Simulated Fiducial Tag Training for Animal Vision Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automated tracking of animals in commercial farming settings are limited by the need for expensive equipment, are prone to errors, and struggle with high accuracy over extended periods. Additionally, existing fiducial systems are not human-readable, not standardized, and fail in environments where fiducials become dirty or obscured. The manual training of machine vision systems is also labor-intensive and costly.
Innovation Solution
The development of a system that uses automatically generated, simulated training data to train a convolutional neural network for identifying fiducial tags associated with animals. This system includes a fiducial tag that is both human and machine readable, designed to withstand environmental challenges, and can be used to track individual animals in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual training methods are used to train machine vision systems, then the system can be trained with real-world data, but the process becomes labor-intensive and expensive
Solution Approach 1:
The patent creates synthetic training images by copying and transforming fiducial tag templates, superimposing them onto background images to generate artificial training data. This eliminates the need for manual collection and annotation of real-world training images, significantly reducing time and labor costs while maintaining training effectiveness
Solution Approach 2:
The patent performs preliminary actions by pre-generating and storing fiducial tag templates and background images in databases before training begins. These pre-prepared elements are then automatically combined through software to create training images, eliminating the need for real-time manual image collection and annotation during the training process
2Extent of automation
If existing fiducial systems are used for tracking, then automated tracking can be implemented, but the systems fail when fiducials become dirty or obscured and are not human-readable
Solution Approach 1:
The fiducial tag design incorporates distinct local quality features including high-contrast color schemes (yellow background with black characters), specific geometric arrangements of detection patterns, and spatially separated functional zones. These localized quality enhancements ensure reliable machine detection even when parts of the tag are obscured, while simultaneously maintaining human readability
Solution Approach 2:
The patent employs specific color changes and high-contrast color combinations in the fiducial tag design, particularly yellow backgrounds with black characters and colored detection patterns. These color choices maximize contrast for machine vision systems while ensuring the tags remain clearly visible and readable by human operators under various lighting conditions
3Extent of automation
If existing fiducial systems are used, then tracking can be implemented, but the systems are not standardized and require expensive equipment
Solution Approach 1:
The fiducial tag design incorporates universal features that serve multiple functions: detection patterns enable machine vision identification, alphanumeric characters provide human readability, and the overall structure works across different camera systems and lighting conditions. This multi-functionality eliminates the need for specialized expensive equipment, as standard cameras and image processing software can successfully read the tags
Data Source
AI summary
The present invention relates to the training of a convolutional neural network by a set of automatically generated simulated training data for use in a machine vision system for identifying fiducial tags associated with entities. Four-character fiducial tags are randomly generated or selected and then randomly transformed and superimposed on images to generate a set of simulated training data to be fed into a convolutional neural network to train the network model. The trained convolutional neural network is used to identify fiducial tags disposed on entities in real-world images and video.