Image Processing Apparatus Using Segmented Training Data for Region Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face challenges in accurately identifying image regions when image features vary due to changes in imaging conditions, such as daytime and evening scenes, as they struggle to distinguish between similar features caused by different lighting conditions.
Innovation Solution
An image processing apparatus that includes a first learning unit for identifying region classes, an evaluation unit for assessing the identification results, a generation unit for creating new training data based on evaluation results, and a second learning unit for generating multiple identifiers, allowing for the selection of suitable region identifiers and determiners to improve image segmentation accuracy across varying conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single region identifier is learned from various training images, then the identifier can process diverse images, but it cannot distinguish between images with similar features caused by different imaging conditions
Solution Approach 1:
The patent segments the training data into multiple subsets based on imaging conditions (daytime, evening, night). Each subset is used to train a dedicated region identifier specialized for that condition. This segmentation allows each identifier to focus on specific feature characteristics of its designated condition, resolving the contradiction between versatility and precision by creating multiple specialized identifiers rather than one general-purpose identifier.
Solution Approach 2:
The patent applies local quality by creating region identifiers with specialized features tailored to specific imaging conditions. Each identifier learns local feature characteristics specific to its training condition (e.g., cloud features in daytime vs. sunset features in evening). This enables each identifier to have high precision for its designated condition while the system as a whole maintains versatility through multiple identifiers.
2Adaptability or versatility
If training data includes images from various imaging conditions, then the identifier can handle diverse scenes, but it becomes difficult to distinguish between similar features from different conditions
Solution Approach 1:
The training data is segmented into multiple subsets based on imaging conditions. Each subset contains images from a specific condition (daytime, evening, night), and a dedicated region identifier is trained on each subset. This segmentation prevents feature confusion by ensuring each identifier learns from homogeneous data, maintaining reliability while achieving versatility through the collection of multiple condition-specific identifiers.
Solution Approach 2:
The patent changes the parameter of training data composition by creating separate training subsets with distinct imaging condition parameters. Each subset is characterized by specific lighting conditions, time of day, and atmospheric features. By training separate identifiers on these parameterized subsets, the system achieves reliable feature distinction for each condition while maintaining overall versatility.
3Device complexity
If the problem is divided into multiple conditions with predetermined thresholds, then recognition is simplified, but it is difficult to determine appropriate division conditions in advance
Solution Approach 1:
The patent applies self-service by using the evaluation unit to automatically assess identification results and generate new training data based on evaluation outcomes. The system self-adjusts by identifying which imaging conditions cause confusion and automatically creating targeted training subsets for those conditions. This eliminates the need for manual predetermined thresholds, as the system discovers and divides conditions based on actual performance feedback.
Solution Approach 2:
The evaluation unit provides feedback on identification accuracy, revealing which imaging conditions cause feature confusion. This feedback loop allows the system to iteratively refine the segmentation of training data by identifying problematic condition boundaries and creating appropriate subsets. The feedback mechanism replaces manual condition determination with automated, performance-driven condition segmentation.
Data Source
AI summary
An image processing apparatus includes a first learning unit configured to learn an identifier for identifying a class of a region formed by segmenting an image based on first training data, an evaluation unit configured to evaluate a result of identification of a class of the first training data by the identifier, a generation unit configured to generate second training data from the first training data based on an evaluation result by the evaluation unit, and a second learning unit configured to learn a plurality of identifiers different from the identifier learned by the first learning unit based on the second training data.