Computer Vision Landmark Localization for GNSS-Denied Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation systems, particularly for unmanned vehicles and environments where satellite signals are unavailable or unreliable, face challenges in determining accurate position and orientation without relying on Global Navigation Satellite Systems (GNSS).
Innovation Solution
A method and system utilizing Computer Vision (CV) with Deep Convolutional Neural Networks (CNN) and Region Proposals to identify and classify landmarks, combined with a Monte Carlo Localization algorithm, allowing for position determination and navigation without GNSS, by using a camera to observe the environment and comparing observed data against a stored dataset of known objects and their parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image processing approaches are used to identify landmarks, then the system can recognize features in images, but the identified features are not sufficiently distinctive to enable position determination over wide areas
Solution Approach 1:
The patent segments the image processing task into multiple stages: initial feature detection using conventional computer vision, followed by classification using a neural network to identify distinctive landmark types. This segmentation allows the system to first find any features, then selectively identify which features are sufficiently distinctive for wide-area positioning, resolving the contradiction between feature recognizability and positional distinctiveness
Solution Approach 2:
The patent changes the parameter of feature classification from simple geometric detection to semantic classification using a neural network. By transforming the feature identification process from detecting basic shapes to classifying objects by type (e.g., building, tree, signpost), the system achieves both accurate position determination and wide area coverage, as different landmark types provide varying levels of distinctiveness across different spatial scales
2Adaptability or versatility
If an unrealistically large dataset of image information is used for matching, then more locations can be identified, but the system becomes impractical due to data requirements
Solution Approach 1:
The patent extracts only the essential distinguishing features and semantic classifications of landmarks rather than storing complete image datasets. By taking out only the critical identification parameters (landmark type, key visual characteristics) needed for position determination, the system achieves wide location coverage while maintaining practical data storage requirements, eliminating the need for unrealistically large datasets
Solution Approach 2:
The patent uses a neural network to learn and copy the essential characteristics of landmarks from training data, then applies this learned knowledge to identify landmarks in real-time without requiring access to the full training dataset. This copying approach allows the system to achieve broad location coverage using a compact model rather than storing vast amounts of reference image data
3Reliability
If satellite navigation systems are used, then position can be determined globally, but the system fails in areas with inadequate satellite coverage such as underground or multi-storey structures
Solution Approach 1:
The patent introduces an intermediary system between the vehicle and the environment for position determination. Instead of directly receiving satellite signals, the system uses a camera to capture images of the environment, processes these images through a neural network to identify landmarks, and determines position based on landmark recognition. This intermediary approach enables navigation in environments where satellite signals are unavailable, achieving both reliability and environmental adaptability
Data Source
AI summary
In relation to the field of vehicle navigation, we describe a method of determining a position of a subject (such as a vehicle, platform or target), comprising the steps of obtaining and storing an object dataset comprising object data indicative of one or more objects in an environment, including an indication of object parameters associated with the or each object, the object parameters including one or more of location, orientation, one or more dimensions, and a type associated with the object, obtaining environment data indicative of a region of the environment from a sensor associated with the subject, determining the presence of an observed object in the environment data, including determining one or more equivalent observed object parameters associated with the observed object, and determining the position of the subject based on a comparison of the observed object parameters with the equivalent object parameters of the objects in the object dataset.