scholarly journals STATE OF THE ART IN DENSE IMAGE MATCHING COST COMPUTATION FOR HIGH-RESOLUTION SATELLITE STEREO

2021 ◽  
Vol XLIII-B2-2021 ◽  
pp. 109-114
Author(s):  
Y. Wang ◽  
D. Gong ◽  
H. Hu ◽  
S. Wang ◽  
Y. Han ◽  
...  
Keyword(s):  
High Resolution ◽  
Image Matching ◽  
Large Scale ◽  
Satellite Image ◽  
Ground Truth ◽  
Surface Model ◽  
Dense Image ◽  
High Resolution Satellite Images ◽  
Matching Cost ◽  
End To End

Abstract. Large-scale Digital Surface Model (DSM) generated with high-resolution satellite images (HRSI) are comparable, cheaper, and more accessible when comparing to Light Detection and Ranging (LiDAR) data and aerial remotely sensed images. Several photogrammetric commercial/open-source software packages are being developed for satellite image-based 3D reconstruction, in which, most of them adopt a modified version of Semi-Global Matching (SGM) algorithm for dense image matching. With the continuous development of matching cost computation methods, the existing methods can be divided into classical (low-level) and learning-based algorithms (non-end-to-end learning and end-to-end learning methods). On Middlebury and KITTI datasets, learning-based algorithms has shown their superiority compared to SGM derived methods. In this context, we assume that matching cost is the key factor of DIM. This paper reviews and evaluates Census Transform, and MC-CNN on a WorldView-3 typical city scene satellite stereo images on the premise that the overall SGM framework remains unchanged, providing a preliminary comparison for academic and industrial. We first compute the cost valume of these two methods, obtains the final DSM after semi-global optimization, and compares their gemetric accuracy with the corresponding LiDAR derived ground truth. We presented our comparison and findings in the experimental section.

scholarly journals STATE OF THE ART IN DIGITAL SURFACE MODELLING FROM MULTI-VIEW HIGH-RESOLUTION SATELLITE IMAGES

2020 ◽  
Vol V-2-2020 ◽  
pp. 351-356 ◽  
Author(s):  
Y. Han ◽  
S. Wang ◽  
D. Gong ◽  
Y. Wang ◽  
Y. Wang ◽  
...  
Keyword(s):  
High Resolution ◽  
Large Scale ◽  
Satellite Images ◽  
Satellite Image ◽  
Ground Truth ◽  
Semantic Interpretation ◽  
Surface Model ◽  
Three Solutions ◽  
Surface Modelling ◽  
High Resolution Satellite Images

Abstract. Data from the optical satellite imaging sensors running 24/7, is collecting in embarrassing abundance nowadays. Besides more suitable for large-scale mapping, multi-view high-resolution satellite images (HRSI) are cheaper when comparing to Light Detection And Ranging (LiDAR) data and aerial remotely sensed images, which are more accessible sources for digital surface modelling and updating. Digital Surface Model (DSM) generation is one of the most critical steps for mapping, 3D modelling, and semantic interpretation. Computing DSM from this dataset is relatively new, and several solutions exist in the market, both commercial and open-source solutions, the performances of these solutions have not yet been comprehensively analyzed. Although some works and challenges have focused on the DSM generation pipeline and the geometric accuracy of the generated DSM, the evaluations, however, do not consider the latest solutions as the fast development in this domain. In this work, we discussed the pipeline of the considered both commercial and opensource solutions, assessed the accuracy of the multi-view satellite image-based DSMs generation methods with LiDAR-derived DSM as the ground truth. Three solutions, including Satellite Stereo Pipeline (S2P), PCI Geomatica, and Agisoft Metashape, are evaluated on a WorldView-3 multi-view satellite dataset both quantitatively and qualitatively with the LiDAR ground truth. Our comparison and findings are presented in the experimental section.

