Indoor Positioning via Visual Feature Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing indoor location technologies face limitations in closed environments due to GPS signal shielding and low precision, requiring pre-installed devices and complex algorithms, and are inaccessible to visually impaired individuals.
Innovation Solution
A method using artificial intelligence that creates a dataset of environment images associated with labels, trains a neural network to identify user positions based on received images, and communicates the location to the user without prior processing, allowing for real-time positioning on simple electronic devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS technology is used for outdoor positioning, then positioning coverage is improved, but positioning precision deteriorates in indoor environments due to signal shielding
Solution Approach 1:
The patent introduces visual features in the environment as intermediary elements for positioning. Instead of directly using GPS signals which are blocked indoors, the system captures images and extracts visual features (edges, corners, textures) that serve as mediators to determine position indirectly through comparison with a pre-built database of environmental features.
2Loss of information
If maps are positioned in determinate zones to indicate user location, then position indication is improved, but user accessibility deteriorates when users are distant from maps or in unstructured environments
Solution Approach 1:
The system enables self-service positioning where the user's mobile device automatically captures images, extracts visual features, compares them with the database, and determines position without requiring the user to manually search for or interact with physical maps. The positioning information comes to the user through the device rather than requiring the user to go to a map.
Solution Approach 2:
The visual feature-based positioning system is universally applicable across both structured environments (with regular layouts) and unstructured environments (irregular layouts, natural settings). The same image processing and feature matching methodology works in diverse settings without requiring environment-specific configuration or pre-installed infrastructure.
3Loss of information
If code-based area identification is used in parking lots, then area identification is improved, but user orientation capability deteriorates as users must memorize codes
Solution Approach 1:
The patent replaces the mechanical/cognitive system of memorizing area codes with an automated visual recognition system. Instead of users mentally processing and remembering codes, the mobile device automatically captures images, extracts visual features, matches them with the database, and provides position information, substituting human cognitive effort with automated image processing.
4Measurement precision
If artificial intelligence is used to recognize reference objects in images for positioning, then positioning accuracy is improved, but computational burden increases
Solution Approach 1:
The patent segments the complex task of positioning into distinct stages: image capture, visual feature extraction (edges, corners, textures), feature database comparison, and position determination. This segmentation allows each stage to be optimized independently, reducing the computational burden at each step while maintaining overall positioning accuracy.
Solution Approach 2:
The system extracts only the essential visual features needed for positioning (edges, corners, textures) rather than performing complete object recognition or processing all image data. This partial action approach focuses computational resources on the most discriminative features for location identification, reducing overall computational complexity while maintaining sufficient accuracy.
Data Source
AI summary
The invention concerns a method for locating users in a determinate indoor environment by means of artificial intelligence, comprising the creation and storage of a data set of images associated with a position of acquisition in said environment, the training of at least one neural network of a processing unit in order to teach it to recognize and determine a relationship between image and position, and the processing of an image received from a user in order to recognize and identify the position of acquisition of the image.