Semantic Visual Landmarks for Mobile Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional navigation systems for mobile platforms face challenges in achieving sub-meter accuracy, especially in GPS-denied environments or where visual landmarks are occluded, due to reliance on costly sensors or complex algorithms that fail to effectively utilize semantic information for navigation.
Innovation Solution
The system employs semantic visual features to differentiate between salient and less-important features, using only the salient features as constraints in a navigation inference engine to generate precise navigation information, leveraging deep learning for semantic segmentation and factor graphs to improve mapping and tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional navigation systems use costly sensors such as differential GPS or laser scanners to achieve sub-meter accuracy, then navigation precision is improved, but device cost and complexity increase
Solution Approach 1:
The patent extracts and utilizes semantic information from visual features obtained by standard cameras, separating the useful semantic content from the raw visual data. This allows the system to achieve high navigation accuracy without relying on costly specialized sensors like laser scanners or differential GPS, effectively extracting the essential navigation-critical information from inexpensive camera inputs
Solution Approach 2:
The patent replaces mechanical/optical sensing systems (laser scanners, differential GPS) with a vision-based system that uses standard cameras combined with semantic segmentation algorithms. The computational processing of semantic information substitutes for the physical measurement capabilities of expensive sensors, achieving comparable or superior accuracy through software-based feature extraction and selection
2Measurement precision
If conventional navigation systems process all visual features to ensure comprehensive coverage, then measurement completeness is improved, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing visual features into distinct semantic categories (e.g., road markings, buildings, vegetation, sky) using semantic segmentation algorithms. This allows the system to process and evaluate features category-by-category, identifying which semantic types provide reliable navigation constraints and which do not, thereby reducing unnecessary computational processing of irrelevant features
Solution Approach 2:
The patent applies local quality by treating different semantic feature types differently based on their reliability for navigation. Instead of uniformly processing all visual features, the system selectively applies constraint-based filtering to specific semantic categories (such as using road markings and building structures as reliable constraints while excluding sky and vegetation), optimizing computational resources for the most valuable features
3Measurement precision
If the system uses all detected visual features for navigation constraints, then feature coverage is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary action by conducting semantic segmentation and feature classification before the navigation constraint optimization step. By pre-categorizing features into semantic types and pre-identifying which categories are reliable for navigation constraints, the system eliminates the need for time-consuming evaluation of all features during real-time navigation processing, significantly reducing processing time while maintaining accuracy
4Adaptability or versatility
If navigation systems operate in GPS-denied environments with occluded landmarks, then adaptability is improved, but measurement reliability deteriorates
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting which semantic feature types are used as navigation constraints based on environmental conditions. In GPS-denied or occluded environments, the system automatically shifts reliance to alternative semantic categories (such as transitioning from GPS-based constraints to visual landmark constraints, or from open-space features to structure-based features), maintaining navigation reliability by adapting the constraint set to available environmental information
Data Source
AI summary
Techniques are disclosed for improving navigation accuracy for a mobile platform. In one example, a navigation system comprises an image sensor that generates a plurality of images, each image comprising one or more features. A computation engine executing on one or more processors of the navigation system processes each image of the plurality of images to determine a semantic class of each feature of the one or more features of the image. The computation engine determines, for each feature of the one or more features of each image and based on the semantic class of the feature, whether to include the feature as a constraint in a navigation inference engine. The computation engine generates, based at least on features of the one or more features included as constraints in the navigation inference engine, navigation information. The computation engine outputs the navigation information to improve navigation accuracy for the mobile platform.