scholarly journals End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++

2019 ◽  
Vol 11 (11) ◽  
pp. 1382 ◽  
Author(s):  
Daifeng Peng ◽  
Yongjun Zhang ◽  
Haiyan Guan
Keyword(s):  
Deep Learning ◽  
High Resolution ◽  
Change Detection ◽  
Land Surface ◽  
Satellite Image ◽  
Feature Representation ◽  
Observation Data ◽  
Feature Maps ◽  
High Resolution Satellite Images ◽  
End To End

Change detection (CD) is essential to the accurate understanding of land surface changes using available Earth observation data. Due to the great advantages in deep feature representation and nonlinear problem modeling, deep learning is becoming increasingly popular to solve CD tasks in remote-sensing community. However, most existing deep learning-based CD methods are implemented by either generating difference images using deep features or learning change relations between pixel patches, which leads to error accumulation problems since many intermediate processing steps are needed to obtain final change maps. To address the above-mentioned issues, a novel end-to-end CD method is proposed based on an effective encoder-decoder architecture for semantic segmentation named UNet++, where change maps could be learned from scratch using available annotated datasets. Firstly, co-registered image pairs are concatenated as an input for the improved UNet++ network, where both global and fine-grained information can be utilized to generate feature maps with high spatial accuracy. Then, the fusion strategy of multiple side outputs is adopted to combine change maps from different semantic levels, thereby generating a final change map with high accuracy. The effectiveness and reliability of our proposed CD method are verified on very-high-resolution (VHR) satellite image datasets. Extensive experimental results have shown that our proposed approach outperforms the other state-of-the-art CD methods.

scholarly journals Automatic Building Footprint Extraction from Multi-Resolution Remote Sensing Images Using a Hybrid FCN

2019 ◽  
Vol 8 (4) ◽  
pp. 191 ◽  
Author(s):  
Philipp Schuegraf ◽  
Ksenia Bittner
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
High Resolution ◽  
Large Scale ◽  
Early Stage ◽  
Satellite Image ◽  
Remote Sensing Data ◽  
Training Data ◽  
Spectral Information ◽  
Surface Model

Recent technical developments made it possible to supply large-scale satellite image coverage. This poses the challenge of efficient discovery of imagery. One very important task in applications like urban planning and reconstruction is to automatically extract building footprints. The integration of different information, which is presently achievable due to the availability of high-resolution remote sensing data sources, makes it possible to improve the quality of the extracted building outlines. Recently, deep neural networks were extended from image-level to pixel-level labelling, allowing to densely predict semantic labels. Based on these advances, we propose an end-to-end U-shaped neural network, which efficiently merges depth and spectral information within two parallel networks combined at the late stage for binary building mask generation. Moreover, as satellites usually provide high-resolution panchromatic images, but only low-resolution multi-spectral images, we tackle this issue by using a residual neural network block. It fuses those images with different spatial resolution at the early stage, before passing the fused information to the Unet stream, responsible for processing spectral information. In a parallel stream, a stereo digital surface model (DSM) is also processed by the Unet. Additionally, we demonstrate that our method generalizes for use in cities which are not included in the training data.

scholarly journals RELATIVE ORIENTATION AND MODIFIED PIECEWISE EPIPOLAR RESAMPLING FOR HIGH RESOLUTION SATELLITE IMAGES

2017 ◽  
Vol XLII-1/W1 ◽  
pp. 579-586 ◽  
Author(s):  
K. Gong ◽  
D. Fritsch
Keyword(s):  
High Resolution ◽  
Satellite Images ◽  
Satellite Image ◽  
Relative Orientation ◽  
Ground Control ◽  
Surface Model ◽  
Control Information ◽  
Rational Polynomial ◽  
High Resolution Satellite Images ◽  
Resampling Method

