Enhances images taken by cameras
The solution is to use the cameras as a sensor. This means converting the image into a status or value that can be interfaced with SCADA/HMI for alarming, trending and even control. In order to reduce bandwidth usage, a snapshot of the equipment or process condition is captured by the camera, and the value or status is recorded in the historian. This information is transmitted to the local or remote SCADA system at the same time so the operator can take the appropriate action.
For example, a standard process camera using deep learning algorithms can help improve ore crushing operations.
Acquires a variety of images
The quality of the data analysis is based on a certain quantity of images taken under various process operations and environmental changes, such as dust levels and lighting changes.
Some manufacturers offer the shelf-vision system that provides similar results. However, most are packaged in a stand-alone solution for each application, with different technologies. Using existing camera infrastructure combined with artificial intelligence can provide an integrated solution. If humans can make decisions based on the image displayed by the camera, a vision system, combined with artificial intelligence, can do the same for all cameras at the same time.
How does it work?
- Acquires the images
- Preprocesses and prepares the images
- Selects an appropriate machine learning model
- Trains the selected model
- Validates and tests the trained model
- Enhances the noise robustness of the model
- Deploys the model and integrates the status and indication with the SCADA/HMI
- Implements closed loop control, when applicable