On-site Anomaly Detection Device Using Local Image Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional anomaly detection systems in manufacturing sites face challenges in efficiently detecting anomalies on-site due to the need for transferring images over networks, which can be time-consuming and risks confidentiality, especially when dealing with sensitive products.
Innovation Solution
An anomaly detection system that performs on-site image analysis by capturing images, preprocessing them, and comparing them to pre-stored normal images to detect anomalies without relying on external networks, using a device with an image capture unit, inference unit, and display unit for real-time feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If images are transferred to an external server for anomaly detection, then anomaly detection can be performed using prestored normal images, but network transfer time increases and confidentiality risks arise
Solution Approach 1:
The image processing device performs multiple functions including anomaly detection, picture correction, and feature quantity calculation within a single on-site device, eliminating the need for external server communication while maintaining comprehensive inspection capabilities
Solution Approach 2:
The device uses an inference unit with pre-trained models as an intermediary to perform anomaly detection locally, acting as a mediator between the captured images and the analysis results without requiring network transfer to external servers
2Measurement precision
If images are transferred to an external server, then anomaly detection can be performed, but confidentiality of inspected pictures is compromised
Solution Approach 1:
The essential anomaly detection functionality is extracted from external servers and implemented within the on-site image processing device, allowing confidential images to remain within the secure facility while maintaining detection accuracy
Solution Approach 2:
The inference unit with pre-trained models serves as a local intermediary that processes confidential images without requiring external communication, protecting information security while enabling accurate anomaly detection
3Measurement precision
If picture correction is performed to improve image quality, then anomaly detection accuracy improves, but processing time increases
Solution Approach 1:
Picture correction and feature extraction are performed as preliminary actions before anomaly detection, with the inference unit pre-trained to handle corrected images efficiently, reducing overall processing time while maintaining accuracy
Solution Approach 2:
The processing pipeline is segmented into distinct stages (picture correction, feature quantity extraction, anomaly detection), allowing each stage to be optimized independently and processed efficiently in sequence
Data Source
AI summary
An anomaly display device provided with a starting information acquisition unit that acquires starting information including information for starting anomaly detection for detecting an anomaly in an image included in an input picture; an input picture acquisition unit that acquires the input picture; an anomaly detection unit that executes the anomaly detection, based on the acquired starting information, by comparing the acquired input picture with information based on a prestored normal picture; and a display unit that displays, in overlay on the input picture, information based on information detected by the anomaly detection unit.