Back to results
US 10,295,338 B2Granted
Method and system for generating map data from an image
AssigneeMagic Leap, Inc.
InventorsRony Abovitz, Brian T. Schowengerdt, Mathew D. Watson
Life of this patent
Estimating the term…
Abstract
A method of generating map data comprises capturing an image of a field of view of a user, extracting a set of map points based on the captured image, identifying respective sets of sparse points and dense points based on the extracted map points, performing point normalization of the respective sets of sparse points and dense points, generating sparse and dense point descriptors for the respective sets of sparse points and dense points, and combining the sparse point descriptors and dense point descriptors to store as map data.
Claims19
A method of generating map data, comprising:
- capturing, by a virtual or augmented display system, an image of a field of view of a user;
- determining a set of map points without comparing first features in the image and second features from existing images at least by:
- determining and positioning a virtual keyframe based at least in part upon performing one or more analyses on the captured image and one or more additional keyframes;
- projecting a plurality of geometric entities from the virtual key frame to a plurality of features or points in the image and the one or more additional keyframes;
- and
- determining the set of map points by using at least the plurality of geometric entities;
- identifying one or more sets of sparse points and one or more sets of dense points based in part or in whole on the set of map points that has been determined;
- generating sparse point descriptors and dense point descriptors respectively for the one or more sets of sparse points and the one or more sets of dense points after performing point normalization on the one or more sets of sparse points and the one or more sets of dense points;
- storing the one or more sets of sparse points and the one or more sets of dense points, the sparse point descriptors, and the dense point descriptors into a single map;
- executing one or more object recognizers on the single map, wherein the one or more object recognizers are configured to recognize respective predetermined objects of a real world based at least in part on the single map;
- and
- re-inserting, by the one or more object recognizers, geometric information and parametric information about the respective predetermined objects into a passable world model.
9 more elements
A method of generating map data, comprising:
- capturing an image of a field of view of a user;
- extracting a set of map points based on the captured image;
- identifying respective sets of sparse points and dense points based on the extracted map points;
- performing point normalization of the respective sets of sparse points and dense points;
- determining pose information of the virtual or augmented display system or a user of the virtual or augmented display system using sparse point descriptors that are generated for the one or more sets of sparse points after performing point normalization on the one or more sets of sparse points and the one or more sets of dense points;
- determining depth information of one or more objects using dense point descriptors that are generated for the one or more sets of dense points after performing the point normalization;
- storing the one or more sets of sparse points and the one or more sets of dense points, the sparse point descriptors and the dense point descriptors into a single map;
- and
- executing one or more object recognizers on the single map, wherein the one or more object recognizers are configured to recognize respective predetermined objects of a real world based at least in part upon on the single map, wherein
- semantic information is utilized to estimate future positions of the respective predetermined objects.
7 more elements
Dependencies are inferred from the claim text.