Indoor 3D Map Generation via Background Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating 3D maps of indoor spaces face challenges such as sensing errors, difficulty in modeling complex structures, and inability to capture data from all areas, especially with moving objects, leading to unrealistic and incomplete maps.
Innovation Solution
A method that distinguishes background and non-background areas in indoor space images, expands the background to include non-background information, generates depth-image associated information, and uses this data to create a stable 3D map, incorporating object and light source information to enhance realism and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor measurement methods are used to obtain indoor space data, then geographic information can be acquired, but sensing errors and position estimation errors occur that lower map realism
Solution Approach 1:
The patent uses image data as a reference copy to correct geographic information. Instead of relying solely on sensor measurements that contain errors, the system captures images of the indoor space and uses image processing to generate corrected geographic data that accurately reflects the real space structure, thereby eliminating sensor measurement errors.
Solution Approach 2:
The patent introduces image processing as an intermediary between sensor measurement and map generation. The image data serves as a mediator that bridges the gap between inaccurate sensor measurements and the need for realistic maps, allowing correction of position estimation errors through image-based geometric verification.
2Manufacturing precision
If complex indoor structures are modeled using current methods, then three-dimensional data can be generated, but one-to-one matching between image and shape becomes impossible due to measurement errors
Solution Approach 1:
The patent uses images as a master copy to create accurate three-dimensional models. By processing images to extract geometric information and using this image-derived data as the reference copy, the system can model complex structures with high precision that matches the visual appearance in images, overcoming sensor measurement limitations.
Solution Approach 2:
The patent transitions from relying on three-dimensional sensor measurements to using two-dimensional images as the primary data source. This dimensionality change allows accurate representation of complex structures because images capture visual reality directly, and through image processing, three-dimensional information can be extracted with higher fidelity than direct 3D sensing.
3Reliability
If complete indoor space data is obtained using complex sensor systems, then all areas can be mapped, but data collection takes a long time and cannot capture moving objects
Solution Approach 1:
The patent replaces complex mechanical sensor systems with image-based processing. Instead of using multiple sensors that require lengthy data collection and processing, the system uses images that can be captured quickly and processed computationally to generate complete three-dimensional maps, significantly reducing data collection time while maintaining completeness.
Solution Approach 2:
The patent performs preliminary image capture that can include moving objects, then processes these images to extract geometric information. By capturing images first and processing them later, the system can record moving objects in their natural positions without requiring them to be stationary during data collection, unlike traditional sensor methods that need prolonged measurement time.
Data Source
AI summary
According to an exemplary embodiment of the present disclosure, a three-dimensional map generating method of an indoor space, includes: obtaining at least one indoor space image which is an image for an indoor space; distinguishing a background area corresponding to a structure of the indoor space from a non-background area corresponding to objects located in the indoor space in the at least one indoor space image; generating at least one expanded indoor space image by expanding the background area to the non-background area in the at least one indoor space image; generating depth-image associated information based on at least one expanded indoor space image and geographic information including information of a depth value for the indoor space; and generating a three-dimensional map for the indoor space using the at least one expanded indoor space image, the geographic information, and the depth-image associated information.