基于單目視覺的微位移精確測量方法研究
發(fā)布時間：2019-04-10 06:30
【摘要】：隨著機(jī)器視覺技術(shù)、圖像處理技術(shù)、光電技術(shù)等技術(shù)的發(fā)展,視覺測量技術(shù)得到了飛速的發(fā)展與應(yīng)用。單目視覺測量作為一種非接觸性的視覺測量方法,在農(nóng)業(yè)、工業(yè)、交通、醫(yī)學(xué)等領(lǐng)域內(nèi)有著重要的應(yīng)用。本文利用單目視覺與激光點復(fù)合進(jìn)行測量,獲取深度信息,此方法具有系統(tǒng)簡約、無需在被測物體上設(shè)置被動標(biāo)記、無需攝像機(jī)內(nèi)外參數(shù)標(biāo)定等優(yōu)點。本文首先介紹了小孔成像模型與透鏡成像模型,分別分析了一般情況下的基于小孔成像與激光點復(fù)合的測量模型和基于透鏡成像與激光點復(fù)合的測量模型。針對基于小孔成像與激光點復(fù)合的測量模型在微位移測量時,精度不高的問題,采用基于透鏡成像與激光點復(fù)合的測量模型。此模型采用透鏡成像原理,在像距等于兩倍的焦距鄰域附近,當(dāng)物距有微小變化,激光點在CCD上的斑點圖像位置變化明顯,能夠提高測量精度。由CCD相機(jī)獲取激光點圖像時,需要先對圖像進(jìn)行預(yù)處理,包含灰度處理,濾波處理、閾值分割和獲取激光輪廓。激光點的中心位置對測量精度有著重要的影響,本文研究了不同的被測物體對閾值的影響,不同的閾值、開機(jī)工作時間、非激光光斑輪廓與光斑輪廓邊緣奇異點對激光中心位置的影響;分析了非激光光斑輪廓與輪廓邊緣奇異點產(chǎn)生的原因,提出了非激光光斑與輪廓邊緣奇異點的去除方法,通過實驗驗證所提方法的可行性。根據(jù)測量模型中各個元器件之間的位置約束關(guān)系,設(shè)計了一種測量裝置,分別對測量裝置中的CCD相機(jī)、激光器、鏡頭、直線電機(jī)、物理基點、驅(qū)動器、鏡頭支架等零部件進(jìn)行選型和設(shè)計。采用VS2010作為軟件開發(fā)平臺,選用面向?qū)ο蟮腃++語言作為開發(fā)語言,根據(jù)界面設(shè)計原則與測量系統(tǒng)的要求,利用MFC與OpenCV相結(jié)合設(shè)計出單目視覺與激光點復(fù)合測量的界面。最后建立了實驗平臺,主要由三坐標(biāo)測量機(jī)、六自由度位姿調(diào)節(jié)平臺、測量裝置、百分表等組成。首先利用三坐標(biāo)測量機(jī)獲得被測物體與物理基點之間的距離,利用圖像處理技術(shù)獲得激光光斑的中心位置,建立一一對應(yīng)關(guān)系,然后多次移動被測物體,在每個位置獲取CCD中激光點的中心位置,得到多組數(shù)據(jù)。利用這些數(shù)據(jù),再根據(jù)理論測量模型,選用合適的擬合函數(shù),采用最小二乘擬合方法得到理論測量模型的修正模型。得到測量模型的修正模型之后進(jìn)行測量并與三坐標(biāo)測量機(jī)測得的數(shù)據(jù)進(jìn)行比較,分析修正模型的合理性,實驗表明該方法能夠進(jìn)行微位移高精度測量,在1mm范圍內(nèi)測量的相對誤差小于0.022%。
[Abstract]:With the development of machine vision technology, image processing technology, photoelectric technology and other technologies, vision measurement technology has been rapidly developed and applied. Monocular vision measurement, as a non-contact visual measurement method, has an important application in agriculture, industry, transportation, medicine and so on. In this paper, monocular vision and laser point are used to measure and obtain depth information. This method has the advantages of simple system, no need to set passive mark on the measured object, and no need to calibrate the inside and outside parameters of the camera. In this paper, the aperture imaging model and the lens imaging model are introduced firstly, and the measurement models based on the combination of small hole imaging and laser point are analyzed, respectively, and the measurement models based on lens imaging and laser point recombination are analyzed respectively. In order to solve the problem that the precision of the measurement model based on the combination of small hole imaging and laser point is not high in the micro displacement measurement, the measurement model based on the combination of lens imaging and laser point is adopted. This model adopts the principle of lens imaging. In the vicinity of the focal length equal to twice of the image length, when the object distance has a small change, the spot image position of the laser point on the CCD is obviously changed, which can improve the measurement precision. In order to obtain laser point image from CCD camera, it is necessary to pre-process the image, including gray-scale processing, filtering processing, threshold segmentation and obtaining laser contour. The central position of the laser point has an important influence on the measurement accuracy. In this paper, the influence of different objects measured on the threshold, the different threshold, the starting time, the working time, and so on, are studied in this paper. The influence of the non-laser spot contour and the singular point of the spot contour on the laser center position; The causes of non-laser spot contour and contour edge singularity are analyzed, and the removal method of non-laser spot and contour edge singularity is put forward. The feasibility of the proposed method is verified by experiments. According to the position constraint relation of each component in the measurement model, a kind of measuring device is designed. For the CCD camera, laser, lens, linear motor, physical base point, driver in the measuring device, respectively. Lens bracket and other components for selection and design. Using VS2010 as software development platform and object-oriented C language as development language, according to the principle of interface design and the requirement of measurement system, the interface of monocular vision and laser point complex measurement is designed by combining MFC and OpenCV. Finally, the experimental platform is established, which is mainly composed of three coordinate measuring machine, six degrees of freedom position and attitude adjustment platform, measuring device and percentile meter. Firstly, the distance between the measured object and the physical base point is obtained by using the coordinate measuring machine, the central position of the laser spot is obtained by using the image processing technology, and the one-to-one correspondence is established, and then the measured object is moved many times. At each position, the central position of the laser point in the CCD is obtained, and many sets of data are obtained. Based on these data and the theoretical measurement model, the modified model of the theoretical measurement model is obtained by using the least square fitting method and choosing the appropriate fitting function. After obtaining the modified model of the measurement model and comparing it with the data measured by the coordinate measuring machine, the rationality of the modified model is analyzed. The experiments show that the method can be used to measure the micro-displacement with high precision. The relative error is less than 0.022% in the 1mm range.
