Targetless 3D Scanner Tracking via Natural Feature Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D scanning systems face challenges in automatic registration of scans taken from different positions, requiring manual intervention and resulting in inefficiencies, incomplete scans, and increased costs, especially in time-sensitive situations like crime or accident scene investigations, due to the need for artificial targets and complex equipment.
Innovation Solution
A system that uses machine learning algorithms to identify natural features and perform simultaneous localization and mapping (SLAM) during scanning, allowing for automatic registration of scans without the need for manual target placement, enabling efficient and accurate 3D mapping of environments with a handheld or movable scanner.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual registration procedures are used with artificial targets, then registration accuracy can be improved, but the time required and operational complexity increase significantly
Solution Approach 1:
The system automatically identifies and matches natural features in the environment without requiring manual placement of artificial targets. The processor autonomously performs feature detection, descriptor extraction, and scan registration, enabling the system to register scans independently without human intervention.
Solution Approach 2:
The system extracts and utilizes naturally occurring features from the scanned environment itself, rather than relying on externally introduced artificial targets. By detecting keypoints and descriptors from existing environmental features, the system eliminates the need for separate target objects.
2Reliability
If artificial targets are used for registration, then reliable feature matching can be achieved, but device complexity and setup requirements increase
Solution Approach 1:
The system is designed to work with any environment that contains natural features, eliminating the need for specialized artificial targets. The same processor and algorithm handle both feature detection and scan registration, making the system versatile across different scanning scenarios without requiring additional equipment.
Solution Approach 2:
The system extracts registration features directly from the scanned environment's natural structures rather than requiring separate artificial target objects. This reduces the number of components needed and simplifies the overall system architecture.
3Manufacturing precision
If multiple scans from different positions are taken, then complete 3D coverage can be achieved, but automatic registration becomes more difficult without sufficient overlap
Solution Approach 1:
The system uses detected natural features and their descriptors as feedback to automatically adjust and refine scan registration. By comparing feature matches across multiple scans, the processor can automatically determine relative positions and orientations, enabling seamless integration of scans with varying overlap.
Data Source
AI summary
Technical solutions are described to track a handheld three-dimensional (3D) scanner in an environment using natural features in the environment. In one or more examples, the natural features are detected using machine learning. Features are filtered by performing a stereo matching between respective pairs of stereo images captured by the scanner. The features are further filtered using time matching between images captured by the scanner at different timepoints.