The Research And Realization Of Graphics Recognition Based On ARM Mobile Platform

The embedded system has the characteristics of miniaturization portability real-time. Because of meeting the needs of the development of smart appliances, the applications become more and more widely. Image recognition systems based on embedded system also have a rapidly development in security monitoring, industrial control, Electronics products. Based on the hardware platform putted up by self, The paper analyzed the characteristics of embedded system, and then discussed the algorithm characteristics, how to transplant it and how to make it better for hardware platform . Finally, there was a simple application of the algorithm with assembling the platform and compact car together.Author researched and realized a pattern recognition algorithm based on ARM platform after analyzing characteristics about embedded algorithm. Those pictures obtained form camera were identified and classified by the recognition algorithm correctly and rapidly. According to the result, S3C2440A controlled the compact car to do appropriate actions. Graphics for experiment are home-made traffic signs. Subjects of this research are color and shape characteristics. The thesis are as follows:(1) This paper expanded the structure and capability of hardware system and researched how to use registers and pins in mentioned chips. Then the construction and connection of hardware platform was completed. This paper introduced the features of WinCE software system briefly. WinCE 5.0 was transplanted to embedded platform. Finally, Camera driver was completed.(2) We segmented the color video-Graphics to a binary image without most of background using color information, extracted edges of binary image using log operator and extracted the single-pixel border of targets using thinning methods based on morphology. According to shape feature of the image which has a circular characteristic, we got the circumference and then extracted the region of interest. The pseudo-target was removed by setting the threshold of closed area.(3) According to conditions of the hardware, the author optimized the algorithms by converting float-point to fixed-point. This optimization could make the algorithm more efficient. And then experiments were finished on ARM mobile platform. Finally the paper presented ideas of technical improvements after analyzing results of experiments.The hardware system involved in this paper is composed of an ARM9 embedded board, which has a S3C2440A chip manufactured by SAMSUNG, a CMOS camera with a OV9650 chip, a motor driver circuit based on L298N chip, and a compact car with two motors. Software system used WinCE 5.0 system and VS2005 development environment.There was a detailed exposition of Graphics recognition used in ARM platform in this paper. Results from this paper could be some valuable references for other recognition systems based on embedded platform.