Rotation Angle Determination via Feature Point Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection algorithms are unable to accurately determine the rotation angle of objects on a conveyor belt in real time, leading to potential damage and limiting their application in factory environments due to significant errors.
Innovation Solution
A processor-based system that acquires images, extracts feature points, matches them with pre-stored templates, filters out data outliers using the Z-score algorithm, and calculates the rotation angle by comparing angle information to determine the object's orientation with high accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional object detection algorithms are used, then processing speed is improved, but measurement precision of rotation angle deteriorates
Solution Approach 1:
The patent segments the object detection task into multiple independent feature points distributed across the object surface. Each feature point is detected and matched separately, allowing parallel processing that maintains high speed while the collective geometric relationships provide precise rotation angle measurement through statistical aggregation of multiple measurements.
Solution Approach 2:
The patent changes the measurement parameter from single-point detection to multi-point feature detection. By extracting and matching multiple feature points (e.g., corners, edges) and computing their angular relationships, the system achieves both real-time processing capability and high rotation angle precision through geometric consensus of multiple features.
2Device complexity
If image moment method is used for rotation angle detection, then processing complexity is reduced, but measurement precision deteriorates significantly
Solution Approach 1:
The patent replaces the mathematical image moment computation with a geometric feature matching approach. Instead of computing complex moment integrals over the entire image, the system detects discrete feature points, matches them between template and target images, and determines rotation angle from the geometric transformation required to align corresponding features, achieving both simplicity and precision.
Solution Approach 2:
The patent uses a pre-stored template containing known feature point positions as a reference copy. By comparing the detected feature points in the target image with the template copy, the system can accurately determine rotation angle through feature correspondence matching, avoiding the need for complex moment calculations while maintaining high precision.
3Device complexity
If conventional detection methods are used, then device simplicity is maintained, but reliability of object handling deteriorates due to significant errors
Solution Approach 1:
The patent implements a feedback mechanism where multiple feature point measurements are aggregated to determine the rotation angle. The system computes individual angle estimates from each matched feature point pair, then applies statistical processing (such as median or mean) to produce a final robust angle measurement that compensates for individual feature detection errors, thereby improving reliability while maintaining practical system complexity.
Data Source
AI summary
The present disclosure relates to a determining method performed by a determining device, the method comprising: acquiring an image; extracting a first feature point from the image; matching the first feature point of an object included in the image with a second feature point of a template by comparing the extracted first feature point with the second feature point within the template in which the second feature point for a predetermined object is previously stored; and determining a rotation angle of the object included in the image by comparing angle information of the first feature point with angle information of the second feature point.