SLAM Landmark Selection for Reliable Mapping of Movable Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SLAM techniques face challenges in generating reliable maps and maintaining accurate camera positioning due to the inclusion of non-stationary landmarks, leading to potential tracking failures.
Innovation Solution
A method and system for selective landmark point selection using machine learning and computer vision techniques to identify stationary objects for mapping, while tracking movable objects separately, thereby enhancing map reliability and processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all detected feature points are used for mapping without classification, then the mapping process is simple and fast, but the reliability of the map deteriorates due to inclusion of non-stationary landmarks
Solution Approach 1:
The patent segments the set of detected feature points into two distinct categories: stationary landmarks and non-stationary landmarks. This segmentation is achieved by classifying each feature point based on its temporal stability across multiple image frames. By dividing the landmarks into stationary and non-stationary groups, the system can selectively use only stationary landmarks for mapping, thereby improving map reliability while managing complexity through automated classification.
Solution Approach 2:
The patent applies local quality by treating different feature points differently based on their individual properties. Each feature point is evaluated for its stationarity, and those deemed stationary are selected for mapping while non-stationary ones are excluded. This selective approach ensures that only high-quality, reliable landmarks contribute to the map, improving overall map reliability without requiring uniform treatment of all features.
2Measurement precision
If non-stationary landmarks are included in the map, then more features are available for localization, but tracking accuracy deteriorates due to movements of these landmarks
Solution Approach 1:
The patent extracts and removes non-stationary landmarks from the set of features used for mapping and localization. By identifying and excluding these movable features, the system ensures that only stable, stationary landmarks remain for constructing the map and performing localization. This extraction process maintains localization accuracy by preventing erroneous position estimates that would result from tracking movable objects, while the system still benefits from having a comprehensive set of landmarks through the inclusion of stationary ones.
3Productivity
If all detected features are processed for mapping, then the mapping is comprehensive, but processing time increases due to the volume of data
Solution Approach 1:
The patent extracts and excludes non-stationary landmarks from the processing pipeline, thereby reducing the quantity of feature points that require full mapping processing. By removing these problematic features early in the process, the system maintains comprehensive mapping of the relevant stationary environment while significantly reducing computational load and processing time.
Solution Approach 2:
The patent segments the feature points into stationary and non-stationary categories, then processes only the stationary subset for mapping. This segmentation allows the system to maintain high productivity by avoiding unnecessary processing of movable features, while still achieving comprehensive coverage of the static environment through efficient processing of the reduced feature set.
Data Source
Figure 1~2
Figure 3a~4
Figure 5~6
AI summary
A method of environment mapping comprising the steps of receiving a plurality of images from respective different viewpoints; using a simultaneous localisation and mapping 'SLAM' process, detect feature points in the plurality of images; classifying one or more objects within the plurality of images as being objects of interest; determining the position of the or each object of interest using feature points corresponding to the objects of interest in the images; and providing an indication of the position of one or more objects of interest to a user of a virtual display; and wherein the step of determining the position of the or each object of interest using their feature points is repeated, more frequently than for other feature points in the images.