Image Recognition System Hardware Software Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems face a tradeoff between software and hardware implementations, where software flexibility increases circuit scale and power consumption, while hardware implementation enhances performance but reduces flexibility and increases cost.
Innovation Solution
An image recognition system that separates processing steps into software and hardware components, with object detection, tracking, and recognition performed by hardware, and tracking point specification and behavior determination by software, utilizing a semiconductor integrated circuit with a CPU, DMA controller, and data bus to optimize circuit scale, power consumption, and processing performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the entire image recognition processing is implemented by software, then the system can flexibly adapt to various types of algorithms, but a high-performance CPU or multiple CPUs are required, increasing overall cost, limiting processing performance, and increasing power consumption
Solution Approach 1:
The image recognition processing is segmented into two parts: hardware processing for object detection, tracking target recognition, and tracking (high power-consuming operations), and software processing for tracking point specification and behavior determination. This segmentation allows power-intensive operations to be offloaded to energy-efficient hardware while maintaining algorithmic flexibility through software control.
Solution Approach 2:
A hardware processor acts as an intermediary between the image sensor and the software system. The hardware processor handles computationally intensive tasks (object detection, tracking) using dedicated circuits, while the software CPU handles higher-level decision-making (tracking point specification, behavior determination), thereby reducing the power burden on the main CPU while preserving flexibility.
2Productivity
If the entire image recognition processing is implemented by hardware, then processing performance is enhanced and power consumption is reduced, but the system loses flexibility and the circuit scale increases
Solution Approach 1:
Processing functions are segmented between hardware and software: hardware executes fixed, high-performance operations (detection, tracking), while software handles flexible, algorithm-dependent operations (tracking point specification, behavior determination). This segmentation achieves high processing performance for routine tasks while preserving flexibility for adaptive decision-making.
Solution Approach 2:
Different parts of the system have different qualities: the hardware processor provides high-speed, fixed-function processing for computationally intensive tasks, while the software CPU provides flexible, programmable processing for adaptive tasks. This local differentiation optimizes both performance and flexibility where needed.
3Measurement precision
If high-definition images are processed using software, then image quality is maintained, but power consumption significantly increases because a high-performance CPU with high frequency is indispensable
Solution Approach 1:
Image processing is segmented into hardware-accelerated operations (object detection, tracking) that handle the computationally intensive analysis of high-definition images, and software-based operations (tracking point specification, behavior determination) that require lower computational power. This segmentation enables high-definition processing without requiring a high-power CPU.
4Productivity
If more hardware components are added to enhance processing performance, then productivity increases, but device complexity and circuit scale increase
Solution Approach 1:
The hardware processor is designed as a multi-functional unit that can perform object detection, tracking target recognition, and tracking operations. This universal hardware component handles multiple processing tasks that would otherwise require separate dedicated circuits, thereby improving processing performance without proportionally increasing circuit scale.
Data Source
AI summary
An image recognition system for detecting and tracking at least an image portion associated with a predefined object from a moving picture is configured to be able to perform: an object detection processing step of detecting the object; a tracking point specification processing step of specifying a predetermined point as a tracking point; a tracking target recognition processing step of recognizing an actual tracking target based on the tracking point; a tracking processing step of tracking the tracking target; and a determination processing step of determining the type of the tracking target's behavior. The tracking point specification processing step and the determination processing step are implemented by software, while the object detection processing step, the tracking target recognition processing step, and the tracking processing step are implemented by hardware.


