Enhanced hermit crabs detection using super-resolution reconstruction and improved YOLOv8 on UAV-captured imagery

Area of Science:

• Marine Biology
• Remote Sensing
• Computer Vision

Background:

• Hermit crabs are crucial for coastal ecosystem health, acting as indicators and ecosystem engineers.
• Traditional survey methods are inefficient and limited by environmental conditions.
• Automated monitoring is needed for effective coastal benthos conservation.

Purpose of the Study:

• To develop an automated, accurate, and cost-effective method for monitoring hermit crab populations.
• To enhance image quality for improved hermit crab detection using Super-Resolution Reconstruction (SRR).
• To evaluate a modified YOLOv8s network (CRAB-YOLO) for hermit crab identification.

Main Methods:

• Utilized Unmanned Aerial Vehicles (UAVs) for remote sensing data acquisition.
• Applied Super-Resolution Reconstruction (SRR) to enhance image resolution and reduce blur.
• Developed and implemented the CRAB-YOLO detection network, a modified YOLOv8s model.

Main Results:

• The Residual Dense Network (RDN) showed superior image reconstruction performance.
• CRAB-YOLO achieved a mean average precision (mAP) of 69.5% on the SRR test set.
• Demonstrated a 40% improvement in detection accuracy compared to the Bicubic method at 4x magnification.

Conclusions:

• The proposed UAV-based SRR and CRAB-YOLO method offers an effective solution for automated hermit crab monitoring.
• This technology supports efficient and cost-effective conservation of coastal benthos.
• The findings highlight the potential of integrating remote sensing and deep learning for ecological surveys.