Research on Cyanobacterial-Bloom Detection Based on Multispectral Imaging and Deep-Learning Method.

Area of Science:

• Environmental Science
• Ecology
• Remote Sensing

Background:

• Cyanobacterial blooms pose significant environmental challenges, impacting water ecosystems and requiring effective monitoring.
• Traditional detection algorithms lack stability, necessitating advanced solutions for accurate identification.

Purpose of the Study:

• To develop a robust and accurate method for detecting cyanobacterial blooms.
• To enhance the recognition of blooms in complex environmental conditions using deep learning.

Main Methods:

• Utilized Unmanned Aerial Vehicle (UAV) multispectral imagery for feature extraction.
• Employed an improved vegetation-index method to capture subtle spectral signatures of blooms.
• Implemented an enhanced transformer model with feature enhancement and pixel-correction fusion for improved accuracy.

Main Results:

• Achieved a detection accuracy exceeding 85% for cyanobacterial blooms in Chinese rivers.
• Provided accurate estimations of bloom contamination area and pollution severity.
• Demonstrated the algorithm's effectiveness in complex scenes with noise.

Conclusions:

• The proposed deep-learning algorithm offers a stable and accurate solution for cyanobacterial bloom detection.
• This method provides a foundation for ecological and environmental agencies in managing and controlling harmful algal blooms.
• Advanced remote sensing and AI integration can significantly improve water quality monitoring.