文档介绍：南京财经大学
硕士学位论文
基于机器视觉的烟支在线检测系统的研究
姓名:吴晓飞
申请学位级别:硕士
专业:计算机应用技术
指导教师:李耀
2011-01-06
摘要
在卷烟厂的香烟包装流水线上,需要对烟包进行缺支、空头检测,剔除不合
格烟包。目前国内卷烟厂大都采用基于光电法的检测设备,存在检测元器件多、
检测信号易受烟丝颜色和周围环境中灰尘的影响等问题,实际使用中存在误判的
现象。
本文提出了一种基于机器视觉技术的烟支在线检测方法,从硬件和软件两方
面对系统进行设计,实现了对烟支缺支、空头的检测。
文中首先根据烟支缺支、空头图像的特点,采用数字图像处理的方法对烟包
缺支、空头图像的判断算法进行了研究。具体给出了图像除噪算法、缺支检测算
法和空头检测算法,并使用数码相机拍摄烟支图片,通过对算法进
行了验证。
其次给出了系统的硬件电路设计。选用基于 ARM7TDMI-S 内核的 32 位 RISC
微控制器 LPC2136;数字图像传感器 OV7620 采集烟支图像;先进先出的数据存
储器 AL422B 缓冲存储图像数据;大功率白光 LED 作为照明光源。
本系统的软件采用嵌入式 C 语言设计,文中给出了软件设计思想和流程图。
本课题来源于某卷烟检测设备制造有限公司,目前本设计已经在某卷烟厂进
行了现场试验,结果表明该系统检测速度快,可以实现缺支、空头判断,检测结
果与国家标准的规定吻合。

关键词:烟支检测;图像处理;LPC2136;OV7620

I
ABSTRACT
On the production line of cigarette packing, it’s need to detect the cigarette
packets whether the cigarette packets or the o is full or not, and remove
defective cigarette packets. At present, in domestic cigarette factories, they mostly use
the method of photoelectric conversion. There are many problems in this method such
as too many detecting devices, it easily influenced by surrounding dust and o’s
color. There is still erroneous judgement in our process of using it.
The paper proposes a method of cigarette detection which is online and based on
machine vision. And it describes the design of the hardware circuit and the software, it
finally achieves the target.
According to the features of the image in which cigarette packets or the o
is not full, this paper firstly studies on the algorithm of these images, which is used to
judge the images whether cigarette packets or the o is full or not. In the article,
it gives the algorithms of removing the noise, detecting cigarette packets and the
o which is not full. Then it uses digital cameras gathering the images, and
proves the algorithms through .
Then, the paper gives the design of the hardware circuit. The paper selects
LP