WO2012071817A1 - 基于小波变换的三维测量方法 - Google Patents

基于小波变换的三维测量方法 Download PDF

Info

Publication number
WO2012071817A1
WO2012071817A1 PCT/CN2011/071441 CN2011071441W WO2012071817A1 WO 2012071817 A1 WO2012071817 A1 WO 2012071817A1 CN 2011071441 W CN2011071441 W CN 2011071441W WO 2012071817 A1 WO2012071817 A1 WO 2012071817A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
point
matrix
map
phase
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2011/071441
Other languages
English (en)
French (fr)
Inventor
达飞鹏
黄昊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to US13/517,226 priority Critical patent/US9163937B2/en
Priority to KR1020117030045A priority patent/KR101214702B1/ko
Publication of WO2012071817A1 publication Critical patent/WO2012071817A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01BMEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00Measuring arrangements characterised by the use of optical techniques
    • G01B11/24Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
    • G01B11/25Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01BMEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00Measuring arrangements characterised by the use of optical techniques
    • G01B11/24Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
    • G01B11/25Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
    • G01B11/254Projection of a pattern, viewing through a pattern, e.g. moiré
    • GPHYSICS
    • G02OPTICS
    • G02BOPTICAL ELEMENTS, SYSTEMS OR APPARATUS
    • G02B27/00Optical systems or apparatus not provided for by any of the groups G02B1/00 - G02B26/00, G02B30/00
    • G02B27/60Systems using moiré fringes
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00Three-dimensional [3D] modelling for computer graphics
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0004Industrial image inspection
    • G06T7/0006Industrial image inspection using a design-rule based approach
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/50Depth or shape recovery
    • G06T7/521Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01JMEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J9/00Measuring optical phase difference; Determining degree of coherence; Measuring optical wavelength
    • G01J9/02Measuring optical phase difference; Determining degree of coherence; Measuring optical wavelength by interferometric methods
    • G01J2009/0234Measurement of the fringe pattern
    • G01J2009/0238Measurement of the fringe pattern the pattern being processed optically, e.g. by Fourier transformation

