Feature Descriptor Matching for Vehicle Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Global positioning systems (GPS) face challenges in determining vehicle location due to inclement weather, urban regions, and occlusions, which obstruct the view of GPS satellites, leading to difficulties in vehicle localization.
Innovation Solution
A system for feature descriptor matching using a memory, feature detector, and descriptor extractor, which learns local feature descriptors from input images based on a trained model and generates geometric transformations between images using a convolutional neural network (CNN), enabling effective image matching and localization even in challenging conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS satellites are used for vehicle localization, then location determination is achieved under clear sky conditions, but localization fails in occluded environments such as urban regions, mountainous terrain, or inclement weather
Solution Approach 1:
The patent introduces image-based feature descriptor matching as an intermediary system between the vehicle and the environment. Instead of directly relying on GPS satellites, the system captures images, extracts features, and matches them against a database to determine location. This intermediary approach allows localization to function in occluded environments where direct satellite communication is blocked.
Solution Approach 2:
The system creates a digital copy of the physical environment through image capture and feature extraction. By storing images and their extracted features in a database, the system establishes a digital representation that can be matched against real-time images to determine vehicle location without requiring direct satellite signals.
2Ease of operation
If traditional GPS methods are used, then location determination is simple under clear conditions, but the system becomes non-functional when satellite view is obstructed
Solution Approach 1:
The patent creates a universal localization system that can operate across multiple environments and conditions. The image-based feature matching system serves as a multi-functional solution that works both in open areas where GPS would normally function and in occluded environments where traditional GPS fails, providing consistent location determination regardless of satellite visibility.
3Measurement precision
If feature descriptor matching with CNN is implemented, then image matching accuracy improves in challenging conditions, but computational complexity and processing time increase
Solution Approach 1:
The system performs preliminary actions by pre-processing images and extracting features during offline phases. Images are captured, features are extracted and stored in a database before actual localization needs occur. This preliminary preparation reduces the computational burden during real-time operation, as the system only needs to perform feature matching rather than full image processing.
Solution Approach 2:
The patent segments the image processing task into distinct components: feature detection, feature description, and feature matching. By dividing the complex task of image comparison into these separate stages, the system can apply specialized algorithms to each segment and optimize processing efficiency while maintaining high matching accuracy.
Data Source
AI summary
Feature descriptor matching described herein may include receiving a first input image and a second input image. A feature detector may detect features from the first and second input images. A descriptor extractor may learn local feature descriptors from the features of the first and second input images based on a feature descriptor matching model trained using a ground truth data set. The descriptor extractor may determine a listwise mean average precision (mAP) rank of a pool of candidate image patches from the second input image with respect to a queried image patch from the first input image based on the feature descriptor matching model, the first set of local feature descriptors, and the second set of local feature descriptors. The descriptor matcher may generate a geometric transformation between the first input image and the second input image based on the listwise mAP and a convolutional neural network.