Development And Research On Image Processing And Interactive Interface Based On Embedded Linux

Image information processing and the application of processing results play an important role in computer information processing systems and their applications. Today, digital image processing attracts more and more attention, many good and mature algorithms have been proposed, but embedded systems of image processing catering to application are not exploited very well and could not meet the need of application, for a instance, the feathers of real time, mobility and integration should been developed further.Firstly, the thesis researches the implemented method of image processing systems based on desktop PC machine and the current development status of embedded image processing system. The Linux interactive compiled environment based on this system and the structure of multithread software have been completed through software and hardware design of embedded processing system and building of development environment.Secondly, resolution of falling-action detection for human based on vision and embedded system is introduced further with R&D of Video4Linux driver programming and software design of image capture subsystem. Capture of two valued foreground image depends on adaptive background subtraction algorithm based on predigested Single Gauss Model and falling action is estimated by proportion between width and height after morphology filter, moreover, false alarms is rejected. The experiment result shows that it has reached the feature of real-time and accuracy for the detection system.Then, in order to research the analysis and recognition of human body morphology using embedded surveillance system further, this paper presents a new algorithm of dynamic erosion contour extraction based on dynamic mathematical morphology. The algorithm, which dynamically selects the operational direction of structural element by defining the evaluating function, extracts the object contour directly with constraint conditions in dynamic mathematical morphology as judgment criterion, furthermore, adaptive threshold for filtering get the algorithm robust against the noise as well. Experimental results based on embedded surveillance system demonstrate that clear contour of people could be obtained through this algorithm.Finally, the thesis completed the research and develops of image processing which aims to Linux operation system using the Qt/Embedded, the structure of R&D of image processing interface include head file, implement file and main function has been designed in detail. This system completed the issue of Qt/Embedded programming on the ARM9 through the application on Virtual frame buffer of image processing interface.