3D Model Alignment Using Artificial Objects for Displacement Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating three-dimensional models from camera images face challenges in accurately detecting displacement between models captured at different times, particularly due to variations in natural objects and sparse point densities, which affect alignment and detection accuracy.
Innovation Solution
A method and device that utilize attribute information of artificial objects to detect corresponding points between three-dimensional models, perform alignment, and instruct image capture in sparse regions to improve accuracy, using a system comprising terminal devices and a processing device for generating and detecting displacements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If three-dimensional models are generated from camera images for displacement detection, then the method is simpler and more versatile than laser measurement, but the accuracy of displacement detection deteriorates due to sparse point densities and variations in natural objects
Solution Approach 1:
The patent segments the three-dimensional model into multiple regions based on point density, identifying sparse regions that require additional imaging. This segmentation allows targeted improvement of model quality in critical areas without requiring complete re-capture of the entire scene, thus balancing simplicity with accuracy.
Solution Approach 2:
The system performs preliminary analysis of the initial three-dimensional model to identify sparse regions before final displacement detection. This preliminary action enables proactive capture of additional images in problematic areas, preventing accuracy issues before they affect the final measurement results.
2Adaptability or versatility
If natural objects are used for alignment between models, then the method works in environments without artificial objects, but alignment accuracy deteriorates due to variations in natural objects between different capture timings
Solution Approach 1:
The patent implements a dynamic selection mechanism that adapts the alignment approach based on the environment. When artificial objects are present, they are used for high-precision alignment; when absent, natural objects are used with adjusted expectations. This dynamic adaptation maintains versatility while managing accuracy requirements appropriately for each scenario.
Solution Approach 2:
The system changes the alignment parameters and tolerance thresholds based on the type of objects available. For artificial objects with stable characteristics, stricter alignment parameters are applied. For natural objects that may vary between captures, more flexible parameters are used, allowing the system to maintain functionality across different environments while optimizing accuracy for each case.
3Measurement precision
If additional images are captured to improve model density in sparse regions, then the accuracy of displacement detection is improved, but the time and resources required for image capture increase
Solution Approach 1:
The patent applies local quality enhancement by capturing additional images only in specific sparse regions rather than uniformly across the entire scene. This localized approach improves model density and displacement detection accuracy where needed while minimizing the additional time and resources required compared to comprehensive re-capture.
Solution Approach 2:
The system performs self-diagnosis of the three-dimensional model quality and automatically identifies which regions require additional imaging. This self-service capability enables the system to optimize its own data collection process, capturing images only where necessary to achieve acceptable accuracy levels, thereby reducing unnecessary time expenditure.
Data Source
AI summary
A displacement detection method, performed by a computer, includes: obtaining (i) a first three-dimensional (3D) model representing a region at a first timing by 3D points and including first objects and (ii) a second three-dimensional model representing the region at a second timing by 3D points and including second objects, the second timing being different from the first timing; detecting, by use of attribute information of the first objects and the second objects, (i) first 3D points included in the first 3D model and (ii) second 3D points included in the second 3D model and associated with the respective first 3D points; performing alignment of the first 3D model and the second 3D model by use of the first 3D points and the second 3D points; and detecting displacement between the first 3D model and the second 3D model after the alignment.