Vehicle Edge Vision for Real-Time Person Trajectory Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer vision systems are unsuitable for real-time, scalable, and cost-effective processing of high-resolution video due to high computational requirements, leading to unscalable and uneconomical solutions, especially in home automation environments, and lack privacy guarantees.
Innovation Solution
A computer-vision system that generates a digital representation of objects from raw image data, determines attributes, and controls networked devices, using an ASIC-based engine for real-time metadata processing without continuous video output, enabling robust tracking and control of multiple objects with varying orientations and gestures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sophisticated video analysis algorithms are used on servers, then analysis accuracy is improved, but scalability and cost-effectiveness deteriorate due to linear scaling of transmission, storage and processing costs
Solution Approach 1:
The patent segments the video processing pipeline into two parts: simple preprocessing at the camera edge device, and sophisticated analysis only when needed at the server. This segmentation allows the system to maintain high accuracy for complex analysis while avoiding the cost of continuously transmitting and processing all video data centrally, thus resolving the scalability contradiction.
Solution Approach 2:
The patent applies preliminary action by performing basic video analysis and filtering at the edge device before transmission. This preprocessing step identifies and filters out irrelevant frames, so that only meaningful data is transmitted to the server for sophisticated analysis, reducing overall processing costs while maintaining accuracy where needed.
2Ease of manufacture
If simple video analysis is performed inside cameras, then processing cost is reduced, but analysis reliability and value deteriorate
Solution Approach 1:
The patent introduces an intermediary approach where the camera performs preliminary filtering and basic analysis, then selectively transmits data to the server for sophisticated analysis. This intermediary processing layer ensures that simple low-cost processing doesn't compromise reliability, as complex analysis is applied to the filtered results.
Solution Approach 2:
The camera performs preliminary action by conducting basic video analysis and filtering before transmission. This preliminary processing reduces the data load while maintaining reliability through selective transmission of meaningful data that will undergo further sophisticated analysis when needed.
3Loss of information
If continuous video transmission is used for analytics, then data availability is improved, but computational requirements and storage costs increase linearly
Solution Approach 1:
The patent extracts only the essential information from continuous video streams at the edge device, transmitting selectively filtered data to the server. This extraction approach maintains data availability for analytics while dramatically reducing computational requirements and storage costs by avoiding transmission of redundant continuous video data.
4Measurement precision
If high-resolution video processing is performed, then image quality is improved, but processing cost and power consumption increase
Solution Approach 1:
The patent segments processing tasks by performing basic high-resolution analysis at the edge and selectively transmitting only relevant high-quality data to the server. This segmentation maintains image quality where needed while distributing computational load to reduce overall power consumption.
Data Source
AI summary
A computer-vision-based monitoring system for a road vehicle, including: (i) a server located external to the road vehicle and forming part of a distributed computing infrastructure; (ii) a first camera positioned in, or configured to be attached to, the road vehicle and configured to capture an image of the environment external to the road vehicle; (iii) a computer vision sub-system connected to the first camera, in which the computer vision sub-system is configured to be locatable in the road vehicle, and forms or includes at least some of an edge layer of the distributed computing infrastructure; (iv) a person tracking sub-system that is connected to, or is part of, the computer vision sub-system, the person tracking sub-system having been trained using machine learning, wherein the person tracking sub-system is configured to use its training using the machine learning to analyse a trajectory of a person approaching the road vehicle.