Camera Pose Tracking Using Points and Planes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SLAM systems face challenges in accurately tracking camera pose in scenes with limited texture and geometric variations, particularly in larger environments, and often fail in textureless or repetitive texture regions, due to reliance on point features and insufficient geometric features.
Innovation Solution
A method that uses both points and planes as primitive features for camera pose tracking, incorporating camera motion prediction and a prediction-and-correction framework, with relocalization and bundle adjustment processes to establish point-to-point and plane-to-plane correspondences, enabling fast and accurate registration and recovery from tracking failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If point features are used for camera pose tracking, then tracking can be performed in textured scenes, but tracking fails in textureless or repetitive texture regions
Solution Approach 1:
The patent combines point features and plane features into a unified SLAM framework. Point features provide robust tracking in textured regions, while plane features provide reliable tracking in textureless or repetitive texture regions. The system merges these two feature types to achieve reliable camera pose tracking across diverse scene types, resolving the contradiction between tracking reliability and adaptability to different scene types.
2Measurement precision
If ICP-based methods are used for depth camera tracking, then geometric variations can be exploited, but accurate registration fails when geometric variations are small
Solution Approach 1:
The patent merges point-based ICP methods with plane-based registration. In scenes with sufficient geometric variations, the system uses point-to-point or point-to-plane ICP correspondence for accurate pose estimation. In planar scenes with limited geometric variations, the system switches to plane-to-plane correspondence methods that exploit the dominant planar structures. This combination resolves the contradiction between measurement precision and adaptability to planar scenes.
3Adaptability or versatility
If only plane features are used for camera pose tracking, then degeneracy issues occur when field of view or sensor range is small
Solution Approach 1:
The patent combines point features and plane features to avoid degeneracy issues. When using plane features in scenes with small field of view or limited sensor range, the system supplements them with point features that provide additional geometric constraints. This combination ensures sufficient geometric diversity for reliable pose estimation while maintaining the ability to handle planar scenes, thus resolving the contradiction between adaptability and reliability.
4Reliability
If relocalization is performed for all frames, then tracking failures can be recovered from, but processing speed decreases to about three frames per second
Solution Approach 1:
The patent uses camera motion prediction to establish preliminary correspondences between consecutive frames before performing full relocalization. This preliminary action reduces the computational burden of relocalization by providing good initial estimates, allowing the system to maintain high processing speeds while still performing relocalization when needed to recover from tracking failures.
Solution Approach 2:
The patent implements a dynamic relocalization strategy where the frequency and intensity of relocalization operations are adjusted based on tracking confidence and scene characteristics. Instead of performing relocalization for all frames at full computational cost, the system dynamically decides when relocalization is necessary, maintaining reliability while optimizing processing speed.
Data Source
AI summary
A method registers data using a set of primitives including points and planes. First, the method selects a first set of primitives from the data in a first coordinate system, wherein the first set of primitives includes at least three primitives and at least one plane. A transformation is predicted from the first coordinate system to a second coordinate system. The first set of primitives is transformed to the second coordinate system using the transformation. A second set of primitives is determined according to the first set of primitives transformed to the second coordinate system. Then, the second coordinate system is registered with the first coordinate system using the first set of primitives in the first coordinate system and the second set of primitives in the second coordinate system. The registration can he used to track a pose of a camera acquiring the data.