Loading [MathJax]/extensions/MathMenu.js

Evaluating saliency scores in point clouds of natural environments by learning surface anomalies

Research output: Contribution to journalArticleResearchpeer review

Authors

  • Reuma Arav
  • Dennis Wittich
  • Franz Rottensteiner

External Research Organisations

  • TU Wien (TUW)
  • University of Natural Resources and Applied Life Sciences (BOKU)

Details

Original languageEnglish
Pages (from-to)235-250
Number of pages16
JournalISPRS Journal of Photogrammetry and Remote Sensing
Volume224
Early online date12 Apr 2025
Publication statusPublished - Jun 2025

Abstract

In recent years, three-dimensional point clouds are used increasingly to document natural environments. Each dataset contains a diverse set of objects, at varying shapes and sizes, distributed throughout the data and intricately intertwined with the topography. Therefore, regions of interest are difficult to find and consequent analyses become a challenge. Inspired from visual perception principles, we propose to differentiate objects of interest from the cluttered environment by evaluating how much they stand out from their surroundings, i.e., their geometric salience. Previous saliency detection approaches suggested mostly handcrafted attributes for the task. However, such methods fail when the data are too noisy or have high levels of texture. Here we propose a learning-based mechanism that accommodates noise and textured surfaces. We assume that within the natural environment any change from the prevalent surface would suggest a salient object. Thus, we first learn the underlying surface and then search for anomalies within it. Initially, a deep neural network is trained to reconstruct the surface. Regions where the reconstructed part deviates significantly from the original point cloud yield a substantial reconstruction error, signifying an anomaly, i.e., saliency. We demonstrate the effectiveness of the proposed approach by searching for salient features in various natural scenarios, which were acquired by different acquisition platforms. We show the strong correlation between the reconstruction error and salient objects. To promote benchmarking and reproducibility, the code used in this work can be found on https://github.com/rarav/salient_anomaly/releases/tag/v1.0.0 while the datasets are published on doi: 10.48436/mps0m-c9n43 and 10.48436/fh0am-at738.

Keywords

    Anomaly detection, Deep neural network, Geomorphological entities, Salient object detection (SOD)

ASJC Scopus subject areas

Cite this

Evaluating saliency scores in point clouds of natural environments by learning surface anomalies. / Arav, Reuma; Wittich, Dennis; Rottensteiner, Franz.
In: ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 224, 06.2025, p. 235-250.

Research output: Contribution to journalArticleResearchpeer review

Arav R, Wittich D, Rottensteiner F. Evaluating saliency scores in point clouds of natural environments by learning surface anomalies. ISPRS Journal of Photogrammetry and Remote Sensing. 2025 Jun;224:235-250. Epub 2025 Apr 12. doi: 10.1016/j.isprsjprs.2025.03.022
Download
@article{e0ca02f0932241c9bd0d4b0714b3f177,
title = "Evaluating saliency scores in point clouds of natural environments by learning surface anomalies",
abstract = "In recent years, three-dimensional point clouds are used increasingly to document natural environments. Each dataset contains a diverse set of objects, at varying shapes and sizes, distributed throughout the data and intricately intertwined with the topography. Therefore, regions of interest are difficult to find and consequent analyses become a challenge. Inspired from visual perception principles, we propose to differentiate objects of interest from the cluttered environment by evaluating how much they stand out from their surroundings, i.e., their geometric salience. Previous saliency detection approaches suggested mostly handcrafted attributes for the task. However, such methods fail when the data are too noisy or have high levels of texture. Here we propose a learning-based mechanism that accommodates noise and textured surfaces. We assume that within the natural environment any change from the prevalent surface would suggest a salient object. Thus, we first learn the underlying surface and then search for anomalies within it. Initially, a deep neural network is trained to reconstruct the surface. Regions where the reconstructed part deviates significantly from the original point cloud yield a substantial reconstruction error, signifying an anomaly, i.e., saliency. We demonstrate the effectiveness of the proposed approach by searching for salient features in various natural scenarios, which were acquired by different acquisition platforms. We show the strong correlation between the reconstruction error and salient objects. To promote benchmarking and reproducibility, the code used in this work can be found on https://github.com/rarav/salient_anomaly/releases/tag/v1.0.0 while the datasets are published on doi: 10.48436/mps0m-c9n43 and 10.48436/fh0am-at738.",
keywords = "Anomaly detection, Deep neural network, Geomorphological entities, Salient object detection (SOD)",
author = "Reuma Arav and Dennis Wittich and Franz Rottensteiner",
note = "Publisher Copyright: {\textcopyright} 2025 The Authors",
year = "2025",
month = jun,
doi = "10.1016/j.isprsjprs.2025.03.022",
language = "English",
volume = "224",
pages = "235--250",
journal = "ISPRS Journal of Photogrammetry and Remote Sensing",
issn = "0924-2716",
publisher = "Elsevier BV",

