Image Processing Anomaly Detection Using Feature Reconstruction Errors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods struggle to accurately identify images with a high proportion of anomaly parts, often misclassifying them as normal due to high reconstruction errors.
Innovation Solution
A method involving a feature quantity extraction process using a convolutional neural network, followed by reconstruction error calculation based on a difference from a predetermined average image, and threshold-based determination to distinguish normal and anomaly images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sparse coding anomaly detection is used to detect images with low anomaly proportion, then detection accuracy for low anomaly images is improved, but detection accuracy for high anomaly images deteriorates
Solution Approach 1:
The patent segments the anomaly detection process into two independent detection paths: one for detecting images with low anomaly proportions and another for detecting images with high anomaly proportions. Each path uses detection models trained on specific data characteristics, allowing accurate detection across the full range of anomaly proportions without the trade-off present in single-model approaches.
2Productivity
If reconstruction error is calculated based on difference from average normal image, then detection speed is improved, but detection accuracy for high anomaly proportion images deteriorates
Solution Approach 1:
The patent performs preliminary classification to determine whether the input image has a low or high anomaly proportion before applying the corresponding detection model. This preliminary action enables the system to select the appropriate detection path in advance, ensuring both fast processing and high accuracy by matching the detection method to the image characteristics.
3Device complexity
If single detection model is used for all anomaly proportions, then device complexity is reduced, but detection precision across varying anomaly proportions deteriorates
Solution Approach 1:
The patent creates a universal anomaly detection system that handles both low and high anomaly proportion images through a unified framework. The system uses a classification module to route images to appropriate detection models, achieving multi-functionality where a single system architecture can accurately detect anomalies across the entire spectrum of anomaly proportions.
Data Source
AI summary
A detection method with high robustness that can determine a feature of an input image regardless of characteristics of a detected part in the image is provided. An operation program causes a computer to perform a feature quantity acquiring step of acquiring a feature quantity which is extracted from an input image, a reconstruction error calculating step of calculating a reconstruction error on the basis of a difference between the acquired feature quantity and an average which is determined in advance on the basis of a normal image, and an output step of outputting a result of calculation in the reconstruction error calculating step.