【作者】 陆志敏；

【导师】 王强；

【作者基本信息】 广西师范大学 ， 无线电电子学， 2000， 硕士

【摘要】 在玻璃制品的质量检测过程中，随着生产规模的扩大、生产速度的提高以及越来越严格的质量要求，传统的人工检测方法已经无法胜任。在这种情况下，国外不少厂家已经开始研制用于玻璃瓶罐质量检测的机器。我国在玻璃瓶罐质量检测机器的研制方面相对落后，目前国内有些厂家也正在研制用于玻璃瓶罐质量检测的机器，他们一般都是仿制国外产品，研制工作尚在进行。从国外已开发的产品来看，在玻璃瓶罐外形尺寸的检测方面，一般都采用机械接触方式，而这种方式要求很高的机械制造工艺水平。 作者设计的玻璃瓶罐外形尺寸的计算机视觉检测系统是广西师范大学电子技术研究所与桂林市玻璃厂协作开发的“玻璃制品的计算机视觉在线检测系统”的一个子系统。该系统避开了我国机械制造工艺水平低这一弱点，采用了非接触的传感方式，利用计算机视觉和图像处理技术来检测玻璃瓶罐外形尺寸。检测的内容为：瓶口的内径、外径，瓶子的高度，瓶的垂直度。 检测系统对一个被检瓶子进行外形尺寸检测时，需要用两台摄像机分别采集两幅图像。一幅是瓶口图像，由工业摄像机垂直对着瓶口拍摄而得，用来检测瓶口的内径、外径和瓶子的垂直度是否合格；另一幅是瓶高图像，由工业摄像机沿水平方向对着瓶子的上半部拍摄而得，用来检测瓶子的高度是否合格。系统采用外触发方式控制摄像机进行图像采集，即当被检瓶到达检测工位时，外部触发电路产生一个触发信号送到图像采集卡。计算机检测到外触发信号，立即控制摄像机进行图像采集。系统采用先定标后检测的方法，即先用标准瓶子的外形尺寸来确定一个标准尺寸，检测时，被检瓶的尺寸与标准尺寸相比较，看其偏差是否在允许的范围之内，以决定该被检瓶子的外形尺寸是否合格。 系统软件由两大功能模块组成，一个是瓶口图像处理模块，另一个是瓶高图像处理模块。瓶口图像处理模块包括瓶口图像的采集，图像的边缘检测，瓶口内径外径所对应的内圆外圆的困检测，瓶口内径外径的尺寸分析及垂直度分析。瓶高图像处理模块包括瓶高图像的采集，瓶子轮廓边缘的检测，瓶口上边缘所在行的确定，高度的合格分析。在对瓶口图像和瓶高图像进行边缘检测时，都采用了灰度阈值分割提取边缘的方法，而不是采用边缘检测算子来检测边缘。在对瓶口图像中的瓶口内圆和外圆进行圆检测时，作者提出了半分法和弦的垂直平分线求圆心法两种方法，并通过实验对比，决定采用半分法来检测瓶口内圆和外圆。在软件开发的整个过程中，作者都从速度和效果两个方面来设计算法和编写程序的。 检测系统的制造代价小，对机械制造的精度要求低。且系统的检测速度可随CPU速度的提高而不断提高。 作者用Visual C++完成了玻璃瓶罐外形尺寸检测的软件开发工作。检测系统在实验阶段较成功地实现了玻璃瓶罐外形尺寸的检测。更多还原

【Abstract】 In the course of the quality detection of the glass products, with the enlarging of the production scale. the enhancing of production speed itnd more and more severe requests of quality, the artificial detection on the production line hasn’t traditionally been competent. In such case, many foreign factories have began to manufacture the machine that can detect the quality of glassware. in the manufacture of the machine, our country drops behind overseas. Now some domestic factories are manufacturing the machine. They commonly copy foreign product, and the jobs are still carrying through. At the side of shape size detection of glass bottles and jars, the foreign products adopt mechanism-touched mode, which request high level of mechanism manufacture technics.The system designed by author is a subsystem of on-line detecting system of glassware with computer vision. We cooperate with Guilin Glass Factory to develop this system. This system adopts none-touched sensors, and makes the use of computer vision and image process to detect the shape size of the glass bottles and jars, avoiding the weakness of mechanism manufacture technics in our country. The detection content includes the inside radius and outside radius of the bottle mouth, the height, and the perpendicularity of the bottle.When detecting the shape size of every bottle, the detection system needs two vidicons to get two images. One is the bottle mouthimage screened by the vidicon facing perpendicularly to the bottle mouth, which is used to detect the inside radius and outside radius of the bottle mouth, and perpendicularity of the bottle. Another is the bottle tall image screened by the vidicon facing to the top part of the bottle, which is used to detect the height of the bottle. The system adopts outer trigger mode to control vidicons to get the images. When the detected bottle reaches the detection place, the outer trigger circuit produces a trigger signal and sends it to the image card. The computer detects the trigger signal and immediately controls the vidicons to get the images. The system uses shape size of a standard bottle to confirm a standard size before detection. The size of the inspected bottle is compared to the standard size when the inspection happens. If the warp is between the allowable ranges, the shape size is qualified. Otherwise, the size is not qualified.The system software is marie up of two function blocks. One is the bottle mouth image processing block. Another is the bottle height image processing block. Bottle mouth image processing block is consisted of getting the bottle mouth image, edge detection, circle detection, and analysis of the size of mouth radius and perpendicularity. Bottle height image processing block is consisted of getting the bottle height image, edge detection of the bottle profile, confirming the row of top edge of the bottle mouth, and the analysis of the height. Segmentation based on Grey-level threshold instead of edge detection operator is used to detect the edge in both the bottle mouth image and the bottle height image. The author brings forward t~o techniques, half-divide technique and circle detection b perpendicular bisector of chord, to detect the inside circle and outside circle of2bottle mouth in bottle mouth image. After the contrast in experiments, half-divide technique is used to detect circles. In the course of software development, the author designs the arithmetic and writes the program in both the aspect of speed and the aspect of effect.This detection system has the virtue of low manufacturing price and low precision to mechanical manufacturing. Detecting speed of the system can be enhanced with the increase of CPU speed. The author has finished the software development job of the shape size detection of bottles and jars with Visual C++. The detection system has successfully realized the shape size detection of bottles and jars.更多还原