Highlights

• This research introduces an edge AI-chip-based system capable of real-time classification of algae species and prediction of Harmful Algal Blooms (HAB). The system’s real-world effectiveness was demonstrated through a case study in Hong Kong’s subtropical waters, showing high accuracy (99.87%) in classifying 25 common HAB species from over 11,000 images.
• By utilizing the Algal Morphology Deep Neural Network (AMDNN) model, this tool provides a low cost and accessible solution for environmental managers and fish farmers, aiding in early warning and risk management of HABs.
• The use of data augmentation techniques such as Resizing with Aspect Ratio Preserved (RAP) on algae images is highlighted. This process significantly improves the classification accuracy of the AMDNN model, which outperforms traditional models like the Random Forest (RF) method.

Summary

This research introduces an innovative, cost-effective edge AI-chip-based system for the real-time classification of algae species and prediction of harmful algal blooms (HABs). The system can potentially be used with low cost field microscopes to greatly reduce the cost for real time HAB monitoring.

Traditional monitoring methods are labor-intensive and inadequate for dynamic algal population tracking, leading to the use of advanced instruments like IFCB for real time, in-situ monitoring at species level.  However, the advanced sophisticated instrumentation are beset with challenges in real-world application and data transmission.

The newly developed system utilizes an on-site edge AI chip with an Algal Morphology Deep Neural Network (AMDNN) model, significantly improving algae classification. The model, tested on a one-month dataset of 11,250 images from Hong Kong’s subtropical waters, achieved 99.87% accuracy.

The edge AI-chip-based system significantly advances real-time algae monitoring and HAB prediction. Due to its affordability and minimal power requirements, it is accessible to a broader audience, including environmental managers and fish farmers.

Yuan, B. Wang, J. Li, and J. H. W. Lee, “A low-cost edge AI-chip-based system for real-time algae species classification and HAB prediction,” Water Research, vol. 233, p. 119727, Apr. 2023, doi: .