Efficient multi-scale Gaussian process regression for massive remote sensing data with satGP v0.1.2

Abstract. Satellite remote sensing provides a global view to processes on Earth that has unique benefits compared to making measurements on the ground, such as global coverage and enormous data volume. The typical downsides are spatial and temporal gaps and potentially low data quality. Meaningful statistical inference from such data requires overcoming these problems and developing efficient and robust computational tools. We design and implement a computationally efficient multi-scale Gaussian process (GP) software package, satGP, geared towards remote sensing applications. The software is able to handle problems of enormous sizes and to compute marginals and sample from the random field conditioning on at least hundreds of millions of observations. This is achieved by optimizing the computation by, e.g., randomization and splitting the problem into parallel local subproblems which aggressively discard uninformative data. We describe the mean function of the Gaussian process by approximating marginals of a Markov random field (MRF). Variability around the mean is modeled with a multi-scale covariance kernel, which consists of Matérn, exponential, and periodic components. We also demonstrate how winds can be used to inform covariances locally. The covariance kernel parameters are learned by calculating an approximate marginal maximum likelihood estimate, and the validity of both the multi-scale approach and the method used to learn the kernel parameters is verified in synthetic experiments. We apply these techniques to a moderate size ozone data set produced by an atmospheric chemistry model and to the very large number of observations retrieved from the Orbiting Carbon Observatory 2 (OCO-2) satellite. The satGP software is released under an open-source license.

Download Full-text

Efficient multi-scale Gaussian process regression for massive remote sensing data with satGP v0.1

10.5194/gmd-2019-156 ◽

2019 ◽

Author(s):

Jouni Susiluoto ◽

Alessio Spantini ◽

Heikki Haario ◽

Youssef Marzouk

Keyword(s):

Remote Sensing ◽

Gaussian Process ◽

Gaussian Process Regression ◽

Remote Sensing Data ◽

Marginal Maximum Likelihood ◽

Computationally Efficient ◽

Covariance Kernel ◽

Multi Scale ◽

Covariance Functions ◽

Sensing Applications

Abstract. Satellite remote sensing provides a global view to processes on Earth that has unique benefits compared to measurements made on the ground. The global coverage and the enormous amounts of data produced come, however, with the price of spatial and temporal gaps and less than perfect data quality. Meaningful statistical inference from such data requires overcoming these problems and that calls for developing efficient computational tools. We design and implement a computationally efficient multi-scale Gaussian process (GP) software package, satGP, geared towards remote sensing applications. The software is designed to be able to handle problems of enormous sizes and is able to compute marginals and sample from a random process with at least over hundred million observations. The mean function of the Gaussian process is described by approximating marginals of a Markov random field (MRF). For covariance functions, Matern, exponential, and periodic kernels are utilized in a multi-scale kernel setting to describe the spatial heterogeneity present in data. We further demonstrate how winds can be used to inform the covariance kernel formulation. The covariance kernel parameters are learned by calculating an approximate marginal maximum likelihood estimate and this is utilized to verify the validity of the multi-scale approach in synthetic experiments. For demonstrating the techniques above, data from the Orbiting Carbon Observatory 2 (OCO-2) satellite is used. The satGP program is released as open source software.

Download Full-text

MULTI-SOURCE MULTI-SCALE HIERARCHICAL CONDITIONAL RANDOM FIELD MODEL FOR REMOTE SENSING IMAGE CLASSIFICATION

ISPRS Annals of Photogrammetry Remote Sensing and Spatial Information Sciences ◽

10.5194/isprsannals-ii-3-w4-293-2015 ◽

2015 ◽

Vol II-3/W4 ◽

pp. 293-300 ◽

Cited By ~ 2

Author(s):

Z. Zhang ◽

M. Y. Yang ◽

M. Zhou

Keyword(s):

Remote Sensing ◽

Random Field ◽

Image Classification ◽

Conditional Random Field ◽

Remote Sensing Images ◽

Cloud Data ◽

Multi Scale ◽

Sensing Applications ◽

Remote Sensing Image Classification ◽

Conditional Random Field Model

Fusion of remote sensing images and LiDAR data provides complimentary information for the remote sensing applications, such as object classification and recognition. In this paper, we propose a novel multi-source multi-scale hierarchical conditional random field (MSMSH-CRF) model to integrate features extracted from remote sensing images and LiDAR point cloud data for image classification. MSMSH-CRF model is then constructed to exploit the features, category compatibility of multi-scale images and the category consistency of multi-source data based on the regions. The output of the model represents the optimal results of the image classification. We have evaluated the precision and robustness of the proposed method on airborne data, which shows that the proposed method outperforms standard CRF method.

