Image Processing System with Topic-Based Message Queue Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image recognition platforms face challenges in dynamically deploying multiple image recognition models, supporting multiple image sources, and ensuring real-time event detection with flexible configuration and scalability.
Innovation Solution
An image arrangement method and system that decodes a video stream into frames, buffers them in a message queue defined as a topic, and transmits the images to subscribed targets, allowing multiple image recognition models to process images in parallel and adapt to changing image sources and detection requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple image recognition models are deployed on a single platform to process different image sources, then the adaptability and versatility of the platform is improved, but the device complexity increases due to the need to manage multiple models and image sources simultaneously
Solution Approach 1:
The system segments the image processing workflow into distinct functional modules: video stream decoding, image buffering in message queues, topic-based routing, and model-specific processing. Each module handles a specific aspect of the pipeline, allowing multiple models to operate independently on the same platform without interfering with each other, thus managing complexity while maintaining versatility
Solution Approach 2:
The patent introduces a message queue system as an intermediary between the video stream decoder and multiple image recognition models. This intermediary component buffers images and routes them to subscribed models based on topics, eliminating the need for direct point-to-point connections between decoders and models, thereby reducing system complexity while enabling flexible adaptability
2Productivity
If multiple image recognition models process images in parallel from the same video stream, then the productivity and real-time processing capability are improved, but the computational load and resource consumption increase
Solution Approach 1:
The system merges the decoding function into a single shared component that decodes the video stream once and produces frames for multiple models. Instead of having each model have its own dedicated decoder, the single decoder serves all subscribed models through the message queue, reducing redundant decoding operations and lowering overall computational load while maintaining parallel processing capability
Solution Approach 2:
The message queue system serves multiple functions simultaneously: buffering images, routing to different models, managing subscriptions, and controlling flow. This universal component handles diverse processing needs without requiring separate specialized systems for each function, thereby improving productivity while optimizing resource utilization
3Adaptability or versatility
If the system supports dynamic deployment of multiple image recognition models with flexible configuration, then the adaptability is improved, but the ease of operation decreases due to the complexity of model deployment and configuration management
Solution Approach 1:
The system implements dynamic deployment capabilities where image recognition models can be subscribed to and unsubscribed from topics in real-time without requiring system restarts or complex reconfiguration. The message queue system dynamically routes images to currently subscribed models, allowing flexible adaptation to changing requirements while maintaining operational simplicity through automated routing
Data Source
AI summary
An image arrangement method and an image processing system are disclosed. In the method, a video stream is decoded into one or more frames of image. The image is buffered in a message queue. The message queue is defined as a first topic. The image in the message queue is transmitted according to a subscribed target of the first topic. Accordingly, the computation burden may be reduced, and the operation efficiency of multiple models may be improved.