3D Shape Determination Using Multi-Angle Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques cannot determine whether a shape surrounded by straight lines in an image represents an actual oblong or parallelogram in a 3D space, regardless of the viewing angle, as they do not account for size variations.
Innovation Solution
An image processing device and method that utilize a standard image input device, a reference image input device, specific shape candidate detection, virtual target object information generation, and shape determination to identify shapes like triangles, rectangles, and parallelograms by comparing images taken at different angles, adjusting size and position, and repeating detection processes to confirm the shape's presence in the actual space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image processing techniques compare shapes from multiple cameras or use neural networks with size information, then object recognition accuracy is improved, but the ability to determine actual shape in 3D space from 2D images remains insufficient due to viewing angle variations
Solution Approach 1:
The patent transforms the 2D image shape analysis problem into a 3D space determination problem by introducing depth information through multiple imaging devices. By capturing images from different viewing angles and reconstructing the object in 3D space, the system can determine the actual shape independent of viewing angle, resolving the contradiction between 2D measurement precision and 3D adaptability.
Solution Approach 2:
The patent introduces an intermediary computational model that maps 2D image coordinates to 3D space coordinates. This intermediary transformation process allows the system to account for viewing angle variations and determine whether observed shapes correspond to actual 3D shapes, thereby maintaining shape determination accuracy across different viewing angles.
2Measurement precision
If the size of the shape is known, then the shape can be determined, but the ability to determine shape when size is unknown in actual space is lost
Solution Approach 1:
The patent segments the shape determination process into two independent components: geometric shape analysis (angles, side relationships) and size analysis. By separating these components, the system can determine the actual shape (rectangle, parallelogram, etc.) based on geometric properties alone, without requiring size information, thus eliminating the loss of information when size is unknown.
Solution Approach 2:
The patent changes the parameters used for shape determination from size-dependent parameters to size-independent geometric parameters. Instead of relying on absolute dimensions, the system uses angles between sides, ratios of adjacent sides, and other geometric relationships that remain invariant to scale, enabling shape identification without size information.
3Measurement precision
If multiple images at different viewing angles are processed, then shape determination in actual space becomes possible, but processing complexity increases
Solution Approach 1:
The patent applies preliminary geometric analysis to each 2D image before 3D reconstruction. By first determining candidate shapes from individual 2D images using geometric properties (angles, side relationships), the system reduces the complexity of subsequent 3D processing. This preliminary action filters out impossible shapes early, reducing overall processing complexity while maintaining accurate actual shape determination.
Data Source
AI summary
It is possible to determine whether or not a specific shape candidate obtained from a captured standard image has a corresponding shape in an actual space. A shape surrounded by straight lines is detected from a standard image as a candidate having a rectangular shape. A target object image obtained by imaging a target object representing the candidate having the rectangular shape at a viewing angle different from the viewing angle of the standard image is generated. The generated target object image is detected from a reference image obtained by imaging the target object at a different viewing angle. If the target object image is detected, the target object represented by the candidate having the rectangular shape is determined to be a rectangle.