AR Depth Detection via Perspective Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality (AR) technologies are limited to room-scale experiences due to hardware constraints, particularly with depth sensors, which restrict the ability to detect depth values of distant objects.
Innovation Solution
A method that determines the depth and relative position of objects in an image based on understanding object geometry in perspective, without relying on hardware such as depth sensors, allowing for world-scale AR experiences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If hardware depth sensors (LIDAR) are used to measure distance for AR experiences, then the depth detection capability is improved, but the device cost and complexity increase, and the depth range is still limited to room scale
Solution Approach 1:
The patent replaces hardware depth sensors (LIDAR) with a software-based perspective understanding system. Instead of using mechanical/optical depth sensing hardware, the system uses image processing algorithms that analyze object geometry, vanishing points, and perspective cues to compute depth and relative positioning, thereby eliminating the need for expensive depth sensing hardware while achieving world-scale depth detection
Solution Approach 2:
The patent creates a virtual copy of depth information through computational methods. By analyzing 2D images and applying perspective geometry principles, the system generates synthetic depth maps and relative position data that replicate what would be obtained from physical depth sensors, but extended to world scale without the hardware constraints
2Reliability
If existing AR occlusion techniques are used, then the AR experience is improved for room scale, but the depth range is limited and cannot cover entire visible distance
Solution Approach 1:
The patent transitions from room-scale AR to world-scale AR by adding a dimensional leap in depth detection capability. Instead of being constrained to room-scale depth maps, the system uses perspective geometry to infer depth relationships across the entire visible scene, effectively adding a new dimension of depth understanding that extends beyond physical sensor limitations
Solution Approach 2:
The patent introduces vanishing points and perspective geometry as intermediary concepts to bridge the gap between 2D image data and 3D depth information. These geometric intermediaries allow the system to infer depth relationships for distant objects without direct depth sensing, enabling world-scale AR occlusion by using perspective cues as a mediator between camera observations and spatial reasoning
3Device complexity
If visual semantic understanding methods are used to define occlusion relationships, then hardware dependency is reduced, but the accuracy and reliability of depth ordering is insufficient
Solution Approach 1:
The patent replaces unreliable visual semantic understanding methods with a geometrically-grounded perspective analysis system. Instead of using heuristic-based semantic segmentation and color/texture analysis, the system uses rigorous perspective geometry principles to compute vanishing points, horizon lines, and depth ordering, thereby achieving both hardware independence and high measurement precision through mathematical geometric reasoning
Data Source
AI summary
A method may include: obtaining at least one semantic parameter associated with the image; segmenting the at least one object based on the at least one semantic parameter to generate a first segmented object and a second segmented object; identifying a camera level of the electronic device; applying a ground mesh to the image based on the camera level; determining placements of the first segmented object and the second segmented object based on the at least one semantic parameter associated with the first segmented object, the second segmented object, and the ground mesh; determining a relative position of the first segmented object with respect to the second segmented object based on the determined placement of the first segmented object and the second segmented object; and displaying the at least one object on a screen of the electronic device based on the determined relative position.