2D Floor Plan Generation From 3D Primitive Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for creating two-dimensional floor plans require manual input and are often imprecise, and rely on depth sensors for 3D point cloud collection, which is time-consuming and inefficient.
Innovation Solution
A method utilizing a camera and machine learning model to detect 2D primitives, register and filter them, cluster similar primitives, and perform graph-based searching to generate a 2D plan without the need for depth sensors, reducing manual input and processing requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to create two-dimensional floor plans, then manual input is required, but the process is time-consuming and imprecise
Solution Approach 1:
The patent replaces manual mechanical input methods with an automated computer vision system that uses camera images and machine learning models to detect 2D primitives and generate floor plans automatically, eliminating the need for manual drawing input while improving precision through algorithmic processing
Solution Approach 2:
The patent creates a digital copy of the physical space by capturing images with a camera and using machine learning to extract geometric primitives, transforming real-world spatial information into a digital representation without requiring manual measurement or drawing
2Reliability
If depth sensors are used for 3D point cloud collection, then three-dimensional data can be obtained, but the process is time-consuming and reduces efficiency
Solution Approach 1:
The patent extracts only the necessary 2D geometric primitives (corners, lines, walls) directly from 2D camera images using machine learning, eliminating the need to capture and process complete 3D point clouds, thus obtaining sufficient spatial information more efficiently
Solution Approach 2:
The patent works entirely in 2D image space to detect primitives and generate floor plans, avoiding the need to convert to 3D space and then back to 2D, thereby simplifying the processing pipeline and improving efficiency while maintaining accuracy
3Ease of manufacture
If manual input methods are used, then floor plans can be created, but the accuracy and precision are compromised
Solution Approach 1:
The system performs automatic detection and processing of spatial features through machine learning algorithms that self-correct and refine primitive detection, eliminating reliance on manual operator skill while maintaining ease of use through automated workflow
Data Source
AI summary
Techniques for two dimensional and three dimensional primitive processing to determine a two dimensional plan are described herein. In an example, first two-dimensional (2D) primitives associated with a room are determined based on an image frame defined in an image space as an input to a machine learning model. The first 2D primitives are of different types. A first three-dimensional (3D) primitive corresponding to the first 2D primitives is determined based on camera pose data associated with the image frame and the first 2D primitives. A 2D projection of the first 3D primitive onto a projection plane is determined based on 2D contour of the room defined in the projection plane. The 2D projection includes second 2D primitives. A 2D plan associated with the room is generated based on the second 2D primitives.