首页期刊导航|International journal of remote sensing
期刊信息/Journal information
International journal of remote sensing
Taylor & Francis
International journal of remote sensing

Taylor & Francis

0143-1161

International journal of remote sensing/Journal International journal of remote sensingSCIISTPEIAHCI
正式出版
收录年代

    Landsat spectral unmixing analysis for mapping herbaceous fractional cover in wildfire-prone Mediterranean-type ecosystems

    Krista R. Lee WestDouglas A. StowDaniel J. SousaDar A. Roberts...
    4079-4106页
    查看更多>>摘要：ABSTRACT Portions of Southern California’s native shrubland communities are being replaced by invasive herbaceous vegetation. These non-native species can increase the risk of wildfire ignition and spread. Expansion of these competitive invasive species in recently burned areas following a wildfire can lead to complete conversion and replacement of native shrubs and trees, which in turn increases the likelihood of future wildfire that spreads rapidly and widely through a positive feedback loop: the grass-fire cycle. Despite the association between herbaceous abundance and wildfire risk, image processing approaches for identification and quantification of fractional herbaceous cover in Southern California shrublands are not well established. The objective of this study is to comparatively assess the accuracy of herbaceous cover estimation and mapping based on three different unmixing models applied to Landsat multispectral data for San Diego County, U.S.A. during 2020. The models included: spectral mixture analysis (SMA) using a single set of spectral endmembers; multiple endmember SMA (MESMA); and temporal mixture model (TMM) analysis of year-long stacks of spectral indices computed from multiple Landsat acquisitions. Feature inputs included single date, multi-date, and spectral reflectance and spectral vegetation index (normalized difference infrared index (NDII) and normalized difference vegetation index (NDVI)) combinations. When compared to reference data generated from aerial imagery, results demonstrated that SMA applied to a date during the summer season (August) estimated unburned and intact herbaceous cover most accurately (mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2) values of 8.85%, 12.02%, and 0.85, respectively). Therefore, Landsat unmixing model results suggest that mapping, reconstructing, and monitoring of herbaceous cover at the 10% accuracy level is appropriate. These methods will enable improved detection of sensitive habitats in Mediterranean-type ecosystems around the world by satellite for wildfire-prone communities and identify target areas for monitoring and mitigating the grass-fire cycle.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Spectral unmixing of a multi-decadal Landsat time sequence to reconstruct herbaceous fractional cover dynamics in wildfire-prone Mediterranean-type ecosystems

    Krista R. Lee WestDouglas A. StowDaniel J. SousaJohn F. O’Leary...
    4107-4136页
    查看更多>>摘要：ABSTRACT Portions of Southern California’s native shrubland communities are being invaded and sometimes replaced by herbaceous vegetation that increases the risk of wildfire ignition and spread in a positive feedback loop called the grass-fire cycle. The objective of this study was to assess the extent to which herbaceous cover has expanded and replaced woody vegetation over the last three decades in San Diego County shrublands. To do this, we reconstructed the spatial-temporal distribution of herbaceous growth form cover using spectral mixture analysis (SMA) applied to Landsat multispectral data from 1988 to 2020. The average error in herbaceous cover maps generated from images captured during four single years within the 33-year study period exhibited a mean absolute error (MAE) = 13.30%, root mean square error (RMSE) = 17.62%, and coefficient of determination (R2) = 0.76 relative to reference data derived from orthoimagery. Error estimates for absolute change in herbaceous cover from the earliest (1988) and recent (2020) dates were MAE = 12.17% and RMSE = 15.57% (assessed using 94 reference sampling grids). Between 1988 and 2020, 26.61% of the full study area exhibited an increase in herbaceous cover >20% and 4.98% experienced a decrease in herbaceous cover <−20%, with the greatest concentration of change occurring in wildland-urban interface (WUI) areas. The factors most strongly associated with a substantial increase in herbaceous cover included fire return interval, drought, proximity to development, and elevation. In addition to the overall expansion of herbaceous cover, we also identified locations with evidence of vegetation-type conversion from woody- to herbaceous-dominated fractional cover. These results suggest that a grass-fire cycle has been established in Southern California. The methods from this work can be applied to Mediterranean-type ecosystems around the world to quantify and monitor herbaceous vegetation change over time.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Spatiotemporal UNet for multi-field spatiotemporal series generation

