AI-Based 3D Scan Registration Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D laser scanners rely on mathematical optimization techniques for registering multiple scans, which can result in inaccurate registration of individual scans even if the overall accuracy meets a predetermined threshold, leading to less trust and incorrect projects.
Innovation Solution
An AI system is trained using supervised learning to identify and correct the registration of landmarks in 3D scans, allowing for the detection of mismatched landmarks and improving the accuracy of scan registration by analyzing patterns and adjusting scan datasets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mathematical optimization techniques are used for registering multiple scans, then the overall accuracy meets a predetermined threshold, but the registration of individual scans becomes inaccurate
Solution Approach 1:
The patent segments the registration verification process into two levels: overall scan accuracy assessment and individual landmark registration verification. By dividing the verification task, the system can ensure both overall accuracy meets thresholds while individually checking each landmark's registration precision through AI-based detection of mismatches between scans.
Solution Approach 2:
The patent implements feedback mechanisms where the AI system continuously monitors and evaluates landmark registration accuracy across multiple scans. The system provides feedback by detecting mismatches between landmarks in different scans and adjusting the registration process accordingly, ensuring both overall and individual accuracy requirements are met.
2Extent of automation
If mathematical optimization techniques are used for scan registration, then processing can be automated, but trust in the generated point cloud models decreases due to incorrect individual registrations
Solution Approach 1:
The patent enables the system to self-verify its own registration accuracy using AI-based landmark detection and mismatch identification. The system automatically detects registration errors without human intervention, maintaining automation while improving reliability by self-monitoring and self-correcting the registration process.
Solution Approach 2:
The AI system provides automated feedback on registration quality by analyzing landmark positions across multiple scans. This feedback loop maintains automation while building trust in the point cloud models by continuously verifying registration accuracy and alerting users to potential issues.
3Manufacturing precision
If AI systems are trained to detect registration accuracy, then individual scan registration can be verified, but the complexity of the system increases
Solution Approach 1:
The patent makes the AI system multi-functional by training it to perform both landmark detection and registration accuracy verification. This universal approach allows a single system component to handle multiple tasks, improving individual registration verification while managing complexity through functional integration rather than adding separate dedicated components.
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 AI system enhances the accuracy of scan registration, ensuring that even if the overall accuracy meets a threshold, individual scans are correctly aligned, thereby increasing trust in the generated point cloud models.
Implementation Method 1
transmitting a beam of light onto the objects and collecting the reflected or scattered light to determine the distance, two-angles
Implementation Method 2
collecting the reflected or scattered light to determine the distance, two-angles
Implementation Method 3
the distance to a target point is determined based on the speed of light in air between the scanner and a target point
Data Source
AI summary
A method for automatically determining quality of registration of landmarks includes training an artificial intelligence (AI) system to detect inaccurate registration of landmarks. Training the AI system uses training data that includes scans of an environment captured by a 3D measuring device from corresponding scan points. A first scan is registered with at least a second scan based on one or more landmarks captured in the first scan and the second scan. Further, a model is created to identify incorrect registration by analyzing the training data. The analysis detects a mismatch in a first instance of a landmark in the first scan and a second instance of said landmark in the second scan. The model is then used to evaluate registration of landmarks in live data, the live data including a set of scans, the result identifying accuracy level of the registration of landmarks.