What is Gemini?
Gemini is a unique two-channel color encoding schema that uses one completely unfiltered channel and a second channel with a specially tailored, high transmission filter. From the data differences between the two channels our algorithms create a full color image.
At all times Gemini uses the full spectral sensitivity of the sensor, resulting in unprecedentedly good low light performance.
Gemini is applicable to digital image sensors and intensifiers operating in the visible, visible/near IR, and out of visual band wavelengths, such as SWIR. In all cases, the colors in the image come from the patterns of spectral reflection of the materials in the scene. They are not ex-post-facto “painted on.”
An important benefit of Gemini encoding is that color rendering is consistent from sunlight to moonlight to starlight. This means that AI systems can use the same image source day and night and still deliver accurate results without the need for expensive retraining of the neural network that is often required by other encoding methods such as RGB/IR. Consistent color rendering in different ambient light levels is important in the defense world for AI/ML tasks such as, threat detection, automated target recognition and mobility. In the civilian world, Gemini can bring increased accuracy in changing light conditions to Automated Driver Assistance Systems (ADAS) and Autonomy.