Self-Checkout Fraud Detection Using Target Image Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection models for self-checkout registers exhibit irregular behaviors such as over-detection and under-detection when deployed in environments different from their training environment, and manually identifying target images for retraining is impractical due to high costs.
Innovation Solution
A system and method for automatically identifying target images using temporal and spatial consistency to adapt the object detection model to new environments, reducing the need for manual labor and costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an object detection model is trained using training data from a specific environment, then the model achieves accurate detection in that environment, but the model exhibits irregular behaviors such as over-detection and under-detection when deployed in different environments
Solution Approach 1:
The system performs preliminary identification of target images that cause irregular detection behaviors before retraining the model. By pre-identifying which images need to be added to training data through automated analysis of detection region positions and appearance probabilities, the system prepares the training data in advance, enabling the model to adapt to new environments without manual intervention.
Solution Approach 2:
The system uses feedback from detection results to automatically identify target images for retraining. By analyzing the statistical information of detection region positions and comparing appearance probabilities against thresholds, the system creates a feedback loop that continuously improves the model's adaptability to different environments through automated data selection for retraining.
2Reliability
If manually identifying target images for retraining is performed, then the model can be updated with accurate training data, but the process becomes impractical due to high costs and time consumption
Solution Approach 1:
The system performs self-service by automatically identifying target images without requiring manual human intervention. The automated process uses statistical analysis of detection region positions and appearance probabilities to select training data, eliminating the need for manual image review and significantly reducing time consumption while maintaining training data quality.
Solution Approach 2:
The system replaces the mechanical process of manual image review and selection with an automated computational process. By using algorithms to calculate statistical information, determine appearance probabilities, and identify target images, the system substitutes human labor with automated mechanisms, making the training process scalable and efficient.
3Reliability
If the object detection model is retrained with additional target images, then the model's detection accuracy improves in varying environments, but the complexity of the training process increases
Solution Approach 1:
The system segments the training process into distinct automated stages: obtaining detection results, calculating statistical information of detection region positions, determining appearance probabilities, identifying target images based on threshold comparison, and retraining the model. This segmentation of the training process into manageable, automated steps reduces overall complexity while improving detection accuracy.
Data Source
AI summary
A computer-readable recording medium having stored therein a fraud detection program causing a computer to execute a process including obtaining a result of object detection by inputting a target image group including a self-checkout-apparatus in an imaging range, into a model trained using a target image and an annotation, and performing fraud detection at the self-checkout-apparatus based on information about an item registered thereto and the result. The target image is identified by calculating statistical information of a position of a detection region of an object in each image in a first group based on positions by inputting the first group into the model, obtaining a position in each image in a second group using the model, and identifying the target image in which a region having an appearance probability equal to or less than a threshold is present, from the second group, based on the statistical information.