UAV Indoor Navigation Using Visual Feature Point Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Unmanned aerial vehicles (UAVs) face challenges in indoor navigation due to GPS interference and the inability to create high-precision maps in complex indoor environments, where existing GPS positioning fails and SLAM algorithms only produce sparse maps with significant errors.
Innovation Solution
An autonomous navigation method and system for UAVs that involves collecting video data, identifying feature points, generating flight paths, and creating high-precision map models using a combination of video analysis and laser scanning data, allowing for autonomous navigation and precise mapping in indoor environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS positioning technology is used for spatial positioning, then positioning can be achieved in open environments, but positioning fails in complex indoor scenes due to building blocking
Solution Approach 1:
The patent introduces visual feature points and map models as intermediary elements between the UAV and the environment. Instead of relying directly on GPS signals that are blocked by buildings, the system uses visual features extracted from camera images and matches them against pre-built map models to determine position and orientation, enabling reliable indoor positioning without direct GPS access
Solution Approach 2:
The patent replaces the GPS satellite-based electromagnetic positioning system with a visual-based positioning system using camera sensors and image processing. This substitution allows the system to operate in indoor environments where GPS signals are unavailable, trading the simplicity of GPS for the adaptability of visual processing
2Extent of automation
If SLAM algorithm is used to establish map model, then autonomous navigation can be achieved, but the map model is sparse with error greater than 5% of whole environment scale
Solution Approach 1:
The patent merges multiple data sources including visual feature information from cameras, depth information from distance measurement devices, and odometry data from motion sensors to construct the map model. This combination of multiple sensing modalities compensates for the limitations of individual methods and achieves both automation and high precision with error less than 1% of the whole environment scale
Solution Approach 2:
The patent creates a composite mapping system that integrates different types of spatial data (2D visual features, 3D depth information, and motion trajectories) into a unified high-precision map model. This composite approach combines the advantages of various sensing methods to achieve superior accuracy and completeness compared to single-method SLAM
3Measurement precision
If laser scanning system is used to establish high precision map model with error less than 1%, then map precision is improved, but the system is not suitable for UAV flying in inner environment
Solution Approach 1:
The patent segments the mapping function across multiple lightweight sensors on the UAV (camera, distance measurement device, motion sensors) rather than using a single complex laser scanning system. This segmentation allows high-precision mapping to be achieved through coordinated data from multiple simpler components, making the system suitable for UAV deployment in indoor environments
Data Source
AI summary
The present disclosure provides an autonomous navigation method and system, and a map modeling method and system. The autonomous navigation method comprises: controlling an unmanned aerial vehicle to take off, and collecting videos of a scene corresponding to the unmanned aerial vehicle at each collection time point; obtaining feature points in the videos of the scene corresponding to each collection time point; generating a flight path of the unmanned aerial vehicle according to the feature points in the videos of the scene corresponding to each collection time point; generating a first map model according to the flight path of the unmanned aerial vehicle and the videos of the scene corresponding to each collection time point; and carrying out autonomous navigation on the unmanned aerial vehicle according to the first map model.