Depth Pose Estimation Using Orthogonal Planes to Limit Drift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual inertial odometry (VIO) and simultaneous localization and mapping (SLAM) systems face challenges in accurately estimating the position and orientation of devices due to drift and complexity, particularly in environments with limited planar features.
Innovation Solution
A depth-based pose estimation method that utilizes orthogonal planes in a scene to determine normal vectors, orientation, and position data, combining this with filter-based and graph-based optimization to enhance accuracy and reduce drift.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If VIO and SLAM systems are used for pose estimation, then position and orientation can be estimated, but drift and complexity increase
Solution Approach 1:
The patent extracts and utilizes orthogonal planes (walls, floor, ceiling) from the environment as key geometric features for pose estimation. By focusing on these specific planar structures with known orthogonal relationships, the system simplifies the pose estimation problem while maintaining accuracy, avoiding the complexity of full VIO/SLAM processing
Solution Approach 2:
The patent introduces normal vectors as an intermediary element to represent the orientation of orthogonal planes. These normal vectors serve as mediators between the raw depth data and the final pose estimation, enabling efficient comparison with orientation candidates and reducing computational complexity
2Reliability
If traditional VIO methods are used, then pose can be estimated, but drift occurs in extended periods
Solution Approach 1:
The patent implements a feedback mechanism by comparing the estimated pose with the geometric constraints of orthogonal planes. The system continuously validates the pose estimation against the known orthogonal relationships in the environment, providing corrective feedback to reduce drift and improve long-term reliability
Solution Approach 2:
The patent performs preliminary identification and characterization of orthogonal planes in the environment before conducting pose estimation. By pre-establishing the geometric model of orthogonal structures and their normal vectors, the system prepares reference information that enables accurate and drift-free pose estimation without requiring complex real-time optimization
3Adaptability or versatility
If feature-based methods are used, then pose estimation can be performed, but performance degrades in environments with limited planar features
Solution Approach 1:
The patent creates a universal pose estimation method that works across different environments by leveraging the common geometric property of orthogonal planes found in most indoor spaces. The approach using normal vectors and orientation candidates provides a multi-functional solution that adapts to various scenarios (rooms, corridors, intersections) while maintaining consistent accuracy
Solution Approach 2:
The patent transforms the pose estimation problem from feature-based matching to parameter-based comparison by using normal vectors and orientation candidates. This parameter transformation enables the system to work effectively in environments with limited visual features by relying on the geometric parameters of orthogonal planes rather than detailed feature descriptions
Data Source
AI summary
A processor-implemented method with pose estimation includes: determining depth data by sensing a depth of a target scene comprising planes orthogonal to each other; determining normal vectors corresponding to depth points of the depth data; determining orientation data of a device by comparing the normal vectors to orientation candidates; and determining position data of the device based on distances from the device to the planes.