计算机视觉实时图像处理检测系统的设计与研究

【作者】 居里锴；

【导师】 赵琦； 皮德富；

【作者基本信息】 南京理工大学 ， 物理电子学， 2004， 硕士

【摘要】 计算机视觉技术的研究与应用随着时代的发展应运而生，并且在现代工农业生产中得到越来越广泛的应用。但对该项技术的要求也越来越高，特别是在速度上为了适应现代化大规模生产的需要，提出了高速实时在线检测的要求。本文是针对“计算机视觉高速在线式食品卫生异物识别与剔除系统”的课题中所需解决的计算机视觉高速实时图像检测系统提出了总体构想与系统框图，并进行了初步的设计与基础研究。 该课题主要解决的是啤酒瓶在灌装前的多项指标的检验，由于现今的灌装的速度可高达60000瓶／小时，单靠人眼是无法满足生产要求的。因此根据实际要求，本文设计了一套基于计算机视觉技术的在线检测系统，系统采用分布式高速视频信号实时采集与处理方法，解决了实时性与精确性的问题。整个检测系统的核心是针对图像处理的可重构并行处理器，该处理器采用了DSP TMS320C6201+FPGA的混合计算结构。其中由FPGA对图像进行预处理，包括图像的滤波、图像的分割、边缘提取等，留出更多的时间由高性能处理器DSP实现复杂的融合算法与图象识别，这样既具有制造完成后的可编程性，又能提供较高的计算性能。更多还原

【Abstract】 With the development of era, the research and application of computer vision technology emerges as the time requires. And it is more and more extensively applied in modern industry and agriculture. Meanwhile, the requisition on this technology is higher. Especially in order to meet the needs of large-scale modernized production, high-speed-online inspection is required. The present paper puts forward the overall idea and systematic framework of that inspection system which needs to be solved in the subject of "high-speed-online food hygiene discern and reject foreign matter system" . The preliminary design and basic research on it has been carried.The present subject is mainly to solve the inspection of the multiple indexes of beer bottles before bottled. Since the present speed of bottling reaches as high as 60000 bph, it is impossible for people to carry out the task with his eyes. For the practical production requirement, a set of online inspection system based on the computer vision technology is designed in this paper . The system has adopted the high-speed collecting and controlling method, which is able to meet the requisition of being accurate, The core of the inspection system is the reconfigurable computation processor for image-processing, which adopts the mixing calculating structure of DSP TMS320C6201+FPGA. During the process, the FPGA pre-treats the pictures, including image filtering, image segmentation, edge detection, etc. Hence, more time is left for the higher performance processor to do the complicated integrative algorithm and image recognition. In this way the production can be programmed and calculated.更多还原