Distributed Vision Processing for Low-Bandwidth Private Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vision systems face challenges in balancing bandwidth requirements, processor costs, and data privacy, particularly in IoT applications, where high-bandwidth transmission of raw image frames lacks privacy and costly processors are needed for complete image analysis.
Innovation Solution
A distributed networked architecture with sensor units performing image pre-processing, object detection, and data feature extraction to generate reduced datasets, which are then transmitted to a gateway processor for further analysis, ensuring data privacy and low bandwidth usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all image processing is performed on a main or cloud server, then bandwidth requirements increase, but processor costs at sensor level decrease
Solution Approach 1:
The vision system divides processing tasks between sensor-level units and central server. Sensor units perform initial image acquisition and generate reduced datasets containing only salient features, while the server performs comprehensive analysis. This segmentation reduces bandwidth consumption by transmitting only essential data rather than complete images.
Solution Approach 2:
Sensor units perform preliminary processing by identifying salient features and generating reduced datasets before transmission to the server. This preliminary action at the edge prepares data in advance, filtering out redundant information and reducing the bandwidth burden on the network infrastructure.
2Reliability
If sensors with built-in processing are used, then data privacy is improved, but processor costs increase
Solution Approach 1:
The system extracts only the essential salient features from images at the sensor level, removing redundant information before transmission. This extraction approach maintains data privacy by ensuring that only necessary data characteristics are transmitted to the server, while keeping the processing requirements at sensor level manageable.
3Measurement precision
If complete image frames are transmitted for processing, then processing accuracy is improved, but bandwidth requirements increase
Solution Approach 1:
The system extracts salient features from complete images at the sensor level, identifying and transmitting only the essential data characteristics needed for accurate object detection. This extraction maintains measurement precision by preserving critical object information while dramatically reducing the volume of data transmitted over the network.
Solution Approach 2:
The processing pipeline is segmented into feature extraction at sensor level and comprehensive analysis at server level. This segmentation allows transmission of compact feature data rather than complete images, reducing bandwidth requirements while maintaining the ability to achieve accurate detection results through subsequent server-side processing.
Data Source
AI summary
The present invention relates to a method 100 for object detection (140), recognition, classification and tracking using a distributed networked architecture comprising one or more sensor units (20) wherein the image acquisition and the initial feature extraction are performed and a gateway processor (30) for further data processing. The present invention also relates to a vision system (10) for object detection (140) wherein the method may be implemented, to the devices of the vision system (10), and to the algorithms implemented in the vision system (10) for executing the method acts.