Visual Self-Localization Key Frame Control for Sparse Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The self-localization process using ORB-SLAM can fail when the number of matching feature points between the camera's sight image and the map is small, leading to loss of self-position and self-posture, necessitating stabilization to maintain accurate navigation.
Innovation Solution
A self-localization device that acquires sight images from a camera, estimates position and posture using feature points, and registers key frames based on movement conditions such as distance and angular velocity, integrating LiDAR-SLAM for higher accuracy and stability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If key frames are registered only when a predetermined number of new feature points are observed, then the map can be kept compact, but key frames may become sparse during straight movement or turning, causing self-localization failure
Solution Approach 1:
The patent changes the registration criteria from a single parameter (number of new feature points) to multiple parameters including angular velocity threshold and distance threshold. By monitoring angular velocity, the system can detect turning movements and register key frames appropriately even when new feature points are scarce, thus preventing sparse key frame distribution and improving self-localization reliability.
2Speed
If the number of matching feature points is small, then processing speed increases, but self-localization fails and position/posture are lost
Solution Approach 1:
The patent introduces feedback mechanisms by continuously monitoring movement state (angular velocity and distance) and using this information to dynamically adjust key frame registration decisions. This feedback loop ensures that key frames are registered at appropriate intervals during movement, maintaining sufficient feature point matching and preventing loss of self-position and self-posture.
3Reliability
If key frames are registered frequently, then self-localization stability improves, but map size and processing load increase
Solution Approach 1:
The patent applies local quality by making key frame registration density adaptive to local movement conditions. During straight movement or turning where feature points are scarce, key frames are registered more frequently. During normal conditions with sufficient feature points, registration is less frequent. This localized adjustment optimizes both stability and map size.
Data Source
AI summary
The self-localization device includes: an acquisition unit configured to acquire a sight image from a camera which is mounted in a mobile object; an estimation unit configured to perform a self-localization process of estimating a position and a posture of the mobile object in a world coordinate system on the basis of positions of feature points in the sight image and a map on which positions of a plurality of feature points in the world coordinate system are recorded; and a registration unit configured to register the sight image as a key frame for adding a new feature point included in the sight image to the map. The registration unit is configured to register the sight image as the key frame when movement of the mobile object satisfies a condition which is prescribed for at least one of the position and the posture of the mobile object.