Edge-Server Image Recognition System for Nursing Care Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems using deep learning for monitoring individuals, such as in nursing-care facilities, face a significant load on server devices, requiring high-performance hardware to process images from multiple sensor devices in real-time, which can be inefficient and costly.
Innovation Solution
An image recognition system where sensor devices perform initial image processing and transmit partial images or feature maps to a server device, allowing the server to perform more accurate deep learning-based recognition only when necessary, thereby reducing the server's workload and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning-based image recognition is performed on all images from multiple sensor devices, then image recognition accuracy increases, but server device load increases significantly
Solution Approach 1:
The system segments the image processing workflow into two stages: first, sensor devices perform initial recognition on captured images; second, only images where recognition results meet a predetermined condition are transmitted to the server for deep learning-based re-recognition. This segmentation reduces the volume of images requiring server processing while maintaining high accuracy for critical cases.
Solution Approach 2:
Different processing qualities are applied to different images based on their characteristics. Images meeting the predetermined condition (e.g., showing specific actions or anomalies) receive high-quality deep learning processing on the server, while other images undergo only basic local recognition at the sensor device, optimizing resource allocation based on local image quality requirements.
2Measurement precision
If deep learning-based image recognition is performed on all images, then recognition accuracy improves, but processing time increases
Solution Approach 1:
The sensor device performs preliminary image recognition and filtering before images are sent to the server. By pre-processing images locally and identifying only those meeting predetermined conditions, the system eliminates unnecessary deep learning processing for images that don't require it, significantly reducing overall processing time while maintaining accuracy for relevant cases.
Solution Approach 2:
Instead of applying full deep learning processing to all images, the system applies partial processing (basic recognition) to most images and reserves excessive/high-level processing (deep learning) only for images where it is truly needed. This selective approach reduces total processing time while maintaining high accuracy for critical recognition tasks.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image recognition system includes a plurality of sensor devices and a server device. Each of the plurality of sensor devices includes a first image recognition unit and a transmission data generation unit. The first image recognition unit performs an image recognition process on time-series images captured in chronological order, and estimates whether a predetermined event has occurred in an imaging target captured in the time-series images. On condition that the first image recognition unit has made the estimation that the predetermined event has occurred in the imaging target, the transmission data generation unit generates, as transmission data to be transmitted to the server device, first time-series images obtained by partial images including the image of the imaging target being cut out from respective images included in the time-series images. The server device includes a second image recognition unit. The second image recognition unit performs, on the transmission data, an image recognition process with image recognition accuracy higher than that of the first image recognition unit, and estimates whether the predetermined event has occurred in the imaging target.