Skip to Main content Skip to Navigation

Géo-référencement précis d'acquisition photogrammétrique de « longues » scènes d'intérieur

Abstract : Over recent decades, many technical advances such as the development and miniaturization of electronic components and digital camera, the advent of civil UAV and mobile mapping and the apparition of new algorithms allow 100% automatic processing yield to a strong development of photogrammetry. It becomes an indispensable measurement and surveillance technique, especially in the contexts where budget and compactness of the equipment matters the most. In this context, IGN and LOEMI team decided to develop the new hardware platforms and software solution allowing enhancements of both productivity and precision. Besides the visualization and demonstration, certain photogrammetry applications require a good measurement precision, for example, change detection for deformation studies in geosciences or very accurate shape reconstruction in industrial metrology. On large areas acquisitions, especially corridors, the photogrammetric process precision is usually limited due to systematic errors, which yield to an insufficiant localization precision of the final result. In the automatic digital photogrammetric processing chain, tie-points extraction is considered as a first step. Therefore, it is one of the important reasons that cause the imprecision of the final result. In this research work, we describe a new method, which reduces the systematic errors to enhance the precision of the existing digital photogrammetric processing chain. We propose a post-processing method for the classical photogrammetric chain. We use the results of the classical chain such as images poses, camera calibration, and mesh of scene. Our method extracts totally new tie-points pack with characteristics optimized for the photogrammetry. These characteristics are: an optimal tie-points distribution in image and object space, a high tie-points multiplicity and a precise points measure on image space. A second bundle adjustment iteration using these new tie-points pack allows refining of external images orientations and camera calibration. In consequence, the localization precision of final triangulated 3D points is enhanced. Evaluated on many different test scenes, the proposed method shows its efficacyand robustness in improving the obtained 3D points accuracy. Computation timeand number of iteration are also discussed in the manuscript. The proposed method converges from the second iteration. It requires additional computation time around 10% of the total time needed by the classical processing chain to reach a significant enhancement
Complete list of metadatas

Cited literature [157 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Wednesday, February 12, 2020 - 4:55:12 PM
Last modification on : Tuesday, May 12, 2020 - 8:28:22 AM
Long-term archiving on: : Wednesday, May 13, 2020 - 6:06:47 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02476477, version 1



Truong Giang Nguyen. Géo-référencement précis d'acquisition photogrammétrique de « longues » scènes d'intérieur. Traitement du signal et de l'image [eess.SP]. Université Paris-Est, 2018. Français. ⟨NNT : 2018PESC2182⟩. ⟨tel-02476477⟩



Record views


Files downloads