A zero-shot anomaly detection method and system based on triple perception learning enhanced visual language model

By employing a triple-perception learning strategy, the spatial consistency and textual description capabilities of the visual language model are improved, generating accurate visual feature alignment. This addresses the limitations of CLIP in zero-shot anomaly detection and achieves efficient anomaly detection and localization.

CN121121767BActive Publication Date: 2026-02-17NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
2 Cites 2 Cited by

Patent Information

Application Number
CN202511658275.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-11-13
Publication Date
2026-02-17
Estimated Expiration
2045-11-13

AI Technical Summary

Technical Problem

The existing visual language model CLIP suffers from problems such as chaotic local feature space modeling, limited semantic description of textual anomalies, and dilution of local anomaly features in zero-shot anomaly detection, resulting in insufficient detection accuracy and robustness.

Method used

A triple-perception learning strategy is adopted, including a spatial awareness attention enhancement module (SAAM) to improve the spatial consistency of visual features, an attribute awareness guidance module (AAGM) to generate fine-grained descriptions of anomalous attributes, and an anomaly awareness reconstruction module (AARM) to dynamically guide the model to focus on potential anomalous regions. This is combined with a large language model (LLM) to generate accurate text prompts and visual feature alignment.

Benefits of technology

It significantly improves the accuracy and robustness of zero-shot anomaly detection, enabling precise anomaly detection and localization even without target domain training data, thus solving the application challenges in data-scarce scenarios.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application discloses a kind of based on triple perception learning enhanced visual language model's zero sample exception detection method and system, it is related to computer vision field, method includes: extracting global and local visual features from input image;Visual coding process is corrected local feature in deep network by spatial perception attention enhancement module, and fine-grained attribute text description is generated for abnormal visual feature;Through attribute perception guide module, attribute text description and general text prompt are deeply semantically aligned;The similarity of enhanced visual feature and optimized text feature is calculated, and pixel-level exception segmentation map is generated;Inference stage converts segmentation map into spatial attention weight by exception perception reconstruction module, and feedback is generated to visual encoder final global feature representation and calculates exception score.The method of the application significantly improves the accuracy and generalization ability of model in industrial defect detection and other scenarios without target domain training samples.
Need to check novelty before this filing date? Find Prior Art