Real-Time Room-Scanning Floorplan Generation from Semantic 3D Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques fail to provide accurate and efficient generation of floorplans and measurements in indoor environments using mobile devices, as sensor data is often incomplete or lacks semantic information, leading to inaccuracies in real-time floorplan generation.
Innovation Solution
The use of semantically labeled 3D representations, combined with 2D lateral domains, facilitates the identification of structural elements and objects within a room, enabling live preview and final floorplan generation through separate detection of wall structures and object bounding boxes, refined using neural networks and algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensor data processing is used for floorplan generation, then the system is simpler to implement, but the accuracy and completeness of the generated floorplan deteriorates due to incomplete sensor data and lack of semantic information
Solution Approach 1:
The patent combines multiple data sources (sensor data, semantic information, 3D representations) and processing techniques (neural networks, algorithms) into an integrated system that generates accurate floorplans. This merging of disparate elements resolves the contradiction by achieving high measurement precision through comprehensive data fusion while managing system complexity through unified architecture.
Solution Approach 2:
The system performs preliminary processing of sensor data to create 3D representations and extract semantic information before final floorplan generation. This preliminary action prepares the data in advance, improving the accuracy of the final output while organizing the complexity into manageable preprocessing and postprocessing stages.
2Measurement precision
If comprehensive post processing techniques are applied to generate final floorplan, then the floorplan accuracy is improved, but the processing time increases and real-time preview capability is compromised
Solution Approach 1:
The patent segments the floorplan generation process into distinct stages: real-time preview generation using lightweight processing, and final floorplan generation using comprehensive post processing techniques. This segmentation allows the system to provide immediate feedback during scanning while applying accurate but time-consuming algorithms only when needed, resolving the time-accuracy tradeoff.
Solution Approach 2:
The system applies partial processing techniques for real-time preview (sufficient for immediate feedback) and reserves excessive/comprehensive processing for final floorplan generation (achieving maximum accuracy). This partial action during scanning and full action during postprocessing resolves the contradiction between speed and precision.
3Reliability
If semantic segmentation and labeling of 3D point clouds is performed, then the identification of structural elements is improved, but the computational complexity and processing requirements increase
Solution Approach 1:
The system performs semantic segmentation and labeling of 3D point clouds as a preliminary step before floorplan generation. This preliminary action extracts and organizes semantic information in advance, improving the reliability of structural element identification while allowing the main floorplan generation algorithm to work with pre-processed, labeled data, thereby managing computational complexity.
4Measurement precision
If multiple neural networks and algorithms are used for detecting wall structures and objects, then the detection precision is improved, but the processing time and computational resources increase
Solution Approach 1:
The patent segments the detection process into specialized neural networks for different tasks (wall structure detection, object detection, semantic labeling). Each network is optimized for its specific function, improving detection precision through specialized processing while managing computational resources by avoiding a single monolithic complex system.
Solution Approach 2:
The system applies multiple neural networks and algorithms selectively: lighter models for real-time preview and more comprehensive models for final floorplan generation. This partial application during scanning and full application during postprocessing improves detection precision while managing processing efficiency through staged computation.
Data Source
AI summary
Various implementations disclosed herein include devices, systems, and methods that generate floorplans and measurements using a three-dimensional (3D) representation of a physical environment generated based on sensor data.