Self Localization Using Parallel Projection Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing self-localization methods for mobile sensors face challenges in maintaining accuracy and computational efficiency, particularly due to errors from image processing and lens non-linearity, especially when multiple reference points are involved, which can lead to inaccuracies in determining the orientation and location of the mobile sensor.
Innovation Solution
A method using a parallel projection model that projects reference objects onto virtual viewable planes, calculating distances and orientations using a zoom factor, and compensating for lens distortion with a calibration table, allowing for iterative calculations to minimize errors and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple reference points are used for self-localization, then the accuracy of determining orientation and location is improved, but computational complexity and error propagation increase
Solution Approach 1:
The patent segments the localization problem into two distinct phases: (1) calculating initial orientation and location using multiple reference points, and (2) compensating for lens non-linearity errors using calibration data. This segmentation allows the system to leverage multiple reference points for accuracy while managing computational complexity through structured error compensation rather than complex real-time calculations.
Solution Approach 2:
The patent performs preliminary error compensation by pre-calculating and storing lens non-linearity correction data in a calibration table during a separate calibration phase. This preliminary action eliminates the need for complex real-time error calculations during actual localization, reducing computational complexity while maintaining high accuracy when multiple reference points are used.
2Loss of information
If image processing is performed to extract features from multiple reference objects, then localization information is obtained, but errors from image processing and lens non-linearity increase
Solution Approach 1:
The patent implements a feedback mechanism where calibration data obtained from preliminary measurements is used to correct errors in subsequent localization measurements. The system uses the known positions of multiple reference points to calculate initial localization, then applies feedback from pre-stored calibration data to compensate for lens non-linearity errors, thereby improving accuracy despite image processing errors.
Solution Approach 2:
The patent converts the harmful effect of lens non-linearity into a beneficial correction process. By deliberately measuring and characterizing lens distortion during calibration using multiple reference points, the system creates compensation data that can be applied to correct future measurements. The same multiple reference points that could propagate errors are instead used to generate correction factors that eliminate errors.
3Productivity
If a pinhole camera model is used with single landmark, then real-time performance is achieved, but only one correspondence can be established limiting localization capability
Solution Approach 1:
The patent makes the localization system universal by enabling it to handle both single-landmark and multi-landmark scenarios using the same parallel projection model and error compensation framework. The system can process multiple reference points simultaneously while maintaining real-time performance through efficient calibration-based error correction, thus achieving both speed and versatility.
Solution Approach 2:
The patent changes the fundamental parameter of the camera model from the traditional pinhole model to a parallel projection model that incorporates lens non-linearity compensation. This parameter change allows the system to process multiple correspondences from multiple reference points while maintaining computational efficiency through pre-calibrated correction factors, achieving both enhanced capability and real-time performance.
Data Source
AI summary
A method of recognizing a self location of an image acquisition device by acquiring an image of two or more reference objects is provided. The method of the present invention comprises setting an actual camera plane, two or more reference object planes, and two or more virtual viewable planes located between the actual camera plane and the reference object planes; projecting the reference objects to a corresponding one of the virtual viewable planes; calculating a distance between a viewing axis and the reference objects and a distance between the viewing axis and images on the actual camera plane, the images corresponding to the reference objects; and sensing the self location of the image acquisition device by using an orientation and a zoom factor of the image acquisition device and coordinates of the reference objects, wherein the zoom factor is a ratio of a length of the reference object plane and a distance between the reference object plane and the virtual viewable plane, and the actual camera plane, the virtual viewable plane, and the reference object plane are perpendicular to the viewing axis.