AR Object Detection via Region-Specific Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AR image processing methods require more frames and time to recognize objects when video quality is poor, leading to increased calculation and display time for augmented reality applications.
Innovation Solution
An image processing apparatus that divides a target space into smaller regions, constructs specific learning models for each region, and selects the appropriate model based on the camera recognition space to detect objects efficiently, reducing the need for multiple frames and improving recognition speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video recognition technique is used to display AR, then detailed information can be overlaid on objects, but more frames are required when video quality is poor, increasing recognition time
Solution Approach 1:
The patent divides the target space into multiple divided spaces and constructs separate learning models for each divided space. This segmentation allows the system to focus on specific regions rather than analyzing the entire frame, reducing the number of frames needed for recognition and decreasing recognition time while maintaining accuracy.
Solution Approach 2:
The patent creates different learning models for different divided spaces, with each model optimized for its specific region. This local quality approach allows the system to use more accurate, specialized models for each area, improving recognition precision without requiring multiple frames across the entire scene.
2Reliability
If the entire image is rescanned to recognize objects, then comprehensive object detection is achieved, but calculation amount increases
Solution Approach 1:
The patent segments the target space into multiple divided spaces and assigns specific learning models to each. This segmentation enables the system to perform localized object detection rather than scanning the entire image, reducing calculation amount while maintaining detection reliability through specialized models for each region.
Solution Approach 2:
The patent extracts and focuses on specific divided spaces that are relevant to the current camera recognition space. By taking out only the necessary regions for analysis rather than processing the entire image, the system reduces calculation amount while maintaining reliable object detection through targeted learning models.
3Measurement precision
If more frames are used to recognize objects in poor video quality, then recognition accuracy is maintained, but time required for AR display increases
Solution Approach 1:
The patent divides the target space into multiple divided spaces with dedicated learning models for each. This segmentation allows the system to maintain recognition accuracy by using specialized models for specific regions while significantly reducing the time required for AR display, as each model can quickly identify objects in its designated space without waiting for multiple frames.
Solution Approach 2:
The patent constructs multiple learning models in advance for different divided spaces before actual object recognition is needed. This preliminary action prepares the system with ready-to-use specialized models, enabling fast object recognition and quick AR display even when video quality is poor, without requiring additional frame processing time.
Data Source
AI summary
An image processing apparatus 10 includes a learning model construction unit 11 that generates a divided space(s) obtained by dividing a target space into one or more spaces and constructs a learning model for recognizing an object(s) included in the divided space, a learning model management unit 12 that manages the learning model and a region forming the divided space including the object recognized by the learning model in association with each other, a space estimation unit 13 that estimates a region forming a camera recognition space captured by a camera provided in a UI device; and a detection unit 14 that selects, from among the managed learning models, a specific learning model associated with the region forming the divided space including the estimated region forming the camera recognition space, and to detect the object included in a space displayed on the UI device using the selected specific learning model.