Panel anomaly detection method and apparatus, electronic device, and storage medium

By preprocessing and clustering the display screen images, structural family prototypes are determined for anomaly scoring, solving the accuracy problem of panel anomaly detection in existing technologies and achieving efficient panel anomaly detection and interpretable detection results.

CN122415445APending Publication Date: 2026-07-17ZHONGJIA 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
2026-03-19
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

The accuracy of existing technologies in detecting panel anomalies during display manufacturing is affected by fluctuations in processes, materials, and structures, leading to false positives and false negatives. In particular, detection methods based on fixed templates are highly sensitive, while methods based on labeled data are difficult to transfer across batches.

Method used

By preprocessing the panel image, standard image patches with the same size, angle, and phase are obtained. Multi-channel structural features are extracted and clustered to determine the prototype of the structural family. Anomaly scoring is performed based on the prototype, and the location information and evidence package of the abnormal region are output to achieve interpretability of the detection results.

Benefits of technology

It improves the accuracy of panel anomaly detection, reduces false detections and missed detections, and avoids detection errors caused by process, material and structural fluctuations by replacing fixed templates with structural family prototypes, thereby improving the interpretability of detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122415445A_ABST
    Figure CN122415445A_ABST
Patent Text Reader

Abstract

本公开提供了一种面板异常检测方法、装置、电子设备和存储介质,该方法包括对面板图像进行预处理,得到尺寸、角度和相位相同的多个标准图像块,提取各标准图像块中多个子区域各自的多通道结构特征,并融合各标准图像块中同一子区域的所述多通道结构特征,得到各标准图像块的多个特征向量,对多个标准图像块的多个特征向量进行聚类,得到多个结构族,并确定各结构族的原型,基于多个结构族和各结构族的原型,对多个待测标准图像块中的每个子区域进行异常评分,将多个待测标准图像块对应的所有异常评分中最高的N个评分对应的子区域,确定为异常区域,并输出检测结果。该方法能够提高面板异常检测的准确性以及检测结果的可解释性。
Need to check novelty before this filing date? Find Prior Art