The Vision System Algorithms In The Dsp-based Design And Realization

With the development of the technology of the artificial intelligence and automaton, computer vision begins to be applied to many technologic fields. This paper introduces a suit of integrated computer vision system. Including the introduction and analysis of kernel algorithm, the design and realization of the whole system. The main works involved in it are as follow:First, as the image segmentation is the base of image manipulation, in this paper, we use Zernike orthogonal moment method to realize edge detection of image, which is helpful to enhance the parameter precision of Hough transform. In use of the above method, we can get available edge points. Then, we deal with these points with deleting the outliers. At last, we combine these points to exactly estimate the curve parameters with Hough transform or Random Hough transform.Second, considering that almost algorithms in computer vision are much more complex and conventional Von Nenmann framework of computer can't process the image in real time. we design and realize computer vision algorithm part based on DSP of Tl. In image capture, we use VIDEO TVP3026-250CPCE chip which is matched with DSP TMS320C6711.DSP TMS320C6711 is chosen as the processing module, on which all program codes are written by C and assembly language. Also, the software is optimized to use the features of DSP. Inline-functions, linear assembly language are used to complete real time realizationTests and experiments demonstrate that the whole system can arrive the anticipate desire performance and have ability to process the image in real time.