Calibration and Comparative Analysis of Forward-Looking Sonar and 3D Sonar for Enhanced Underwater Object Recognition

Aditya Penumarti; Khanh Dong; Zi-Hao Zhang; Yongkyoon Park; Zhenqi Wu; Trung Dong; Shahriar Negahdaripour; Xiaomin Lin

cs.CVAug 29, 2026

Calibration and Comparative Analysis of Forward-Looking Sonar and 3D Sonar for Enhanced Underwater Object Recognition

Authors: Aditya Penumarti, Khanh Dong, Zi-Hao Zhang, Yongkyoon Park, Zhenqi Wu, Trung Dong, Shahriar Negahdaripour, Xiaomin Lin,

Organizations: College of Engineering at the University of Florida, Gainesville, FL 32611, USA · Naval Architecture and Ocean Engineering at the Seoul National University, Seoul 08826, South Korea · Electrical and Computer Engineering, University of South Florida, Tampa, FL 33620, USA · University of Miami, Coral Gables, FL 33146, USA

Abstract

Sonars generate a significant amount of noise. With the advent of new technology capable of producing full 3D point clouds, the noise is amplified in sparse point clouds, making it challenging to recognize features for navigation, recognition, or reconstruction. To address this challenge, we propose using two different sonar modalities: one that produces a 2D intensity image and another that generates a 3D point cloud. By implementing auto-calibration, we can filter out noisy features between the modalities to enhance feature extraction. Experiments demonstrate that auto-calibration improves performance over manual calibration by 5% and that filtering enhances feature extraction by more than 40% relative to the raw point cloud. Code and datasets are given at https://theaprilab.org/fls-3d-calibrator

Figures & tables

Fig. 1 : A representative scenario of an ROV scanning a set of 4 cinder blocks. The left frames show the two modalities with their raw data, with the right frame showing the calibrated and segmented objects that were used for auto-calibration.

Fig. 2 : Blue Robotics Blue ROV equipped with a forward-looking sonar and 3D sonar.

Calibration Type	Dice Score (F1) $\uparrow$	IoU $\uparrow$	Chamfer ( $\mathrm{px}$ ) $\downarrow$	Reproj. Error ( $\mathrm{px}$ ) $\downarrow$
Manual	$0.761$	$0.64$	$4.11$	$\mathbf{7.17}$
Auto	$\mathbf{0.795}$	$\mathbf{0.68}$	$\mathbf{3.67}$	$7.46$

TABLE I : Comparative results for the automatic calibration versus the manually measured transformation using manually labeled data.

Fig. 3 : Cinder block placed on a deck underwater at Blue Grotto, Williston, FL, USA. The cinderblock was chosen as the target for calibration due to its bright return on both modalities. Several tests were conducted in different configurations to find the best for calibration.

Calibration Type	Coverage $\uparrow$	IoU $\uparrow$	Mutual Information $\uparrow$	Hausdorff ( $\mathrm{px}$ ) $\downarrow$
Manual	$0.59$	$0.45$	$0.29$	$35.39$
Auto	$\mathbf{0.64}$	$\mathbf{0.50}$	$\mathbf{0.33}$	$\mathbf{34.89}$

TABLE II : Statistical data for the metrics for the automatic and manual calibration using contour extraction.

Filtering Type	Dice Score $\uparrow$	IoU $\uparrow$
No Filter (Raw)	$0.59$	$0.42$
SOR	$0.78$	$0.63$
FLS-Filtered	$\mathbf{0.85}$	$\mathbf{0.73}$

TABLE III : Object recognition performance on denoised data and cropped raw data. The raw, unfiltered point cloud, the statistical outlier removal (SOR) filter, and the FLS-filtered point clouds are compared.

Panel of Figure 7 — Fig. 4 : Confusion matrices for the different filtering methods used prior to feature extraction using PointNet. (a) Shows the raw point cloud, which was cropped to the FLS range, (b) is the SOR filtered point cloud, and (c) is the calibrated FLS-3D sonar filtering.