High resolution, optical satellite sensors are boosted to a new era in the last few years, because satellite stereo images at half meter or even 30cm resolution are available. Nowadays, high resolution satellite image data have been commonly used for Digital Surface Model (DSM) generation and 3D reconstruction. It is common that the Rational Polynomial Coefficients (RPCs) provided by the vendors have rough precision and there is no ground control information available to refine the RPCs. Therefore, we present two relative orientation methods by using corresponding image points only: the first method will use quasi ground control information, which is generated from the corresponding points and rough RPCs, for the bias-compensation model; the second method will estimate the relative pointing errors on the matching image and remove this error by an affine model. Both methods do not need ground control information and are applied for the entire image. To get very dense point clouds, the Semi-Global Matching (SGM) method is an efficient tool. However, before accomplishing the matching process the epipolar constraints are required. In most conditions, satellite images have very large dimensions, contrary to the epipolar geometry generation and image resampling, which is usually carried out in small tiles. This paper also presents a modified piecewise epipolar resampling method for the entire image without tiling. The quality of the proposed relative orientation and epipolar resampling method are evaluated, and finally sub-pixel accuracy has been achieved in our work.

scholarly journals Continuous high-resolution midlatitude-belt simulations for July–August 2013 with WRF

2017 ◽  
Vol 10 (5) ◽  
pp. 2031-2055 ◽  
Author(s):  
Thomas Schwitalla ◽  
Hans-Stefan Bauer ◽  
Volker Wulfmeyer ◽  
Kirsten Warrach-Sagi
Keyword(s):  
High Resolution ◽  
Land Surface ◽  
Large Scale ◽  
Temperature Fields ◽  
Surface Model ◽  
Limited Area ◽  
Data Set ◽  
Latitude Belt ◽  
Simulated Precipitation ◽  
The Impact

Abstract. Increasing computational resources and the demands of impact modelers, stake holders, and society envision seasonal and climate simulations with the convection-permitting resolution. So far such a resolution is only achieved with a limited-area model whose results are impacted by zonal and meridional boundaries. Here, we present the setup of a latitude-belt domain that reduces disturbances originating from the western and eastern boundaries and therefore allows for studying the impact of model resolution and physical parameterization. The Weather Research and Forecasting (WRF) model coupled to the NOAH land–surface model was operated during July and August 2013 at two different horizontal resolutions, namely 0.03 (HIRES) and 0.12° (LOWRES). Both simulations were forced by the European Centre for Medium-Range Weather Forecasts (ECMWF) operational analysis data at the northern and southern domain boundaries, and the high-resolution Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) data at the sea surface.The simulations are compared to the operational ECMWF analysis for the representation of large-scale features. To analyze the simulated precipitation, the operational ECMWF forecast, the CPC MORPHing (CMORPH), and the ENSEMBLES gridded observation precipitation data set (E-OBS) were used as references.Analyzing pressure, geopotential height, wind, and temperature fields as well as precipitation revealed (1) a benefit from the higher resolution concerning the reduction of monthly biases, root mean square error, and an improved Pearson skill score, and (2) deficiencies in the physical parameterizations leading to notable biases in distinct regions like the polar Atlantic for the LOWRES simulation, the North Pacific, and Inner Mongolia for both resolutions.In summary, the application of a latitude belt on a convection-permitting resolution shows promising results that are beneficial for future seasonal forecasting.

scholarly journals THE POSSIBILITY OF USING IMAGES OBTAINED FROM THE UAS IN CADASTRAL WORKS

2016 ◽  
Vol XLI-B1 ◽  
pp. 909-915 ◽  
Author(s):  
Z. Kurczynski ◽  
K. Bakuła ◽  
M. Karabin ◽  
M. Kowalczyk ◽  
J. S. Markiewicz ◽  
...  
Keyword(s):  
Image Matching ◽  
Bundle Adjustment ◽  
Aerial Photographs ◽  
Surface Model ◽  
Digital Surface Model ◽  
Reference Field ◽  
Alternative Source ◽  
Dense Image ◽  
Low Altitude ◽  
Very High

