Monitoring Device Dynamic Model Selection for Abnormality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional monitoring systems require separate devices for different types of abnormalities, making them complex and inefficient, as each device is designed for specific detection tasks such as traffic accidents or fires, necessitating multiple devices and increased memory capacity to store various models.
Innovation Solution
A monitoring device that captures images and determines the type of monitoring target using image classification, applying the appropriate abnormality detection model from a server, allowing for automatic detection of abnormalities without pre-storing multiple models, and enabling detection of multiple types of abnormalities within a single device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate devices are used for detecting different types of abnormalities, then detection accuracy for specific abnormalities is improved, but device complexity increases
Solution Approach 1:
The patent applies a single monitoring device that can detect multiple types of abnormalities (traffic accidents, smoke, fires) by using image classification to identify the type of monitoring target and selecting appropriate detection models dynamically. This multi-functional approach eliminates the need for separate dedicated devices for each abnormality type while maintaining detection accuracy through model specialization.
Solution Approach 2:
The system dynamically adapts its detection model based on the identified type of monitoring target in real-time. The determination unit classifies the image content, and the abnormality detection unit selects and applies the corresponding detection model accordingly. This dynamic model selection allows one device to perform multiple detection functions with high accuracy for each specific type.
2Adaptability or versatility
If multiple monitoring models are stored in the device, then the ability to detect various abnormalities is improved, but memory capacity requirements increase
Solution Approach 1:
The patent extracts the detection models from the device and stores them on an external server instead. The monitoring device only holds lightweight identification information about available models, while the actual detection models are retrieved from the server as needed. This significantly reduces the memory capacity requirements of the monitoring device while maintaining the ability to detect various abnormalities.
Solution Approach 2:
The server acts as an intermediary that stores the detection models and provides them to the monitoring device on demand. This intermediary architecture allows the monitoring device to access multiple specialized detection models without storing them locally, thereby reducing local storage requirements while maintaining versatile detection capability.
3Reliability
If dedicated devices are prepared for each monitoring purpose, then detection reliability is improved, but ease of operation deteriorates
Solution Approach 1:
The monitoring device provides a unified interface for monitoring multiple types of abnormalities through a single device. Users interact with one device that can detect traffic accidents, smoke, fires, and other abnormalities, eliminating the need to operate multiple separate dedicated devices. The system maintains reliability by selecting appropriate specialized models based on the detected situation type.
Data Source
AI summary
The monitoring device includes a captured image acquisition unit that captures a captured image of a monitoring target, a determination unit that determines a type of the monitoring target included in the captured image, an abnormality detection unit that detects an abnormality by applying the captured image to a monitoring model corresponding to the type of the monitoring target determined by the determination unit, the monitoring model being used to detect an abnormality related to the monitoring target included in the captured image, and an output unit that, when the abnormality is detected by the abnormality detection unit, performs an output related to detection of the abnormality. With such a configuration, it is possible to detect an abnormality using the monitoring model corresponding to the type of the monitoring target included in the captured image, and it is possible to perform abnormality detection according to the actually captured monitoring target.