The use of living organisms or enzymatic reactions as toxicity indicators is gaining prominence in the application of drinking water security. Relevant prior arts can be referred to the works by. By monitoring adverse biological effects on test organisms, end users are able to confirm the presence of toxins in the water supply prior to performing further analysis of determining the exact nature of threat. The key advantages of monitoring adverse biological effects are its rapid response and ability to be used at critical locations, such as at downstream of a water distribution system, before the biological process in a waste-water treatment plant, etc.
Throughout the years, the Public Utilities Board (PUB), Singapore's national water agency, has used fish to monitor the toxicity of the water in the entire water loop, from collection to treatment and distribution of the treated water. These fish monitoring systems are deployed at source water intakes, within the waterworks and at key control points such as along trunk mains and service reservoirs. CCTV with telemetry link to 24/7 Operation Centres were employed to remotely monitor the well-being of fishes in the tanks. Recently, more advanced Fish Biosensors were deployed which monitors the breathing pattern of the fishes using electrical signals. These systems are not only capable of monitoring the fish automatically but also alerting operators at the 24/7 centres whenever the fish is agitated by any toxicity present in the water. These Fish Biosensors systems are more elaborate to setup and expensive to install and maintain. Institute for Infocomm Research and PUB has thus come together in this current project to develop a system using image profiling to monitor the activity of the fish in conventional fish tanks for detecting water contamination.
We name the system: Fish Activity Monitoring System (FAMS). The proposed system performs real time image processing of a group of fishes and counts the numbers of active and nonactive fishes, where these fishes are exposed to water from selected sources. It is a computer vision based solution, in which cameras are used to provide 24/7 visual monitoring. Software algorithm based on background modelling and subtraction are developed to perform automated segmentation of fishes. Since our problem deals with counting a group of fishes within a confined area, occlusion which results from some fish hiding behind others from the camera’s view has posed a challenge. To address
such a challenge, a splitting method to resolve overlapping in the view based on fish shape model is developed. Criteria are imposed to ensure the estimation of fish count achieves a close approximate to the actual count. To analyze fish activity, attributes that describe movement of fish, such as motion trajectory, velocity and direction of movement are extracted. Based on these attributes, an alert notifying scheme is designed and tested based on realistic experiments with an addition of the contaminant to the water.
Typically, a water supply network will have various critical control points such as surface water reservoirs, waterworks, pumping stations and service reservoirs located in a wide geographical area. To cater for the setting in an actual water supply network, a network communication module is also developed to allow two-ways communication between the server at the control centre and the individual monitoring unit at the remote site. This project has been motivated by the needs of PUB’s water supply network. The proposed FAMS aims to enhance the security measure of drinking water by providing 24/7 monitoring and triggering alarm when detecting anomalies in fish activity.