Updating the cadastre requires much work carried out by surveying companies in countries that have still not solved the problem of updating the cadastral data. In terms of the required precision, these works are among the most accurate. This raises the question: to what extent may modern digital photogrammetric methods be useful in this process? The capabilities of photogrammetry have increased significantly after the introduction of digital aerial cameras and digital technologies. For the registration of cadastral objects, i.e., land parcels’ boundaries and the outlines of buildings, very high-resolution aerial photographs can be used. The paper relates an attempt to use an alternative source of data for this task - the development of images acquired from UAS platforms. Multivariate mapping of cadastral parcels was implemented to determine the scope of the suitability of low altitude photos for the cadastre. In this study, images obtained from UAS with the GSD of 3 cm were collected for an area of a few square kilometres. Bundle adjustment of these data was processed with sub-pixel accuracy. This led to photogrammetric measurements being carried out and the provision of an orthophotomap (orthogonalized with a digital surface model from dense image matching of UAS images). Geometric data related to buildings were collected with two methods: stereoscopic and multi-photo measurements. Data related to parcels’ boundaries were measured with monoplotting on an orthophotomap from low-altitude images. As reference field surveying data were used. The paper shows the potential and limits of the use of UAS in a process of updating cadastral data. It also gives recommendations when performing photogrammetric missions and presents the possible accuracy of this type of work.

scholarly journals UNSUPERVISED MULTI-CONSTRAINT DEEP NEURAL NETWORK FOR DENSE IMAGE MATCHING

2020 ◽  
Vol XLIII-B2-2020 ◽  
pp. 163-167
Author(s):  
W. Yuan ◽  
Z. Fan ◽  
X. Yuan ◽  
J. Gong ◽  
R. Shibasaki
Keyword(s):  
Neural Network ◽  
Object Detection ◽  
Image Matching ◽  
Deep Neural Network ◽  
Aerial Images ◽  
Surface Model ◽  
Learning Technology ◽  
Ground Sample ◽  
Volume Calculation ◽  
Dense Image

Abstract. Dense image matching is essential to photogrammetry applications, including Digital Surface Model (DSM) generation, three dimensional (3D) reconstruction, and object detection and recognition. The development of an efficient and robust method for dense image matching has been one of the technical challenges due to high variations in illumination and ground features of aerial images of large areas. Nowadays, due to the development of deep learning technology, deep neural network-based algorithms outperform traditional methods on a variety of tasks such as object detection, semantic segmentation and stereo matching. The proposed network includes cost-volume computation, cost-volume aggregation, and disparity prediction. It starts with a pre-trained VGG-16 network as a backend and using the U-net architecture with nine layers for feature map extraction and a correlation layer for cost volume calculation, after that a guided filter based cost aggregation is adopted for cost volume filtering and finally the soft Argmax function is utilized for disparity prediction. The experimental conducted on a UAV dataset demonstrated that the proposed method achieved the RMSE (root mean square error) of the reprojection error better than 1 pixel in image coordinate and in-ground positioning accuracy within 2.5 ground sample distance. The comparison experiments on KITTI 2015 dataset shows the proposed unsupervised method even comparably with other supervised methods.

scholarly journals AN INTEGRATED PHOTOGRAMMETRIC AND PHOTOCLINOMETRIC APPROACH FOR PIXEL-RESOLUTION 3D MODELLING OF LUNAR SURFACE

2018 ◽  
Vol XLII-3 ◽  
pp. 1117-1122
Author(s):  
W. C. Liu ◽  
B. Wu
Keyword(s):  
High Resolution ◽  
Lunar Surface ◽  
Image Matching ◽  
Point Cloud ◽  
3D Modelling ◽  
Conjugate Points ◽  
Stereo Pair ◽  
Pixel Resolution ◽  
Dense Image ◽  
Lunar Reconnaissance Orbiter

