Real-Time Identification of Cyanobacteria Blooms in Lakeshore Zone Using Camera and Semantic Segmentation: A Case Study of Lake Chaohu (Eastern China)

The surface water in the lakeshore zone is primary area where cyanobacteria bloom floats intensively. In lake environment monitoring, it has become pressing to accurately identify distribution and accumulation coverage of blooms zone. This study proposes a real-time dynamic monitoring technology for using shore-based camera network. specific work as follows: Chaohu Lake, large eutrophic China, selected research object. multithreading used dynamically obtain hourly video images 43 cameras around Lake. semantic segmentation method images, calculate blooms, draw spatial map To improve accuracy recognition, we use ResNet-50 network integrate three models, namely FCN, U-net, DeeplabV3+. By comparing results identified by methods, found that boundary DeeplabV3+(ResNet-50) clear, which more consistent with real information suitable changes Lake demonstrated time requirement on an basis could be met utilizing uses multiple threads. OA (Overall Accuracy), MPA (Mean Pixel IOU (Intersection Over Union) values were highest, 0.83, 0.82, 0.71, 0.74, RMSE between predicted cyanobacterial was 6.65%. above show this technology’s most model. technique can provide technical support scientific development management plan calculating drawing