remotesensingimages.
Accuracy and computational time are two crucial parameters influencing the efficacy of classification algorithms for remotesensing applications. Machine learning algorithms are known for achieving notable success for...
详细信息
Accuracy and computational time are two crucial parameters influencing the efficacy of classification algorithms for remotesensing applications. Machine learning algorithms are known for achieving notable success for several classification problems in various domains, including remotesensing. However, they are well-recognized and considered accurate and efficient for closed-set recognition (CSR) but may provide suboptimal and erroneous results for open-set recognition (OSR) tasks. Many practical image-driven and computer vision applications have open-set and dynamic scenarios with unknown data where classification algorithms have not yet achieved significant prediction performance. This paper presents a group of class-aware (CA) classification algorithms based on a supervised cascaded classifier system ((SCS)-S-2), called CA-(SCS)-S-2, which is accurate for OSR and CSR tasks. We evaluate the prediction accuracy of the proposed methods against the state-of-the-art methods in a multiclass setting using multiple image classification scenarios of OSR and CSR. The test case scenarios use six multispectral and hyperspectral datasets from different sensing platforms. And to assess the computational performance of the methods, we designed various field-programmable gate array (FPGA) architectures of the proposed methods. We evaluated their real-time performance on a low-cost, low-power Artix-7 35 T FPGA.
Road damage assessment holds tremendous potential in evaluating damages and reducing disaster risks to human lives during emergency responses to unforeseen events. The Change Detection (CD) method detects changes in t...
详细信息
ISBN:
(纸本)9789819984619;9789819984626
Road damage assessment holds tremendous potential in evaluating damages and reducing disaster risks to human lives during emergency responses to unforeseen events. The Change Detection (CD) method detects changes in the land surface by comparing bi-temporal remotesensingimageries. Using the CD method for post-disaster assessment, existing research mainly focuses on building, while in terms of road, both the dataset and methodology need to be improved. In response to this, we propose an innovative multi-tasking network that combines Vision Transformer and UNet (BiTransUNet) for identifying road change areas and damage assessments from bi-temporal remotesensingimageries before and after natural disasters, moreover, propose the first road damage assessmentmodel. Notably, our BiTransUNet comprises three efficient modules: Multi-scale Feature Extraction (MFE) module for extracting multi-scale features, Trans and Res Skip Connection (TRSC) module for modeling spatial-temporal global information, and Dense Cased Upsample (DCU) module for change maps reconstruction. In addition, to facilitate our study, we create a new remotesensing Road Damage Dataset, RSRDD, thoughtfully designed to contain 1,212 paired imageries before and after disasters, and the corresponding road change masks and road damage levels. Our experimental results on the proposed RSRDD show that our BiTransUNet outperforms current state-of-the-art approaches. BiTransUNet is also applied on the LEVIR-CD building change detection dataset and achieved the best performance, which demonstrates its compatibility in detecting changes of different important ground objects.
This paper unfolds a new approach to detect items of interest in remotesensingimages by enhancing them with deep learning techniques such as You Only Look Once (YOLO) architecture. Therefore, in order to improve the...
详细信息
Semantic segmentation has crucial importance in various domains due to its ability to recognize and categorize objects within an image at a pixel level. This task enables a wide range of applications, such as autonomo...
详细信息
ISBN:
(纸本)9781510673991;9781510673984
Semantic segmentation has crucial importance in various domains due to its ability to recognize and categorize objects within an image at a pixel level. This task enables a wide range of applications, such as autonomous vehicles, environmental monitoring, and remotesensing (RS). In RS, semantic segmentation plays a crucial role, acting as the basis for applications including land cover classification. Following the success of deep learning (DL) methods in computer vision, our paper addresses the intersection between DL and RS imagery. We focus on improving the efficiency of some baseline and backbone models to ensure their adaptability to the challenges posed by RS imagery. Therefore, we evaluate state-of-the-art models on two datasets and investigate their ability to accurately segment objects in RS imagery. Our research aims to open the way for more accurate and reliable semantic segmentation methods in geospatial analysis.
In-sensor broadband convolutional processing (BCP) holds great significance to the advancement of high-precision imagerecognition for remotesensing and environmental monitoring. Two-dimensional (2D) heterostructures...