High-resolution 3D modelling of lunar surface is important for lunar scientific research and exploration missions. Photogrammetry is known for 3D mapping and modelling from a pair of stereo images based on dense image matching. However dense matching may fail in poorly textured areas and in situations when the image pair has large illumination differences. As a result, the actual achievable spatial resolution of the 3D model from photogrammetry is limited by the performance of dense image matching. On the other hand, photoclinometry (i.e., shape from shading) is characterised by its ability to recover pixel-wise surface shapes based on image intensity and imaging conditions such as illumination and viewing directions. More robust shape reconstruction through photoclinometry can be achieved by incorporating images acquired under different illumination conditions (i.e., photometric stereo). Introducing photoclinometry into photogrammetric processing can therefore effectively increase the achievable resolution of the mapping result while maintaining its overall accuracy. This research presents an integrated photogrammetric and photoclinometric approach for pixel-resolution 3D modelling of the lunar surface. First, photoclinometry is interacted with stereo image matching to create robust and spatially well distributed dense conjugate points. Then, based on the 3D point cloud derived from photogrammetric processing of the dense conjugate points, photoclinometry is further introduced to derive the 3D positions of the unmatched points and to refine the final point cloud. The approach is able to produce one 3D point for each image pixel within the overlapping area of the stereo pair so that to obtain pixel-resolution 3D models. Experiments using the Lunar Reconnaissance Orbiter Camera - Narrow Angle Camera (LROC NAC) images show the superior performances of the approach compared with traditional photogrammetric technique. The results and findings from this research contribute to optimal exploitation of image information for high-resolution 3D modelling of the lunar surface, which is of significance for the advancement of lunar and planetary mapping.

scholarly journals TEXTURE-AWARE DENSE IMAGE MATCHING USING TERNARY CENSUS TRANSFORM

2016 ◽  
Vol III-3 ◽  
pp. 59-66 ◽  
Author(s):  
Han Hu ◽  
Chongtai Chen ◽  
Bo Wu ◽  
Xiaoxia Yang ◽  
Qing Zhu ◽  
...  
Keyword(s):  
Image Matching ◽  
Urban Areas ◽  
Point Clouds ◽  
Parameter Tuning ◽  
Census Transform ◽  
Urban Scenes ◽  
Dense Image ◽  
Global Matching ◽  
Matching Cost ◽  
Sharp Edges

Textureless and geometric discontinuities are major problems in state-of-the-art dense image matching methods, as they can cause visually significant noise and the loss of sharp features. Binary census transform is one of the best matching cost methods but in textureless areas, where the intensity values are similar, it suffers from small random noises. Global optimization for disparity computation is inherently sensitive to parameter tuning in complex urban scenes, and must compromise between smoothness and discontinuities. The aim of this study is to provide a method to overcome these issues in dense image matching, by extending the industry proven Semi-Global Matching through 1) developing a ternary census transform, which takes three outputs in a single order comparison and encodes the results in two bits rather than one, and also 2) by using texture-information to self-tune the parameters, which both preserves sharp edges and enforces smoothness when necessary. Experimental results using various datasets from different platforms have shown that the visual qualities of the triangulated point clouds in urban areas can be largely improved by these proposed methods.

scholarly journals High Resolution Satellite Images for Instantaneous Shoreline Extraction Using New Enhancement Algorithms

Geosciences ◽  
2019 ◽  
Vol 9 (3) ◽  
pp. 123 ◽  
Author(s):  
Donatella Dominici ◽  
Sara Zollini ◽  
Maria Alicandro ◽  
Francesca Della Torre ◽  
Paolo Buscema ◽  
...  
Keyword(s):  
High Resolution ◽  
Large Scale ◽  
Satellite Images ◽  
Human Life ◽  
Essential Element ◽  
Action Planning ◽  
High Resolution Satellite Images ◽  
High Resolution Images ◽  
Territorial Government ◽  
Very High

Knowledge of a territory is an essential element in any future planning action and in appropriate territorial and environmental requalification action planning. The current large-scale availability of satellite data, thanks to very high resolution images, provides professional users in the environmental, urban planning, engineering, and territorial government sectors, in general, with large amounts of useful data with which to monitor the territory and cultural heritage. Italy is experiencing environmental emergencies, and coastal erosion is one of the greatest threats, not only to the Italian heritage and economy, but also to human life. The aim of this paper is to find a rapid way of identifying the instantaneous shoreline. This possibility could help government institutions such as regions, civil protection, etc., to analyze large areas of land quickly. The focus is on instantaneous shoreline extraction in Ortona (CH, Italy), without considering tides, using WorldView-2 satellite images (50-cm resolution in panchromatic and 2 m in multispectral). In particular, the main purpose of this paper is to compare commercial software and ACM filters to test their effectiveness.

Sign in / Sign up

Export Citation Format

Share Document