Study On Embedded Real-time Image Processing Based On DM642

With the fast development of modern information technology, the information to be processed increases immensely. The researches and applications of image processing, especially real-time image processing are drawn great attention all over the world. In recent years, the development of DSP technology has put the theory achievement of digital signal processing field into practical application systems and pushed the progress of new theory and application. At the same time, it plays an important role in the field of image processing. The image processing system based on DSP has been applied in many kinds of practical fields.Firstly, the dissertation presents a real-time image processing system based on DM642, after introducing the schemes and characteristics of traditional image processing system, and analyzing the strongpoint of DSP and DSP system. This system can overcome the shortage of the traditional image processing system, improve its performance in the field of real-time, reduce the cost, and enhance the system's all-purpose application property. Image acquisition,image processing and image storage are integrated on one PCB: Image acquisition is implemented by ultra-low power video decoder and CF card; image processing is implemented by DM642; image storage is implemented by CF card; mass Flash and SDRAM are adopted in this system. Thereby, this system can be applied on almost all the occasion of surface defect detection, and all purpose property is incarnated completely.By the characteristics of bridge cable surface image, we designed area cropping to locate processing area, designed 3×3 window fast median filter to denoise, thereby, overcame time-complication of traditional median filter, after that, the target was separated from background by improved SOBEL edge detection, dynamic threshold and image morphology sequentially. Finally, reach the anticipated goal of bridge cable surface defect detection.In order to improve real-time property of the system, we adopted DSP/BIOS for real-time task scheduling, which has the character of tailor. According to the characteristics of application and kernel structure of DM642, we optimized code as much as possible. Finally, the summary and forecast of this system and the real-time image processing system based on DSP are narrated.