Inertial Measurement Assisted Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image object tracking technologies face challenges such as insufficient valid feature points when objects with similar appearances are present, environmental brightness issues, and image blurring, leading to interrupted tracking and determination errors.
Innovation Solution
An image object tracking method and apparatus that utilizes inertial measurement information to assist in feature point identification, combining feature point extraction algorithms with optical flow methods and inertial measurement data to improve tracking accuracy and efficiency by calculating the lens rotation angle and determining corresponding prediction points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature point extraction algorithm is used to determine target object, then object identification is achieved, but tracking accuracy deteriorates when objects with similar appearances are present
Solution Approach 1:
The patent introduces inertial measurement information as an intermediary to bridge the gap between feature point extraction and optical flow methods. The inertial measurement unit provides rotation angle data that serves as a mediator to disambiguate feature points with similar appearances, enabling reliable tracking even when visual features are insufficient or ambiguous.
2Productivity
If optical flow method is used to track feature points, then motion estimation is achieved, but tracking accuracy deteriorates when valid feature points are insufficient
Solution Approach 1:
The patent changes the parameter space from purely visual features to a combination of visual features and inertial measurement parameters. By incorporating rotation angle information from the inertial measurement unit, the system can maintain accurate feature point matching even when visual features are insufficient, blurred, or ambiguous, thus improving both efficiency and precision.
3Speed
If image capturing device moves rapidly causing blurred images, then motion capture is achieved, but tracking reliability deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the inertial measurement unit continuously provides rotation angle information that feeds back into the feature point matching process. This feedback allows the system to compensate for image blurring and maintain tracking reliability even when the image capturing device moves rapidly, as the inertial data provides a reference frame that remains stable despite motion.
4Duration of action of stationary object
If target object disappears in image area for short time, then image capture continuity is maintained, but tracking reliability deteriorates
Solution Approach 1:
The patent applies beforehand cushioning by using inertial measurement information to predict the position and state of the target object during brief disappearances. The inertial data acts as a cushion that maintains tracking reliability during transient interruptions, allowing the system to recover quickly and continuously track the object without significant loss of accuracy when it reappears in the image area.
Data Source
AI summary
An image object tracking method and apparatus are provided. The image object tracking method includes the steps of: determining a feature point of a target object in a first frame, determining a prediction point of the feature point in a second frame, calculating an estimated rotation angle of an image capturing device according to a distance between the coordinate of the prediction point and the coordinate of the feature point and a distance between the image capturing device and the target object, calculating a lens rotation angle of the image capturing device rotated from a time point that the first frame is captured to a time point that the second frame is captured according to a piece of inertial measurement information provided by an inertial measurement unit, and determining whether the prediction point corresponds to the feature point by comparing the estimated rotation angle and the lens rotation angle.