Hybrid 2D-3D Vision Camera Reducing Computational Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D imaging technologies in industrial automation lack efficient and cost-effective solutions for real-time object recognition and machine vision applications, particularly in scenarios where high accuracy is not mandatory, due to high computational load and power consumption.
Innovation Solution
A compact, lightweight hybrid 2D/3D vision system combining a high-resolution 2D camera with a low-resolution 3D Time of Flight (TOF) camera, utilizing an Intelligent Segmentation Engine (ISE) that learns and updates object recognition through incremental history and feedback, reducing computational overload and power consumption, and employing shape descriptors and reinforcement mechanisms for efficient object classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional 3D imaging system is used for object recognition, then depth information is obtained, but computational load and power consumption increase significantly
Solution Approach 1:
The patent divides the image processing into two segments: a low-resolution 3D camera captures depth information to identify regions of interest, while a high-resolution 2D camera captures detailed images. This segmentation allows the system to process only relevant regions in high detail, reducing overall computational load and power consumption while maintaining measurement precision for depth-critical applications.
Solution Approach 2:
The patent extracts only the essential depth information from the 3D camera to identify regions of interest, rather than processing the entire 3D image at high resolution. By taking out only the necessary depth data for ROI detection and combining it with 2D image processing, the system reduces computational overhead and energy consumption while preserving measurement accuracy where needed.
2Measurement precision
If high-resolution 3D imaging is used for real-time processing, then detailed depth information is obtained, but processing speed decreases due to computational complexity
Solution Approach 1:
The system segments the imaging task by using the low-resolution 3D camera for rapid ROI detection and the high-resolution 2D camera for detailed analysis. This segmentation enables real-time processing by avoiding the computational burden of processing high-resolution 3D data while maintaining the ability to achieve high depth resolution when needed through selective processing.
Solution Approach 2:
The patent applies partial action by processing only the regions of interest identified by the 3D camera at high resolution, rather than processing the entire scene. This allows the system to achieve high processing speed by limiting detailed analysis to small portions of the image while maintaining measurement precision for those specific regions.
3Loss of information
If a full 3D vision system is implemented for object recognition, then comprehensive spatial information is obtained, but device size and cost increase
Solution Approach 1:
The patent merges the capabilities of 3D and 2D cameras into a hybrid system where the low-resolution 3D camera provides spatial depth information and the high-resolution 2D camera provides detailed visual information. By combining these two complementary systems, the patent achieves comprehensive spatial information completeness while keeping individual camera resolutions lower, thus reducing overall device complexity and size.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system achieves improved performance and speed-up in object recognition, reduces computational and memory requirements, and enables real-time processing in low-cost applications like inspection, barcode scanning, and augmented reality, while maintaining reduced size and weight.
Implementation Method 1
a low-resolution 3D Time of Flight (TOF) camera
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Various embodiments of the invention are implemented for an entry level, compact and lightweight single package apparatus combining a conventional high-resolution two-dimensional (2D) camera with a low-resolution three-dimensional (3D) depth image camera, capable to learn, through depth information, how to improve the performance of a set of 2D identification and machine vision algorithms in terms of speed-up (e.g. through regions of interests (ROIs)) and raw discriminative power. The cameras simultaneously capture images that are processed by an Intelligent Segmentation Engine in the system to facilitate object recognition.