Feature Balancing for Head-Mounted Display Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computer vision algorithms struggle to effectively select keypoints from images with varying lighting conditions, often prioritizing background features over foreground objects, leading to insufficient keypoint selection and reduced performance in object recognition and tracking, especially in outdoor environments with bright lighting and high contrast.
Innovation Solution
A balanced algorithm that divides the image into cells and selects keypoints based on both location diversity and quality, ensuring an even distribution of feature points across the image, thereby improving object recognition and tracking accuracy without excluding strong features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional keypoint selection algorithms are used, then keypoints with highest feature strength are selected, but background keypoints are selected disproportionately over foreground object keypoints
Solution Approach 1:
The image is divided into multiple regions or zones, and keypoint selection is performed separately for each region. This segmentation allows the algorithm to control the distribution of selected keypoints across different spatial areas, ensuring that foreground objects receive adequate keypoint selection even when their feature strength is lower than background areas.
2Quantity of substance
If image is divided into regions with equal keypoint selection, then even distribution of keypoints is achieved, but good keypoints are excluded and poor keypoints are selected
Solution Approach 1:
Different regions of the image are assigned different keypoint selection criteria based on their local characteristics. Regions containing foreground objects are given higher priority or different selection thresholds compared to background regions, allowing the algorithm to maintain both uniform distribution and high keypoint quality by adapting to local image properties.
3Device complexity
If more keypoints are selected from background, then algorithm simplicity is maintained, but object recognition and tracking performance is reduced
Solution Approach 1:
The algorithm performs preliminary identification of foreground objects and their spatial locations before keypoint selection. This preliminary action allows subsequent keypoint selection to be biased toward foreground regions, ensuring adequate keypoint coverage on objects of interest while maintaining a relatively simple overall algorithm structure.
Data Source
AI summary
A head-mounted display includes a camera that obtains an image of an object within a field of view. The head-mounted display further includes a processor configured to determine a plurality of feature points from the image and calculate a feature strength for each of the plurality of feature points. The processor is further configured to divide the image into a plurality of cells and select feature points having the highest feature strength from each cell and which have not yet been selected. The processor being further configured to detect and track the object within the field of view using the selected feature points.