Definitions

  • the invention belongs to the field of three-dimensional information reconstruction. Based on the black-and-white sinusoidal grating projection, the wavelet transform is used to analyze the deformed fringe image to solve the relative phase, and the mass map is used to guide the phase unwrapping to obtain the exact absolute phase. Background technique
  • Three-dimensional measurement technology can describe the three-dimensional features of objects and obtain three-dimensional information on the surface of objects. It has a wide range of applications in product inspection and processing control, medical field, cultural relic protection, aerospace, and cultural fields.
  • Grating projection method is an important three-dimensional measurement technique.
  • the raster image of the surface of the object is obtained by CCD, and a specific algorithm is used.
  • the fringe image is processed to extract the phase therein, thereby establishing three-dimensional information of the object.
  • Commonly used methods for solving the phase of a fringe image include a phase shift method, a Fourier transform method, a window Fourier transform method, and a wavelet transform method.
  • the Fourier transform method can perform phase measurement by collecting only one stripe image, which can realize dynamic measurement and is widely used.
  • the Fourier transform is a global signal analysis tool that cannot extract the characteristics of local signals.
  • the spectral aliasing problem existing in the transform affects the accuracy of phase measurement.
  • wavelet transform has been introduced into the field of optical three-dimensional measurement, and wavelet transform is used to analyze the stripe image to realize three-dimensional measurement of the object, namely wavelet transform profilometry.
  • Wavelet analysis is an effective tool for analyzing non-stationary signals. Wavelet transform is more advantageous in analyzing local features of signals than traditional Fourier transform methods, which have been widely used in signal processing. Wavelet transform has the characteristics of multi-resolution analysis.
  • the wavelet transform is used to solve the phase of the fringe image to avoid the spectral aliasing problem in the Fourier transform, and the measurement accuracy is higher.
  • the phase values obtained by the wavelet transform method are all between 0 and 2 , so the phase unwrapping process is required.
  • the simple scan line unwrapping phase method is very fast, but not robust enough to cause error propagation.
  • the quality map guides the phase unwrapping algorithm to select an optimal phase unwrapping path by constructing a quality map reflecting the reliability of each pixel point on the fringe image. This method is robust and can unravel the phase relatively accurately, but the operation is relatively accurate. Long time, not suitable for real-time measurements.
  • the key step in the quality-guided phase unwrapping algorithm is the creation of a quality map.
  • the quality map mainly has the following methods of establishment.
  • the phase gradient method the maximum value of the phase gradient of each point and four neighbors is taken as the mass value of each point, and the larger the mass value, the worse the quality of the point.
  • the surface vector mass map method uses the inner product of the normal vector of each point and the negative unit vector of the CCD projection direction as the mass value of the point, and the normal vector can be calculated from the phase value of the point and the four adjacent points.
  • the above two methods do not involve the parameters of the wavelet transform, and are not suitable for wavelet transform profilometry.
  • Wavelet transform ridge amplitude method based on the principle that the modulus of the wavelet transform represents the similarity between the local signal and the wavelet function, the mass map of the wavelet transform ridge is used to establish the mass map, which guides the initial phase expansion.
  • This method makes full use of the wavelet. Transform the resulting matrix, but this quality map creation method does not consider the scale factor at the wavelet transform ridge. When the scale factor at the ridge is too large or too small, the modulus value cannot accurately indicate the quality of the local signal; and when the local signal When the amplitude changes, the amplitude of the wavelet ridge in the wavelet transform matrix also changes, so the sinusoidality and reliability of the local signal cannot be accurately described. Summary of the invention
  • the object of the present invention is to improve the speed of the phase expansion process of the wavelet transform and the accuracy of solving the absolute phase.
  • the method makes full use of the useful information in the wavelet transform matrix, and combines the advantages of fast direct phase unwrapping speed and high precision of flood phase unwrapping method. It only needs to project a stripe image to realize the acquisition of three-dimensional information on the surface of a more complex object. High accuracy and good real-time performance.
  • Step 1 Project a black and white stripe image onto the surface of the object to be measured, and use the CCD to photograph the surface of the object to be measured, a deformation stripe with a height of width Image g ( ) :
  • Step 2 Performs wavelet transform on the deformed fringe image line by line to obtain the relative phase distribution map of the deformed fringe image, the specific process as follows:
  • Step 2.1 Treat y as a constant, and use the one-dimensional continuous wavelet transform to process the yth line of the deformed fringe image.
  • the processing is:
  • Step 2.2 Find the optimal scale factor distribution map and relative phase distribution map of the fringe image
  • Step 3 Create a quality map 2 ( ⁇ ,
  • Step 3.1 Make a wavelet transform on the one-dimensional sinusoidal signal of frequency ⁇ :
  • Step 3.2 Find the quality map 2( ⁇ ):
  • Step 4 According to the mass map 2 ( ⁇ ), divide the relative phase distribution map ( ⁇ into two parts, step 4.1: establish a height c, a width r 0 value matrix D,
  • Step 4.2 Traverse each point of the mass map, find the element value ft with the highest frequency of occurrence of the mass map 2( ⁇ ), and obtain the threshold.
  • Step 4.3 Traverse each point of the mass map 2( ⁇ ). When the value in 2( ⁇ ) is greater than the threshold: ⁇ , the corresponding position element in matrix D is set to 1,
  • Step 5.1 Establish a height c, width / "0 value matrix S,
  • Step 5.2 Select a point in the 20x 20 field of the central pixel of the relative phase profile to select a point of the corresponding position of the D matrix as the starting point of the expansion.
  • the absolute phase value of the starting point is recorded as the relative phase distribution of the point ( The value in ⁇ , along the row direction, expands the phase of the point where the element value of the corresponding position of the D matrix is 0 to the sides of the image.
  • the value of the element in the corresponding position of the S matrix is set to 1, in the process of direct expansion. If the previous point of the same line of some points is not processed, the adjacent point of the previous line or the next line of the point can be regarded as the previous point and then expanded.
  • Each line is as follows:
  • Step 6 Expand the phase of the relative phase distribution. The phase of the point where the corresponding position of the S matrix is 0,
  • Step 6.1 Mark the connected field of 0 value in the S matrix. For each connected field, do the following: Step 6.2 Select any boundary point in the connected domain as the starting point, enter the current empty stack, and step 6.3 find the starting point. If the point of the corresponding position of the S matrix in the four neighbors is 0, if no such point is found in the four neighbors, skip to step 6.5. If such a point can be found, expand the phase of the points and place the points in the S matrix. Set the corresponding position element value to 1, and store the points on the stack. Step 6.4: Sort the points in the stack according to their corresponding values in the stripe image quality map 2( ⁇ ). The maximum value of the mass value is placed on the stack. the top of,
  • Step 6.5 Take the point at the top of the stack as the starting point, and judge whether the stack is empty. If it is empty, the connected domain expansion process ends; if it is not empty, go to step 6.3.
  • Step 7 After step 5 and step 6, the absolute phase map ( ⁇ ) of the fringe image is obtained. According to the phase-to-height conversion formula of the classical grating projection, the three-dimensional information of the measured object is finally obtained.
  • the present invention Compared with the prior art, the present invention has the following advantages: First, the present invention uses a wavelet transform method to solve the relative phase of a fringe image, which is avoided compared to the Fourier transform three-dimensional measurement method. The spectral aliasing phenomenon can obtain more accurate relative phase results. Compared with the phase shift method, the present invention only needs to project a black and white stripe image to realize dynamic measurement. Secondly, the method for establishing a quality map of the present invention makes full use of the method.
  • the information in the wavelet transform matrix compared with other methods, using the quality map established by the method of the present invention in phase unwrapping, can obtain a more reliable phase unwrapping path, and reduce the probability of error in phase unwrapping, which can be more Accurate absolute phase results;
  • the present invention uses a mass map to divide the relative phase map into two parts, and uses different algorithms for phase unwrapping of the two parts, which is greatly reduced compared to the conventional flood phase unwrapping algorithm.
  • the computation time required for phase unwrapping while maintaining high robustness.
  • the invention can quickly and accurately acquire the three-dimensional height information of the measured object, and has good real-time and robustness.
  • Figure 2 is a flow chart showing the specific process of using the flood algorithm to expand the phase of the "lower quality” point in step 6.
  • Figure 3 is a deformed fringe image of the object measured by the CCD using the foam as the object to be measured.
  • Figure 4 is a relative phase distribution diagram obtained by wavelet transform.
  • Figure 5 is a plot of the best scale factor obtained.
  • Figure 6 is a binary image obtained by threshold segmentation of the optimal scale factor distribution map based on the threshold calculated by the mass map.
