节点文献
数字图像处理技术在尺寸公差检测中应用研究
【作者】 李军;
【导师】 李培芳;
【作者基本信息】 浙江大学 , 电气工程, 2005, 硕士
【摘要】 本课题是根据企业与市场的需求而进行的应用性研究,主要任务是采用计算机图像处理技术,根据CCD采集的物体图像,求取被测物体的有关尺寸与公差。 对采集的被测物体图像,首先通过直方图修正、小波去噪等技术,减少图像噪声,然后对处理后的图像进行阈值分割、数学形态处理、边缘提取等方法进行图像分割,最后通过Hough变换提取物体边缘包含的直线与圆方程,据此求出相关的尺寸公差。 课题分别以玻璃药瓶和电机机座作为研究对象,因为这二种对象,在工农业生产中具有代表性。对药瓶图像,通过Hough变换的方法,检测出边缘几条直线,据此求出尺寸公差与垂直度。对于电机机座图像,采用Hough变换方法,检测出图像包含的圆,进而求出同心度等公差。 论文的创新点是,提出了传统机械产品的非接触式测量方法,特别针对图像采集环节最容易出现的畸变现象,提出了一种严密的直接线性变换模型(DLT),该算法既不需要在影像上有框标,也不需要有摄影机内外方位参数的起始近似值,具有显著优点;提出了利用曲线的方程来确定公差的方法,对传统的Hough圆检测方法进行了改进,提高了检测速度,同时将圆检测用于存在多圆的图像,使多圆的同心度检测成为可能。
【Abstract】 Study to determine the dimension and tolerance of the measured object by using computer image processing technology was done.For the collected matter image processing was divided in steps: firstly, methods including Histogram Equilibrium, Wavelet Transform etc. were used to eliminate the noise in the image; secondly methods including Threshold Division, Mathematic Morphology, Edge Extraction were used to obtain division matter image; thirdly Hough Transform was used to detect the line or circle in the brim of the measured matter and finally the dimensions and tolerances could be calculated based on the line or circle equations.Glass medicine bottle and electric motor frame were chosen to be studied as they are typical representatives in industry. For bottle image, lines in the matter brim was detected by means of Hough Transform and then dimensions and perpendicularity tolerance were obtained. For motor frame, circles in the matter brim was detected by means of Hough Transform and then concentricity tolerance was determined.A novel no touch measuring method for traditional products was put forward and a precise model of Direct Linear Transform (DLT) was presented for the aberration in the images occurred at every turn. Rectangle mark tag and location parameter are no longer needed in this algorithms. A method of tolerance reckon by curve equation was also provided. In order to raise the operation speed Hough circle detection method was improved. In the meantime circle detection method to images with more than one circle was applied, so that the concentricity calculation is possible.
【Key words】 Image Processing; Tolerance Detection; Wavelet Transform; Hough Transform;
- 【网络出版投稿人】 浙江大学 【网络出版年期】2005年 05期
- 【分类号】TG801;TP391.41
- 【被引频次】13
- 【下载频次】501