| In the applications of computer image visual processing, the machine vision and imageinformation processing applications take very large proportion. In recent years, digital imageprocessing becomes the study hotspot which obtains the great attention and tremendousdevelopment. In the communication, management, medical, earthquake, meteorology,aerospace as well as education, it also plays a more and more important role. Because ofthat, many excellent theories and algorithms were created. But as for theapplication-oriented image processing system, it is not so satisfactory, which means it cannotfulfill the image application requirements. Traditional PC image processing system is hugevolume, lack of portability. Desktop operating system can not respond in real-time. So we cansay the traditional image processing method cannot satisfy the real-time and miniaturizationrequirements of today’s image application.For this, the thesis explores an image processing solution based on ARM9embeddedplatform. The result was applied in the field of artificial intelligence, automatically searchfield. The Embedded platform has high integration, good real-time performance and supportsmulti-tasking, which accords with our pursuit. It can face the increasingly complex imageapplications. Image processing system based on embedded platform is the future developmenttrend. Research on the combination of embedded platform and image processing has aguiding significance on rapid image application development.The function and designation of the whole system is analyzed on paper. The paperanalyzes some key problems on ARM embedded platform, such as the image segmentationand machine visual learning. It also included the system framework design, the developmentprocess design and the development principle designation, which could lead somesignificance to machine vision in embedded platform.This system is composed by hardware and software. For hardware part, TQ2440wasadopted as an experimented development platform. Some unnecessary hardware parts werecut according to the embedded system function’s demands. The procedure of softwaredevelopment includes completing the analysis of the boot process, compiling embeddedoperating system ARM-Linux according to the configuration of the image processingapplication, establishing cross-compile development environment, finishing the kernelconfiguration, BootLoader compiling, camera driver procedure and QT program. Then asegmentation algorithm (the weighted k-means clustering algorithm) was analyzed. At last,the Rubik’s cube reduction algorithm was successfully implemented. At the end of the paper, we pointed out the unlimited potential application based on ARMplatform of the digital image and machine vision and also pointed out the need of furtherresearch and exploration because of the time limitation. Finally, we put forward an outlook forfuture research. |