Download Full-text

MSST-Net: A Multi-Scale Adaptive Network for Building Extraction from Remote Sensing Images Based on Swin Transformer

Remote Sensing ◽

10.3390/rs13234743 ◽

2021 ◽

Vol 13 (23) ◽

pp. 4743

Author(s):

Wei Yuan ◽

Wenbo Xu

Keyword(s):

Neural Network ◽

Remote Sensing ◽

Convolutional Neural Network ◽

Network Model ◽

Remote Sensing Images ◽

Feature Maps ◽

Global Features ◽

Adaptive Network ◽

Data Set ◽

Multi Scale

The segmentation of remote sensing images by deep learning technology is the main method for remote sensing image interpretation. However, the segmentation model based on a convolutional neural network cannot capture the global features very well. A transformer, whose self-attention mechanism can supply each pixel with a global feature, makes up for the deficiency of the convolutional neural network. Therefore, a multi-scale adaptive segmentation network model (MSST-Net) based on a Swin Transformer is proposed in this paper. Firstly, a Swin Transformer is used as the backbone to encode the input image. Then, the feature maps of different levels are decoded separately. Thirdly, the convolution is used for fusion, so that the network can automatically learn the weight of the decoding results of each level. Finally, we adjust the channels to obtain the final prediction map by using the convolution with a kernel of 1 × 1. By comparing this with other segmentation network models on a WHU building data set, the evaluation metrics, mIoU, F1-score and accuracy are all improved. The network model proposed in this paper is a multi-scale adaptive network model that pays more attention to the global features for remote sensing segmentation.

Download Full-text

Remote Sensing in Urban Forestry: Recent Applications and Future Directions

Remote Sensing ◽

10.3390/rs11101144 ◽

2019 ◽

Vol 11 (10) ◽

pp. 1144 ◽

Cited By ~ 7

Author(s):

Xun Li ◽

Wendy Y. Chen ◽

Giovanni Sanesi ◽

Raffaele Lafortezza

Keyword(s):

Remote Sensing ◽

Data Processing ◽

Urban Forest ◽

Scale Effects ◽

Urban Forestry ◽

Urban Forests ◽

Accurate Information ◽

Multi Scale ◽

Sensing Applications ◽

Multi Temporal

Increasing recognition of the importance of urban forest ecosystem services calls for the sustainable management of urban forests, which requires timely and accurate information on the status, trends and interactions between socioeconomic and ecological processes pertaining to urban forests. In this regard, remote sensing, especially with its recent advances in sensors and data processing methods, has emerged as a premier and useful observational and analytical tool. This study summarises recent remote sensing applications in urban forestry from the perspective of three distinctive themes: multi-source, multi-temporal and multi-scale inputs. It reviews how different sources of remotely sensed data offer a fast, replicable and scalable way to quantify urban forest dynamics at varying spatiotemporal scales on a case-by-case basis. Combined optical imagery and LiDAR data results as the most promising among multi-source inputs; in addition, future efforts should focus on enhancing data processing efficiency. For long-term multi-temporal inputs, in the event satellite imagery is the only available data source, future work should improve haze-/cloud-removal techniques for enhancing image quality. Current attention given to multi-scale inputs remains limited; hence, future studies should be more aware of scale effects and cautiously draw conclusions.

Download Full-text

Deriving Particulate Organic Carbon in Coastal Waters from Remote Sensing: Inter-Comparison Exercise and Development of a Maximum Band-Ratio Approach

Remote Sensing ◽

10.3390/rs11232849 ◽

2019 ◽

Vol 11 (23) ◽

pp. 2849

Author(s):

Trung Kien Tran ◽

Lucile Duforêt-Gaurier ◽

Vincent Vantrepotte ◽

Daniel Schaffer Ferreira Jorge ◽

Xavier Mériaux ◽

...

