3D Image Data Visualization Using Depth-Limited Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current visualization methods for 3D image data from 3D sensors, such as depth maps and point clouds, are not intuitive, making it difficult to understand spatial environments and distinguish between objects and background, especially in industrial safety applications where accurate detection of protection zones is critical.
Innovation Solution
A method using Δz-limited Connected Components Labeling (CCL) to form connected segments from laterally adjacent 3D points with similar depth values, generating surface meshes and shadowing meshes, and applying texturing to enhance visualization, allowing for intuitive understanding of spatial relationships and object boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If depth maps are displayed with gray values or false colors, then distance information is encoded, but the visualization is not intuitive and spatial environment understanding is difficult
Solution Approach 1:
The patent transforms 2D depth map data into a 3D point cloud representation, adding spatial depth dimension to the visualization. This allows the system to preserve distance information while presenting it in an intuitively understandable 3D spatial format that maintains the camera's perspective view of the environment.
Solution Approach 2:
The patent applies different visualization qualities to different spatial regions by using depth-based segmentation. Points at different depth ranges are grouped into separate connected components, allowing local optimization of visualization properties such as point density, coloring, and rendering intensity based on depth distance from the camera.
2Ease of operation
If point clouds are used for visualization, then three-dimensional spatial representation is achieved, but objects and background are indistinguishable and shadow areas are hard to locate
Solution Approach 1:
The patent segments the point cloud into connected components based on depth similarity and spatial proximity. This segmentation groups points belonging to the same object or surface together, creating distinct visual regions that represent different objects, backgrounds, and shadow areas, making them distinguishable from one another.
Solution Approach 2:
The patent applies color coding to different connected components based on their depth characteristics and spatial relationships. By assigning different colors or color intensities to points in different depth ranges or connected components, the system enhances the visual distinction between objects, backgrounds, and shadow areas while preserving the 3D spatial representation.
3Shape
If surface mesh is created from all points, then continuous surface with texturing is achieved, but object boundaries and background are visualized indistinguishably
Solution Approach 1:
The patent avoids creating a single continuous surface mesh from all points instead. Rather, it maintains the segmented point cloud structure where connected components represent distinct objects and regions. This prevents the merging of object boundaries with background that would occur in a unified mesh, preserving the distinction between different spatial elements.
Solution Approach 2:
The patent extracts and emphasizes boundary points and depth transition regions from the point cloud. By identifying and highlighting points where depth values change significantly or where connected components meet, the system makes object boundaries and shadow areas visually distinct without requiring a continuous surface mesh.
4Ease of operation
If vertical surfaces are displayed with central perspective, then three-dimensional visualization is achieved, but vertical surfaces appear distorted
Solution Approach 1:
The patent applies different rendering qualities to points based on their orientation and depth characteristics. For vertical surfaces, the system adjusts point density, size, and spacing locally to compensate for perspective distortion. Points on vertical surfaces are rendered with properties that maintain their apparent vertical alignment and spacing, reducing the distortion effect while preserving the 3D visualization.
Data Source
AI summary
A method for visualizing 3D image data of a 3D sensor (10) with a plurality of 3D points which form a lateral 2D arrangement with a respective depth value, wherein connected segments (32) are formed from connected 3D points and the segments (32) are displayed, and wherein two respective 3D points are connected in the same segment (32) if they are laterally adjacent and also differ in their depth value by at most a depth threshold (z).