Image Processing Apparatus Stochastic Target Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately tracking targets in video frames, particularly in continuously captured videos, as they struggle to efficiently adjust focus and maintain clear frames due to stochastic movement prediction methods.
Innovation Solution
An image processing apparatus and method that calculates position offsets for candidate areas using a feature regression matrix and determines a final selected area by allocating weights based on texture values and similarities, employing sparse subspace clustering models to enhance target tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a particle filter is used to predict stochastic movement of each point for target tracking, then the tracking can handle stochastic movements, but the computational complexity and time consumption increase significantly
Solution Approach 1:
The patent segments the target tracking problem into multiple candidate areas within a bounding box, evaluating each area independently with different weights. This divides the complex stochastic movement prediction into smaller, manageable sub-problems that can be processed more efficiently than a full particle filter approach.
Solution Approach 2:
Instead of applying full particle filter computation to all possible target positions, the patent applies partial action by only evaluating a limited set of candidate areas with calculated weights. This reduces computational load while still capturing the essential stochastic movement characteristics through selective evaluation rather than exhaustive search.
2Productivity
If selective search is performed on a predetermined bounding box to track target, then the search area is limited and processing is faster, but the tracking accuracy decreases when target moves stochastically
Solution Approach 1:
The patent dynamically changes the parameters of candidate areas (positions, weights, sizes) based on the basis image and frame differences. By adjusting these parameters adaptively, the system maintains high processing speed through limited candidate areas while improving tracking accuracy through weight-based selection that accounts for stochastic movements.
Solution Approach 2:
The system uses feedback from frame differences and similarity calculations to adjust the weights and positions of candidate areas in subsequent frames. This feedback mechanism allows the bounding box search to adapt to stochastic movements, maintaining both speed and accuracy by concentrating computational resources on the most likely target positions.
3Measurement precision
If multiple candidate areas are evaluated with weights to determine final selected area, then the target tracking accuracy is improved, but the computational time increases
Solution Approach 1:
The patent performs preliminary actions by pre-calculating weights for candidate areas based on the basis image before full evaluation. This preliminary weighting reduces the computational time for subsequent frame comparisons, as the most promising candidate areas are already identified and prioritized, allowing accurate tracking with reduced processing time.
Data Source
AI summary
An image processing apparatus includes a calculator configured to calculate a respective position offset for each of a plurality of candidate areas in a second frame based on a position of a basis image in a first frame and a determiner configured to determine a final selected area that includes a target in the second frame based on a respective weight allocated to each of the plurality of candidate areas and the calculated respective position offset.