AR Object Recognition via Image Difference Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional augmented reality systems face challenges in accurately identifying physical objects, especially when they are partially contained in a capture image, and in distinguishing similar objects, leading to user confusion and the need for complex and time-consuming manual development processes.
Innovation Solution
An augmented reality system that employs object and spatial recognition techniques to assign unique identifiers based on image differences in additional recognition regions, allowing for accurate identification of physical objects and dynamic adjustment of virtual object information and positioning to prevent overlap and enhance user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional object recognition methods are used, then the system can identify fully visible objects, but it fails to accurately identify partially contained objects leading to user confusion
Solution Approach 1:
The system performs preliminary actions by predicting the complete position and boundaries of objects that are only partially visible in the capture image. It uses the visible portion to infer and mark the full object location before displaying the virtual object, ensuring accurate identification even when the object is not fully contained in the image frame.
2Measurement precision
If conventional recognition methods are used, then the system can process simple scenes, but it cannot distinguish similar objects leading to user confusion
Solution Approach 1:
The system applies local quality by extracting and comparing specific local features from additional recognition regions of objects. Instead of analyzing entire objects, it focuses on distinctive local characteristics to generate unique identifiers for similar objects, enabling accurate differentiation while maintaining manageable system complexity.
3Adaptability or versatility
If manual development processes are used for AR contents, then the contents can be customized, but the development time and cost increase significantly
Solution Approach 1:
The system enables self-service by automatically generating AR contents through object recognition and matching. It autonomously identifies objects, retrieves corresponding virtual objects from databases, and displays them without requiring manual development or customization, dramatically reducing development time while maintaining adaptability to different objects.
4Ease of operation
If virtual objects are displayed without precise positioning, then the display process is simple, but virtual objects may overlap causing user confusion
Solution Approach 1:
The system replaces manual positioning mechanisms with automated image processing and coordinate transformation algorithms. It automatically calculates the positions of virtual objects based on recognized object locations in the capture image, ensuring precise positioning and preventing overlap without requiring complex manual adjustment processes.
Data Source
AI summary
In an augmented reality method capable of identifying a plurality of similar objects when the similar objects are detected from actual image information captured by a terminal, an augmented reality method with space and object recognition according to the present invention comprises obtaining the actual image information capturing a plurality of objects; detecting a plurality of objects of the actual image information; classifying at least two or more objects with a predetermined similarity among the plurality of detected objects as similar objects; determining a region with an image difference learned by subtracting actual image information of the two or more similar objects from each other as an additional recognition region; assigning a unique identifier to the similar object based on the image difference of the additional recognition region; matching the similar object to which the unique identified has been assigned to a virtual object assigned to the additional recognition region, unique identifier, and the similar object and storing a matching result to a database; and identifying similar objects of the plurality of objects based on unique identifiers of additional recognition regions stored in the database if the plurality of objects are captured after a predetermined time period.