Benthic bivalves, such as clams and mussels, are commonly used as ecological indicators to assess the condition of marine ecosystems. Underwater Visual Census (UVC) of these organisms often relies on diver surveys, fixed cameras, or manual inspection of underwater images, which limits spatial coverage and requires substantial human effort. This paper studies the feasibility of a low-cost intelligent robotic system based on a compact AI camera to support ecological monitoring through automated UVC of benthic organisms. The proposed module integrates synchronized image acquisition and onboard neural inference within a pressure-rated device designed for deployment on underwater robotic platforms. A lightweight YOLOv8 Nano detector is exported to TensorFlow Lite for deployment on the embedded cameras and used to identify bivalves in the acquired images. In addition to annotated images with bounding boxes, the system generates structured reports that record the number of detected clams and mussels and their image coordinates. Experimental evaluation is conducted on a representative underwater dataset comprising scenes with natural substrates, suspended particles, and non-uniform illumination. The results show that the system achieves a [email protected] of 86.9% and an F1-score of 85.2%, confirming reliable detection and localization of bivalves in challenging underwater conditions.

Feasibility Analysis of an Embedded AI Vision System for Automated Underwater Bivalve Visual Census / Gioiello, F., Bartolucci, V., Beni, F., Dimitri, R., Pizzuto, A., Scaradozzi, D.. - (2026), pp. 902-907. (34th Mediterranean Conference on Control and Automation, MED 2026 Ancona, IT 23 - 26 June 2026) [10.1109/med70602.2026.11598205].

Feasibility Analysis of an Embedded AI Vision System for Automated Underwater Bivalve Visual Census

Gioiello, Flavia
Primo
;
Bartolucci, Veronica;Scaradozzi, David
Ultimo
2026-01-01

Abstract

Benthic bivalves, such as clams and mussels, are commonly used as ecological indicators to assess the condition of marine ecosystems. Underwater Visual Census (UVC) of these organisms often relies on diver surveys, fixed cameras, or manual inspection of underwater images, which limits spatial coverage and requires substantial human effort. This paper studies the feasibility of a low-cost intelligent robotic system based on a compact AI camera to support ecological monitoring through automated UVC of benthic organisms. The proposed module integrates synchronized image acquisition and onboard neural inference within a pressure-rated device designed for deployment on underwater robotic platforms. A lightweight YOLOv8 Nano detector is exported to TensorFlow Lite for deployment on the embedded cameras and used to identify bivalves in the acquired images. In addition to annotated images with bounding boxes, the system generates structured reports that record the number of detected clams and mussels and their image coordinates. Experimental evaluation is conducted on a representative underwater dataset comprising scenes with natural substrates, suspended particles, and non-uniform illumination. The results show that the system achieves a [email protected] of 86.9% and an F1-score of 85.2%, confirming reliable detection and localization of bivalves in challenging underwater conditions.
2026
9798319547460
File in questo prodotto:
File Dimensione Formato  
Gioiello_Feasibility-Analysis-Embedded-AI-Vision_2026.pdf

Solo gestori archivio

Tipologia: Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza d'uso: Tutti i diritti riservati
Dimensione 3.88 MB
Formato Adobe PDF
3.88 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11566/360914
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact