A Generic Bundle Adjustment Methodology for Indirect RPC Model Refinement of Satellite Imagery

Marí, Roger; Franchis, Carlo de; Meinhardt-Llopis, Enric; Anger, Jérémy; Facciolo, Gabriele

doi:10.5201/ipol.2021.352

Cited by 15 publications

(16 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…-A study of the advantages of correcting RPC inconsistencies before training, e.g. by means of a bundle adjustment [31]. We show that eluding this step leads to a drop in the performance of the model.…”

Section: Our Contributions Consist Ofmentioning

confidence: 97%

“…Bundle adjustment approaches are a common good practice in remote sensing to correct inconsistencies between a collection of RPC models observing the same scene [21,31,32,38]. In particular, bundle adjustment methods correct the RPCs by minimizing the reprojection error of a set of reference points seen across the images [50].…”

Section: Rpc Refinement For Improved Performancementioning

confidence: 99%

“…This would cause a systematic loss of accuracy in any NeRF methodology for satellite imagery, because rays traced from corresponding pixels of different views would not intersect at an exact point in the object space. To prevent this situation, before training Sat-NeRF, we apply the bundle adjustment method described in [31], which performs a relative correction of the RPC models of all input images. The reference points used by the bundle adjustment are derived from correspondences of SIFT keypoints [29].…”

Section: Rpc Refinement For Improved Performancementioning

confidence: 99%

“…DS-NeRF [13] discussed how the training of a NeRF can benefit from a sparse set of previously known 3D points, under the idea that such points can be easily produced using SfM pipelines. In the case of satellite imagery, the bundle adjustment produces an equivalent set of sparse 3D points derived from image features [10,31,39,45]. Based on this idea, we explore the benefits of adding the depth-supervision term proposed in [13] to the loss of our Sat-NeRF model:…”

Section: Rpc Refinement For Improved Performancementioning

confidence: 99%

“…We empirically find λ SC = 0.1 and λ DS = 1000 to provide good results, to keep the secondary terms sufficiently relevant but below the magnitude of L RGB . For depth supervision, we used ∼2k-10k bundle adjustment points depending on the output of [31] for each area of interest. R, R SC and R DS have the same batch size.…”

Section: Multi-task Loss and Network Architecturementioning

confidence: 99%

See 4 more Smart Citations

Sat-NeRF: Learning Multi-View Satellite Photogrammetry With Transient Objects and Shadow Modeling Using RPC Cameras

Marí¹,

Facciolo²,

Ehret³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

We introduce the Satellite Neural Radiance Field (Sat-NeRF), a new end-to-end model for learning multi-view satellite photogrammetry in the wild. Sat-NeRF combines some of the latest trends in neural rendering with native satellite camera models, represented by rational polynomial coefficient (RPC) functions. The proposed method renders new views and infers surface models of similar quality to those obtained with traditional state-of-the-art stereo pipelines. Multi-date images exhibit significant changes in appearance, mainly due to varying shadows and transient objects (cars, vegetation). Robustness to these challenges is achieved by a shadow-aware irradiance model and uncertainty weighting to deal with transient phenomena that cannot be explained by the position of the sun. We evaluate Sat-NeRF using WorldView-3 images from different locations and stress the advantages of applying a bundle adjustment to the satellite camera models prior to training. This boosts the network performance and can optionally be used to extract additional cues for depth supervision.

show abstract

Section: Our Contributions Consist Ofmentioning

confidence: 97%

Section: Rpc Refinement For Improved Performancementioning

confidence: 99%

Section: Rpc Refinement For Improved Performancementioning

confidence: 99%

Section: Rpc Refinement For Improved Performancementioning

confidence: 99%

Section: Multi-task Loss and Network Architecturementioning

confidence: 99%

See 3 more Smart Citations

Sat-NeRF: Learning Multi-View Satellite Photogrammetry With Transient Objects and Shadow Modeling Using RPC Cameras

Marí¹,

Facciolo²,

Ehret³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

An Analytical Study about Evaluation the Accuracy of Topographic Maps and Digital Elevation Models from Stereo Satellite Images

Madi,

Ali Al-Baghdadi,

Hamdoon

2024

IOP Conf. Ser.: Earth Environ. Sci.

