發(fā)布時間：2019-03-05 15:30

【摘要】：隨著機械加工行業(yè)自動化發(fā)展進程的不斷推進,傳統(tǒng)的測量技術(shù)已不能充分適應(yīng)當前的發(fā)展需求,機械加工行業(yè)對加工工件的尺寸測量也提出了更高的要求。本文所研究的基于機器視覺的小型工件測量系統(tǒng),主要解決人工測量速度慢,長時間測量易疲勞,測量精度不高,以及難以實現(xiàn)測量自動化等問題。本文首先根據(jù)課題需要確定系統(tǒng)硬件方案,分析研究相機、鏡頭、光源等主要硬件的工作原理。為了突出被測工件的輪廓邊界,同時避免金屬工件表面的反光,本系統(tǒng)選用了 LED環(huán)形無影光源照亮工件的輪廓。另外確定了相機、鏡頭等硬件設(shè)備的具體型號,構(gòu)建了測量系統(tǒng)的整體硬件平臺,完成了相機標定和被測工件的圖像拍攝工作。利用Matlab 8.0進行程序開發(fā),整體流程分為五個主要部分。一是圖像濾波,經(jīng)過對比分析,使用了中值濾波算法,在不破壞工件邊緣信息的前提下較好的濾出了圖像中的部分噪聲。二是圖像二值化處理,本系統(tǒng)采用了直方圖閾值法確定閾值,將圖像的邊緣和背景清晰的區(qū)分開來。三是圖像邊緣提取,本系統(tǒng)采用了 Canny算子進行邊緣提取,有效的提取出了圖像的邊緣信息,包括工件真實的邊緣信息和噪聲的邊緣信息。四是噪聲邊緣地消除,在這部分中本文提出了提出了二值圖像連續(xù)亮點區(qū)域聚合算法。該算法先將各個邊緣像素點,包括噪聲邊緣像素點和工件有效邊緣像素點,分別聚合在各自集合中,再跟據(jù)噪聲邊緣和工件有效邊緣之間的像素個數(shù)差別,判斷并消除噪聲邊緣點,最終實現(xiàn)了工件真實邊緣的有效提取。五是邊緣分段擬合,本文提出了映射分段算法,該算法利用圖像拐點前后點的坐標斜率差值較大的特點來尋找邊緣的拐點,準確有效的實現(xiàn)了連續(xù)邊緣的分段。最后本文所涉及到的邊主要是直線和圓弧,直接用直線和圓對各邊進行擬合,再求取各條邊的長度,完成測量工作。從實驗結(jié)果可以看出,本文所設(shè)計組建的基于機器視覺的小型工件測量系統(tǒng)能夠?qū)崿F(xiàn)實時,自動化,高精度測量。
[Abstract]:With the development of automation in machining industry, the traditional measurement technology can no longer fully adapt to the current development needs, and the machining industry also put forward higher requirements for the size measurement of machined workpiece. The small workpiece measurement system based on machine vision studied in this paper mainly solves the problems of slow manual measurement, easy fatigue in long time measurement, low measurement precision, and difficulty to realize measurement automation and so on. In this paper, the hardware scheme of the system is determined according to the need of the project, and the working principle of the main hardware such as camera, lens and light source is analyzed. In order to highlight the contour boundary of the measured workpiece and avoid the reflection of the surface of the metal workpiece, the LED ring shadowless light source is selected to illuminate the contour of the workpiece. In addition, the specific models of the camera, lens and other hardware equipment are determined, the whole hardware platform of the measurement system is constructed, and the camera calibration and the image photographing of the measured workpiece are completed. Using Matlab 8.0 to develop the program, the whole process is divided into five main parts. The first is image filtering, after comparison and analysis, the median filtering algorithm is used to filter the partial noise in the image without destroying the edge information of the workpiece. The second is image binarization. In this system, histogram threshold method is used to determine the threshold value, and the edge and background of the image are clearly distinguished. Third, image edge extraction, the system uses the Canny operator for edge extraction, effective extraction of image edge information, including the real edge information of the workpiece and noise edge information. Fourth, the edge of noise is eliminated. In this part, a binary image continuous bright spot region aggregation algorithm is proposed. Firstly, each edge pixel, including noise edge pixel and workpiece effective edge pixel, is aggregated in each set, and then the number of pixels between the noise edge and the effective edge of the workpiece is different from the number of pixels between the noise edge and the effective edge of the workpiece. Finally, the real edge of the workpiece can be extracted effectively by judging and eliminating the noise edge points. Fifth, the edge segment fitting, this paper proposes a mapping segmentation algorithm, this algorithm uses the image inflection point before and after the coordinates of the slope difference is large to find the edge of the inflection point, accurate and effective realization of the continuous edge segmentation. Finally, the edges involved in this paper are mainly straight lines and arcs, which are fitted directly by straight lines and circles, and then the length of each edge is obtained to complete the measurement work. From the experimental results, it can be seen that the small workpiece measurement system based on machine vision designed in this paper can realize real-time, automatic and high-precision measurement.
【學位授予單位】：西南交通大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TG806;TP391.41

