Environmental pollution is always a matter of great concern in society. In Vietnam, there are policies to monitor emissions and discharge water remotely and transmit it to the Department of Natural Resources and Environment of the City/Province for monitoring. Sensor parameters and water quality images are brought back and monitored by humans; however, with many images from the camera, humans can't control them. From there, we propose a solution using AI to identify anomalies and pollution, helping to reduce the manager's load and easily manage multiple stations simultaneously. In this paper, we share a local experimental result from Vietnam.