Multi-2D Camera Object Recognition via Sensor Fusion and Fallback Linking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object recognition systems face challenges in quickly identifying objects from a large number of known objects, especially in uncontrolled environments where objects can be positioned variably, and are limited by the use of expensive 3D cameras and complexity in barcode scanning systems.
Innovation Solution
A system utilizing multiple 2D cameras to capture images from various angles, concatenating them into a single image, and employing sensor fusion with transformation matrices to synchronize data and improve object recognition accuracy, while maintaining a fallback hierarchy for obstructed views.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D cameras are used for object recognition, then object recognition accuracy is improved, but system cost increases
Solution Approach 1:
The system divides the object recognition task into multiple 2D camera views captured at different angles and times, processing each view separately and combining results through sensor fusion to achieve accurate identification without requiring expensive 3D cameras
Solution Approach 2:
The system transitions from using 3D spatial information captured by 3D cameras to using temporal dimension by capturing multiple 2D images at different time points and angles, then fusing these temporal and angular perspectives to reconstruct sufficient 3D understanding for accurate object recognition
2Measurement precision
If 3D cameras are used for object recognition, then object recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The system segments the complex 3D object recognition problem into multiple simpler 2D image capture tasks performed by standard 2D cameras at different angles, with each camera capturing a specific viewpoint that is then processed independently before fusion
Solution Approach 2:
The system replaces direct 3D spatial capture with temporal sequencing of 2D captures, using time-based multi-angle photography to achieve 3D understanding through sensor fusion of multiple 2D perspectives rather than relying on 3D camera hardware
3Productivity
If barcode scanning systems are used, then object identification is achieved, but adaptability to uncontrolled environments deteriorates
Solution Approach 1:
The system creates a universal object recognition solution that works across diverse uncontrolled environments by using multiple 2D cameras to capture objects from various angles and positions, making the system adaptable to different object placements, lighting conditions, and environmental scenarios without requiring controlled settings or specific object orientations
Solution Approach 2:
The system overcomes the limitations of 2D barcode scanning by adding temporal and angular dimensions through multi-camera captures at different time points and viewpoints, enabling 3D understanding that allows identification of objects regardless of their position, orientation, or environmental conditions
4Measurement precision
If multiple 2D cameras are used with sensor fusion, then object recognition accuracy is improved, but data processing complexity increases
Solution Approach 1:
The system performs preliminary actions by capturing all necessary multi-angle 2D images before processing, using temporal sequencing to ensure complete data collection, then processes these pre-captured images through sensor fusion to achieve accurate object recognition without requiring complex real-time processing
Data Source
AI summary
A system and method for synchronizing two-dimensional (“2D”) camera data for object recognition. An object recognition kiosk includes a plurality of 2D cameras and a stage for placement of one or more items. The plurality of 2D cameras, one of which acts as a reference camera, capture images of items on the stage from multiple angles. The system synchronizes image data from each camera in order to accurately identify items present on the stage. In situations where items in camera images fail to be linked to items identified in the reference camera perspective, the system maintains a fallback hierarchy of camera pairs to utilize to link items found in different camera images. The use of different camera pairs to link item detections continues down the fallback hierarchy until either all the item detections are linked or the last camera pair in the fallback hierarchy is reached.