Optical cable fusion splice quality detection method and system based on multi-modal data

By processing multimodal data, a multimodal teacher model and a lightweight student model are constructed. By combining the Transformer fusion layer and a deep separable convolutional network, the problems of high precision and low cost in optical cable splice quality inspection are solved, and efficient inspection in complex environments is achieved.

CN122153806APending Publication Date: 2026-06-05GUANGDONG ALTRATEK COMM TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGDONG ALTRATEK COMM TECH CO LTD
Filing Date
2026-04-29
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing methods for testing the quality of optical cable splices cannot simultaneously guarantee high accuracy and low cost. Traditional OTDR test data cannot intuitively identify physical defects such as end-face contamination, cracks, or bubbles, while simple image analysis is difficult to quantify the dynamic changes in splice loss, leading to frequent misjudgments.

Method used

A method for detecting optical cable splice quality based on multimodal data is adopted. By constructing a multimodal teacher model and a lightweight student model, the splice end face image, OTDR test data and environmental detection data are used for training. Combined with Transformer fusion layer and deep separable convolutional network, knowledge distillation is performed to improve detection accuracy and reduce hardware cost.

Benefits of technology

It has achieved high-accuracy detection of optical cable splice quality in complex environments, reduced the hardware cost and deployment difficulty of on-site testing equipment, and improved the robustness of testing and on-site operation efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122153806A_ABST
    Figure CN122153806A_ABST
Patent Text Reader

Abstract

The application relates to the field of multi-modal data processing, and more particularly to a multi-modal data-based optical cable fusion quality detection method and system. The method comprises the following steps: acquiring a multi-modal fusion sample set, each sample containing a fusion end face image, OTDR test data and environment detection data of a corresponding same optical fiber fusion point; training a preset multi-modal teacher model based on the sample set, and freezing the parameters of the teacher model after the training is completed; constructing a lightweight student model, receiving and processing the image and OTDR test data through the student model; using the frozen teacher model to perform knowledge distillation on the student model, minimizing the difference between the two models in the intermediate representation space to obtain a distilled student model; collecting fusion end face images and OTDR test data of a to-be-detected optical fiber fusion point and inputting the distilled student model to obtain a fusion quality prediction result. The application is used for improving the accuracy of optical cable fusion quality detection.
Need to check novelty before this filing date? Find Prior Art