Orientation-Aware Object Tracking Using Rotated Template Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems face increased processing costs and memory requirements when handling variations in image capturing orientation, such as upright, vertical, and upside-down capturing, due to the need for rotation processing of input images.
Innovation Solution
An image processing apparatus and method that sets filter coefficients based on orientation information to generate feature maps, registers template features, and performs correlation calculations efficiently, allowing detection of target objects regardless of image capturing orientation without requiring extensive rotation of input images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If rotation processing of the input image is executed to handle variations in image capturing orientation, then the tracking processing can be continued regardless of orientation, but the processing time increases and a large buffer memory becomes necessary
Solution Approach 1:
Instead of rotating the input image to match the template orientation, the patent inverts the approach by rotating the template feature to match the input image orientation. This is achieved by generating multiple template features corresponding to different orientations (0°, 90°, 180°, 270°) and selecting the appropriate template based on the input image orientation, thereby avoiding expensive image rotation operations
Solution Approach 2:
The patent performs preliminary action by pre-generating template features for all possible orientations (0°, 90°, 180°, 270°) before the actual tracking process. This allows the system to quickly select the pre-computed template that matches the current input image orientation without performing rotation processing during tracking, thus reducing processing time
2Adaptability or versatility
If rotation processing of the input image is executed to handle variations in image capturing orientation, then the tracking processing can be continued regardless of orientation, but a large buffer memory for processing becomes necessary
Solution Approach 1:
Instead of rotating the input image which requires large buffer memory, the patent inverts the approach by rotating only the template feature. The template feature is much smaller than the full input image, so rotating and storing multiple orientation versions of the template requires significantly less memory while achieving the same adaptability
3Adaptability or versatility
If rotation processing of the input image is executed to handle variations in image capturing orientation, then the tracking processing can be continued regardless of orientation, but an increase in processing costs becomes a problem in low-cost systems
Solution Approach 1:
The patent inverts the conventional approach by rotating the template feature instead of the input image. Since the template feature is much smaller in size than the full input image, this inversion significantly reduces the computational complexity and processing costs, making the solution suitable for low-cost systems while maintaining adaptability to different orientations
Solution Approach 2:
The patent segments the template feature into multiple orientation-specific versions (0°, 90°, 180°, 270°), allowing the system to select and use only the relevant segment that matches the current input image orientation. This segmentation approach reduces the overall processing burden compared to rotating the entire image
Data Source
AI summary
An apparatus sets coefficients in a first array based on first information indicating an image capturing orientation of a first image, generates a first map by applying the coefficients to the first image, acquires a template feature corresponding to an object based on the first map, registers the template feature in an array based on the first information, sets coefficients in a second array based on second information indicating an image capturing orientation of a second image, generates a second map by applying the coefficients set in the second array to the second image, sets the template feature in a feature array based on the second information, performs a correlation calculation between the template feature set in the feature array and the second map, and detects the object from the second image based on a result of the correlation calculation.