Panel of Figure 8 — Fig. 5 : Segmentation results from PointNet. (a) Ground truth label for the PointNet segmentation validation, (b) PointNet results for the SOR filtering, and (c) PointNet results on the FLS-3D sonar filtering.

Underwater Imaging

Underwater imaging studies how to capture, restore, interpret, and reconstruct visual and sonar data in challenging aquatic environments.

See all 72 papers

uScenes: A Multimodal RGB and 3D Sonar Dataset for Underwater Robot Perception
Aug 28, 2026Trung Tien Dong, Zhenqi Wu, Aditya Penumarti +4Underwater Robotics 3D Scene Understanding
Embodied Robotics and Automation Lab, University of South Florida, Tampa, FL 33610, USA · Mechanical and Aerospace Engineering Department, University of Florida, Gainesville, FL 32611, USA · Naval Architecture and Ocean Engineering at the Seoul National University, Seoul 08826, South Korea
SHRUMS: Sensor Hallucination for Real-time Underwater Motion Planning with a Compact 3D Sonar
Oct 21, 2025Susheel Vadakkekuruppath, Lauritz R. Fosso, Michail Chatzispyrou +4Autonomous Underwater Navigation Motion Planning
Susheel Vadakkekuruppath and Jason M. O’Kane are with the Department of Computer Science and Engineering, Texas A&M University, USA · Herman B. Amundsen and Marios Xanthidis are with the Aquaculture Robotics and Automation, SINTEF Ocean
ISOPoT: Imaging Sonar Odometry by Point Tracking
Jun 22, 2026Jaša Samec, Vid Rijavec, Marko Peljhan +4Point Tracking Relative Pose Estimation
Faculty of Computer and Information Science, University of Ljubljana, Večna pot 113, 1000 Ljubljana, Slovenia · OZON Research Group, TIMTEC, Goriska cesta 6c, Vipava, 5271, Slovenia · MAT Systemics, University of California, Santa Barbara, 3309 Phelps Hall, Santa Barbara, 93106-6065, California, United States of America +1

Aug 28, 2026J/K move · Enter open · S save

uScenes: A Multimodal RGB and 3D Sonar Dataset for Underwater Robot Perception

Trung Tien Dong, Zhenqi Wu, Aditya Penumarti, Zi-Hao Zhang+3

Embodied Robotics and Automation Lab, University of South Florida, Tampa, FL 33610, USA · Mechanical and Aerospace Engineering Department, University of Florida, Gainesville, FL 32611, USA · Naval Architecture and Ocean Engineering at the Seoul National University, Seoul 08826, South Korea

Underwater Robotics 3D Scene Understanding Multimodal Sensor Fusion

Open paper AskPDF arXiv

Robust perception is essential for the deployment of autonomous underwater robots. However, optical cameras become unreliable under poor illumination and backscatter. Forward looking (2D) acoustic sensors remain effective under these conditions, but they measure range and bearing while leaving elevation unresolved, creating an ambiguity that prevents individual sonar returns from being localized in three dimensional (3D) space. This complicates the sensor use for 3D scene understanding and precise object detection. We introduce uScenes, a multimodal underwater dataset containing synchronized 3D multibeam sonar point clouds and RGB imagery. The dataset contains 110 scenes and 95,834 synchronized observation, representing 277.6 minutes of data collected across multiple field sessions. uScenes establishes a foundation for underwater sensor fusion, cross modal representation learning and 3D scene understanding. Code and datasets are given at https://github.com/era-research-lab/uScenes.

Abstract

Underwater Imaging

Explore similar work

uScenes: A Multimodal RGB and 3D Sonar Dataset for Underwater Robot Perception

SHRUMS: Sensor Hallucination for Real-time Underwater Motion Planning with a Compact 3D Sonar

ISOPoT: Imaging Sonar Odometry by Point Tracking