  • Figure 7 is an unfolded phase image after step 5.
  • Figure 8 is the resulting absolute phase profile.
  • Figure 9 is a point cloud image representing the three-dimensional information of the object based on the phase-to-height conversion formula of the classical raster projection. detailed description
  • VC++6.0 is selected as the programming tool to process the deformed fringe image collected by the CCD.
  • This example uses foam as the object to be measured, and finally obtains a more accurate full-field absolute phase distribution containing the three-dimensional information of the shield.
  • Figure 2 is a flow chart showing the specific process of step 6 using the flood algorithm to develop the phase of the "lower quality" point.
  • the present invention firstly uses the wavelet transform to solve the relative phase distribution of the acquired deformed fringe image, and uses the scale factor at the wavelet transform ridge to establish a mass map to guide the quality of the relative phase map.
  • the relative phase distribution is divided into two parts according to the quality map according to the quality map. The part with better quality is directly phase-expanded by the scanning line method, and the quality is generally poor. Part of the method of using quality guidance to select the best path.
  • This method eliminates some unnecessary steps of the global quality guidance algorithm, improves the efficiency of the phase unwrapping algorithm, and maintains high measurement accuracy. After obtaining the accurate absolute phase distribution, according to the phase-to-height conversion formula of the classical grating projection, the three-dimensional information of the measured object is finally obtained.
  • the invention is based on a three-dimensional measurement method of wavelet transform, and the specific implementation steps are as follows:
  • Step 1 Project the black and white stripe image onto the surface of the object to be measured, and use the CCD to shoot the surface of the object to be measured.
  • a deformed stripe image g ( ) with a height of width is :
  • a (A> is the background light intensity distribution
  • ⁇ ( , > is the surface reflectance of the object
  • the two parameters A(U), ⁇ ( ,) can be treated as constants
  • ⁇ ⁇ ) is the relative phase distribution to be sought
  • ( ⁇ represents the two-dimensional coordinates of the deformed fringe image
  • Figure 3 shows the deformed fringe image
  • Step 2 Perform a wavelet transform on the deformed fringe image line by line to obtain a relative phase distribution map of the deformed fringe image.
  • the specific process is as follows:
  • Step 2.1 Treat y as a constant, and use the one-dimensional continuous wavelet transform to process the first y-line of the deformed fringe image g( ⁇ , the process is:
  • the value range is 10 to 50, every 0.2 is a value, is the translation factor, the value range is 1 to the stripe image width every 1 value, the unit is pixel, the obtained W ( , b) is the row label of the element in the array W ( , b) of a 200- row r- column, called the y-row wavelet transform matrix,
  • the conjugate function, M is a wavelet function, and the expression is as follows:
  • is the bandwidth of the wavelet function
  • is the center frequency of the wavelet function, which is a unit complex number.
  • Step 2.2 Find the best scale factor distribution and relative phase distribution of the fringe image
  • Step 3 Establish the mass map 2( ⁇ ).
  • the theoretical basis for establishing the mass map is that the modulus of the wavelet transform matrix represents the similarity between the local signal of the fringe image and the wavelet function. Since the wavelet function selected in the present invention is a morlet function, It has better sinusoidality, and according to step 2.2, the modulus of the wavelet transform matrix reaches the maximum at the wavelet transform ridge, so the scale factor at the wavelet transform ridge can be used to judge the sinusoidality of the local signal of the fringe image.
  • the optimal scale factor of a certain point of the fringe image When the optimal scale factor of a certain point of the fringe image is too large, it indicates that the local signal near the pixel point is consistent with the low-frequency sinusoidal signal, which may be a stripe-free area, so the "quality" of the point, that is, the reliability can be determined as Low; When the optimal scale factor at this point is too small, it means that the local signal near the pixel is more consistent with the high frequency sinusoidal signal, which may be interfered by noise, and the "quality" of the point can also be judged as low. Therefore, we should first find such a value.
  • step 3.1 is the process of finding such a value
  • step 3.2 is the process of establishing the quality map, quality.
  • the meaning of the quality of each point in the graph is actually the "distance" between the scale factor of the point and the optimal scale factor.
  • the mass map obtained in this step is used to divide the relative phase distribution map into two parts in step 5.
  • step 6 the phase unwrapping process is guided to obtain the best path for phase unwrapping.
  • Step 3.1 The frequency is /.
  • the one-dimensional sinusoidal signal is wavelet transformed: Find the modular matrix ⁇ ( ⁇ b) of the two-dimensional complex matrix, search for the largest value of the elements in each column, record the row labels, and obtain the average of these row labels ⁇ to find the optimal scale factor:
  • Step 3.2 Find the mass map ⁇ , :
  • Step 4 According to the quality map 2 ( ⁇ , the relative phase distribution map ( ⁇ is divided into two parts, step 4.1: establishing Height c, width r 0 value matrix D, D matrix and stripe image size are the same, used to distinguish the two parts of the relative phase map according to the mass map,
  • Step 4.2 Traverse the quality graph 2 ( ), and get the most valued value ft to get the threshold value.
  • Step 4.3 When the value in the traversal mass map is greater than the threshold r, the corresponding position element in the matrix D is set to 1.
  • Figure 6 is the image mode of the D array, where the white area represents the point where the quality is better and the priority is expanded, and the black area is required to adopt the flood. The area of the point where the algorithm is expanded,
  • Step 5 Unfold the phase of the relative phase distribution.
  • the points developed in this step are all points with better quality.
  • the mass values are similar, according to the traditional flood phase. Algorithm, these points still have to participate in the sorting, but in fact, the phase unwrapping path after sorting and before sorting has little effect on the precision, so this part uses the simplest scan line phase unwrapping algorithm to get more accurate results, and the operation And very fast,
  • Step 5.1 Create a height c, width r 0 value matrix S, S matrix is the same size as the stripe image, used to mark the relative phase distribution map ( ⁇ > has been expanded,
  • Step 5.2 Select a point in the relative phase distribution map (20 x 20 field of the central pixel of ⁇ ) to select the position of the corresponding position of the D matrix as the starting point of the expansion.
  • the absolute phase value of the starting point is recorded as the relative phase of the point.
  • the value in the distribution map > expands the phase of the point where the element value of the corresponding position of the D matrix is 0 along the row direction. For each pixel point, the value of the element in the corresponding position of the S matrix is set to 1. If the previous point of the same line of some points is not processed during the process, the adjacent point of the previous line or the next line of the point may be regarded as the previous point and then expanded. If the two points are not processed, leave it to step 6. Processing, every line is calculated like this,
  • Step 6 Unfold the phase of the relative phase distribution.
  • the flood phase expansion algorithm is used.
  • Flood algorithm The flow chart is shown in Figure 2. Since the number of points is small, the calculation time is also faster.
  • Step 6.1 Mark the connected domain of the 0 value in the S matrix.
  • a connected domain of the binary matrix refers to a set of elements in the binary matrix that are connected to each other and have the same value.
  • Each closed black area in Figure 6 is A connected domain, for each connected domain, do the following:
  • Step 6.2 Select any boundary point in the connected domain as the starting point, and enter the current empty stack.
  • Step 6.3 Find the point where the corresponding position of the S matrix in the four neighbors of the starting point is 0. If no four neighbors are found, To such a point, skip to step 6.5; if you can find such a point, expand the phase of these points, set the values of the points in the corresponding position in the S matrix, and store the points on the stack, step 6.4: The points in the order are sorted according to the value of the corresponding position in the stripe image quality map 2 ( ⁇ , the maximum value of the mass value is placed on the top of the stack,
  • Step 6.5 Take the point at the top of the stack as the starting point, and judge whether the stack is empty. If it is empty, the connected domain expansion process ends; if it is not empty, go to step 6.3.
  • Step 7 After processing in steps 5 and 6, the relative phase distribution of the fringe image (all points of the fringe image is obtained, and the absolute phase map of the fringe image is obtained ( ⁇ , as shown in Fig. 8 is the absolute phase map of the measured object)
  • the phase operation time of the invention is 27s, and the full-field phase expansion operation time is nearly 90min. It can be seen that the method of the invention has better real-time performance, according to the phase-to-height conversion formula of the classical grating projection, and finally The three-dimensional information of the measuring object is obtained, and FIG. 9 is a representation of the three-dimensional information of the object in the form of a point cloud.
  • the distance from the CCD camera to the projector, A c, y) - % c, indicates the amount of phase change, ( ) is the unwrapped phase result, % ( ⁇ is the initial phase result, determined by the measurement reference plane, ". is the projection grating
  • the angular frequency is available from the system calibration.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Optics & Photonics (AREA)
  • Quality & Reliability (AREA)
  • Computer Graphics (AREA)
  • Geometry (AREA)
  • Software Systems (AREA)
  • Length Measuring Devices By Optical Means (AREA)
  • Image Processing (AREA)