}

Download

TY - JOUR

T1 - Evaluating saliency scores in point clouds of natural environments by learning surface anomalies

AU - Arav, Reuma

AU - Wittich, Dennis

AU - Rottensteiner, Franz

N1 - Publisher Copyright: © 2025 The Authors

PY - 2025/6

Y1 - 2025/6

N2 - In recent years, three-dimensional point clouds are used increasingly to document natural environments. Each dataset contains a diverse set of objects, at varying shapes and sizes, distributed throughout the data and intricately intertwined with the topography. Therefore, regions of interest are difficult to find and consequent analyses become a challenge. Inspired from visual perception principles, we propose to differentiate objects of interest from the cluttered environment by evaluating how much they stand out from their surroundings, i.e., their geometric salience. Previous saliency detection approaches suggested mostly handcrafted attributes for the task. However, such methods fail when the data are too noisy or have high levels of texture. Here we propose a learning-based mechanism that accommodates noise and textured surfaces. We assume that within the natural environment any change from the prevalent surface would suggest a salient object. Thus, we first learn the underlying surface and then search for anomalies within it. Initially, a deep neural network is trained to reconstruct the surface. Regions where the reconstructed part deviates significantly from the original point cloud yield a substantial reconstruction error, signifying an anomaly, i.e., saliency. We demonstrate the effectiveness of the proposed approach by searching for salient features in various natural scenarios, which were acquired by different acquisition platforms. We show the strong correlation between the reconstruction error and salient objects. To promote benchmarking and reproducibility, the code used in this work can be found on https://github.com/rarav/salient_anomaly/releases/tag/v1.0.0 while the datasets are published on doi: 10.48436/mps0m-c9n43 and 10.48436/fh0am-at738.

AB - In recent years, three-dimensional point clouds are used increasingly to document natural environments. Each dataset contains a diverse set of objects, at varying shapes and sizes, distributed throughout the data and intricately intertwined with the topography. Therefore, regions of interest are difficult to find and consequent analyses become a challenge. Inspired from visual perception principles, we propose to differentiate objects of interest from the cluttered environment by evaluating how much they stand out from their surroundings, i.e., their geometric salience. Previous saliency detection approaches suggested mostly handcrafted attributes for the task. However, such methods fail when the data are too noisy or have high levels of texture. Here we propose a learning-based mechanism that accommodates noise and textured surfaces. We assume that within the natural environment any change from the prevalent surface would suggest a salient object. Thus, we first learn the underlying surface and then search for anomalies within it. Initially, a deep neural network is trained to reconstruct the surface. Regions where the reconstructed part deviates significantly from the original point cloud yield a substantial reconstruction error, signifying an anomaly, i.e., saliency. We demonstrate the effectiveness of the proposed approach by searching for salient features in various natural scenarios, which were acquired by different acquisition platforms. We show the strong correlation between the reconstruction error and salient objects. To promote benchmarking and reproducibility, the code used in this work can be found on https://github.com/rarav/salient_anomaly/releases/tag/v1.0.0 while the datasets are published on doi: 10.48436/mps0m-c9n43 and 10.48436/fh0am-at738.

KW - Anomaly detection

KW - Deep neural network

KW - Geomorphological entities

KW - Salient object detection (SOD)

UR - http://www.scopus.com/inward/record.url?scp=105002282736&partnerID=8YFLogxK

U2 - 10.1016/j.isprsjprs.2025.03.022

DO - 10.1016/j.isprsjprs.2025.03.022

M3 - Article

AN - SCOPUS:105002282736

VL - 224

SP - 235

EP - 250

JO - ISPRS Journal of Photogrammetry and Remote Sensing

JF - ISPRS Journal of Photogrammetry and Remote Sensing

SN - 0924-2716

ER -