Indoor Positioning via Image Dimensionality Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional indoor positioning technologies face challenges with high computational loads and large map data sizes, leading to slow positioning speeds and increased burden on user devices and map data integration servers, especially in large indoor spaces.
Innovation Solution
A distributed indoor positioning system that includes a map data loading module, an image capturing module, an image dimensionality reduction module, and a comparison module, which captures indoor images, reduces their dimensionality, and uses the reduced data to determine user positions, thereby reducing the size of map data files and computational loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If large quantity of map data is loaded for positioning in large indoor spaces, then positioning accuracy is improved, but computational load and positioning speed deteriorate
Solution Approach 1:
The patent divides the large indoor space into multiple sub-regions or floors, and loads only the map data corresponding to the current location. This segmentation allows the system to maintain high positioning accuracy within each sub-region while keeping the loaded data volume small, thus avoiding the trade-off between accuracy and speed.
Solution Approach 2:
The patent performs preliminary dimensionality reduction processing on map data during the data preparation phase. By pre-compressing the map data into a compact representation, the system can load reduced data quickly while still enabling accurate positioning through the dimensionality-reduced feature space comparison.
2Measurement precision
If large quantity of map data is stored on user device, then positioning accuracy is improved, but device memory and computation resources are overwhelmed
Solution Approach 1:
The patent transforms the map data from high-dimensional original images to low-dimensional feature representations through dimensionality reduction. This parameter change in data representation maintains the essential positioning information while dramatically reducing the data size and computational complexity required for processing on user devices.
Solution Approach 2:
The patent creates a simplified copy of the map data in the form of dimensionality-reduced features. Instead of storing and processing complete high-resolution map images, the system uses these compact feature copies that retain positioning information but require minimal computational resources to handle.
3Measurement precision
If complete map data is processed for positioning, then positioning accuracy is improved, but energy consumption increases
Solution Approach 1:
The patent extracts only the essential positioning information from complete map data by performing dimensionality reduction. This extraction process removes redundant information while preserving the key features needed for accurate positioning, thereby reducing the computational energy required for processing.
4Productivity
If dimensionality reduction is applied to map data, then data size and computational load are reduced, but positioning accuracy may deteriorate
Solution Approach 1:
The patent carefully selects and adjusts the dimensionality reduction parameters to maintain positioning accuracy. By optimizing the reduction rate and feature selection, the system achieves a balance where data size is significantly reduced while the essential positioning information is preserved in the reduced feature space.
Data Source
AI summary
A distributed indoor positioning system and a method thereof are disclosed. In the method, indoor map data corresponding to a place where a user is located is loaded first, and an indoor image of the place where the user is located is captured to generate an image stream. Next, the image stream is compressed to reduce dimensionality thereof, so as to generate a dimensionality-reduced image. The indoor map data corresponding to the dimensionality-reduced image is obtained from the loaded indoor map data, and a position of the user is determined according to the obtained indoor map data.