Our Technology
We are developing multispectral image analysis systems that can image objects at different wavelengths. Acquired images are enhanced using various image processing algorithms and meaningful information is extracted using enhanced image analysis and computer vision techniques. Extracted information is used to produce machine learning models that enables accurate and precise prediction. Finally, artificial intelligence models are trained for high performance classification or synthesize realistic images. Detailed explanation of these steps are described below. To inquire more information, please email info@ergovisio.com.
Multispectral Imaging
Every high quality imaging system requires a solid optical system composed of low-distortion lens and sharp-response filters with uniform illumination. We are using patented technology (patent no) that packs multiple lenses and filters to build a compact and ergonomic optical system. Illumination and filters at various wavelengths of light can form high-contrast images of objects. Using various wavelengths enables an improved imaging that can lead to high performance vision applications.
Multispectral imaging combined with light polarization and filtering enables to image objects that are under the surface at different depths. This technology can be used to reduce the undesired glare on wet surfaces, etc.
Multispectral imaging has important applications in healthcare, wellness and cosmetics.
Image Processing
High quality images can require pre-processing to further enhance the acquired images. Pre-processing will also optimize the images for the following processing and analysis steps.
Existing and novel image processing algorithms are used to extract features (statistical, texture etc.) and machine learning algorithms are used to segment and to classify objects of interest. High accuracy and precision are critically important for segmentation and classification.
Besides using existing image processing algorithms, we are designing new algorithms or adapt existing algorithms for each task to obtain the best performance.