Research On Image Processing And Feature Recognition Method Based On Machine Vision

Basically, the current machine vision system is built on top of the computer system, with high-speed, high accuracy, high degree of automation features. However, due to computer vision system require higher hardware and software environment, larger bulk, higher cost, the range of applications subject to certain restrictions.With the development of new microprocessor, the new technology used in general computer was transplanted to the embedded system, that brings the rapid development of the further improvement of embedded hardware and software platform. Embedded systems have some advantages such as low power consumption, small size, low price, and flexible usage. The machine vision system combined embedded technology and machine vision technology, not only expands the application of machine vision, but also trends a new direction of machine vision's research and development.The purpose of this paper is to design a real-time, hardware and software reconfigurable embedded image processing system in a single FPGA chip using SOPC technology. This paper bases on the basic digital image processing algorithms, combines parallel processing technology, completes the original image edge detection with FPGA logic resources, thus achieving the hardware acceleration of image processing. With the usage of Nios II soft-core processor, this paper completes the region labeling and feature recognition based on the edge of image, ultimately display the processing result through the LCD.This paper build a real-time image processing platform based on Altera FPGA's SOPC technology, detail the various functional modules'design, structure design, work flow and the process of system integration and build. To improve real-time, this paper combines hardware acceleration, algorithm optimization, structural improvements and frequency adjustment, which improves system performance.The results presented in this paper to analyze the system's processing power, and the function of embedded real-time image processing and feature recognition is ultimately achieved.