Non-rigid Object Tracking via Region Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting and tracking non-rigid, slightly textured objects in video streams, such as hands, are inefficient due to reliance on complex hardware and lack of precision in augmented reality applications, especially in real-time environments with varying lighting and noise.
Innovation Solution
A method involving the reception of image streams, identification of regions of interest, comparison of these regions between frames to determine movement, and triggering actions based on detected changes, which can handle low-quality images and limited processing resources, allowing for the tracking of deformable objects like hands without specific color dependency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex hardware and specialized sensors are used for tracking non-rigid objects, then tracking precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces complex mechanical sensing hardware with a software-based image processing system. The tracking is achieved through algorithmic analysis of video frames using region-of-interest comparison and feature point matching, eliminating the need for specialized sensors or complex hardware setups while maintaining tracking capability for non-rigid objects like hands
Solution Approach 2:
The patent creates a virtual model of the non-rigid object by identifying and tracking feature points across video frames. Instead of physically measuring the object, the system creates a digital representation through image processing that can be manipulated and tracked computationally, achieving precise tracking through software rather than hardware
2Reliability
If traditional object tracking algorithms are used, then rigid object tracking is effective, but tracking of non-rigid, slightly textured objects fails
Solution Approach 1:
The patent applies different processing strategies to different parts of the image. Regions of interest are identified and processed with specific algorithms suited for non-rigid objects, while other parts of the image receive standard processing. This localized approach allows the system to handle the unique characteristics of non-rigid, slightly textured objects like hands without compromising overall system performance
Solution Approach 2:
The patent implements dynamic tracking that adapts to the deformation of non-rigid objects. The system continuously updates the region of interest and feature point locations based on object movement and shape changes, allowing reliable tracking of objects that change their geometry over time rather than assuming rigid structure
3Measurement precision
If high-quality image sensors and complex processing are used, then tracking accuracy is improved, but computational demands and processing time increase
Solution Approach 1:
The patent divides the image processing task into segments by identifying and focusing only on regions of interest containing the non-rigid object. Instead of processing the entire image frame, the system isolates specific areas for detailed analysis, significantly reducing computational requirements while maintaining tracking accuracy for the target object
Solution Approach 2:
The patent applies partial processing by focusing computational resources only on the necessary regions of the image. The system performs full-detail processing only on regions containing the object of interest, while other areas receive minimal or no processing, achieving accurate tracking with reduced overall computational power requirements
4Measurement precision
If skin color detection and motion history analysis are used, then hand detection is improved, but the system becomes dependent on specific color ranges and lighting conditions
Solution Approach 1:
The patent changes the detection parameters from color-based to feature-point-based tracking. Instead of relying on skin color ranges that are sensitive to lighting variations, the system identifies and tracks geometric feature points that remain detectable across different lighting conditions, making the tracking robust to environmental changes while maintaining precision
Data Source
Figure 1a~1b
Figure 2a~2d
Figure 3~5
AI summary
The method involves receiving two images from an image sensor, and identifying two interest regions respectively in the images, where one region corresponds to one image and another region corresponds to the former region. The regions are compared (440), and an interest mask characterizing a variation of a characteristic of points in the regions is determined (445). Movement of an object is determined from the mask, where the object is partially represented in one of the regions. The movement is analyzed (470), and a preset action is triggered or non-triggered based on the movement analysis. Independent claims are also included for the following: (1) a computer program comprising a set of instructions for performing an interaction detecting method (2) a device comprising units for implemented steps of an interaction detecting method.