    Lianlei LinSheng GaoZongwei ZhangHanqing Zhao...
    4137-4166页
    查看更多>>摘要：ABSTRACT Building multi-field spatiotemporal virtual environments is of great importance for industrial applications. However, due to the calculation complexity and the lack of consideration for spatiotemporal correlation, existing methods cannot meet the real-time and accuracy requirements. In this paper, we propose a novel Spatiotemporal UNet based on the three-dimensional convolutional neural network for the generation of multi-field spatiotemporal series. The MultiScale Attention Head is proposed to learn the multiscale spatiotemporal information. The Time InfoFusion module is proposed to mine the temporal correlation from spatiotemporal series. Moreover, considering the long-term periodicity and short-term stochasticity of spatiotemporal series, we propose the absolute time encoding strategy and spatiotemporal moving averages mean square error to optimize network learning. Experiments on three different regions of different physical fields show that the proposed method can generate virtual spatiotemporal environments in different tasks, with the RMSE decreasing 0.96–49.59% in 3-day lead time task, 0.16–49.97% in 5-day lead time task, and the MAE decreased up to 54.33% in 3-day lead time task and 55.30% in 5-day lead time task. The inference speed of Spatiotemporal UNet is 4.66 times of ConvLSTM, indicating its accuracy and real-time performance.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Image segmentation and knowledge graph based prototype for ground object interpretation on coastal high resolution remote sensing imagery

    Zilu WangJianyu Chen
    4167-4192页
    查看更多>>摘要：ABSTRACT Recently, the methodologies for analysing remote sensing images have been iteratively refined, evolving from pixel-by-pixel analytical paradigms to Geographic Object-Based Image Analysis. Nonetheless, there remains an area for advancement concerning the transparency of the interpretation process and the rational incorporation of domain knowledge. This study introduces a framework, which hinges on image segmentation and knowledge graph, leveraging a scoring mechanism and a thresholding approach to recognize target objects. It computes the scores for all segmented patches, considering attribute weights, and classifies them as target or non-target objects based on their score thresholds. To validate the proposed framework, this paper conducted an experimental evaluation using 10 GF-2 remote sensing images from Zhejiang Province. These images were partitioned into sample and test sets. The sample image segmentation outcomes were utilized to derive attribute weights and score thresholds for eight object categories. Some object datasets, such as WHU-RS19, complemented the aforementioned data. These data elements were employed to formulate decision rules for the specified object categories. Subsequently, 10 image slices were extracted from the test images, and the above eight categories within these slices were interpreted according to the established decision rules. The interpretation outcomes revealed that both recall and precision for the eight target object categories exceeded 90%. Additionally, the overall accuracy for the interpretation of all patches was notably high at 0.88, with a Kappa coefficient of 0.99. The outcomes are also compared with the outcomes using eCognition software, which is found to be superior to eCognition’s interpretation results. The analysis of these evaluation results concluded that the proposed method enables high-precision interpretation with minimal dataset requirements, while ensuring transparency in the interpretation process.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Mapping bathymetry on Tibetan Plateau lakes using ICESat-2 laser altimetry

    Hao HeJun ChenHui ShengDrolma Lhakpa...
    4193-4214页
    查看更多>>摘要：ABSTRACT Lake water storage variations on the Tibetan Plateau (TP) serve as crucial indicators of regional hydrological dynamics and climate changes, providing more comprehensive insights than discrete measurements of lake area or water level alone. While accurate bathymetric data is fundamental for quantifying lake water storage, conventional bathymetric surveys are often constrained by logistical challenges and high operational costs in the remote region like the TP. The high altitude and minimal human activity on the TP result in exceptional lake water clarity, allowing laser altimetry to penetrate water depths of several tens of metres. In this study, we used data from Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) laser altimetry data collected from 2019 to 2023 to map five shallow, elongated lakes on the TP. First, we applied the DBSCAN denoising algorithm to eliminate anomalous photons and then fitted polynomial functions to the lakebed elevation profiles for individual tracks. Subsequently, we merged the profiles from all valid tracks within each lake area to derive comprehensive lakebed topography and depth estimates. Comparative analysis with depth measurements from previous studies revealed strong agreement in both absolute depths and spatial patterns of bottom topography. Our results showed that the water depths of the five studied lakes range from 0 to 47 m, with Puma Yumco identified as the deepest (maximum depth of 47 m) and Pelrap Tso as the shallowest (maximum depth of 26 m. The shoreline of Puma Yumco exhibited steeper topography compared to the other four lakes. This study demonstrated the capability of ICESat-2 laser altimetry as a cost-effective and reliable tool for lake bathymetry estimation on the TP. The approach presented in this study holds promise for broader applications in other regions with optically clear water bodies, thereby contributing to improve monitoring of lake dynamics and understanding of regional water storage dynamics and climate change impacts.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    MDSTA: masked diffusion spatio-temporal autoencoder for multimodal remote sensing image classification

