Process Video Analysis for Context-Rich Anomaly Review
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to efficiently analyze performed processes from moving images captured by cameras, particularly in identifying anomalies and their associated contexts, which hinders effective management and optimization of production processes.
Innovation Solution
An image analysis device and method that utilize a controller and storage to detect anomaly periods in processes, extract relevant moving images, and present them for analysis, allowing for efficient identification and review of anomalies in production processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all video information is distributed to managers for review, then complete monitoring of production processes is achieved, but time and effort required for analysis increases significantly
Solution Approach 1:
The system extracts only abnormal video segments from the complete production process recordings. The anomaly detection unit identifies periods where process status deviates from predetermined references, and only these extracted abnormal portions are presented to managers for review, eliminating the need to analyze normal process videos.
Solution Approach 2:
An automated anomaly detection and extraction system acts as an intermediary between the complete video recording system and the manager. This intermediary automatically analyzes all videos, detects anomalies using multiple criteria (work time deviation, process status deviation, recognition errors), and presents only relevant segments to managers, resolving the contradiction between complete monitoring and time efficiency.
2Loss of time
If only abnormal condition videos are displayed, then analysis time is reduced, but context information about surrounding processes may be lost
Solution Approach 1:
The system performs preliminary action by automatically detecting and extracting abnormal video segments before presenting them to managers. The extraction process includes not only the abnormal segment itself but also surrounding context videos from adjacent process areas and time periods, preparing comprehensive contextual information in advance for manager review.
Solution Approach 2:
The video data is segmented into multiple components: the core abnormal video segment, context videos from neighboring process areas, and context videos from preceding and subsequent time periods. This segmentation allows the system to present organized, context-rich information packages to managers, maintaining temporal and spatial relationships of the anomaly within the production process.
3Area of stationary object
If multiple cameras capture process areas, then comprehensive coverage is achieved, but data processing and anomaly detection complexity increases
Solution Approach 1:
The anomaly detection unit is designed with multi-functionality to handle multiple data sources simultaneously. It can detect anomalies based on work time information, process status information, and recognition results from multiple cameras, using a unified detection framework that processes diverse input types through consistent anomaly criteria.
Solution Approach 2:
The system merges data from multiple cameras and multiple information sources (work time, process status, recognition results) into a unified anomaly detection process. By combining these diverse data streams and processing them through integrated anomaly detection criteria, the system achieves comprehensive monitoring while managing complexity through data integration rather than separate processing for each source.
Data Source
AI summary
An image analysis device for analyzing a moving image captured by one or more cameras in a site where a plurality of processes is performed, including: a storage that stores moving image data indicating a captured moving image of a process area in which each process is performed; and a controller that detects an anomaly period where a progress status of the processes deviates from a predetermined reference set for each process. The controller extracts, from the captured moving image, a first moving image of a first process area where a first process is performed in the anomaly period, and a second moving image related to the first moving image, based on the detected anomaly period. The second moving image includes one or both of image capturing results of a second process area located near the first process area and a second process performed before and/or after the first process.