Adaptive Template Regeneration for Robust Image Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition methods face challenges in accurately detecting objects in images due to disturbances and variations, leading to potential missed detections.
Innovation Solution
An image processing device that regenerates a representative image based on a training image group containing more unused images than used images, allowing it to better adapt to disturbances and variations, thereby improving detection robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If template matching is used for object recognition, then the method is simple and easy to implement, but detection accuracy deteriorates due to disturbances and variations
Solution Approach 1:
The patent applies dynamics by making the template image adaptive and updateable. Instead of using a fixed template, the system dynamically regenerates the template image based on newly acquired images and detected objects, allowing it to adapt to changing conditions, disturbances, and variations in the environment, thereby maintaining high detection accuracy while remaining relatively simple to implement
Solution Approach 2:
The patent implements feedback by using detected objects and newly acquired images to regenerate the template image. The system continuously monitors detection results and uses this feedback information to update the template, creating a closed-loop system that improves accuracy over time while maintaining implementation simplicity through automated regeneration
2Productivity
If a fixed template image is used for similarity determination, then computational efficiency is maintained, but detection robustness deteriorates due to inability to adapt to disturbances
Solution Approach 1:
The system dynamically regenerates the template image based on newly acquired images and detected objects, allowing it to adapt to changing conditions and disturbances. This dynamic approach maintains computational efficiency by using image processing techniques while significantly improving detection robustness through continuous adaptation to environmental variations
Solution Approach 2:
The patent applies preliminary action by pre-processing and storing multiple images and detected object information before they are needed for regeneration. The system prepares and stores training data in advance, enabling efficient template regeneration when needed without significant computational delay, thus maintaining productivity while improving robustness
3Reliability
If the template image is regenerated frequently to adapt to variations, then detection robustness is improved, but computational load increases
Solution Approach 1:
The system performs self-service by automatically regenerating its own template image using newly acquired images and detected objects. This self-updating mechanism improves detection robustness through continuous adaptation while minimizing external computational intervention, as the system manages its own regeneration based on internal detection results and stored training data
Solution Approach 2:
The patent applies parameter changes by adjusting the regeneration frequency and parameters of the template image based on detection conditions. The system changes parameters such as when regeneration occurs and how many new images to process, optimizing the balance between improving robustness through frequent updates and reducing computational load by updating only when necessary
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing device includes a communication unit and a controller. The communication unit acquires a first image. The controller performs a similarity determination between a second image and at least a portion of the first image. The second image is generated based on a training image group. The training image group includes multiple detection target images different from the first image. The controller is capable of regenerating the second image. The controller regenerates the second image based on a training image group including more unused images than used images used to generate the second image.