Automatic 2D Floorplan Annotation via Neural Network Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing measurement systems for creating digital 2D floorplans of buildings are complex, require specialized personnel, and are time-consuming, especially in sensitive situations like crime or accident scene investigations, due to the need for manual annotation and the bulkiness of scanning equipment.
Innovation Solution
A portable system that uses a scanner and vision-based sensors to automatically generate and annotate 2D floorplans by applying neural networks for room segmentation and labeling, allowing for accurate room identification and annotation without human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing measurement systems are used to create digital 2D floorplans, then the floorplans can be generated, but the process is time-consuming and requires manual annotation by specialized personnel
Solution Approach 1:
The system performs automatic room segmentation and annotation without human intervention. The neural network automatically processes the point cloud data to identify walls, doors, windows, and room types, eliminating the need for specialized personnel to manually annotate the floorplans while maintaining measurement precision
Solution Approach 2:
The patent replaces the manual mechanical annotation process with an automated computational system. Instead of specialized personnel manually labeling features, a neural network processes the scanning data to automatically identify and annotate architectural elements, significantly reducing annotation time while preserving accuracy
2Measurement precision
If existing scanning equipment is used, then coordinate measurements can be obtained, but the equipment is bulky and requires specialized personnel
Solution Approach 1:
The system uses a portable computing device with integrated sensors that can perform multiple functions: capturing images, collecting point cloud data, and processing floorplan generation. This multi-functional device eliminates the need for specialized personnel and bulky dedicated scanning equipment while maintaining measurement precision
Solution Approach 2:
The patent replaces bulky mechanical scanning equipment with a portable computing device using vision-based sensors and cameras. The system uses computational photography and image processing to obtain coordinate measurements, substituting complex mechanical scanning systems with more portable and easier-to-operate devices
3Loss of information
If manual labeling is performed to add context to digital 2D floorplans, then accurate annotations can be achieved, but the process is time-consuming
Solution Approach 1:
The neural network automatically extracts context information from the scanning data and generates annotations for room types, architectural features, and spatial relationships. The system serves itself by automatically processing the data to produce complete context information without requiring manual labeling, reducing labeling time while maintaining information completeness
Solution Approach 2:
The system performs preliminary processing of the point cloud data to identify walls, doors, windows, and room characteristics before final floorplan generation. This preliminary automated analysis prepares the context information in advance, eliminating the need for time-consuming manual labeling while ensuring complete and accurate context is captured
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 system enables rapid and accurate generation of segmented and annotated 2D maps, reducing the time and complexity of documenting environments, and improving the precision of room identification and annotation.
Implementation Method 1
a scanning device that determines coordinates of surfaces in the environment by emitting a light and capturing a reflection to determine a distance
Implementation Method 2
a vision-based sensor to facilitate automatic room segmentation for 2D floorplan annotation
Data Source
AI summary
A system for generating an automatically segmented and annotated two-dimensional (2D) map of an environment includes processors coupled to a scanner to convert a 2D map from the scanner into a 2D image. Further, a mapping system categorizes a first set of pixels from the image into one of room-inside, room-outside, and noise by applying a trained neural network to the image. The mapping system further categorizes a first subset of pixels from the first set of pixels based on a room type if the first subset of pixels is categorized as room-inside. The mapping system also determines the room type of a second subset of pixels from the first set of pixels based on the first subset of pixels by using a flooding algorithm. The mapping system further annotates a portion of the 2D map to identify the room type based on the pixels corresponding to the portion.