Sequential Cycle Consistency Inference for Eliminating Incorrect Relative Orientations in Structure from Motion

Xiao, Teng; Wang, X.; Deng, Fei; Heipke, Christian

doi:10.1007/s41064-021-00152-1

Cited by 3 publications

(3 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following that, the merging of point clouds was manually combined using paper target marks homogeneously positioned on the building surfaces, as reflective target marks were not employed. Next, photogrammetric processes were performed using the Agisoft PhotoScan software, which uses Structure From Motion (SfM) based techniques [10,[30][31][32] CloudCompare software enables various combining alternatives. This research integrated point data obtained from different sources with manual point selection (Figure 4).…”

Section: Data-processing and Accuracy Assessmentmentioning

confidence: 99%

Accuracy Assessment Toward Merging of Terrestrial Laser Scanner Point Data and Unmanned Aerial System Point Data

Karasaka

Makineci

ERDAL

2023

Konya Journal of Engineering Sciences

View full text Add to dashboard Cite

Terrestrial Laser Scanning (TLS) techniques are widely preferred for 3D models of small and large objects, buildings, and historical and cultural heritages. However, sometimes relying on a single method for 3D modelling an object/structure is insufficient to arrive at a solution or meet expectations. For example, Unmanned Aerial Systems (UAS) provide perspective for building roofs, while terrestrial laser scanners provide general information about building facades. In this research, several facades of a selected building could not be modelled using terrestrial laser scanning, and UAS was used to complete the missing data for 3D modelling. The transformation matrix, a linear function, is created to merge different data types. In the transformation matrix, the scale was found to be 1:1.012. The accuracy analysis of the produced 3D model was also made by comparing the spatial measurements taken from different building facades and the differences in the measurement values obtained from the 3D model and calculating statistically. According to the accuracy analysis results, the Root Mean Square Error (RMSE) value is approximately 3 cm. The results of the accuracy research, which are within the 95% confidence interval with the three-sigma rule, are approximately 2 cm as RMSE. As a result of the study, it was determined that the data obtained from UAV photogrammetry and the data obtained by the TLS technique could be combined, and the integrated 3D model obtained can be used more efficiently.

show abstract

Section: Data-processing and Accuracy Assessmentmentioning

confidence: 99%

Accuracy Assessment Toward Merging of Terrestrial Laser Scanner Point Data and Unmanned Aerial System Point Data

Karasaka

Makineci

ERDAL

2023

Konya Journal of Engineering Sciences

View full text Add to dashboard Cite

show abstract

“…VisSat proposed a method to approximate the RPC camera model with a pinhole camera model, enabling the application of widely used vision-based techniques to satellite imagery. This advancement allows for the rapid integration of structure from motion (SfM) [43][44][45], multi-view stereo (MVS) [46][47][48], and NeRF into satellite image analysis. Building upon this approximation, S-NeRF successfully applied the pinhole camera-based NeRF method to satellite imagery, facilitating shadow detection, albedo synthesis, and transient object filtering.…”

Section: Perspective Approximate For Rpc Cameramentioning

confidence: 99%

Sat-Mesh: Learning Neural Implicit Surfaces for Multi-View Satellite Reconstruction

Qu,

Deng

2023

Remote Sensing

View full text Add to dashboard Cite

Automatic reconstruction of surfaces from satellite imagery is a hot topic in computer vision and photogrammetry. State-of-the-art reconstruction methods typically produce 2.5D elevation data. In contrast, we propose a one-stage method directly generating a 3D mesh model from multi-view satellite imagery. We introduce a novel Sat-Mesh approach for satellite implicit surface reconstruction: We represent the scene as a continuous signed distance function (SDF) and leverage a volume rendering framework to learn the SDF values. To address the challenges posed by lighting variations and inconsistent appearances in satellite imagery, we incorporate a latent vector in the network architecture to encode image appearances. Furthermore, we introduce a multi-view stereo constraint to enhance surface quality. This constraint minimizes the similarity between image patches to optimize the position and orientation of the SDF surface. Experimental results demonstrate that our method achieves superior visual quality and quantitative accuracy in generating mesh models. Moreover, our approach can learn seasonal variations in satellite imagery, resulting in texture mesh models with different and consistent seasonal appearances.

show abstract

“…Structure from Motion (SfM, Xiao et al, 2021) can simultaneously obtain image orientation parameters and 3D point coordinates of the scene. Because of its high degree of automation, SfM has been a powerful tool used in UAV mapping, image-based 3D reconstruction, and robot vision positioning.…”

Section: Introductionmentioning

confidence: 99%

A Criterion of Radial Distortion on Structure From Motion With Consumer-Grade Cameras

Xiao

Guo²,

Wu³

et al. 2023

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. Images from consumer-grade cameras typically contain significant radial distortions, resulting in less accurate photogrammetry results generated using Structure from Motion (SfM). It is prone to reconstruct inaccurate scene points, becoming the so-called doming effect. The main reason is that in bundle adjustment, the last step of SfM, radial distortion is insufficiently estimated even though self-calibrating bundle adjustment was carried out. This paper designed a criterion to measure the radial distortion before SfM calculation, called Criterion on Radial Distortion (CRD). Firstly, the feature points were extracted and matched on the image; then, the image was divided into circular blocks, and the fundamental matrices were estimated by the matching points in different ring sub-blocks; finally, the difference between these fundamental matrices was used to reflect the radial distortion. Experiments on simulation data show that CRD can accurately reflect the radial distortion, which verifies its validity. Using actual UAV image data to conduct SfM experiments, CRD is consistent with the self-calibration results of Agisoft Metashape. After evaluating the accuracy of the elevation of scene points, it verifies its necessity for SfM.

show abstract

Sequential Cycle Consistency Inference for Eliminating Incorrect Relative Orientations in Structure from Motion

Cited by 3 publications

References 25 publications

Accuracy Assessment Toward Merging of Terrestrial Laser Scanner Point Data and Unmanned Aerial System Point Data

Accuracy Assessment Toward Merging of Terrestrial Laser Scanner Point Data and Unmanned Aerial System Point Data

Sat-Mesh: Learning Neural Implicit Surfaces for Multi-View Satellite Reconstruction

A Criterion of Radial Distortion on Structure From Motion With Consumer-Grade Cameras

Contact Info

Product

Resources

About