Recognition System Of The Moving Number Label Implement And Optimization Based On DM642

With the vigorous development of DSP technology,DSP chips used in a variety of embedded intelligent systems more and more widely. The most typical application is to be used in real-time video processing system. The thesis is to design the identification system of the movement digital lab based on a high-performance DSP digital signal processor DM642 of TI Company.The main respects of the subject is listed in the following:Firstly, the drive program of captureing image used the structure of Micro-driver and Class-driver . In order to improve development efficiency, in the DSP / BIOS operating system, Through compiling the module program of the micro-drive channel, I2C-bus configuration,SAA7115configuration,EDMA configuration and VP0 configuration,the subject achieve the Micro-driver; Through compiling the FVID module function, the subject achieve Class-driver and the real-time image capturing. Secondly, after Analyzing the fuzzy images captured, the subject restitute the fuzzy images by using direct inverse filter algorithm, based on the comparison of several image restoration algorithm and recognition algorithm. the subject use improved BP algorithm to train samples, and recognize the numbers in the label by trained network. Finally, the subject transplant the whole recognition system to RF5 structure, making the structure of the whole program more standardized. The subject optimize the algorithm mainly using dual-channel balance between data resources, cycle expansion, loop mergence and so on, make the running speed of the whole system improve greatly.Debuging and simulation experiment after the whole system designed, Through analyzing the experimental data. The results show that the system can real-time recognize the numbers in moving label, achieve the desired goal. The system of the moving number label recognition can process 2 number label images per second.