Fusion Splicer Fiber-Type Recognition for Accurate Splicing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fusion splicers face challenges in accurately discriminating the types of optical fibers to be spliced due to variations in features such as optical characteristics and structural features, leading to limited discrimination accuracy even with machine learning alone.
Innovation Solution
A fusion splicing system employing a combination of two discrimination algorithms: a first algorithm based on feature correlations and a second algorithm using machine learning, with a discrimination model created by a model creation device, enhances discrimination accuracy by selecting the most accurate algorithm based on feature thresholds or variations, and classifying splicers into groups with similar imaging data tendencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning is used for optical fiber type discrimination, then discrimination capability is enhanced, but discrimination accuracy is limited due to variations in optical characteristics and structural features
Solution Approach 1:
The patent combines two different discrimination algorithms: a first algorithm based on feature correlations (traditional method) and a second algorithm using machine learning. The system selectively applies either algorithm based on which one provides higher discrimination accuracy for the given optical fiber types, thereby merging the advantages of both approaches to overcome the limitation of using machine learning alone.
2Device complexity
If a single discrimination algorithm is used, then device complexity is reduced, but discrimination accuracy is limited
Solution Approach 1:
The patent implements a dynamic selection mechanism that chooses between two discrimination algorithms based on real-time evaluation of which algorithm provides higher accuracy for the current set of optical fibers. This dynamic approach allows the system to adapt its complexity level according to the specific discrimination task, using the more complex machine learning algorithm only when necessary.
3Measurement precision
If multiple discrimination algorithms are implemented, then discrimination accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the discrimination task into two distinct algorithmic approaches: a first algorithm based on feature correlations and a second algorithm using machine learning. By dividing the overall discrimination function into these separate segments, the system can evaluate and select the most appropriate algorithm for each specific case, managing complexity through functional segmentation.
4Adaptability or versatility
If machine learning model is created from sample data, then discrimination capability is enhanced, but calculation time increases
Solution Approach 1:
The patent performs preliminary evaluation to determine which discrimination algorithm is more suitable before executing the actual discrimination task. By预先 assessing the characteristics of the optical fibers and selecting the appropriate algorithm in advance, the system avoids unnecessary computational overhead from using the more complex machine learning model when the simpler first algorithm would suffice, thus reducing calculation time while maintaining discrimination capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A fusion splicer according to the disclosure includes an imaging unit, a discrimination unit, and a splicing unit. The imaging unit images a pair of optical fibers and generates imaging data. The discrimination unit discriminates a type of each of a pair of optical fibers based on a plurality of feature amounts obtained from imaging data provided from the imaging unit. The discrimination unit adopts a discrimination result by any of first and second discrimination algorithms. The first discrimination algorithm is predetermined by a method, other than machine learning, based on a correlation between a plurality of feature amounts obtained from the imaging data of the optical fibers and the type of optical fiber. The second discrimination algorithm includes a discrimination model. The discrimination model is created by machine learning using sample data indicating a correspondence relationship between a plurality of feature amounts obtained from the imaging data of the optical fibers and the type of optical fiber. The splicing unit fusion-splices the pair of optical fibers to each other under a splicing condition according to a combination of the types of pair of optical fibers based on a discrimination result in the discrimination unit.