HANDHELD PORTABLE OPTICAL SCANNER AND METHOD OF USING
First Claim
1. A method comprising:
- capturing an image in digital data with a camera;
storing digital data of the image on a non-transitory computer readable medium;
extracting gray values from image color channels of the image digital data;
creating image pyramids from the grey values and from depth data of the digital data;
computing a scene fitness value using the image pyramids;
predicting a camera pose;
aligning the image with each element of a first subset of selected keyframes, yielding an aligned pose and also yielding a quantity of poses and a corresponding quantity of overlap values with respect to the selected keyframes;
computing a new camera pose estimate using the aligned image and the quantity of poses and the quantity of overlap values;
creating a keyframe from the digital data of the image using the new camera pose estimate when desired;
after the keyframe is created, selecting a second subset of keyframes different from the first subset of keyframes and repeating the step of aligning with each keyframe of the second subset of keyframes to yield a plurality of pose values and overlap values;
deciding for each element of the selected first subset of keyframes and the selected second subset of keyframes whether new links are required to the keyframe in a keyframe pose graph; and
linking the keyframes.
1 Assignment
0 Petitions
Accused Products
Abstract
A system and method for real-time or near-real time processing and post-processing of RGB-D image data using a handheld portable device and using the results for a variety of applications. The disclosure is based on the combination of off-the-shelf equipment (e.g. an RGB-D camera and a smartphone/tablet computer) in a self-contained unit capable of performing complex spatial reasoning tasks using highly optimized computer vision algorithms. New applications are disclosed using the instantaneous results obtained and the wireless connectivity of the host device for remote collaboration. One method includes steps of projecting a dot pattern from a light source onto a plurality of points on a scene, measuring distances to the points, and digitally reconstructing an image or images of the scene, such as a 3D view of the scene. A plurality of images may also be stitched together to re-position an orientation of the view of the scene.
146 Citations
21 Claims
-
1. A method comprising:
-
capturing an image in digital data with a camera; storing digital data of the image on a non-transitory computer readable medium; extracting gray values from image color channels of the image digital data; creating image pyramids from the grey values and from depth data of the digital data; computing a scene fitness value using the image pyramids; predicting a camera pose; aligning the image with each element of a first subset of selected keyframes, yielding an aligned pose and also yielding a quantity of poses and a corresponding quantity of overlap values with respect to the selected keyframes; computing a new camera pose estimate using the aligned image and the quantity of poses and the quantity of overlap values; creating a keyframe from the digital data of the image using the new camera pose estimate when desired; after the keyframe is created, selecting a second subset of keyframes different from the first subset of keyframes and repeating the step of aligning with each keyframe of the second subset of keyframes to yield a plurality of pose values and overlap values; deciding for each element of the selected first subset of keyframes and the selected second subset of keyframes whether new links are required to the keyframe in a keyframe pose graph; and linking the keyframes. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
Specification