EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Long‐Range Depth Estimation Using Learning‐Based Hybrid Distortion Model for CCTV Cameras

Ami Pandat, Rajasekhar Punna, G. Aravamuthan, Gopika Vinod, Rohit Shukla

Journal of Sensors · 2026-01 · 2026 (1)

DOI: 10.1155/js/9911829 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Sensors, online erschienen am 06.08.2026. Verfasst von Ami Pandat und 4 weiteren Autor:innen. Die Arbeit nennt 59 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate camera models are essential for photogrammetry applications such as 3D mapping and object localization, particularly for long distances. Various stereo‐camera‐based 3D localization methods are available but are limited to a few hundreds of meters’ range. This is majorly due to the limitation of the distortion models assumed for the nonlinearities present in the camera lens. This article presents a framework for modeling a suitable distortion model that can be used for localizing the objects at longer distances. It is well known that neural networks (NNs) can be a better alternative to model a highly complex nonlinear lens distortion function; on the contrary, our experiments show that a purely NN‐based distortion model fails to converge or remains unstable when directly applied to camera calibration for long‐range stereo vision. To resolve the same, this article presents a hybrid distortion modeling framework that combines the stability and physical interpretability of classical distortion models with the flexibility of learning‐based refinement. The conventional distortion model is systematically extended using higher‐order polynomial terms to better capture long‐range distortion effects, and a lightweight NN is employed only to learn residual corrections to intrinsic and distortion parameters, rather than replacing the analytical model entirely. The proposed framework is validated on a large‐scale stereo CCTV setup with a 10 m baseline and evaluated for 3D localization at distances ranging from 100 m up to 5 km. The estimated 3D coordinates are transformed to GIS coordinates and are plotted on a GIS map for visualization. Experimental results demonstrate that the hybrid approach significantly outperforms both standard and extended classical distortion models, enabling stable and accurate kilometer‐scale localization using medium‐quality surveillance cameras.

Zitieren

APA

Pandat, A., Punna, R., Aravamuthan, G., Vinod, G., & Shukla, R. (2026). Long‐Range Depth Estimation Using Learning‐Based Hybrid Distortion Model for CCTV Cameras. Journal of Sensors, 2026(1). https://doi.org/10.1155/js/9911829

Vancouver

Pandat A, Punna R, Aravamuthan G, Vinod G, Shukla R. Long‐Range Depth Estimation Using Learning‐Based Hybrid Distortion Model for CCTV Cameras. Journal of Sensors. 2026;2026(1). doi:10.1155/js/9911829

BibTeX
@article{Pandat2026,
  author = {Pandat, Ami and Punna, Rajasekhar and Aravamuthan, G. and Vinod, Gopika and Shukla, Rohit},
  title = {Long‐Range Depth Estimation Using Learning‐Based Hybrid Distortion Model for CCTV Cameras},
  journal = {Journal of Sensors},
  year = {2026},
  volume = {2026},
  number = {1},
  doi = {10.1155/js/9911829}
}
Kostenlos starten →