详细信息
In-sensor broadband convolutional processing (BCP) holds great significance to the advancement of high-precision imagerecognition for remotesensing and environmental monitoring. Two-dimensional (2D) heterostructures offer abundant band alignment configurations with electrical tunability, which is promising to implement the in-sensor BCP at hardware level. Huge efforts have been devoted to developing 2D heterostructures based intelligent edge devices for BCP, which however either lack the potential for scalability and reproducibility, or require high processing temperature. Here, we demonstrate a PtSe2/WSe2 heterostructure based Schottky diode fabricated by using thermal-assisted conversion (TAC) technique, which converts the pre-deposited Pt films into PtSe2 via the controllable selenization process with complementary metal-oxide-semiconductor (CMOS) back-end-of-line (BEOL) compatible thermal budget. The TAC-PtSe2 in various thicknesses demonstrate high crystalline nature and low contact resistance (425 Omega center dot mu m), facilitating an atomically sharp interface and gate-tunable band alignment. Such characteristics give rise to polarity-changeable built-in electric field, resulting in a remarkable rectification ratio approaching similar to 10(5). Moreover, the positive-negative switching of the photoresponse with linear intensity dependence is achieved across a wide spectrum from ultraviolet to near-infrared light, which is desirable in constructing various convolution kernels for BCP tasks. A hyperspectral remotesensingimage is adopted to demonstrate the BCP operations including edge detection and sharpness, where the outputs are of comparable quality with the simulated results. Our work envisions a CMOS-compatible approach for fabricating 2D Schottky diodes with tunable band alignment, offering a viable route for the scalable hardware implementation of broadband imageprocessing units.
Self-supervised contrastive learning (CL) methods can utilize large-scale label-free data to mine discriminative feature representations for vision tasks. However, most existing CL-based approaches focus on image-leve...
详细信息
ISBN:
(纸本)9789819984619;9789819984626
Self-supervised contrastive learning (CL) methods can utilize large-scale label-free data to mine discriminative feature representations for vision tasks. However, most existing CL-based approaches focus on image-level tasks, which are insufficient for pixel-level prediction tasks such as change detection (CD). This paper proposes a multi-scale CL pre-training method for CD tasks in remotesensing (RS) images. Firstly, unlikemost existing methods that rely on random augmentation to enhance model robustness, we collect a publicly available multi-temporal RS dataset and leverage its temporal variations to enhance the robustness of the CD model. Secondly, an unsupervised RS building extraction method is proposed to separate the representation of buildings from background objects, which aims to balance the samples of building areas and background areas in instance-level CL. In addition, we select an equal number of local regions of the building and background for the pixel-level CL task, which prevents the domination caused by local background class. Thirdly, a position-based matching measurement is proposed to construct local positive sample pairs, which aims to prevent the mismatch issues in RS images due to the object similarity in local areas. Finally, the proposed multi-scale CL method is evaluated on benchmark OSCD and SZTAKI databases, and the results demonstrate the effectiveness of our method.
Unsupervised domain adaptation (UDA) is a challenging open problem in land cover mapping. Previous studies show encouraging progress in addressing cross-domain distribution shifts on remotesensing benchmarks for land...
详细信息
ISBN:
(纸本)9798350365474
Unsupervised domain adaptation (UDA) is a challenging open problem in land cover mapping. Previous studies show encouraging progress in addressing cross-domain distribution shifts on remotesensing benchmarks for land cover mapping. The existing works are mainly built on large neural network architectures, which makes them resource-hungry systems, limiting their practical impact for many real-world applications in resource-constrained environments. Thus, we proposed a simple yet effective framework to search for lightweight neural networks automatically for land cover mapping tasks under domain shifts. This is achieved by integrating Markov random field neural architecture search (MRF-NAS) into a self-training UDA framework to search for efficient and effective networks under a limited computation budget. This is the first attempt to combine NAS with self-training UDA as a single framework for land cover mapping. We also investigate two different pseudo-labelling approaches (confidence-based and energy-based) in self-training scheme. Experimental results on two recent datasets (OpenEarthMap & FLAIR #1) for remotesensing UDA demonstrate a satisfactory performance. With only less than 2M parameters and 30.16 G FLOPs, the best-discovered lightweight network reaches state-of-the-art performance on the regional target domain of OpenEarthMap (59.38% mIoU) and the considered target domain of FLAIR #1 (51.19% mIoU). The code is at https://***/cliffbb/UDA-NAS.
暂无评论