Behavior Recognition Device Abnormality Transmission via Segmented Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for identifying work performed by individuals in video images are infrequently updated, leading to difficulties in improving identification accuracy due to parallel development with Web applications, resulting in inefficient monitoring and analysis.
Innovation Solution
A system comprising a behavior recognition device and a cloud server that acquires video images from cameras, analyzes them using a hidden semi-Markov model to detect abnormal behaviors, and transmits identified abnormal frames with corresponding categories, enabling real-time monitoring and improving identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are developed in parallel with Web applications, then both systems can be provided consistently, but the machine learning models are infrequently updated and identification accuracy cannot be improved
Solution Approach 1:
The system divides the machine learning model into two independent components: a behavior recognition model that detects abnormal behaviors and a Web application for monitoring. This segmentation allows the behavior recognition model to be updated independently and more frequently without being constrained by the Web application development cycle, thereby improving identification accuracy while maintaining consistent service provision.
Solution Approach 2:
The patent introduces a behavior recognition device as an intermediary between video input and the Web application. This device continuously analyzes video images using the behavior recognition model and transmits abnormal behavior information to the Web application. This intermediary architecture enables the model to be updated independently and more frequently, resolving the contradiction between update frequency and identification accuracy.
2Measurement precision
If the machine learning model is updated infrequently, then system stability is maintained, but identification accuracy of work performed by persons cannot be improved
Solution Approach 1:
The system segments the model update process from the Web application deployment process. The behavior recognition model can be trained and updated independently with new data, and these updates are applied to the behavior recognition device without requiring a full system redeployment. This enables more frequent model updates while maintaining system stability.
Solution Approach 2:
The patent implements a dynamic model update mechanism where the behavior recognition model can be continuously improved through retraining with new video data. The updated model is then applied to the behavior recognition device, allowing the system to adapt to new patterns and improve identification accuracy over time without disrupting the stable Web application service.
3Measurement precision
If video images are analyzed in detail to improve identification accuracy, then abnormal behaviors can be detected more accurately, but processing time and computational resources increase
Solution Approach 1:
The system extracts and transmits only the essential information about abnormal behaviors to the Web application, rather than transmitting entire video images or detailed analysis results. The behavior recognition device identifies abnormal behaviors, extracts key features, and sends condensed information including abnormal behavior type and location. This extraction approach maintains high detection accuracy while significantly reducing processing time and communication overhead.
Solution Approach 2:
The patent applies partial action by focusing computational resources only on detecting abnormal behaviors rather than analyzing all video content in detail. The behavior recognition model is trained to identify specific abnormal patterns, allowing the system to achieve high detection accuracy for abnormal behaviors while using minimal processing time. Normal behaviors are not subjected to detailed analysis, reducing overall computational load.
Data Source
AI summary
A behavior recognition device acquires a video image in which a person is captured, and determines, by analyzing the acquired video image, whether or not an elemental behavior performed by the person is abnormal for each section that is obtained by dividing the video image. When the behavior recognition device determined that the elemental behavior is abnormal, the behavior recognition device extracts, from the acquired video image, the video image included in the section in which the elemental behavior is determined to be abnormal. The behavior recognition device transmits, in an associated manner, the extracted video image included in the section and a category of the elemental behavior that is determined to be abnormal.