Description

基于小波变换的三维测量方法
技术领域
本发明属于三维信息重构的领域。 基于黑白正弦光栅投影, 采用小波变 换法分析变形条纹图像求解相对相位, 建立质量图指导相位展开, 得到精确 绝对相位的过程。 背景技术
三维测量技术能够描述物体的三维特征和获取物体表面的三维信息, 在 产品检测和加工控制、 医疗领域、 文物保护领域、 航空航天领域、 文化领域 等有着广泛的应用。
光栅投影法是一种重要的三维测量技术, 通过向物体表面投射正弦光栅, 将物体的高度信息以相位的形式隐含在光栅中, 利用 CCD获得物体表面的光 栅条纹图像, 并使用特定算法对条纹图像进行处理, 提取其中的相位, 从而 建立物体的三维信息。 常用的求解条纹图像相位的方法有相移法、 傅里叶变 换法、 窗口傅里叶变换法、 小波变换法等。
傅里叶变换法由于只需采集一幅条纹图像即可完成相位的测量, 可以实 现动态测量, 应用非常广泛。 但是傅里叶变换是一种全局的信号分析工具, 不能提取局部信号的特征, 在变换中存在的频谱混叠问题影响着相位测量的 精度。 近年来, 随着小波分析理论的迅速发展, 小波变换被引入光学三维测 量领域, 使用小波变换分析条纹图像进而实现物体的三维测量, 即小波变换 轮廓术。 小波分析是一种分析非平稳信号的有效工具, 相比于已经在信号处 理领域得到广泛应用的传统傅里叶变换法, 小波变换在分析信号的局部特征 时更有优势。 小波变换具有多分辨率分析的特性, 不仅能够像傅里叶变换那 样得到信号的整体特征, 而且能够对局部信号的细节进行分析, 具有较好的 时间 /空间局部性。 所以采用小波变换求解条纹图像的相位避免了傅里叶变换 中出现的频谱混叠问题, 测量精度更高。
小波变换法求解得到的相位值都是介于 0_2 的, 所以需要相位展开过 程。 为了实现动态测量, 一般不采用添加辅助图像增加信息协助相位展开。 简单的扫描线展开相位方法运算速度非常快, 但鲁棒性不高, 容易出现错误 传递现象。 质量图指导相位展开算法通过对条纹图像建立一幅反映每一像素 点可靠性的质量图选取一条最佳的相位展开路径, 这种方法鲁棒性较高, 能 够相对精确地展开相位, 但运算时间较长, 不适用于实时测量。 质量指导相位展开算法的关键步骤是质量图的建立。 质量图主要有以下 几种建立方法。 相位梯度法, 每个点与四邻的相位梯度的最大值作为每个点 的质量值, 质量值越大, 说明该点的质量越差。 表面向量质量图法, 使用每 一点的法向量和 CCD投影方向的负单位向量的内积作为该点的质量值, 法向 量可由该点和四邻点的相位值计算得出。 以上两种方法没有涉及到小波变换 的参数, 并不适用于小波变换轮廓术。 小波变换脊幅值法, 根据小波变换脊 处的模值表示局部信号和小波函数相似程度的原理, 利用小波变换脊的模值 建立质量图, 指导初始相位的展开, 这种方法充分利用了小波变换得到的矩 阵, 但这种质量图建立方法没有考虑小波变换脊处的尺度因子, 当脊处的尺 度因子过大或者过小时, 模值不能准确说明局部信号的质量好坏; 并且当局 部信号的幅值变化时, 小波变换矩阵中小波脊处的幅值也会变化, 故不能准 确描述局部信号的正弦性和可靠性。 发明内容
技术问题: 针对小波变换轮廓术中相位展开的精确性和实时性问题, 本 发明的目的在于提高小波变换相位展开过程的速度和求解绝对相位的精度。 本方法充分利用了小波变换矩阵中的有用信息, 结合了直接相位展开速度快 和洪水相位展开法精度高的优点, 只需投影一幅条纹图像, 实现对较复杂物 体表面三维信息的获取, 具有较高的精确性, 同时实时性较好。
技术方案: 一种基于小波变换的三维测量方法, 具体步骤如下: 步骤 1 : 将黑白条纹图像投影到被测物体表面, 使用 CCD对被测物体表 面进行拍摄, 一幅高度为 宽度为 的变形条纹图像 g ( ) :
Figure imgf000004_0001
其中, , 是背景光强分布, 是物体表面反射率, /。是正弦条 纹频率, 是待求的相对相位分布, (U)表示变形条纹图像的二维坐标, 步骤 2: 对变形条纹图像逐行做小波变换, 得到变形条纹图像的相对相位 分布图, 具体过程如下:
步骤 2.1 :将 y视为常数,采用一维连续小波变换对变形条纹图像 的 第 y行进行处理, 处理过程为:
W(ai,b) = g (x, y)M*a,b (x) ^
其中, "是尺度因子, 取值范围为 10到 50, 每隔 0.2取一个值, 是平 移因子, 取值范围为 1到条纹图像宽度 , 每隔 1取一个值, 单位为像素, 获 得的 , 是一个 200行 r列的二维复数矩阵, A是矩阵 , 中元素的行 标号, 称第 y行小波变换矩阵, 0 )= ^^^), (^是 Λ^ 的 共轭函数, Μ (4是小波函数,
表达式如下:
Figure imgf000005_0001
其中, Λ是小波函数的带宽, Λ是小波函数的中心频率, 为单位复数, 步骤 2.2:求取条纹图像的最佳尺度因子分布图 和相对相位分布图
Φ、χ, y),求取 a r (^ 和 Φ、χ, y)在坐标 (^, y)处数值的方法如下:
求出 W {a^b)的对应的模矩阵 A(fll b)和角度矩阵 , 搜索矩阵 A(fll b)第 列中值最大的元素, 求出该元素在矩阵 A(fll b)中的行标号为 " 则 =10 + 0·2χΩΐΜΧ, 为条纹图像的最佳尺度因子分布图 "^, 在坐标 (^) 处数值, 矩阵 W " )中第 列中行标号为 " 的元素数值为条纹图像的相对相 位分布图 ίΚ^ 在坐标 处数值,
遍历条纹图像所有坐标点, 求得条纹图像的最佳尺度因子分布图 (^ 和相对相位分布图 (^ ,
步骤 3: 建立质量图 2(^ ,
步骤 3.1: 对频 ^的一维正弦信号做小波变换:
Figure imgf000005_0002
求出二维复数矩阵 , 的模矩阵 ^(^b) , 搜索每列元素中数值最大 的元素, 记录其行标号, 求得这些行标号的平均值^ 求得最优尺度因子 :
Figure imgf000005_0003
步骤 3.2: 求质量图 2(^) :
Figure imgf000005_0004
步骤 4: 根据质量图 2(^) 将相对相位分布图 (^ 分成两个部分, 步骤 4.1:建立高度 c, 宽度 r的 0值矩阵 D,
步骤 4.2:遍历质量图 的每一个点,求出质量图 2(^) 出现频率最高 的元素数值 ft, 得到阈值
Figure imgf000005_0005