    Zongqin YueJindong XuZiyi LiHaihua Xing...
    4215-4240页
    查看更多>>摘要：ABSTRACT Deep learning methods have significantly advanced multimodal remote sensing data classification recently. However, challenges in data acquisition frequently lead to missing modalities, which substantially hinder the performance of these models. Diffusion models have shown tremendous potential in generative tasks, demonstrating a superior ability to model complex data distributions. However, in the remote sensing domain, which incorporates diverse modalities, each modality possesses unique information, and diffusion models may need help to effectively recover critical information when one modality is missing, resulting in unsatisfying performance. To address this, we propose the Masked Diffusion Spatio-Temporal Autoencoder (MDSTA) network for the joint classification of remote sensing data under arbitrary modalities. MDSTA consists of three main components: a Conditional Masked Diffusion Process (CMDP), a Reverse Diffusion Reconstruction Process (RDRP), and an Attention Multi-Layer Perceptron (MLP). The CMDP progressively adds noise to the remote sensing data to prepare for subsequent denoising and reconstruction. The RDRP extracts features from multimodal data through our designed Spatio-Temporal Fusion (STF) Encoder, mapping them into a shared-parameter modality-mixed space to capture multimodal shared features. The Masked Reconstruction (MR) Decoder then utilizes these features for independent reconstruction of each modality, which helps to learn the unique characteristics of each modality. Then, we designed an attention MLP to fuse multimodal classification tokens and obtain the final classification result. Furthermore, we introduce masked training in the conditional masked diffusion block to alleviate memory consumption. Comprehensive experimental findings on four datasets indicate that the proposed MDSTA model surpasses leading models in performance.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Multi-angle analysis of shading-based modelling: extracting the urban building height based on ZY-3 three-line-array camera

    Siqi LuHeli LuYi ChenChuanrong Zhang...
    4241-4273页
    查看更多>>摘要：ABSTRACT Building height is a crucial parameter for urban environment analysis and an essential component in urban microclimate modelling. Its rapid and accurate acquisition is fundamental for applications such as urban environmental management and planning. Shading-based modelling from high-resolution imagery is a common method used to estimate building height. Current new ZY-3 three-line-array camera, capable of simultaneously capturing multi-angle high-resolution images from orthometric and forward/backward cameras, offers enhanced opportunities for improving building height extraction. In this study, we developed and tested the effects of using multi-angle images on building height extraction across various urban residential district scenes. The results showed that shading-based modelling achieves higher accuracy in low height-unevenness district scene and lower accuracy in high height-unevenness district scene. Further analysis revealed that the dual-view imagery model significantly improves accuracy in high height-unevenness districts, while the three-view imagery model did not provide additional benefits. These findings highlight the importance of selecting appropriate multi-angle imagery models tailored to specific district scene to achieve meaningful improvements in building height estimation accuracy.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Ecological quality assessment of large-scale regions using RSEI improved with YTT temperature rectification

