Cotton foreign fiber identification method and system based on improved RT-DETR

By improving the RT-DETR model and utilizing CA-SPM, CAPN modules, and ASD Loss, the problem of foreign fiber identification in cotton was solved, achieving efficient and accurate foreign fiber detection and reducing yarn breakage rate and fabric defects.

CN122176361APending Publication Date: 2026-06-09HUZHOU QUALITY & TECH SUPERVISION & INSPECTION INST (HUZHOU FIBER QUALITY MONITORING CENT) +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUZHOU QUALITY & TECH SUPERVISION & INSPECTION INST (HUZHOU FIBER QUALITY MONITORING CENT)
Filing Date
2026-02-03
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing technologies struggle to effectively identify foreign fibers such as plastic film and human hair mixed into cotton, leading to high yarn breakage rates and numerous fabric defects during spinning. Furthermore, existing deep learning models suffer from deficiencies in real-time performance and recognition accuracy.

Method used

An improved RT-DETR model is adopted, which optimizes the cotton foreign fiber detection dataset by embedding the cross-attention strip pooling module CA-SPM and the channel attention pyramid module CAPN in the backbone network and combining it with the adaptive scale dynamic loss ASD Loss, thereby improving the model's ability to extract and identify foreign fiber features.

Benefits of technology

It significantly reduced the false negative and false positive rates, improved the accuracy of identifying low-contrast foreign fibers, ensured the computational efficiency and generalization performance of the model, and achieved accurate and efficient identification of difficult-to-detect foreign fibers.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122176361A_ABST
    Figure CN122176361A_ABST
Patent Text Reader

Abstract

This invention discloses a cotton foreign fiber identification method and system based on an improved RT-DETR. The improved RT-DETR cotton foreign fiber identification method includes: Step S1. Obtaining a cotton foreign fiber detection dataset; Step S2. Constructing an improved RT-DETR detection model, wherein the improved RT-DETR detection model uses RT-DETR-ResNet50 as the baseline model, and replaces the original feature pyramid network structure by embedding a cross-attention strip pooling module (CA-SPM) in the backbone network, designing a channel attention pyramid module (CAPN) in the neck network, and replacing the original loss function with an adaptive scale dynamic loss (ASD Loss); Step S3. Training the improved RT-DETR detection model using the dataset to obtain the optimal detection model; Step S4. Inputting the cotton image to be detected into the optimal detection model for detection, and the optimal detection model outputs the recognition result.
Need to check novelty before this filing date? Find Prior Art