Design Of Embedded Smart Camera Based On ZYNQ SoC

With the development of embedded technology,the traditional equipment designed by PC because of its large size,high cost,long hardware upgrade cycle,which makes it replaced by the solutions based on embedded processor.The development of semiconductor technology ensures that the embedded processing chip has enough system resources and high performance to meet the different needs of users for functional interfaces and processing speed.To this end,this article relies on embedded technology designing a image processing system with intelligent and good interactivity feature which is called smart cameraIn this paper,ZYNQ SoC is chosen as the hardware processing platform of smart camera system.According to the system design requirements,the whole system function structure is divided into three levels: user space,kernel space and underlying hardware.In the hardware layer,it's mainly to achieve the system modular design including USB and OV7725 dual channel image acquisition module,HDMI display module.In the kernel layer,it's mainly to transplant embedded Linux operating system and Qt graphical interface library,develop the underlying hardware device driver which provides interface functions for he user layer.In the user layer,it's mainly to achieve the target tracking and image classification module and design intelligent camera top application software interactive system.Finally,the smart camera system is used in face recognition tracking and industrial parts damage classification.The real-time tracking function based on the space-time context is started after the face recognition is successful,which effectively solves the problem of tracking error when the target is partially blocked.The real-time tracking rate is 8 frames/sec.Through the construction of the defect classification system based on BOW,it provides a new idea for defect classification.After testing,the classification system has strong robustness to the changes of scale and angle for the measured object.The accuracy of classification is more than 80%.