    Xingxing LiuRuibo ChenLijuan HeRundong Liu...
    4274-4294页
    查看更多>>摘要：ABSTRACT The remote sensing based ecological index (RSEI), with its inherent advantages and without subjective intervention, has gained widespread usage as an ecological assessment model. However, the traditional RSEI has many restrictions (e.g. for assessing the ecological quality of large-scale regions, the heat component will be invalid). This paper presents Yesterday-Today-Tomorrow (YTT) temperature rectification model that can address the critical stitching problem caused by temporal disparities in image acquisition over large-scale regions (i.e. apply the heat component to large-scale regions for ecological quality assessment). The improved RSEI is named as L-RSEI (RSEI for Large-scale region). Furthermore, L-RSEI adopts an Enhanced Normalized Difference Vegetation Index (ENDVI) that leverages land cover/use information to enhance assessment accuracy. Applying L-RESI for the Guangxi section of the Xijiang River, the statistical results indicate that (1) The YTT model significantly enhances stitching accuracy compared to raw data, showing improvements of 62.85% in 2000, 58.97% in 2005, 66.12% in 2010, 31.96% in 2015, and 64.12% in 2020. These results affirm its superior performance for large areas. (2) The analysis of ENDVI demonstrates significant improvements over traditional NDVI. Specifically, the percentage of the D class (NDVI value in the range of 0.6–0.8) increases from 15.61% to 29.62%, while the E class (NDVI value in the range of 0.8–1.0) decreases from 60.56% to 47.82%. These findings highlight the finer segmentation achieved through ENDVI; (3) The proportion of middle-to-upper classes (a combination of Normal, Good, and Very Good classes) ranges from 91.88% in 2000 to 92.56% in 2020. Temporal analysis using L-RSEI reveals an initial deterioration followed by recovery in the eco-logical quality of the region, due to technological advancements, government interventions, and the shift in human-nature relationships. All of the aforementioned evidence demonstrates that our L-RSEI method is effective for assessing and monitoring ecological quality in large-scale regions.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    An adaptive superpixelwise coordination factor analysis approach for feature extraction of hyperspectral images

    Xianyue WangZifan ShiLongxia Qian
    4295-4318页
    查看更多>>摘要：ABSTRACT Dimensionality reduction of hyperspectral image (HSI) is crucial in improving detector performance. However, employing heterogeneous and homogeneous spatial regions in dimensionality reduction necessitates more comprehensive consideration. Therefore, we propose an adaptive superpixelwise coordination factor analysis (ASCFA) framework for hyperspectral classification. An adaptive denoising strategy is developed using superpixel segmentation based on entropy rates. This approach removes noise from each superpixelwise block by employing an enhanced median filtering technique. It adaptively adjusts the filter window size, effectively preserving edges and fine details while eliminating noise. After the denoising process, a novel unsupervised dimensionality reduction method, grounded in superpixelwise coordination factor analysis, is utilized to estimate the parameters of linear low-dimensional manifolds. These manifolds are then aligned parametric, transforming the denoised HSI into an optimal low-dimensional subspace. The resulting low-dimensional features are not only discriminative and compact but also robust against noise, significantly improving classification performance. To validate the effectiveness of ASCFA, we conducted extensive experiments on three benchmark datasets: Indian Pines, Pavia University, and WHU-Hi-Longkou. ASCFA maintains the highest overall accuracy values on all datasets, even with various added noise. These results underscore the robustness of ASCFA as an effective tool for hyperspectral image analysis, offering improved classification performance and reduced computational complexity.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis

    Research on all-day water vapour profile retrieval method based on lidar data and machine learning algorithm

    Xiao ChengHuige DiQimeng LiNing Chen...
    4319-4344页
    查看更多>>摘要：ABSTRACT Raman lidar can achieve high spatial and temporal resolution retrieval of atmospheric water vapour vertical profiles. However, it is difficult to effectively solve the problem of limited daytime water vapour retrieval distances owing to the influence of the solar background light. To enhance the daytime water vapour retrieval capability of lidar, this paper proposes a technique for retrieving the water vapour vertical profile by integrating lidar and ground meteorological parameters based on the backpropagation neural network algorithm. This study constructed a neural network training model, maximized its retrieval accuracy, and achieve the retrieval of daytime water vapour profiles. Under strong background light conditions at noon in summer, the proposed method increases the maximum retrieval height by 2.5 km compared to traditional lidar retrieval methods. A regression analysis was conducted between the neural network retrieval method proposed in this study and traditional lidar retrieval methods within the effective daytime detection height. The results demonstrate that the proposed method exhibits high accuracy, achieving a correlation coefficient, a coefficient of determination, and a root mean square error of 0.951, 0.904, and 0.889 g·kg− 1, respectively.
      原文链接:
    • NETL
    • NSTL
    • Taylor & Francis