【參考文獻】

相關(guān)期刊論文 前10條

1 黃曉波;;機械加工工藝對加工精度的影響[J];裝備制造技術(shù);2012年09期

2 向大學;劉明生;;機械加工誤差產(chǎn)生的原因及改進措施[J];制造業(yè)自動化;2012年12期

3 吳銳;黃劍華;唐降龍;劉家鋒;;基于灰度直方圖和譜聚類的文本圖像二值化方法[J];電子與信息學報;2009年10期

4 馬俊青;宋愛國;甘英俊;;一種基于ARM的圖像處理系統(tǒng)的設(shè)計[J];電氣電子教學學報;2009年04期

5 盧洋;張旭秀;;圖像邊緣檢測算法的對比分析與研究[J];太原科技;2009年03期

6 王斌;劉興龍;劉永福;張立國;;機器視覺技術(shù)在農(nóng)業(yè)應(yīng)用中的研究[J];農(nóng)機化研究;2008年09期

7 楊敏,葉邦彥,牟麗,朱若磊;基于機器視覺的圓軸直徑精密檢測算法[J];計算機工程與應(yīng)用;2004年31期

8 唐宇慧,楊敏,葉邦彥,劉其洪;基于計算機視覺的圓軸直徑精密檢測[J];現(xiàn)代制造工程;2004年09期

9 張永宏,胡德金,徐俊杰;基于視覺圖像的機加工零部件亞像素邊緣定位[J];機械工程學報;2004年06期

10 馬強 ,唐文彥 ,陳金存 ,王慶;自由環(huán)狀零件尺寸圖像測量系統(tǒng)研究[J];航天制造技術(shù);2004年01期

相關(guān)會議論文 前1條

1 宋愛群;黃元慶;;CCD圖像傳感器的應(yīng)用技術(shù)與發(fā)展趨勢[A];第十七屆全國測控計量儀器儀表學術(shù)年會（MCMI'2007）論文集（上冊）[C];2007年

相關(guān)博士學位論文 前1條

1 卞曉東;基于機器視覺的車輛幾何尺寸測量系統(tǒng)研究[D];東南大學;2005年

相關(guān)碩士學位論文 前5條

1 臧祖望;工業(yè)產(chǎn)品表面缺陷檢測研究與開發(fā)[D];北京交通大學;2012年

2 杜紅茹;基于機器視覺的齒輪尺寸測量問題研究[D];蘭州理工大學;2011年

3 劉長春;一種新型超精密光學元件瑕疵檢測裝置開發(fā)[D];重慶大學;2010年

4 郎東明;CCD數(shù)字相機圖像采集卡的研究與實現(xiàn)[D];南京航空航天大學;2010年

5 高飛;基于機器視覺技術(shù)測量軸類零件尺寸的研究[D];吉林大學;2007年

，

本文編號：2435055