Optimized & Customizable
CleanDet Solution at a Glance
We have developed an optical imaging system that can visualize biofilms invisible to the naked eye. The imaging system consists of a hyperspectral camera, a light source emitting wavelengths suitable for excitation of observation targets (e.g., biofilms), an adjustable bandpass filter and software that uses artificial intelligence to analyze imaging results. The results can be presented to the user in a clear way showing the exact location of the contaminant, which can be identified. The system allows real-time monitoring and cloud-based services for the end user.
The hyperspectral imaging technology from CleanDet can be customized for several different usage areas
Benefits with hyperspectral imaging technology
Our cleanliness measurement is cost-effective, objective and instantly available for everyone. This leads to a revolution in the cleaning sector as cleaning changes from scheduled and reactive to proactive and data driven.
Example of an application
There used to be no cost-effective way to measure actual cleanliness levels on surfaces. Because of this, cleaning has been based on timed schedules rather than cleaning on demand. Our cost-efficient way to measure surface cleanliness allows 20 % savings by reallocating cleaners. In addition, the existing ways of measuring cleanliness are prone to human error and are not suitable for scanning large areas. Our solution can provide real-time feedback on large areas.
It works in real life
The results of testing the product in real-life circumstance pilot test have been successful. A hospital was chosen as the customer for final measurements due to the high need for automatic contamination detection. With the selected device parameters, algorithms and database, contamination detection of stains both visible and invisible to the human eye was successful.
Who is it for?
The primary users are firstly the facility managers and cleaning service providers that would benefit from a cost-efficient way to verify that a certain cleanliness level is achieved and need a way to optimize resource allocation, all the while improving the overall cleanliness level by utilizing objective data. The environment testing market is huge and growing as the demand for hygiene increases.
The key aspect of CleanDet HSI system is an interplay between artificial intelligence and hyperspectral imaging. The traditional way of detecting organic compounds relies on fluorescence, i.e., when illuminating a sample with short wavelength light (ultraviolet or blue), excitation radiation with longer red wavelengths can be observed. For example, irradiating milk with blue light shows the milk in red wavelengths due to this phenomenon.
The new CleanDet method extends this technique further: the energy for the red spectrum has to come from somewhere. As the blue light is the only source, milk must absorb some of the blue wavelengths reducing the reflection of blue light. Thus, milk appears darker in blue light than it would in red light.
Two bacterial colonies are visible in Figure a), but invisible in Figure b). The only things differing between these two figures are the illumination and detection wavelengths. Feeding this image to a neural network allows characterization of the surface contaminants.
An example of automated data clustering shows measured intensity values for a clean surface area as purple and bacterial colony as yellow. The axis are computer units of a multidimensional hypercube used in data clustering by CleanDet machine vision software.
The final step is to generalize the measurement procedure. Figure 1 shows an example of this for two bacterial colonies. Just by changing the illumination and detection conditions, the colonies can be made to vanish or appear on the surface. In practice, we want to illuminate and record the sample with all technically possible wavelength combinations and feed these input and output parameters to a modern classification neural network. This fully automated method can reveal what is on a surface. Figure 2 shows an example of data clustering (signal intensity values) on clean part of a studied area (purple) and contaminated area (yellow).