3D Computer Vision Processing Engine for Real-Time Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D computer vision processing methods are inefficient for real-time applications, particularly on mobile and embedded platforms, due to high computational requirements and the need for manual post-processing, and are hindered by noisy and incomplete scans, occlusions, and limited processing power.
Innovation Solution
A fully automated 3D computer vision processing method that uses priori information and is tolerant to noise and occlusions, capable of generating accurate 3D models on low-power platforms like ARM processors, by analyzing images to extract 3D point clouds and matching them to reference models for feature detection and measurement calculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing 3D computer vision processing methods are used, then 3D models can be generated, but the processing time and computational power requirements increase significantly
Solution Approach 1:
The system uses prior knowledge about object structures and scenes to guide the 3D reconstruction process before receiving complete scan data. This preliminary action allows the system to make intelligent guesses about object geometry and fill in missing information, significantly reducing the processing time needed to generate accurate 3D models from incomplete scan data.
Solution Approach 2:
The patent introduces an intermediary representation layer between the raw scan data and the final 3D model. This intermediary structure processes and pre-processes the scan data, extracting key features and organizing them in a way that accelerates the subsequent 3D reconstruction, thereby reducing overall processing time while maintaining model accuracy.
2Manufacturing precision
If manual post-processing is performed to clean up 3D models, then model quality improves, but the workflow complexity and time consumption increase
Solution Approach 1:
The system performs self-service by automatically cleaning up and optimizing the 3D models through integrated algorithms that handle noise removal, mesh repair, and quality enhancement. This eliminates the need for manual post-processing steps, reducing workflow complexity while maintaining high model quality through automated intelligent processing.
Solution Approach 2:
The patent merges multiple processing functions (scanning, reconstruction, cleaning, optimization) into a single integrated workflow. By combining these previously separate steps into one unified process, the system reduces workflow complexity and eliminates manual intervention requirements while delivering high-quality 3D models.
3Ease of manufacture
If low-cost scanners are used, then device cost decreases, but the scan quality becomes noisy and incomplete
Solution Approach 1:
The system converts the harmful effects of noise and incompleteness in low-cost scanner data into beneficial outcomes. By using intelligent algorithms that treat noise as informative signals and leverage the limited scan data through creative reconstruction techniques, the system produces high-quality 3D models from what would traditionally be considered poor-quality input data.
Solution Approach 2:
The patent applies parameter changes by transforming the representation of scan data from raw sensor values to enhanced feature representations. This parameter transformation allows the system to extract meaningful geometric information even from noisy and incomplete data, effectively improving scan quality without requiring expensive hardware.
4Speed
If real-time processing is implemented, then application responsiveness improves, but the computational power requirements increase
Solution Approach 1:
The system segments the 3D reconstruction process into multiple independent stages that can be processed in parallel or at different rates. By dividing the complex computation into manageable segments, the system achieves real-time responsiveness for critical operations while using lower computational power for less time-sensitive tasks, balancing speed and power consumption requirements.
Data Source
AI summary
Methods and systems are described for generating a three-dimensional (3D) model of a fully-formed object represented in a noisy or partial scene. An image processing module of a computing device receives images captured by a sensor. The module generates partial 3D mesh models of physical objects in the scene based upon analysis of the images, and determines a location of at least one target object in the scene by comparing the images to one or more 3D reference models and extracting a 3D point cloud of the target object. The module matches the 3D point cloud of the target object to a selected 3D reference model based upon a similarity parameter, and detects one or more features of the target object. The module generates a fully formed 3D model of the target object using partial or noisy 3D points from the scene, extracts the detected features of the target object and features of the 3D reference models that correspond to the detected features, and calculates measurements of the detected features.