Keyword(s):

Remote Sensing ◽

Coastal Waters ◽

Model Performance ◽

Chemical Compositions ◽

Band Ratio ◽

Data Set ◽

Near Surface ◽

Percentage Difference ◽

The Mean

Recently, different algorithms have been developed to assess near-surface particulate organic matter (POC) concentration over coastal waters. In this study, we gathered an extensive in situ dataset representing various contrasted bio-optical coastal environments at low, medium, and high latitudes, with various bulk particulate matter chemical compositions (mineral-dominated, 50% of the data set, mixed, 40%, or organic-dominated, 10%). The dataset includes 606 coincident measurements of POC concentration and remote-sensing reflectance, Rrs, with POC concentrations covering three orders of magnitude. Twelve existing algorithms have then been tested on this data set, and a new one was proposed. The results show that the performance of historical algorithms depends on the type of water, with an overall low performance observed for mineral-dominated waters. Furthermore, none of the tested algorithms provided satisfactory results over the whole POC range. A novel approach was thus developed based on a maximum band ratio of Rrs (red/blue, red/yellow or red/green ratio). Based on the standard statistical metric for the evaluation of inverse models, the new algorithm presents the best performance. The root-mean square deviation for log-transformed data (RMSDlog) is 0.25. The mean absolute percentage difference (MAPD) is 37.48%. The mean bias (MB) and median ratio (MR) values are 0.54 μg L−1 and 1.02, respectively. This algorithm replicates quite well the distribution of in situ data. The new algorithm was also tested on a matchup dataset gathering 154 coincident MERIS (MEdium Resolution Imaging Spectrometer) Rrs and in situ POC concentration sampled along the French coast. The matchup analysis showed that the performance of the new algorithm is satisfactory (RMSDlog = 0.24, MAPD = 34.16%, MR = 0.92). A regional illustration of the model performance for the Louisiana continental shelf shows that monthly mean POC concentrations derived from MERIS with the new algorithm are consistent with those derived from the 2016 algorithm of Le et al. which was specifically developed for this region.

Download Full-text

Gaussian process regression-based robust free space detection for autonomous vehicle by 3-D point cloud and 2-D appearance information fusion

International Journal of Advanced Robotic Systems ◽

10.1177/1729881417717058 ◽

2017 ◽

Vol 14 (4) ◽

pp. 172988141771705 ◽

Cited By ~ 4

Author(s):

Zhipeng Xiao ◽

Bin Dai ◽

Hongdong Li ◽

Tao Wu ◽

Xin Xu ◽

...

Keyword(s):

Random Field ◽

Gaussian Process ◽

Free Space ◽

Feature Fusion ◽

Conditional Random Field ◽

Autonomous Vehicle ◽

Gaussian Process Regression ◽

Bayesian Framework ◽

Source Information ◽

Data Set

Free space detection is crucial to autonomous vehicles while existing works are not entirely satisfactory. As cameras have many advantages on environment perception, a stereo vision-based robust free space detection method is proposed which mainly depends on geometry information and Gaussian process regression. In this work, in order to improve the performance by exploiting multiple source information, we apply Bayesian framework and conditional random field inference to fuse the multimodal information including 2-D image and 3-D point geometric information. Particularly, a Bayesian framework is used for multiple feature fusion to provide a normalized and flexible output. Gaussian process regression is used to automatically and incrementally regress the data, resulting enhanced performance. Finally, conditional random field with color and geometry constrains is applied to make the result more robust. In order to evaluate the proposed method, quantitative experiments on popular KITTI-road data set and qualitative experiments on our own campus data set are tested. The results show satisfactory and inspiring performance compared to the outstanding works and even are competitive to some relevant Lidar-based methods.

Download Full-text

Fusion Feature Multi-Scale Pooling for Water Body Extraction from Optical Panchromatic Images

Remote Sensing ◽

10.3390/rs11030245 ◽

2019 ◽

Vol 11 (3) ◽

pp. 245 ◽

Cited By ~ 3

Author(s):

Baogui Qi ◽

Yin Zhuang ◽

He Chen ◽

Shan Dong ◽

Lianlin Li

Keyword(s):

