Computer Vision Object Tracking with Human-in-the-Loop Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking methods in industrial complexes face high computational demands, human error, and unreliable data due to the use of labels or RFID tags, while fully computer vision-based approaches require excessive computation and costly installations.
Innovation Solution
A method employing a hybrid edge-cloud system with human-in-the-loop correction, using multiple cameras and local processors to process data with virtual sensors, reducing computational workload and ensuring high reliability through human intervention for error correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If physical tracking labels or RFID tags are used on objects, then object identification and tracking become easier, but the labels can be damaged during transport or handling and may not be visible from distance or specific angles, greatly reducing the effectiveness of the tracking method
Solution Approach 1:
The patent replaces physical tracking labels and RFID tags with a computer vision-based optical recognition system. Instead of attaching mechanical/electronic tags to objects, the system uses cameras to capture images and processes these images to identify and track objects based on their visual characteristics, thereby eliminating the reliability issues associated with physical labels.
Solution Approach 2:
The patent creates a digital representation (digital twin) of the physical object by capturing its visual appearance through camera images. This digital copy contains all the necessary information for tracking and identification, replacing the need for physical labels while maintaining tracking effectiveness.
2Adaptability or versatility
If RFID tags are used for tracking objects, then optical tracking is not required, but the movement of the object cannot be reliably tracked along its entire path and RFID technology does not work reliably in metallurgy environments
Solution Approach 1:
The patent substitutes RFID technology with a computer vision system that uses cameras to track objects. This optical approach works reliably in metallurgy environments where RFID fails and provides continuous visual tracking of object movement along the entire path rather than intermittent detection at specific points.
3Reliability
If 3D modeling the space using photogrammetry is used for tracking, then a fully computer vision-based approach is achieved, but it requires camera images from a number of views filling an entire warehouse and costs a high amount of computation
Solution Approach 1:
The patent extracts only the essential visual features from camera images that are necessary for tracking objects, rather than performing complete 3D photogrammetric modeling. This selective feature extraction approach maintains tracking reliability while significantly reducing the computational workload and hardware requirements.
Solution Approach 2:
The patent applies partial computer vision processing by focusing on key visual attributes of objects rather than complete 3D reconstruction. This partial action approach provides sufficient and reliable information for tracking without the excessive computational cost of full photogrammetry.
4Ease of manufacture
If human workers constantly monitor and control the inventory, then tracking can be performed without technology, but the information on objects is outdated and only updated when inventory is carried out and completed, and there is distinct possibility of human error
Solution Approach 1:
The patent replaces human workers with an automated computer vision system that continuously monitors and tracks objects. This automated system eliminates human error and provides real-time accurate information about object positions and movements, updating data continuously rather than only during inventory completion.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
The disclosed invention relates to a method of tracking objects using computer vision with reduced computational workload and human-in-the-loop scheme ensuring high reliability and fast setup. The invention is particularly suitable for tracking larger objects of known shape or appearance and employs use of a finite number of well-positioned stationary cameras, wherein the tracking is performed, for example, in large industrial halls, warehouses without fixed shelf systems, container docks, train docks, car parks etc. The method provides a reliable data source for obtaining digital twin of objects to be tracked, representing real-time information on each stock keeping unit (SKU) in a warehouse (WH), which brings significant benefits to WH managers and thus saving human labor spent on looking for material, improving management flow, reducing the overall equipment effectiveness (OEE) of vehicles, improving throughput, quality etc. The main barrier to installing digital twins is the problem of funneling the real-time data into the WH object model.