Depth Sensor Point Cloud Enhancement via Synthetic Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Depth sensors, such as time-of-flight sensors, often capture point clouds with discontinuities and artifacts due to factors like surface angle, distance, and color, which impede accurate object dimensioning in applications like transport and logistics.
Innovation Solution
A method and apparatus that enhance the point cloud by generating synthetic points based on image data and plane equations, correcting under-sampled regions and multipath artifacts, to create a more comprehensive and accurate representation of object surfaces for dimensioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a depth sensor captures point cloud data, then object surfaces can be detected, but discontinuities and artifacts occur in the point cloud under certain conditions
Solution Approach 1:
The patent uses image data as an intermediary to bridge the gap where depth sensor data is insufficient. By combining the image data with plane equations derived from the point cloud, the system generates synthetic depth points that fill in discontinuities and artifacts, thereby improving point cloud completeness without sacrificing measurement precision of the captured points.
Solution Approach 2:
The patent creates synthetic depth points by copying and extrapolating information from existing captured points through plane equations. These synthetic points replicate the surface geometry in regions where direct sensor measurements are missing or artifact-ridden, effectively copying the underlying surface structure to restore point cloud completeness.
2Manufacturing precision
If the point cloud contains discontinuities, then object dimensioning becomes impeded, but the discontinuities arise from inherent sensor limitations
Solution Approach 1:
The patent introduces image data as an intermediary source to recover surface information lost in the point cloud. By using the image data in conjunction with plane equations, the system reconstructs missing surface regions, thereby reducing information loss and enabling accurate dimensioning even when the original point cloud contains discontinuities.
Solution Approach 2:
The patent transforms the problem by changing from relying solely on direct depth measurements to using a combined approach of image data and plane equation parameters. This parameter transformation allows the system to infer surface geometry in regions where direct measurements are missing, thereby recovering lost surface information for accurate dimensioning.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The enhanced point cloud allows for accurate determination of object dimensions, overcoming limitations imposed by sensor artifacts and environmental factors, enabling flexible positioning and handling of various object sizes and colors.
Implementation Method 1
A depth sensor, such as a time-of-flight sensor, can be implemented in conjunction with any of a variety of computing devices to capture a point cloud depicting an object
Data Source
AI summary
A method in a computing device includes: obtaining (i) a point cloud having captured points depicting a plurality of surfaces of a target object, and (ii) an image depicting the plurality of surfaces of the target object; determining, from the point cloud, respective planes corresponding to each of the plurality of surfaces; determining, from the image, respective boundaries of each of the plurality of surfaces; for each of the plurality of surfaces, generating synthetic points, at least some of the synthetic points having positions different from positions of the captured points in the point cloud, the positions of the synthetic points disposed on a corresponding plane of the target object and within a corresponding surface boundary; and providing an enhanced point cloud including (i) at least a portion of the captured points, and (ii) the synthetic points.