Back to results
US 12,644,847 B2Granted
Single-pass 3D reconstruction of internal surface of pipelines using depth camera array
AssigneeNUtech Ventures
InventorsZhigang Shen, Zhexiong Shang
Life of this patent
Estimating the term…
Abstract
A system for inline inspection of a pipeline where the system includes: an inline vehicle to transport the system along the pipeline; a depth camera array (DCA) of multiple RGBD cameras located on an end of the vehicle, circular lights attached around the cameras to illuminate inside the pipeline, and a controller connected to the DCA.
Claims17
A system for inline inspection of a pipeline, the system comprising:
- an inline vehicle comprising a frame and configured to transport the system along the pipeline;
- a depth camera array (DCA) comprising plurality of RGBD cameras, the DCA disposed on an end of the frame, each of the plurality of RGBD cameras being mounted on sliding brackets of the frame that are configured to allow adjustment of a relative distance between each of the plurality of RGBD cameras, each of the plurality of RGBD cameras configured to rotate toward and away from an axis of symmetry of the pipeline;
- a plurality of circular lights attached around the plurality of cameras and configured to illuminate inside the pipeline with diffuse light in a uniformly illuminated ring around the periphery of the pipe;
- a controller disposed on the frame and operatively connected to the DCA;
- and
- an additional light disposed on the end of the frame and facing outward,
- wherein the plurality of RGBD cameras provide overlapping oblique views of an inner surface of the pipeline.
4 more elements
A method for reconstructing a digital 3D model of an interior surface of a pipeline, the method comprising:
- calibrating a depth camera array (DCA);
- preprocessing data from input videos from the calibrated DCA, the preprocessing comprising:
- synchronizing the cameras of the DCA;
- and
- capturing data;
- projecting a central camera;
- applying visual odometry frame-to-frame, the applying visual odometry comprising:
- computing initial transformations with colored iterative closest point;
- optimizing camera poses, the optimizing comprising:
- refining transformations with graph optimization;
- querying successive frames
- reconstructing a 3D surface, the reconstructing comprising:
- integrating images with a volumetric truncated signed distance function (TSDF);
- and
- outputting a mesh model.
12 more elements
A method for automated pipeline surface reconstruction, comprising:
- synchronizing video sequences from a plurality of depth cameras using a temporal filter;
- registering color and depth frames at each time stamp using a 3D lookup table;
- outputting a sequence of time-stamped pipe point clouds collected as the plurality of depth cameras travels along a pipeline;
- generating, for each pipe point cloud, a virtual front-viewed and center-positioned RGBD camera;
- calculating a color/depth value at each pixel as the median of the neighbor points within a pre-defined distance;
- converting input point clouds into a sequence of front-viewed RGBD images;
- calculating motion of the plurality of cameras over timestamps based on a rendered RGBD image sequence;
- and
- integrating the RGBD sequence with computed camera poses into a global system.
6 more elements
Dependencies are inferred from the claim text.