The dense and toxic blooms formed by cyanobacteria in the aquatic environment have caused serious damage to public health and aquatic ecosystems. It is crucial to monitor and prevent cyanobacterial blooms in freshwater waters in a timely manner. Although the object detection method has been applied to the field of algae identification, there are some issues to be solved in the existing research. The main challenge is that the setting of the training set is too idealistic, which is separated from the actual water quality environment and cannot guide the actual algae identification and water quality monitoring. In this paper, we collect 2,024 algae microscopic images of algae from a reservoir in southern China and a data set for object detection. Aiming at the above problems, we proposed a morphable attention based algal tiny object detection strategy to sample and extract features of algae small targets more effectively. It is hoped that the model training and improvement results of this paper can help to establish a future algae real-time monitoring and water quality evaluation system.
原文链接
分享
引用
