Robots and autonomous cars are emerging in our lives. Yet these solutions are have their niches but are not fully autonomous in all environments. They still struggle with the interpretation and full perception of their environment. also because motion estimation and in consequence 3D resolution by SLAM or Structure from Motion is still a difficult task to do. Most solutions try to resolve by detecting features with strong contrast in the image. But this leads to risk of confusion when the environment has highly redundant patterns. Also surfaces tend to have holes when there is no contrast in the texture. We have another approach solving this problem

Meet Implicit Kalman a solution which aims to disassemble the image in a resolution pyramid looking not for contrast rich image features but to find the phase shift of the entire elements in the picture. Relying on phase shift can resolve not only contrast rich parts of an image, but also smooth areas with no texture simply by their phase shift.

This computer vision algorithm builds just the fundament of a sophisticated perception system which creates a full semantic hierarchy to detect moving objects in an environment and categorize them properly for a fully autonomous system of all kind. Weather full self driving cars or humanoid robots supporting humans in daily tasks. But all this starts with a proper perception of the environment that is validated and structured in a context.

More about the theory behind the solution is coming soon under XIXUM – Acabemy