View full text Add to dashboard Cite

Topographic maps (TMs) play an essential role in spatial data collection. In remote and complex terrains, the importance of the data collected from topographic maps in various applications has led to the search for new techniques and accurate software to enhance mapping accuracy and efficiency. The accuracy of topographic maps and digital elevation models (DEMs) have been evaluated according to available techniques and modern instruments that served the success of this study, such as using the real-time kinematic technique (RTK), which combines with differential ground positioning system (DGPS), using remote sensing technique, by focusing on Leica photogrammetry suit (LPS) software to generate and analysis the results. This study addressed two case studies in the Makhoul Hills to Salah Din Governorate in northeast Iraq because it’s one of the most critical regions characterized by rugged terrain, and it is challenging to reach and get data from field surveying. Many factors highlight the significance of this study. The first factor used Pleiades stereo satellite images with very high resolution (VHR), with spatial resolution reaching (0.5m) for panchromatic images and (2 m) for multispectral images. The second is that overlapping between the adjacent images equals 90 %, in addition to all data available. Taking into account the Archive topographic maps that were published in 1957 by “Royal Dutch Airlines Koninklijke Luchtvaart Maatschappij” (KLM) were used as reference data when needed to compare the elevations of the surface extracted from DEM by using stereo images, the datum of KLM maps was universal transverse Mercator (UTM). the contour lines which were drawn from KLM maps with contour interval (CI) equal to (3m) and map scale 1:25,000 while photo scale 1:30,000, C. I represented the distance between the lines and indicated the accuracy of the maps on a large scale. As much as possible, the reasons behind collecting all data from multiple sources are as follows: To achieve the aim of this study and evaluate the topographic maps from stereo satellite images, (2). extracting high-accuracy DEMs from stereo satellite images, (3). assessing the accuracy of topographic maps and DEMs using ground control points (GCPs) collected in the field by the DGPS (4). the challenge of evaluating the accuracy of high-quality maps for hard-to-reach regions. Finally, A statistical analysis was performed to measure the differences in terrain elevations between rugged and flat areas to evaluate the accuracy of DEMs and TMs. This study concluded the differences were (0.992m) in flat regions and (1.482 m) in rugged areas, according to the computed values for the root mean square error (RMSE). Also, rugged and flat regions have standard deviations (STD) of (1.643 and 1.071) m, respectively. The results were within the allowable RMSE values of (1 m) for plain surfaces and (5 m) for rough terrain. That was extracted from stereo satellite images when running the LPS software. These results proved the work’s reliability and the maps’ accuracy. It is possible to verify the accuracy of the data by increasing the number of checkpoints for the study area so that they are used as GCPS to ensure the results and are not operated inside the block file while using the LPS program.

show abstract

Sat-NeRF: Learning Multi-View Satellite Photogrammetry With Transient Objects and Shadow Modeling Using RPC Cameras

Marí

Facciolo

Ehret

2022

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

A Generic Bundle Adjustment Methodology for Indirect RPC Model Refinement of Satellite Imagery

Cited by 15 publications

References 33 publications

Sat-NeRF: Learning Multi-View Satellite Photogrammetry With Transient Objects and Shadow Modeling Using RPC Cameras

Sat-NeRF: Learning Multi-View Satellite Photogrammetry With Transient Objects and Shadow Modeling Using RPC Cameras

An Analytical Study about Evaluation the Accuracy of Topographic Maps and Digital Elevation Models from Stereo Satellite Images

Sat-NeRF: Learning Multi-View Satellite Photogrammetry With Transient Objects and Shadow Modeling Using RPC Cameras

Contact Info

Product

Resources

About