Research And Implementation Of Smart Camera System Based On Davinci Technology

IC(Integrated Circuit) industry is the fundamental and strategic industry of the national economy, whose technology level and industry scale is an important symbol of a national science and technological strength. As a kind of non-contact and high-precision intelligent sensing technology, machine vision is being widely used in a variety of IC packaging equipments, and its performance directly affects the production efficiency and the product quality of IC packaging equipment. The traditional PC-based machine vision system has many defaults, such as large volume, complex struture, poor stability, low degree of automation, which demands professional machine vision engineers to develop the vision system. According to the application requirements of intelligent, high-speed and networked vision system of IC packaging equipment, this paper proposed an image matching algorithm of vision positioning method based on phase correlation, and developed a kind of industrial smart camera based on Davinci technology, and carried out some application tests in RFID(Radio Frequency Identification) package equipment.(1) According to the functional requirements of the vision positioning system of typical IC packaging equipment, a kind of industrial smart camera scheme was proposed, in which the CMOS image sensor as image acquistion unit, DM6467 T chip as the main processing unit and Gigabit Ethernet as transmission unit ware adopted. The industrial smart camera hardware system for IC package equipment was invented, and the relevant debugging and tests work were carried out. This smart camera has realized real-time acquisition, processing and transmission function of 200 W pixels image at 30 fps.(2) In order to solve the time-consuming problem of the traditional spatial correlation image matching algorithm, a kind of image matching algorithm based on phase correlation,which could meet the visual positioning requirements of IC packaging equipment, was proposed in this paper. And some experiments were conducted to test the speed, accuracy and rubustness of the algorithm. The experiments result indicated that the average maching time is less than 20 ms when the model image is 128×128 pixels and the object image is 512×512pixels, and the positioning accuracy is less than 0.5 pixel, and the error in rotation is less than0.2 degree. The proposed algorithm can estimate position and rotation angle of IC chips with high accuracy and fast speed even under noise and defocus blur influence.(3) With the Linux-based industrial smart camera's software system architecture designed, the migration process of Bootloader and Linux operating systems were described,the development of video capture program based on V4L2 framework was completed, the visual positioning algorithm flow based on Codec Engine architecture was analysed, a network communication protocol based on GigE Vision was established, and the remoteoperation of smart camera was realized finally.