AI-based seagrass morphology measurement
Year: 2024
Type: article
Abstract: Seagrass meadows are crucial in the Great Barrier Reef (GBR) ecosystem. They filter nutrients and sediment between catchment runoffs and the Coral Sea, serve as nurseries for fish and shellfish, and play a vital role in capturing atmospheric carbon as blue carbon. However, they are currently identified as degraded across Australia. Seagrasses are known to exhibit phenotypic plasticity, acclimating their morphology to environmental stressors, which can act as an early warning of the possible risk of larger-scale seagrass loss. Changes in morphology can also influence the delivery of ecosystem services such as carbon sequestration. However, studies examining these relationships are limited and often poorly replicated due to the time-consuming nature of measuring plant morphological metrics from large numbers of plants. Seagrass morphologies are observed to differ with varying environmental conditions. The growth of seagrass can be measured by measuring the morphological parameters (e.g., Length and width of leaves, rhizomes and roots) of the seagrass. Furthermore, morphological changes may make seagrass more vulnerable to light stress. Thus, the measurement of morphological parameters can give information for proper restoration management and utilization of seagrass species. The proposed research would examine whether it is possible to utilise machine learning techniques to automate the process. Specifically, the study intends to develop an algorithm that leverages image processing and artificial intelligence for the automated measurement of morphological parameters from digital imagery. The research adopts a deep learning model, specifically YOLO-v6, to classify three distinct seagrass object types and determine their dimensions. The empirical findings underscore the efficacy of the proposed model, with an average recall of 97.5\%, average precision of 83.7%, and an average F1-score of 90.1%.
Authors: Sajal Halder, Nahina Islam, Biplob Ray, Elizabeth Andrews, Pushpika Hettiarachchi, Emma Jackson
Institutions: Data61, Central Queensland University
Language: English
Subfield: Oceanography
Field: Earth and Planetary Sciences
Domain: Physical Sciences
SDG: Life below water
Open Access status: hybrid
Estimated APC paid: $3,850
OpenAlex ID: W4402311409