Binary Image Descriptor for Fast Visual Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for vehicle localization in autonomous driving, particularly in urban environments, face challenges due to GPS errors, inertial sensor drift, and the computational inefficiency of current feature vectors, which are not adapted for fast inference methods.
Innovation Solution
A binary image descriptor based on the histogram of vertical lines in urban environments is used to create a compact and efficient feature representation, allowing for fast and accurate localization by comparing query images with a database using Hamming distance, and employing neural networks for turn-classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS is used for vehicle localization, then location information can be obtained, but GPS suffers from drift due to accumulation of errors and cannot provide accurate positioning in urban environments
Solution Approach 1:
The patent introduces visual features (vertical lines in images) as an intermediary to establish correspondence between query images and database images, enabling location determination without relying on GPS signals. This intermediary approach allows the system to overcome GPS limitations in urban environments by using image-based localization instead.
2Measurement precision
If traditional feature vectors are used for image comparison, then location information can be extracted, but the computational cost is high and inference speed is slow
Solution Approach 1:
The patent changes the parameter representation from traditional real-valued feature vectors to binary descriptors. This parameter transformation enables the use of Hamming distance for comparison, which is computationally much faster than traditional distance metrics, thereby significantly improving inference speed while maintaining location accuracy.
Solution Approach 2:
The patent extracts only the essential visual feature (vertical lines) from images to create the binary descriptor, ignoring other non-essential features. This extraction of key elements reduces computational complexity while preserving the most important location-discriminative information.
3Measurement precision
If inertial sensors are used for localization, then position can be tracked, but the sensors are prone to drift
Solution Approach 1:
The patent creates a visual copy (binary descriptor) of the urban environment from images and compares it with stored database entries to determine location. This copying approach replaces inertial sensing with visual feature matching, eliminating drift accumulation while maintaining position tracking capability.
4Measurement precision
If wheel odometry is used for localization, then position can be estimated, but it is unreliable in terrains where wheels slip or sink
Solution Approach 1:
The patent replaces the mechanical wheel odometry system with a visual-based localization system using image processing and binary descriptor comparison. This substitution eliminates dependence on wheel-terrain interaction, making the system reliable in terrains where wheels slip or sink.
Data Source
AI summary
Systems and methods for an imaging system having a memory with a historical localization dictionary database having geo-located driving image sequences, such that each reference image is applied to a threshold to produce a binary representation. A sensor to acquire a sequence of input images of a dynamic scene. An encoder to determine, for each input image in the sequence, a histogram of each input image indicating a number of vertical edges at each bin of the input image and to threshold the histogram to produce a binary representation of the input image. A visual odometer to compare the binary representations of each input image and each reference image, by matching an input image against a reference image. Wherein the visual odometer determines a location of the input image based on a match between the input image and the reference image.