Assessing the Effect of UAV Data Acquisition Time and Camera Angle for Mapping Shallow Water Benthic Habitats in Dulanga Beach, Gorontalo Regency, Indonesia

Abstract

Aims: This study aimed to evaluate the quality and accuracy of benthic habitat classification results based on drone imagery, verified through field data (ground truthing) based on drone flight time and sensor angle. Study Design: A stratified random sampling method was employed to collect benthic habitat data across six substrate-based strata, yielding 189 observation points. Of these, 95 were used for training and 94 for validation during classification and accuracy assessment. Place and Duration of Study: The study was conducted at Dulanga Beach, Gorontalo Regency, Indonesia. Data collection occurred from June to July 2025, with UAV imagery acquired on 14–15 June 2025. Methodology: UAV data were acquired using an Autel Evo II Pro V3 drone at 100 m altitude with 80% front-lap and side-lap, across four time intervals (08:00, 10:00, 12:00, and 14:00 Ante Meridiem (AM) local time) and two camera angle configurations (90° nadir and 45° oblique). Image pre-processing involved orthophoto generation via Agisoft Metashape using Structure from Motion (SfM), followed by mosaicking and orthorectification. Image analysis applied a two-level object-based image analysis (OBIA) workflow. Level-1 segmentation (scale 200) distinguished broad habitat zones, while Level-2 segmentation (scale 50) delineated benthic classes. Classification combined contextual rule-based editing and Support Vector Machine (SVM) algorithms implemented in eCognition Developer 8.7. Accuracy assessment was conducted using a confusion matrix to calculate overall, producer, and user accuracy, along with the Kappa coefficient. Results: The optimal acquisition was achieved at 10:00 AM local time with a 90° camera angle. Under these conditions, SVM classification with a segmentation scale of 50 yielded an overall accuracy of 91.75% and a Kappa coefficient of 0.89, indicating almost perfect agreement with field validation. Conclusion: UAV-based object-oriented classification offers an accurate and efficient approach for benthic habitat mapping in shallow waters. The findings highlight the significance of acquisition time and camera angle and offer a methodological framework for improving spatial planning, ecosystem monitoring, and the management of marine conservation areas in Gorontalo Province.