【學(xué)位授予單位】：揚州大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41
本文編號：2455554
[Abstract]:With the development of machine vision technology, image processing technology, photoelectric technology and other technologies, vision measurement technology has been rapidly developed and applied. Monocular vision measurement, as a non-contact visual measurement method, has an important application in agriculture, industry, transportation, medicine and so on. In this paper, monocular vision and laser point are used to measure and obtain depth information. This method has the advantages of simple system, no need to set passive mark on the measured object, and no need to calibrate the inside and outside parameters of the camera. In this paper, the aperture imaging model and the lens imaging model are introduced firstly, and the measurement models based on the combination of small hole imaging and laser point are analyzed, respectively, and the measurement models based on lens imaging and laser point recombination are analyzed respectively. In order to solve the problem that the precision of the measurement model based on the combination of small hole imaging and laser point is not high in the micro displacement measurement, the measurement model based on the combination of lens imaging and laser point is adopted. This model adopts the principle of lens imaging. In the vicinity of the focal length equal to twice of the image length, when the object distance has a small change, the spot image position of the laser point on the CCD is obviously changed, which can improve the measurement precision. In order to obtain laser point image from CCD camera, it is necessary to pre-process the image, including gray-scale processing, filtering processing, threshold segmentation and obtaining laser contour. The central position of the laser point has an important influence on the measurement accuracy. In this paper, the influence of different objects measured on the threshold, the different threshold, the starting time, the working time, and so on, are studied in this paper. The influence of the non-laser spot contour and the singular point of the spot contour on the laser center position; The causes of non-laser spot contour and contour edge singularity are analyzed, and the removal method of non-laser spot and contour edge singularity is put forward. The feasibility of the proposed method is verified by experiments. According to the position constraint relation of each component in the measurement model, a kind of measuring device is designed. For the CCD camera, laser, lens, linear motor, physical base point, driver in the measuring device, respectively. Lens bracket and other components for selection and design. Using VS2010 as software development platform and object-oriented C language as development language, according to the principle of interface design and the requirement of measurement system, the interface of monocular vision and laser point complex measurement is designed by combining MFC and OpenCV. Finally, the experimental platform is established, which is mainly composed of three coordinate measuring machine, six degrees of freedom position and attitude adjustment platform, measuring device and percentile meter. Firstly, the distance between the measured object and the physical base point is obtained by using the coordinate measuring machine, the central position of the laser spot is obtained by using the image processing technology, and the one-to-one correspondence is established, and then the measured object is moved many times. At each position, the central position of the laser point in the CCD is obtained, and many sets of data are obtained. Based on these data and the theoretical measurement model, the modified model of the theoretical measurement model is obtained by using the least square fitting method and choosing the appropriate fitting function. After obtaining the modified model of the measurement model and comparing it with the data measured by the coordinate measuring machine, the rationality of the modified model is analyzed. The experiments show that the method can be used to measure the micro-displacement with high precision. The relative error is less than 0.022% in the 1mm range.
