Expert agent recipe management method for automated optical inspection models

CN122391742APending Publication Date: 2026-07-14ZHONGJIA MICROVISION (SHENZHEN) SEMICONDUCTOR TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHONGJIA MICROVISION (SHENZHEN) SEMICONDUCTOR TECHNOLOGY CO LTD
Filing Date
2026-04-30
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing automated optical inspection equipment has poor adaptability in its defect detection models, making it difficult to be compatible with different customers, different production line rhythms, and different process risks.

Method used

By generating multiple candidate expert agent formulations, and based on the current state of the automatic optical detection model, the optimal expert agent formulation is selected from the set beat index, set risk index, and set drift index, and the detection task is performed on the image to be detected.

Benefits of technology

The adaptive adjustment capability of the automated optical inspection model has been improved, enabling it to adapt to and be compatible with inspection scenarios of different customers, different production line rhythms, and different process risks, thereby enhancing the adaptability and compatibility of the inspection model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122391742A_ABST
    Figure CN122391742A_ABST
Patent Text Reader

Abstract

The present disclosure provides an expert agent recipe management method of an automatic optical detection model, and relates to the field of defect detection, which comprises: acquiring a detected image and a detection task; generating a plurality of candidate expert agent recipes based on an expert pool of the automatic optical detection model; based on a current state of the automatic optical detection model, selecting one candidate expert agent recipe as an optimal expert agent recipe from the plurality of candidate expert agent recipes for the detection task of the current state from a set beat index, a set risk index and a set drift index; the set beat index is used to evaluate the current task processing speed of the automatic optical detection model, the set risk index is used to evaluate the current task processing quality cost of the automatic optical detection model, and the set drift index is used to evaluate the current task processing stability of the automatic optical detection model; and performing the detection task based on the optimal expert agent recipe for the detected image. The adaptation compatibility of the automatic optical detection model is improved.
Need to check novelty before this filing date? Find Prior Art