Object Tracking With Defocus Maps for Similar Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking techniques struggle to distinguish between a tracking target object and similar objects within a set distance range, leading to potential misidentification.
Innovation Solution
An information processing apparatus that utilizes a neural network to perform correlation calculations using reference and search images, along with defocus amount maps, to accurately track objects by incorporating depth information through defocus amount maps and reliability value maps, and adapts parameters based on learning from ground truth data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching processing is used to track objects based on pixel pattern and color histogram similarity, then object tracking can be performed, but there is a possibility of tracking the wrong object when another similar object exists in the video
Solution Approach 1:
The patent introduces a new dimension for object identification by incorporating defocus amount information alongside traditional pixel pattern and color histogram features. By adding this depth-related dimension, the system can distinguish between similar objects that appear identical in 2D image space but have different focal characteristics, thereby resolving the tracking ambiguity without sacrificing tracking speed
2Reliability
If defocus amount is used as distance information to narrow down image region, then tracking target object and similar object can be distinguished, but device complexity increases
Solution Approach 1:
The patent introduces a defocus amount map as an intermediary data structure that captures depth information without requiring complex hardware modifications. This map serves as a mediator between the imaging system and the tracking algorithm, enabling the system to utilize defocus information in a computationally efficient manner that avoids direct complexity increases in the optical system
3Speed
If correlation calculation is performed using only reference and search images, then processing speed is maintained, but tracking precision decreases when similar objects are present
Solution Approach 1:
The patent merges multiple feature types (pixel pattern, color histogram, and defocus amount) into a unified correlation calculation framework. By combining these diverse features in the correlation computation, the system maintains processing speed through efficient mathematical operations while achieving superior tracking precision that can differentiate between similar objects based on their combined feature signatures
Data Source
AI summary
There is provided with an information processing apparatus. A first obtaining unit obtains a reference image including a tracking target object and a search image including the tracking target object. A second obtaining unit obtains a reference image defocus amount map and a search image defocus amount map. An extracting unit extracts a feature from each of the reference image, the search image, the reference image defocus amount map, and the search image defocus amount map. A correlation calculating unit performs correlation calculation regarding a feature of the reference image and the search image and feature of the reference image defocus amount map and the search image defocus amount map. A tracking result calculating unit calculates a tracking result including a position of the tracking target object by using a result of the correlation calculation.