Pose Estimation Using 3D Map Updates with Ray Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pose estimation technologies face challenges in accuracy due to insufficient feature points from images with low texture, scale issues, and ambiguity in three-dimensional maps, especially when the environment is dynamic or the recognition target is blocked.
Innovation Solution
An image processing apparatus and method that generates and updates a three-dimensional map using a shape model of the recognition target, registering landmarks based on feature points and ray vectors to improve pose estimation accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pose estimation uses object recognition with feature point matching, then pose can be estimated through dictionary matching, but accurate pose estimation becomes impossible when the recognition target includes less image texture or does not have a certain size in the input image
Solution Approach 1:
The patent transitions from two-dimensional image feature matching to three-dimensional spatial reasoning by introducing ray vectors that extend from camera positions through feature points into 3D space. This dimensional change allows the system to estimate pose even when 2D image features are insufficient due to low texture or scale issues, as the 3D geometric constraints provide additional information.
Solution Approach 2:
The patent introduces ray vectors as an intermediary element that connects 2D feature points with 3D spatial positions. These ray vectors serve as mediators between the image data and the pose estimation process, enabling the system to bridge the gap when direct 2D-2D matching fails due to insufficient image texture or scale.
2Measurement precision
If pose estimation uses environment recognition with three-dimensional map learning, then pose can be estimated from the learned map and detected feature points, but the three-dimensional map may collapse when the recognition target moves or is blocked, making pose estimation impossible
Solution Approach 1:
The patent makes the pose estimation system dynamic by allowing ray vectors to be continuously regenerated from current camera positions and detected feature points, rather than relying on a static pre-built three-dimensional map. This dynamic approach allows the system to adapt to moving targets and blocked views, as the ray vectors are computed in real-time based on current observations rather than assuming a fixed environmental map.
3Adaptability or versatility
If three-dimensional map is learned online in real time, then the map can be sequentially updated, but cumulative errors occur in pose estimation results due to ambiguity in the three-dimensional map
Solution Approach 1:
The patent performs preliminary action by establishing ray vectors from current camera positions and detected feature points before pose estimation is performed. By pre-computing these geometric constraints based on current observations, the system avoids accumulating errors that would occur from sequentially updating an ambiguous three-dimensional map, as each pose estimation is grounded in current rather than historical data.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present technology relates to an image processing apparatus and method and a program which enable more accurate estimation of a pose. A matching unit performs matching processing on the basis of a detection result of a feature point from an input image and a three-dimensional map indicating a position of a landmark in real space, and a pose estimating unit estimates a pose on the basis of a result of the matching processing. A three-dimensional map updating unit updates the three-dimensional map on the basis of a three-dimensional shape model indicating a position and a shape of a recognition target, a detection result of the feature point from the input image and an estimation result of the pose so that only a portion of the recognition target is registered as the landmark. The present technology can be applied to a wearable information terminal apparatus.