Method for accurately identifying and evaluating quality of AI seed variety based on attention mechanism

CN122090150APending Publication Date: 2026-05-26山东电子职业技术学院
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
山东电子职业技术学院
Filing Date
2026-02-11
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Existing seed identification technologies cannot effectively extract structured information within seed batches, resulting in insufficient ability to distinguish similar varieties. Furthermore, the quality assessment process suffers from low system integration and computational redundancy.

Method used

An AI-based seed variety identification method based on attention mechanism is adopted. Individual seed images are extracted through instance segmentation algorithm, feature vectors are constructed by combining morphological and texture features, a seed association graph structure is built, and graph attention network is used to adaptively learn the importance weights of neighboring seeds to achieve variety classification and quality assessment.

Benefits of technology

It achieves full-process intelligent and automated seed testing, improves testing efficiency and accuracy, eliminates the influence of human factors, ensures the objectivity and repeatability of test results, and has good adaptability and generalization ability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122090150A_ABST
    Figure CN122090150A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of seed detection technology and discloses an AI-based method for accurate seed variety identification and quality assessment based on an attention mechanism. The method includes: acquiring a group image of a target batch of seeds and extracting an individual image of each seed; sequentially extracting morphological and texture features from each individual image, calculating the feature similarity between any two seeds, constructing a seed association graph structure, and determining the neighboring seeds of each seed; adaptively learning the importance weights of each neighboring seed, generating an enhanced seed representation for each seed, and classifying all seeds into different variety categories; sequentially extracting the quality features of each seed from the enhanced seed representation, identifying inferior seeds in each variety category, calculating batch quality indicators, and generating a quality assessment report. This invention can effectively ensure the objectivity and repeatability of the detection results and has good adaptability and generalization ability for different varieties and batches of seeds.
Need to check novelty before this filing date? Find Prior Art