ANOMALITY DETECTION DEVICE AND ANOMALITY DETECTION METHOD

The anomaly detection device addresses false positives and delayed detection by using an inspection coefficient based on uninspected periods to accurately identify anomalies in security inspections.

DE112023006793T5Undetermined Publication Date: 2026-06-25MITSUBISHI ELECTRIC CORP
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
DE112023006793P0
Authority / Receiving Office
DE · DE
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-11-14
Filing Date
2023-11-14
Publication Date
2026-06-25

AI Technical Summary

Technical Problem

Existing anomaly detection methods in security inspections suffer from high recall rates leading to false positives, such as detecting unnecessary objects like fallen leaves, and can result in delayed detection due to abrupt changes in features between images.

Method used

An anomaly detection device that includes an object detection unit for identifying defect areas, calculates an inspection value using an inspection coefficient based on the length of uninspected periods, and determines anomalies based on a cumulative inspection value exceeding a threshold.

Benefits of technology

Reduces false detection of defects and effectively detects anomalies by adjusting the inspection coefficient based on uninspected periods, improving detection accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

An object detection unit (120) identifies an area where a defect in an inspection target is inferred to occur as a current defect area by performing object detection on a current target image acquired by imaging the inspection target, and derives an accuracy of the inference for the current defect area as a current-area confidence level. An inspection value calculation unit (140) calculates a current inspection value using an inspection coefficient, which is used to calculate an inspection value, and the current-area confidence level. An anomaly detection unit (150) determines, based on the current inspection value, whether or not an anomaly has occurred in the inspection target.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Target object recognition device

    JP2016012226A