Matterport has been granted a patent for systems and methods that allow arbitrary visual features to be registered as fiducial elements. The method involves aligning a reference object and a visual feature, capturing an image, identifying feature points, deriving a coordinate system, providing measures to each feature point, and placing the registered visual feature in a locale. This technology can be used for various applications such as augmented reality and virtual reality. GlobalData’s report on Matterport gives a 360-degree view of the company including its patenting strategy. Buy the report here.
According to GlobalData’s company profile on Matterport, 3D modelling and rendering was a key innovation area identified from patents. Matterport's grant share as of September 2023 was 63%. Grant share is based on the ratio of number of grants to total number of patents.
Registering visual features as fiducial elements for localization
A recently granted patent (Publication Number: US11775788B2) describes a computer-implemented method for registering and identifying visual features using fiducial elements. The method involves aligning a reference object and a visual feature, capturing an image of both, and identifying a set of feature points in the visual feature. A coordinate system is derived from the image using the reference object, and a set of measures is provided to each feature point using the coordinate system. The registered visual feature can then be placed in a specific location.
The method can also include capturing an image of the visual feature in the desired location and deriving the pose of the visual feature or the imager in that location using the set of measures. Additionally, a unique identifier for the visual feature can be generated using data associated with the set of feature points, and this identifier can be determined from the image of the visual feature in the desired location.
The patent further describes different scenarios for aligning the reference object and the visual feature. In one scenario, a two-dimensional fiducial tag and a two-dimensional visual feature are placed near each other on a surface. In another scenario, a three-dimensional object and a three-dimensional visual feature are placed near each other on a surface, with alignment features located on a plane normal to the surface. The method can also involve capturing the image using multiple visible light cameras and including depth information derived from these cameras.
Various types of visual features are mentioned in the patent, including two-dimensional pictures with nonrecurrent texture maps, two-dimensional asymmetrical pictures, and regular two-dimensional arrays of fiducial tags. The set of feature points can be sub-pixel locations relative to the pixels of the captured image.
The patent also discusses the possibility of registering multiple visual features using the reference object and generating unique identifiers for each feature. In this case, images of both features can be captured in the desired location, and the unique identifiers can be identified from the image.
Overall, this patent presents a computer-implemented method for registering and identifying visual features using fiducial elements, providing a potential solution for various applications such as augmented reality, object recognition, and localization.