Remote Sensing ◽

Water Body ◽

Water Bodies ◽

Optical Remote Sensing ◽

Markov Modeling ◽

Complex Scene ◽

Evaluation Indexes ◽

Multi Scale ◽

Sensing Applications ◽

Fusion Feature

Water body extraction is a hot research topic in remote sensing applications. Using panchromatic optical remote sensing images to extract water bodies is a challenging task, because these images have one level of gray information, variable imaging conditions, and complex scene information. Refined water body extraction from optical panchromatic images often experiences serious under- or over- segmentation problems. In this paper, for producing refined water body extraction results from optical panchromatic images, we propose a fusion feature multi-scale pooling for Markov modeling method. Markov modeling includes two aspects: label field initialization and feature field establishment. These two aspects are jointly created by the fusion feature multi-scale pooling process, and this process is proposed to enhance the feature difference between water bodies and land cover. Then, the greedy algorithm in the iteration conditional method is used to extract refined water bodies according to the rebuilt Markov initial label and feature fields. Finally, to prove the effectiveness of proposed method, extensive experiments were used with collected 2.5m SPOT 5 and 1m GF-2 optical panchromatic images and evaluation indexes (precision, recall, overall accuracy, kappa coefficient and boundary detection ratios) to demonstrate that our proposed method can produce more refined water body extraction results than the state-of-the-art methods. The global and local refined indexes are improved by about 7% and 10%, respectively.

Download Full-text

Accurate Detection of Built-Up Areas from High-Resolution Remote Sensing Imagery Using a Fully Convolutional Network

Photogrammetric Engineering & Remote Sensing ◽

10.14358/pers.85.10.737 ◽

2019 ◽

Vol 85 (10) ◽

pp. 737-752

Author(s):

Yihua Tan ◽

Shengzhou Xiong ◽

Zhi Li ◽

Jinwen Tian ◽

Yansheng Li

Keyword(s):

Remote Sensing ◽

State Of The Art ◽

Automatic Extraction ◽

Feature Maps ◽

Convolutional Network ◽

Data Set ◽

Fully Convolutional Network ◽

Sensing Applications ◽

Wide Range ◽

Remote Sensing Applications

The analysis of built-up areas has always been a popular research topic for remote sensing applications. However, automatic extraction of built-up areas from a wide range of regions remains challenging. In this article, a fully convolutional network (FCN)–based strategy is proposed to address built-up area extraction. The proposed algorithm can be divided into two main steps. First, divide the remote sensing image into blocks and extract their deep features by a lightweight multi-branch convolutional neural network (LMB-CNN). Second, rearrange the deep features into feature maps that are fed into a well-designed FCN for image segmentation. Our FCN is integrated with multi-branch blocks and outputs multi-channel segmentation masks that are utilized to balance the false alarm and missing alarm. Experiments demonstrate that the overall classification accuracy of the proposed algorithm can achieve 98.75% in the test data set and that it has a faster processing compared with the existing state-of-the-art algorithms.

Download Full-text

LOCAL ALGORITHM FOR MONITORING TOTAL SUSPENDED SEDIMENTS IN MICRO-WATERSHEDS USIN DRONES AND REMOTE SENSING APPLICATIONS. CASE STUDY: TEUSACÁ RIVER, LA CALERA, COLOMBIA

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ◽

10.5194/isprsarchives-xl-1-w4-159-2015 ◽

2015 ◽

Vol XL-1/W4 ◽

pp. 159-165 ◽

Cited By ~ 1

Author(s):

N. A. Sáenz ◽

D. E. Paez ◽

C. Arango

Keyword(s):

Remote Sensing ◽

Empirical Relationship ◽

Digital Camera ◽

Large Data ◽

Suspended Sediments ◽

Superficial Layer ◽

Local Algorithm ◽

Data Set ◽

Sensing Applications ◽

Total Suspended Sediments

An empirical relationship of Total Suspended Sediments (TSS) concentrations and reflectance values obtained with Drones’ aerial photos and processed using remote sensing tools was set up as the main objective of this research. A local mathematic algorithm for the micro-watershed of the Teusacá River at La Calera, Colombia, was developed based on the computing of four component of bands from consumed-grade cameras obtaining from each their corresponding reflectance values from procedures for correcting digital camera imagery and using statistical analysis for study the fit and RMSE of 25 regressions. The assessment was characterized by the comparison of reflectance values and 34 in-situ data measurements concentrations between 1.6 and 33 mg L−1 taken from the superficial layer of the river in two campaigns. A large data set of empirical and referenced algorithm from literature were used to evaluate the accuracy and precision of the relationship. For estimation of TSS, a higher accuracy was achieved using the Tassan’s algorithm with the BAND X/ BANDX ratio. The correlation coefficient with R2 = X demonstrate the feasibility of use remote sensed data with consumed-grade cameras as an effective tool for a frequent monitoring and controlling of water quality parameters such as Total Suspended Solids of watersheds, these being the most vulnerable and less compliance with environmental regulations.

Download Full-text