不规则成像机器视觉系统探索
Study on a Machine Vision System with an Irregular Imaging Function
作者：苏真伟(四川大学 制造科学与工程学院 ,四川 成都 610065)；王瑾玮(四川大学 制造科学与工程学院 ,四川 成都 610065)；黄明飞(四川大学 制造科学与工程学院 ,四川 成都 610065)
Author：(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)；(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)；(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)
收稿日期：2007-12-20 年卷（期）页码：2008,40(6):165-169
期刊名称：工程科学与技术
Journal Name：Advanced Engineering Sciences
关键字：机器视觉系统；不规则图像；多分辨率；选择性视觉注意
Key words：machine vision system；irregular image；multi-resolution； visual attention
基金项目：国家自然科学基金资助项目(0875022/F030410 )
中文摘要
传统的机器视觉系统在检测中按照预先设定的采样频率与固定的分辨率采集图像。在大背景小目标的机器视觉检测中，图像中存在大量的冗余数据。提出了一种具有不规则成像功能的机器视觉系统，重点讨论了不规则图像的数据格式、不规则图像的拼接和内存处理三个问题。通过预注视机制，这种具有不规则成像功能的机器视觉系统可以使缺陷区域分辨率高，背景区域分辨率低，显著地减少图像处理的数据量，提高检测的精度和速度。
英文摘要
he existing machine vision systems sample images at a fixed resolution and a presetting speed in an inspection. In detection of small defects in a large background, there is usually a large amount of redundancy in the image data. Based on the development of low cost CMOS cameras and IEEE1394 data transmission technology, a machine vision system with an irregular imaging function is presented in the paper, including the irregular images’ data format, its mosaicing algorithm and memory distribution method of the image data. The system can irregularly sample images with a high resolution in the defect area and a low resolution in the background area, which reduces the total data in image processing and increases the precision and speed of the inspection significantly.
【关闭】