步骤 4.3: 遍历质量图 2(^ ) 的每一个点,当 2(^ ) 中的数值大于阈值: Γ 时, 矩阵 D中相应位置元素置 1,
步骤 展开相对相位分布图 (^ 中 D矩阵相应位置元素值为 0的点的 相位,
步骤 5.1 : 建立高度 c, 宽度/ "的 0值矩阵 S ,
步骤 5.2: 在相对相位分布图 的中心像素的 20x 20领域内选择一个 D矩阵相应位置元素值为 0的点作为展开的起始点,起始点的绝对相位值记为 该点在相对相位分布图 (^ 中的数值, 沿着行方向向图像两侧展开 D矩阵 相应位置元素值为 0的点的相位, 每展开一个像素点, 该点在 S矩阵相应位 置元素值置 1, 在直接展开过程中如遇到某些点的同一行的前一点并未处理, 可以将该点的上一行或者下一行的相邻点视为前一点然后展开, 每行都如此 每一点具体展开过程如下:
1
2π 式中, 是前一点的绝对相位值, 是当前点的相对相位值, υ 求得的当前点的绝对相位值, 是取整函数,
步骤 6:展开相对相位分布图 ) 中 S矩阵相应位置元素值为 0的点的 相位,
步骤 6.1 在 S矩阵中标记 0值的连通域, 对于每个连通域做如下处理: 步骤 6.2 选择连通域中的任一边界点做为起始点, 入当前为空的堆栈, 步骤 6.3 查找起始点四邻点中 S矩阵相应位置元素值为 0的点, 如果四 邻点中找不到这样的点,跳到步骤 6.5; 如果可以找到这样的点, 展开这些点的 相位, 将这些点在 S矩阵中相应位置元素值置 1, 并将这些点存入堆栈, 步骤 6.4: 对堆栈中的点按照其在条纹图像质量图 2(^ ) 中相应位置的数 值大小进行排序, 质量值最大点放置于堆栈的顶部,
步骤 6.5: 将堆栈的顶部的点取出做为起始点, 判断堆栈是否为空, 如果 为空, 则该连通域展开过程结束; 如果不为空, 转到步骤 6.3,
步骤 7: 经过步骤 5和步骤 6处理, 得到条纹图像的绝对相位图 (^), 根据经典光栅投影的相位到高度的转换公式, 最终求得测量物体的三维信息。
有益效果: 与现有技术相比, 本发明具有如下优点: 首先, 本发明采用 小波变换法求解条纹图像的相对相位, 相比于傅里叶变换三维测量法, 避免 了频谱混叠现象, 可以得到更加精确的相对相位结果, 相比于相移法, 本发 明只需投影一幅黑白条纹图像, 可以实现动态测量; 其次, 本发明建立质量 图的方法充分利用了小波变换矩阵中的信息, 相比其他方法, 在相位展开中 使用本发明方法所建立的质量图, 能够得到一条可靠性更高的相位展开路径, 降低了相位展开中出错的概率, 可以得到更加精确的绝对相位结果; 最后, 本发明利用质量图将相对相位图分成两个部分, 对这两个部分分别采用不同 的算法进行相位展开, 相比于传统的洪水相位展开算法, 极大减少了相位展 开所需的运算时间, 同时保持了较高的鲁棒性。 综上, 本发明能够快速准确 地获取被测物体的三维高度信息, 具有较好的实时性和鲁棒性。 附图说明
图 1是本发明的整个过程的流程图。
图 2是步骤 6中采用洪水算法展开 "质量"较低的点的相位的具体过程 流程图。
图 3是 CCD采集到的以泡沫塑料为被测物体的变形条纹图像。
图 4是采用小波变换法求解得到的相对相位分布图。
图 5是得到的最佳尺度因子分布图。
图 6 是根据质量图计算的阈值对最佳尺度因子分布图做阈值分割得到的 二值图像。
图 7是步骤 5过后的展开相位图像。
图 8是最终得到的绝对相位分布图。
图 9 是根据经典的光栅投影的相位到高度的转换公式, 得到的表示物体 三维信息的点云图像。 具体实施方式
下面结合附图对本发明的具体实施方式做进一步说明。 在 windows操作 系统下选用 VC++6.0作为编程工具,对 CCD采集到的变形条纹图像进行处理。 该实例采用泡沫塑料作为被测物体, 最终得到比较精确的含有护板三维信息 的全场绝对相位分布。
图 1是本发明的整个过程的流程图。
图 2是步骤 6采用洪水算法展开 "质量"较低的点的相位的具体过程流 程图。 针对小波变换轮廓术中的精确性和快速性的问题, 本发明首先采用小波 变换求解采集到的变形条纹图像的相对相位分布, 利用小波变换脊处的尺度 因子建立质量图指导相对相位图的质量展开, 在此基础上, 为了提高相位展 开算法的速度, 根据质量图将相对相位分布分成质量好坏的两个部分, 质量 普遍较好的部分采用扫描线法直接相位展开, 质量普遍较差的部分采用质量 指导选取最佳路径的方法。 这种方法省去了全局质量指导算法的一些不必要 的步骤, 提高了相位展开算法的效率, 而且保持了较高的测量精度。 在得到 精确的绝对相位分布后, 根据经典光栅投影的相位到高度转换公式, 最终得 到测量物体的三维信息。
本发明基于小波变换的三维测量方法, 具体实施步骤如下:
步骤 1 : 将黑白条纹图像投影到被测物体表面, 使用 CCD对被测物体表 面进行拍摄, 一幅高度为 宽度为 的变形条纹图像 g ( ) :
Figure imgf000008_0001
其中, A (A> 是背景光强分布, β ( ,> 是物体表面反射率, 由于环境光 是不变的, 所以这 A(U)、 β( ,) 两个参数可以当做常数处理, /。是正弦条 纹频率, 可取/。= 32,^ ^)是待求的相对相位分布, (^ 表示变形条纹图像 的二维坐标, 图 3为变形条纹图像,
步骤 2: 对变形条纹图像逐行做小波变换, 得到变形条纹图像的相对相位 分布图, 具体过程如下:
步骤 2.1 :将 y视为常数,采用一维连续小波变换对变形条纹图像 g(^ 的 第 y行进行处理, 处理过程为:
W(ai,b) = g (x, y)M*a,b (x) ^
其中, "是尺度因子, 取值范围为 10到 50, 每隔 0.2取一个值, 是平 移因子, 取值范围为 1到条纹图像宽度 每隔 1取一个值, 单位为像素, 获 得的 W ( , b)是一个 200r列的 阵 W ( , b)中元素的行 标号, 称第 y行小波变换矩阵,
Figure imgf000008_0002
共轭函数, M 是小波函数, 表达式如下:
Figure imgf000008_0003
其中, Λ是小波函数的带宽, Λ是小波函数的中心频率, 为单位复数, 本发明采用 Λ Λ = ο·845,进行小波变换,
步骤 2.2:求取条纹图像的最佳尺度因子分布图 和相对相位分布图
Φ、χ, y),求取 a r (^, 和 Φ、χ, y)在坐标 (^, y)处数值的方法如下:
求出 w(A b)的对应的模矩阵 和角度矩阵 ^("»,
Λ ?) =小 mag2 + )
Figure imgf000009_0001
