Back to results
US 12,047,550 B2Granted
Three-dimiensional point cloud generation using machine learning
AssigneeFaro Technologies, Inc.
InventorsGeorgios Balatzis, Francesco Bonarrigo, Andrea Riccardi
Life of this patent
Estimating the term…
Abstract
An example method for training a machine learning model is provided. The method includes receiving training data collected by a three-dimensional (3D) imager, the training data comprising a plurality of training sets. The method further includes generating, using the training data, a machine learning model from which a disparity map can be inferred from a pair of images that capture a scene where a light pattern is projected onto an object.
Claims19
A method comprising:
- receiving training data collected by three-dimensional (3D) imagers, the training data comprising a plurality of training sets having pairs of training images of scenes where light patterns are projected onto objects;
- and
- generating, using the training data, a machine learning model from which disparity maps are generated from pairs of images;
- receiving a captured pair of images comprising a left camera image and a right camera image, the captured pair of images captured by a 3D imager while projecting a light pattern onto an object;
- and
- generating, using the machine learning model, a disparity map for the captured pair of images.
3 more elements
A system comprising:
- a first scanner comprising a first left camera, a first right camera, and a first projector, the first scanner capturing a pair of training images by projecting a first light pattern onto a first object and capturing a first left image of the first object using the first left camera and a first right image of the first object using the first right camera;
- a processing system to generate a machine learning model based at least in part on the first left image, the first right image, and a first disparity map calculated using the left image and the right image;
- and
- a second scanner comprising a second left camera, a second right camera, and a second projector, the second scanner capturing a pair of images by projecting a second light pattern onto a second object and capturing a second left image of the second object using the second left camera and capturing a second right image of the second object using the second right camera, the second scanner generating, using the machine learning model, a second disparity map from the second left image and the second right image.
1 more element
A method comprising:
- receiving a pair of images, comprising a left camera image and a right camera image, that capture a scene of a light pattern projected on an object;
- generating, using a machine learning model, a disparity map for the pair of images, the machine learning model trained to create point clouds from pairs of images having light patterns projected onto objects;
- and
- generating a point cloud by applying a triangulation technique to the disparity map for the pair of images.
1 more element
Dependencies are inferred from the claim text.