Mixed Reality Collision Detection via Self-Area Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mixed reality systems face challenges in accurately detecting obstacles in the real world, particularly with moving objects and the user's own body parts, leading to potential collisions and hindering the mixed reality experience due to limitations in field of view and object recognition.
Innovation Solution
An image processing apparatus comprising a location estimation unit, recognition unit, area decision unit, and determination unit that estimates the self-location and self-area based on video input, allowing for real-time recognition of physical objects and determination of warnings without pre-registration of obstacles, using methods such as SLAM, index images, color detection, and depth movement analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pre-registration of obstacles is used for collision detection, then false warnings are reduced, but the system cannot detect moving objects or unregistered obstacles
Solution Approach 1:
The system performs preliminary registration of safe areas (self-areas) where the user is located, rather than pre-registering obstacles. This allows the system to recognize and exclude the user's body parts from obstacle detection before actual collision detection occurs, enabling both reliability and adaptability
Solution Approach 2:
Instead of directly detecting and registering obstacles as in conventional systems, this patent inverts the approach by first defining safe areas (self-areas) and then determining that objects outside these areas are potential obstacles. This inversion allows the system to handle moving objects and unregistered obstacles effectively
2Area of stationary object
If the field of view is expanded in MR display, then virtual object visibility is improved, but the view of the real world is blocked
Solution Approach 1:
The display area is segmented into multiple regions: a first area displaying the real world video and a second area displaying virtual objects. This segmentation allows both real world visibility and virtual object visibility to coexist by allocating specific display regions for each purpose
Solution Approach 2:
The patent overlays virtual objects in the second area on top of the real world video in the first area, utilizing the depth dimension of display to present both real and virtual content simultaneously without completely blocking either view
Data Source
AI summary
In order to make it possible for the user to perceive the possibility of a collision with an object in the real world, an image processing apparatus comprises: a location estimation unit configured to, based on a video obtained by an image capturing unit for capturing a physical space, estimating a self-location of the image capturing unit in the physical space; a recognition unit configured to recognize a physical object existing within a certain distance from the self-location based on the video; an area decision unit configured to decide a predetermined area in the physical space in relation to the video; and a determination unit configured to determine whether or not a warning is given in accordance with whether or not a physical object recognized by the recognition unit is included in the predetermined area.