搜索矩阵 ^ )第 列中值最大的元素, 称为条纹图像 (^, 坐标处的点 的小波变换脊, 求出该元素在矩阵 ^ b)中的行标号为 Ω , 则 = 10 + 0·2 X Ω ,这个式子实际上是根据行标号为 Ω 寻找小波变换脊处的尺 度因子的数值, 为条纹图像的最佳尺度因子分布图 (^, 在坐标 ( 处数 值, 矩阵 , 中第 列中行标号为 " 的元素数值为条纹图像的相对相位分 布图 (^ > 在坐标(D 处数值,
遍历条纹图像所有坐标点, 求得条纹图像的最佳尺度因子分布图 y) 和相对相位分布图 A > ,实验效果分别见图 5和图 4,
步骤 3 : 建立质量图 2(^), 建立质量图的理论基础是小波变换矩阵的模 值代表了条纹图像的局部信号和小波函数的相似程度, 由于本发明中选用的 小波函数是 morlet函数, 具有较好的正弦性, 而根据步骤 2.2, 小波变换矩阵 的模值在小波变换脊处达到最大, 所以可以采用小波变换脊处的尺度因子判 断条纹图像局部信号的正弦性好坏。 当条纹图像某一点的最佳尺度因子过大 时, 说明该像素点附近的局部信号和低频率的正弦信号比较吻合, 可能是无 条纹区域, 故该点的 "质量" 即可靠性可以判定为低; 当该点的最佳尺度因 子过小时, 说明该像素点附近的局部信号与高频率的正弦信号比较吻合, 可 能是受到噪声干扰, 该点的 "质量"也可以判定为低。 故首先应该找到这样 一个数值, 当像素点的最佳尺度因子等于该数值时, 像素点的 "质量"最好, 步骤 3.1为找到这样一个数值的过程, 步骤 3.2为建立质量图的过程, 质量图 中每一点质量的含义实际上是该点的尺度因子与最优尺度因子的 "距离"。 本 步骤中求取的质量图用于在步骤 5 中将相对相位分布图分成两个部分, 在步 骤 6中指导相位展开过程, 得到相位展开的最佳路径,
步骤 3.1: 对频率为 /。的一维正弦信号做小波变换: 求出二维复数矩阵 , 的模矩阵 ^ (^ b) , 搜索每列元素中数值最大 的元素, 记录其行标号, 求得这些行标号的平均值^ 求得最优尺度因子 :
Figure imgf000010_0001
步骤 3.2 : 求质量图 ^ , :
Figure imgf000010_0002
本发明所建立的质量图中, 质量值越大, 说明该像素点的可靠性越低, 步骤 4: 根据质量图 2 (^, 将相对相位分布图 (^ 分成两个部分, 步骤 4.1:建立高度 c,宽度 r的 0值矩阵 D, D矩阵和条纹图像大小相同, 用于区分根据质量图将相对相位图分成的两个部分,
步骤 4.2:遍历质量图 2 ( ) , 统计得到取值最多的质量值 ft , 得到阈值
Figure imgf000010_0003
步骤 4.3 : 遍历质量图 当 中的数值大于阈值 r时,矩阵 D中 相应位置元素置 1, 图 6为 D阵的图像模式,其中白色区域代表质量较好优先 展开的点, 黑色区域为需要采用洪水算法展开的点的区域,
步骤 5 :展开相对相位分布图 ) 中 D矩阵相应位置元素值为 0的点的 相位, 该步骤展开的点都是 "质量"较好的点, 且质量值相差不多, 按照传 统的洪水相位展开算法, 这些点仍然要参与排序, 但实际上排序后和排序前 的相位展开路径对于精度的影响不大, 故这一部分采用最简单的扫描线相位 展开算法即可得到比较精确的结果, 而运算且速度非常快,
步骤 5.1 : 建立高度 c, 宽度 r的 0值矩阵 S, S矩阵与条纹图像大小相同, 用于标记相对相位分布图 (Α> 的点是否已经展开,
步骤 5.2 : 在相对相位分布图 (^ 的中心像素的 20 x 20领域内选择一个 D矩阵相应位置元素值为 0的点作为展开的起始点,起始点的绝对相位值记为 该点在相对相位分布图 > 中的数值, 沿着行方向向图像两侧展开 D矩阵 相应位置元素值为 0的点的相位, 每展开一个像素点, 该点在 S矩阵相应位 置元素值置 1, 在直接展开过程中如遇到某些点的同一行的前一点并未处理, 可以将该点的上一行或者下一行的相邻点视为前一点然后展开, 如果这两点 也没有处理, 留待步骤 6处理, 每行都如此运算,
每一点具体展开过程如下: Punwrp
Figure imgf000010_0004
式中, 是前一点的绝对相位值, ^^是当前点的相对相位值, U 求得的当前点的绝对相位值, 是取整函数, 图 7为经过步骤 5过程后的 相位分布图,
步骤 6:展开相对相位分布图 ) 中 S矩阵相应位置元素值为 0的点的 相位, 这一部分为 "质量"较差的点, 为达到较高展开精度故采用洪水相位 展开算法处理, 洪水算法的流程图如图 2所示, 由于这部分点数量较少, 所 以运算时间也较快,
步骤 6.1 : 在 S矩阵中标记 0值的连通域, 二值矩阵的一个连通域是指二 值矩阵中相互连通且取值相同的元素的集合, 如图 6 中每个封闭的黑色区域 都是一个连通域, 对于每个连通域做如下处理:
步骤 6.2: 选择连通域中的任一边界点做为起始点, 入当前为空的堆栈, 步骤 6.3: 查找起始点四邻点中 S矩阵相应位置元素值为 0的点, 如果四 邻点中找不到这样的点,跳到步骤 6.5; 如果可以找到这样的点, 展开这些点的 相位, 将这些点在 S矩阵中相应位置元素值置 1, 并将这些点存入堆栈, 步骤 6.4: 对堆栈中的点按照其在条纹图像质量图 2(^ 中相应位置的数 值大小进行排序, 质量值最大点放置于堆栈的顶部,
步骤 6.5: 将堆栈的顶部的点取出做为起始点, 判断堆栈是否为空, 如果 为空, 则该连通域展开过程结束; 如果不为空, 转到步骤 6.3,
步骤 7: 经过步骤 5和步骤 6处理, 已经遍历了条纹图像的相对相位分布 图 (^ 的所有点, 得到条纹图像的绝对相位图 (^ , 如图 8所示为被测 物体的绝对相位图, 在个人计算机上, 基于 VC++平台, 本发明展开相位运算 时间为 27s, 而全场相位展开运算时间近 90min,可见本发明方法实时性较好, 根据经典光栅投影的相位到高度转换公式, 最终求得测量物体的三维信息, 图 9为将物体的三维信息以点云的形式表现出来。
由条纹图像的绝对相位图 (^ 到物体高度的公式如下:
( 、 ΙΑφ
h x, y) =
Αφ - ω0ά 其中, 是测量系统的几何参数, z是投影仪到测量平面的距离, 是
CCD摄像头到投影仪的距离, A = c,y) - % c, 表示相位变化量, ( ) 为展开相位结果, %(^ 为初始相位结果, 由测量参考面决定, 《。为投影光 栅的角频率, 由系统标定可得。

