Target Part Position Estimation Using Pretrained Image Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques struggle to accurately estimate the position of target parts, such as the pupil, in images, particularly in cases where template matching and eye identification methods are insufficient for precise localization.
Innovation Solution
An information processing apparatus and method that utilizes a target image acquisition unit to acquire images, extracts target part features using a learning result from reference and target images, and estimates the position of the target part based on these features, employing machine learning models like convolutional neural networks and similarity degree information to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If template matching is used to detect and track specific objects, then the tracking function can be implemented, but the measurement precision of target part position is insufficient
Solution Approach 1:
The patent performs preliminary learning training offline to build a position estimation model using reference images with known target part positions. This pre-trained model is then applied to estimation target images to achieve high-precision position estimation without complex real-time processing, thus improving measurement precision while maintaining simple online processing complexity.
Solution Approach 2:
The patent introduces a position estimation model as an intermediary between the image processing system and the target part position detection. This model, trained offline with reference images, acts as a mediator that transforms the complex task of precise position detection into a simplified inference process, achieving high precision without increasing online processing complexity.
2Measurement precision
If machine learning models are used to extract target part features and estimate positions, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent performs the complex machine learning model training in advance during an offline learning phase using reference images with annotated target part positions. The trained model is then stored and reused for multiple estimation target images. This approach achieves high measurement precision through sophisticated modeling while keeping the online processing complexity low since the model application is straightforward inference.
3Measurement precision
If reference images and target images are used for learning, then the accuracy of target part position estimation is improved, but the processing time increases
Solution Approach 1:
The patent performs the time-consuming learning processing using reference images and target images in advance during an offline training phase. The learned position estimation model is then applied to estimation target images for rapid position estimation. This separates the heavy computational burden from the time-critical application phase, achieving high accuracy while minimizing the time loss during actual operation.
Data Source
AI summary
An information processing apparatus (100) includes a target image acquisition unit (101) and a position estimation unit (110). The target image acquisition unit (101) acquires an estimation target image including an image of a target part in an estimation target. The position estimation unit (110) extracts a target part feature related to a portion associated with the target part from a feature of the estimation target image, based on a learning result acquired by performing learning by using a reference image including an image of the target part and a target image including an image of the target part, and estimates a position of the target part, based on the target part feature.