In-Vehicle Occupant Vision Analysis With Edge Event Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer vision systems for home environments are either too simple and unpredictable or too complex and uneconomical, failing to provide accurate, scalable, and privacy-guaranteed real-time data analytics for detected people or objects.
Innovation Solution
A computer-vision system that generates a digital representation of people or objects from pixel streams, determines attributes, and controls networked devices, using an ASIC-based engine for real-time metadata processing without continuous video output.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sophisticated video analysis is performed on servers, then analysis accuracy is improved, but scalability and economic viability deteriorate due to linear scaling of storage and computational costs
Solution Approach 1:
The system segments video analysis into two parts: simple motion detection performed by individual cameras, and sophisticated analysis performed selectively by servers only when needed. This segmentation allows accurate analysis to be achieved without continuously processing all video streams on servers, thereby improving scalability while maintaining analysis accuracy for relevant events.
Solution Approach 2:
The system performs preliminary motion detection and filtering at the camera level before transmitting data to servers. By pre-processing video streams to identify only relevant motion events, the system reduces the computational burden on servers, enabling sophisticated analysis to be performed economically while maintaining high accuracy for detected events.
2Ease of manufacture
If simple video analysis is performed in cameras, then processing cost is reduced, but reliability and predictive capability deteriorate
Solution Approach 1:
The system introduces an intermediary layer that selectively transfers data between cameras and servers based on detected events. Simple cameras perform initial analysis and filter out normal activity, while the intermediary mechanism triggers server-based sophisticated analysis only for unusual or relevant events, thereby maintaining reliability without incurring continuous high processing costs.
3Loss of information
If full-frame video is transmitted to remote servers, then data completeness is improved, but scalability deteriorates due to linear growth of storage and computational requirements
Solution Approach 1:
The system extracts only the essential information from full video frames - specifically motion events and their characteristics - and transmits only this extracted data to servers. This extraction approach maintains data completeness for analysis purposes while dramatically reducing storage requirements and improving scalability by avoiding transmission of complete video frames.
4Speed
If continuous video streaming is implemented, then real-time monitoring capability is improved, but energy consumption and computational load increase
Solution Approach 1:
Instead of continuous video processing, the system implements periodic motion detection at camera level, triggering sophisticated analysis only when motion events occur. This periodic action approach maintains real-time monitoring capability for relevant events while significantly reducing energy consumption and computational load by avoiding continuous processing of all video streams.
Data Source
AI summary
A computer-vision-based monitoring system for monitoring an occupant of a road vehicle, including: a server located external to the vehicle and forming part of a distributed computing infrastructure; a first camera in the vehicle to capture an image of the environment external to the vehicle; a second camera in the vehicle to capture an image of occupants in the vehicle; a computer vision sub-system connected to the cameras in the vehicle, having an edge layer of the infrastructure; a vehicle occupant analysis sub-system trained using machine learning, wherein the vehicle occupant analysis sub-system uses its training to analyse the pose of a person and to detect if the person's head is directed towards the second camera to indicate that the person is giving attention to the camera.