Claims

权 利 要 求 书
1 种基于小波变换的三维测量方法, 具体步骤如下:
步骤 1: 将黑白条纹图像投影到被测物体表面, 使用 CCD对被测物体表 面进行拍摄, 得到一幅高度为 宽度为 的变形条纹图像
_ ) = A( , _ ) + 5( , _ )cos 2π/0χ + φ(^χ, _ ) 其中, A >是背景光强分布, s(^是物体表面反射率, /。是正弦条 纹频率, (^是待求的相对相位分布图, 表示变形条纹图像的二维坐 标,
步骤 2: 对变形条纹图像逐行做小波变换, 得到变形条纹图像的相对相位 分布图, 具体过程如下:
步骤 2.1:将 y视为常数,采用一维连续小波变换对变形条纹图像 g( >的 第 y行进行处理, 处理过程为:
W(ai,b) = g (x, y)M*a,b (x) ^
其中, "是尺度因子, 取值范围为 10到 50, 每隔 0.2取一个值, 是平 移因子, 取值范围为 1到条纹图像宽度 每隔 1取一个值, 单位为像素, 获 得的 W(fll b)是一个 200r列的 W(fll b)中元素的行 标号, 称第 y行小波变换矩阵,
Figure imgf000012_0001
^C 是 的 共轭函数, Μ (4是小波函数, 表达式如下:
Figure imgf000012_0002
其中, Λ是小波函数的带宽, Λ是小波函数的中心频率, 为单位复数, 步骤 2.2:求取条纹图像的最佳尺度因子分布图 (^和相对相位分布图
Φ、χ, y),求取 a r (^ 和 Φ、χ, y)在坐标 (^, y)处数值的方法如下:
求出 W {ai,b)的对应的模矩阵 A ( b)和角度矩阵 φ ,ΐ , 搜索矩阵 A(fll b)第 c列中值最大的元素, 求出该元素在矩阵 A(fll b)中的行标号为 Ω , 则^=10 + 0.2>^ 为条纹图像的最佳尺度因子分布图 "^, 在坐标 ( ) 处数值, 矩阵 , 中第 列中行标号为 " 的元素数值为条纹图像的相对相 位分布图 (^在坐标 (^处数值,
遍历条纹图像所有坐标点, 求得条纹图像的最佳尺度因子分布图 (^ 和相对相位分布图 (^ , 步骤 3: 建立质量图 2(^),
步骤 3.1: 对频率为 ^的一维正弦信号做小波变换:
W1(al,b)=^cos^f0x)M*ah(x)dx:
求出二维复数矩阵 , 的模矩阵 ^(^b) , 搜索每列元素中数值最大 的元素, 记录其行标号, 求得这些行标号的平均值^ 求得最优尺度因子 :
步骤 3.2: 求质量图
Figure imgf000013_0001
步骤 4: 根据质量图 将相对相位分布图 ίΚ^分成两个部分, 步骤 4.1:建立高度 c, 宽度 r的 0值矩阵 D,
步骤 4.2:遍历质量图 β( )的每一个点,求出在质量图 β( )中出现频率 最高的元素数值 ft,得到阈值: Tzl.OSxQ ,
步骤 4.3: 遍历质量图 β(^的每一个点,当 2(^中的数值大于阈值: Γ 时, 矩阵 D中相应位置元素置 1,
步骤 5:展开相对相位分布图 (^中 D矩阵相应位置元素值为 0的点的 相位,
步骤 5.1: 建立高度 c, 宽度 r的 0值矩阵 S,
步骤 5.2: 在相对相位分布图 的中心像素的 20x20领域内选择一个 D矩阵相应位置元素值为 0的点作为展开的起始点,起始点的绝对相位值记为 该点在相对相位分布图 (^中的数值, 沿着行方向向图像两侧展开 D矩阵 相应位置元素值为 0的点的相位, 每当展开一个像素点, 该点在 S矩阵相应 位置元素值置 1, 在直接展开过程中如遇到某些点的同一行的前一点并未处 理, 将该点的上一行或者下一行的相邻点视为前一点然后展开, 每行都如此 每一点具体展开过程如下:
Figure imgf000013_0002
式中, 是前一点已经解出的绝对相位值, ^是当前点的相对相位 值, 其数值为当前点在相对相位分布图 中的数值, 是求得的当前 点的绝对相位值, 是取整函数, 步骤 6:展开相对相位分布图 (^ 中 S矩阵相应位置元素值为 0的点的 相位,
步骤 6.1 : 在 S矩阵中标记 0值的连通域, 对于每个连通域做如下处理: 步骤 6.2: 选择连通域中的任一边界点做为起始点, 入当前为空的堆栈, 步骤 6.3 : 查找起始点四邻点中 S矩阵相应位置元素值为 0的点, 如果四 邻点中找不到这样的点,跳到步骤 6.5; 如果可以找到这样的点, 展开这些点的 相位, 将这些点在 S矩阵中相应位置元素值置 1, 并将这些点存入堆栈, 步骤 6.4: 对堆栈中的点按照其在条纹图像质量图 2(^ 中相应位置的数 值大小进行排序, 质量值最大点放置于堆栈的顶部,
步骤 6.5: 将堆栈的顶部的点取出做为起始点, 判断堆栈是否为空, 如果 为空, 则该连通域展开过程结束; 如果不为空, 转到步骤 6.3,
步骤 7: 经过步骤 5和步骤 6处理, 得到条纹图像的绝对相位图 (^), 根据经典光栅投影的相位到高度的转换公式, 最终求得测量物体的三维信息。
PCT/CN2011/071441 2010-11-30 2011-03-02 基于小波变换的三维测量方法 Ceased WO2012071817A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US13/517,226 US9163937B2 (en) 2010-11-30 2011-03-02 Three-dimensional measurement method based on wavelet transform
KR1020117030045A KR101214702B1 (ko) 2010-11-30 2011-03-02 웨이블릿 변환에 기반한 3차원 측정 방법

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2010105657651A CN102032877B (zh) 2010-11-30 2010-11-30 基于小波变换的三维测量方法
CN201010565765.1 2010-11-30

Publications (1)

Publication Number Publication Date
WO2012071817A1 true WO2012071817A1 (zh) 2012-06-07

Family

ID=43886081

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2011/071441 Ceased WO2012071817A1 (zh) 2010-11-30 2011-03-02 基于小波变换的三维测量方法

Country Status (4)

Country Link
US (1) US9163937B2 (zh)
KR (1) KR101214702B1 (zh)
CN (1) CN102032877B (zh)
WO (1) WO2012071817A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104897083A (zh) * 2015-06-11 2015-09-09 东南大学 一种基于投影仪散焦解相位的光栅投影三维快速测量方法
CN112556602A (zh) * 2020-12-02 2021-03-26 深圳大学 一种相位快速展开方法及系统