【學(xué)位授予單位】：揚州大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41
【參考文獻(xiàn)】
相關(guān)期刊論文 前10條
1 侯震;許燕玲;黃色吉;楊雪君;韓瑜;陳善本;;視覺傳感技術(shù)在機(jī)器人焊接中的應(yīng)用[J];上海交通大學(xué)學(xué)報;2016年S1期
2 陳莉;尹鵬和;;MEMS制造中精確測量薄膜厚度的方法研究與比較[J];傳感器與微系統(tǒng);2015年10期
3 吳慧利;唐義;白廷柱;蔣玉蓉;蔣靜;;紫外薄膜光學(xué)常數(shù)的多層模型橢偏測量[J];光電工程;2015年09期
4 王運永;錢進(jìn);韓森;張齊元;;激光干涉儀引力波探測器的基本光學(xué)結(jié)構(gòu)[J];光學(xué)儀器;2015年04期
5 李兵;孫彬;陳磊;魏翔;;激光位移傳感器在自由曲面測量中的應(yīng)用[J];光學(xué)精密工程;2015年07期
6 孫彬;李兵;;一種量化的激光位移傳感器傾角誤差補償模型[J];儀器儀表學(xué)報;2015年05期
7 薛媛;陳青山;劉力雙;呂勇;;基于雙頻激光干涉儀的顯微模板精密測量[J];應(yīng)用光學(xué);2015年02期
8 宋麗梅;覃名翠;楊燕罡;李宗艷;習(xí)江濤;朱騰達(dá);;激光視覺方法用于檢測齒輪加工誤差[J];光電工程;2015年01期
9 張婷;;人機(jī)交互界面設(shè)計在產(chǎn)品可用性中的應(yīng)用研究[J];包裝工程;2014年20期
10 韓瑞珍;何勇;;基于計算機(jī)視覺的大田害蟲遠(yuǎn)程自動識別系統(tǒng)[J];農(nóng)業(yè)工程學(xué)報;2013年03期
相關(guān)博士學(xué)位論文 前1條
1 馮春;基于單目視覺的目標(biāo)識別與定位研究[D];南京航空航天大學(xué);2013年
相關(guān)碩士學(xué)位論文 前6條
1 劉偉;鐵鉆工夾緊缸同步技術(shù)與鉆具自動找心方法的研究[D];揚州大學(xué);2016年
2 張得明;割草機(jī)器人草地視覺識R%與路徑規(guī)劃技術(shù)研究[D];揚州大學(xué);2016年
3 陳瑾藝;移動醫(yī)療產(chǎn)品App界面設(shè)計研究[D];南昌大學(xué);2015年
4 陸偉男;基于VS2010的PLC程序編輯系統(tǒng)的研究與設(shè)計[D];廣西科技大學(xué);2015年
5 高尚;基于激光環(huán)主動標(biāo)記的單目視覺試驗平臺研究[D];揚州大學(xué);2012年
6 尹楠;基于單目視覺及主動激光標(biāo)記的空間曲面重構(gòu)方法研究[D];揚州大學(xué);2012年
，本文編號：2455554
本文鏈接：http://sikaile.net/kejilunwen/ruanjiangongchenglunwen/2455554.html
最近更新
教材專著
