News
– This project has developed a surveillance system to detect pollutant discharges at wastewater treatment plants and alert the operator
– The system uses advanced computer vision techniques to identify industrial discharges, the presence of foam, hydrocarbons and grease
The DeepFlow project has successfully digitalised water quality monitoring at wastewater treatment plants for the detection of contaminants using computer vision. A hybrid technology approach was chosen: classical vision algorithms have proven to be an effective solution for detecting colour changes caused by discharges, while Deep Learning-based Artificial Intelligence has demonstrated itself to be the only effective means of detecting complex textures such as the iridescence of hydrocarbons.
The tool developed by DeepFlow manages the complete contaminant detection cycle: from image capture through to alarm notification to the operator. The aim is to monitor visual characteristics that provide indicators of water quality through the use of conventional RGB cameras, transforming them into intelligent sensors capable of operating continuously, with a system validated both in a laboratory and in real operational environments.
In the laboratory, discharge conditions were recreated in controlled tanks using sunscreen and real fuels in water with varying levels of turbidity. Validation in real environments has made it possible to identify the system’s limitations, for instance, that detecting thin oil films requires an artificial lighting system in the monitoring zones.
Within the framework of this project, led by Cetaqua – Water Technology Centre with the participation of CVC Computer Vision Center and validated through the operational expertise of the Veolia group, a continuous, non-intrusive, low-cost, and scalable surveillance solution has been developed for industrial environments and natural bodies of water. It enables problems in the water at wastewater treatment plants (WWTPs) to be detected before the water is returned to the natural environment.
Among the contaminants that reach WWTPs untreated chemical loads and hydrocarbons such as diesel, mineral oils, and petrol represent an operational threat. If these substances enter the biological treatment stages, they can destroy the bacterial biomass responsible for purification.
The arrival of discharges not only halts the treatment process and forces costly technical shutdowns, but also jeopardises compliance with discharge regulations. Furthermore, the release of these contaminants into the natural environment causes damage to aquatic ecosystems and affects their biodiversity as well as public health.
With this technology, DeepFlow aims to overcome the limitations of traditional detection systems. Historically, the detection of such events has relied on two approaches: visual inspection carried out by operators during surveillance rounds —which left large windows of time without monitoring— and traditional analytical instrumentation, such as hydrocarbon analysers, which require high acquisition and maintenance costs.
For more information about the project, visit the DeepFlow website.