Families Citing this family (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102628676B (zh) * 2012-01-19 2014-05-07 东南大学 一种光学三维测量中的自适应窗口傅里叶相位提取法
CN103063160A (zh) * 2012-12-09 2013-04-24 黑龙江科技学院 基于复Morlet小波变换的光学三维轮廓术参数的确定方法
CN103267496B (zh) * 2013-05-20 2016-01-27 东南大学 一种基于小波变换的改进窗口傅里叶三维测量法
CN103471552A (zh) * 2013-09-04 2013-12-25 陈慧群 碳纤维复合材料切削加工表面形貌表征方法
TWI524050B (zh) * 2014-04-15 2016-03-01 聚晶半導體股份有限公司 取像裝置、影像深度產生裝置及其方法
CN104964657B (zh) * 2015-06-30 2018-02-09 东莞市盟拓光电科技有限公司 一种基于质量图的多队列快速相位展开方法及系统
CN105066905B (zh) * 2015-07-20 2018-01-12 中国科学院上海光学精密机械研究所 小波变换轮廓术抑噪方法
CN105783785B (zh) * 2016-04-11 2018-05-15 重庆理工大学 一种小波脊相位提取方法
CN106643558A (zh) * 2017-03-06 2017-05-10 中国科学院光电技术研究所 一种基于相位纵向拼接的宽光谱干涉形貌检测方法
CN107014313B (zh) * 2017-05-16 2020-02-07 深圳大学 基于s变换脊值的加权最小二乘相位展开的方法及系统
CN107578385A (zh) * 2017-09-01 2018-01-12 中国空气动力研究与发展中心低速空气动力研究所 基于特征边缘提取的组合特征定位方法
DE102018211371A1 (de) * 2018-07-10 2020-01-16 Sirona Dental Systems Gmbh Optisches Messverfahren sowie optische Messvorrichtung
CN109472834B (zh) * 2018-10-23 2023-04-14 桂林电子科技大学 一种基于小波变换的卡尔曼滤波相位展开方法
CN109978948A (zh) * 2019-03-25 2019-07-05 国网上海市电力公司工程建设咨询分公司 一种基于视觉的距离测量方法
CN110057321B (zh) * 2019-04-28 2021-01-15 西安理工大学 基于x-f-k变换快速实现频域解相的三维物体面形测量方法
US11035665B2 (en) * 2019-07-30 2021-06-15 Kla Corporation System and method for enhancing data processing throughput using less effective pixel while maintaining wafer warp coverage
CN111412867A (zh) * 2020-05-06 2020-07-14 上海艾克森集团有限公司 一种板式热交换器板片结构测量装置及测量方法
CN112561861A (zh) * 2020-11-25 2021-03-26 中科融合感知智能研究院(苏州工业园区)有限公司 一种实现双目相位计算算法的方法及装置
CN113074668B (zh) * 2021-03-24 2022-03-01 四川大学 一种基于新型2d复小波的三维面形测量方法
CN113218337B (zh) * 2021-04-29 2022-08-19 苏州天准软件有限公司 条纹图像能量提取方法、存储介质和能量提取系统
CN114739322B (zh) 2022-06-09 2022-09-16 广东工业大学 一种三维测量方法、设备和存储介质
CN116105632B (zh) * 2023-04-12 2023-06-23 四川大学 一种结构光三维成像的自监督相位展开方法及装置
CN116883276B (zh) * 2023-07-10 2026-01-30 南京信息工程大学 一种基于Morlet小波变换的莫尔条纹相位信息提取方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1758020A (zh) * 2005-11-18 2006-04-12 北京航空航天大学 基于自适应正弦条纹投射的立体视觉检测系统
CN101074869A (zh) * 2007-04-27 2007-11-21 东南大学 基于相位法的三维轮廓测量方法
CN101666631A (zh) * 2009-09-07 2010-03-10 东南大学 基于正反码彩色编码条纹的三维测量方法
CN101806587A (zh) * 2010-04-29 2010-08-18 浙江师范大学 一种采用绝对相位测量的光学三维测量方法
CN101825443A (zh) * 2010-03-09 2010-09-08 深圳大学 一种三维成像方法及系统

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5953448A (en) * 1996-03-01 1999-09-14 Textile/Clothing Technology Corporation Contour measurement of an object having a discontinuous surface using block point identification techniques
KR100492517B1 (ko) * 2002-03-07 2005-06-03 장형수 닥나무 인피섬유를 이용한 인피섬유지 제조장치, 제조방법및 그 방법에 의해 제조된 인피섬유지
US7412103B2 (en) * 2003-10-20 2008-08-12 Lawrence Livermore National Security, Llc 3D wavelet-based filter and method
KR100837365B1 (ko) 2007-07-27 2008-06-12 광운대학교 산학협력단 룩업 테이블을 이용한 컴퓨터 형성 홀로그램 생성과 재생방법 및 그 장치
CN101650164B (zh) * 2009-09-07 2011-02-09 东南大学 三维扫描系统中基于错位条纹的相位展开方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1758020A (zh) * 2005-11-18 2006-04-12 北京航空航天大学 基于自适应正弦条纹投射的立体视觉检测系统
CN101074869A (zh) * 2007-04-27 2007-11-21 东南大学 基于相位法的三维轮廓测量方法
CN101666631A (zh) * 2009-09-07 2010-03-10 东南大学 基于正反码彩色编码条纹的三维测量方法
CN101825443A (zh) * 2010-03-09 2010-09-08 深圳大学 一种三维成像方法及系统
CN101806587A (zh) * 2010-04-29 2010-08-18 浙江师范大学 一种采用绝对相位测量的光学三维测量方法

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104897083A (zh) * 2015-06-11 2015-09-09 东南大学 一种基于投影仪散焦解相位的光栅投影三维快速测量方法
CN104897083B (zh) * 2015-06-11 2017-04-26 东南大学 一种基于投影仪散焦解相位的光栅投影三维快速测量方法
CN112556602A (zh) * 2020-12-02 2021-03-26 深圳大学 一种相位快速展开方法及系统
CN112556602B (zh) * 2020-12-02 2022-03-22 深圳大学 一种相位快速展开方法及系统

Also Published As

Publication number Publication date
US20120278036A1 (en) 2012-11-01
CN102032877A (zh) 2011-04-27
KR20120075441A (ko) 2012-07-06
KR101214702B1 (ko) 2012-12-21
US9163937B2 (en) 2015-10-20
CN102032877B (zh) 2012-05-23

Similar Documents

Publication Publication Date Title
WO2012071817A1 (zh) 基于小波变换的三维测量方法
CN106485740B (zh) 一种结合稳定点和特征点的多时相sar图像配准方法
US5081689A (en) Apparatus and method for extracting edges and lines
Qiao et al. A single-shot phase retrieval method for phase measuring deflectometry based on deep learning
CN103267496B (zh) 一种基于小波变换的改进窗口傅里叶三维测量法
CN104812288B (zh) 图像处理装置和图像处理方法
CN101832756B (zh) 测量图像中目标位移以及应变与应变率成像的方法和装置
CN111461295A (zh) 基于多尺度生成对抗神经网络的单帧条纹分析方法
WO2023284576A1 (zh) 深度图生成方法、装置、电子设备
CN102032875A (zh) 一种基于图像处理的电缆护套厚度测量方法
CN110223384B (zh) 一种白光干涉三维形貌重建方法、装置、系统及存储介质
Yin et al. Initializing and accelerating Stereo-DIC computation using semi-global matching with geometric constraints
Ghosh et al. ADP: Automatic differentiation ptychography
CN115131413A (zh) 基于多尺度三维频域卷积特征的多模遥感图像配准方法
CN110686652A (zh) 一种基于深度学习和结构光相结合的深度测量方法
Grifoni et al. 3D multi-modal point clouds data fusion for metrological analysis and restoration assessment of a panel painting
Feng et al. Stereo-DICNet: An efficient and unified speckle matching network for stereo digital image correlation measurement
Ye et al. Comparison and evaluation of feature matching methods for multisource planetary remote sensing imagery
CN118072054A (zh) 基于多尺度卷积特征的通用多模态影像匹配方法及装置
CN115100446A (zh) 一种sar与可见光遥感图像匹配的相似性度量方法
CN118279330B (zh) 一种多线程的正则化相位跟踪单幅干涉条纹解算方法
JP5656018B2 (ja) 球体の検出方法
CN110476043A (zh) 用于采集电磁场的波前的断层成像分布的方法和光学系统
CN113503832B (zh) 一种基于物体横向尺寸辅助的绝对相位恢复方法
Wang et al. CFOG-like image registration algorithm based on 3D-structural feature descriptor for suburban optical and SAR

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 20117030045

Country of ref document: KR

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 13517226

Country of ref document: US

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11844328

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11844328

Country of ref document: EP

Kind code of ref document: A1