A method and device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace

Abstract

The present invention relates to a method and device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace, wherein the device comprises a binocular image acquisition module (11), an identification module (12) and a detection module (13); the method comprises acquiring left and right images; respectively identifying left and right image points corresponding to the same boundary point in the left and right images, calculating the spatial position coordinates according to a trigonometric method, and then converting the coordinates into world coordinates with a liquid surface of Z=0 through matrix conversion, substituting the corresponding x and y coordinates into an equation, and calculating a, i.e. the crystal rod diameter. The detection method and device directly measure the physical diameter, thus avoiding system errors.

Description

Technical Field

[0001] The invention relates to image recognition and processing technology, and in particular to a method and a device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace based on machine vision. Background Art

[0002] The automatic detection method of the diameter of the crystal rod of the CZ silicon single crystal furnace reduces the need for production personnel and improves production efficiency and inspection results. The Chinese invention patent "Diameter measurement and control system of CZ germanium single crystal", publication number CN102912429B, discloses a diameter measurement and control system of CZ germanium single crystal, characterized in that: the system has a camera, an image processing device and an automatic control unit, the camera collects images of the equal-diameter growth of the germanium single crystal, the image processing device identifies the position of the aperture of the solid-liquid interface and calculates the curvature radius of the aperture of the solid-liquid interface, so that the automatic control unit can achieve the equal-diameter growth of the germanium single crystal.

[0003] This technology calculates the diameter of the crystal rod by collecting the intersection points between the light spot position and three straight lines on the solid-liquid interface image. The calculation principle is:

[0004] According to the sine formula 2R=a / sinA=b / sinB=c / sinC, the image processing device calculates the radius of the circle determined by the three points where the aperture of the solid-liquid interface intersects with the three preset straight lines as the radius of curvature R, where a represents the opposite side of point A in the triangle formed by points A, B, and C, and A also represents the inner angle of the triangle with A as the vertex; b represents the opposite side of point B in the triangle formed by points A, B, and C, and B also represents the inner angle of the triangle with B as the vertex; c represents the opposite side of point C in the triangle formed by points A, B, and C, and C also represents the inner angle of the triangle with C as the vertex. However, because the camera axis is not absolutely perpendicular to the liquid surface, the edge of the bright spot at the solid-liquid interface obtained by imaging is not an absolute circle, and the closer to the lower edge, the more serious the deformation, which should be an ellipse, so the calculation formula for the circle is not applicable.

[0005] In addition, these detection methods are all based on a single camera, and the pixel diameter calculated by a single camera is multiplied by a manually input scaling factor to obtain an estimated value of the diameter. The image collected on the plane is different from the actual shape of the three-dimensional space, so the system error caused is inevitable. Summary of the invention

[0006] The technical problem to be solved by the present invention is how to provide a method and device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace, which can obtain the actual physical diameter of the crystal rod and avoid the systematic error interference caused by the estimated value.

[0007] The first technical problem of the present invention is solved as follows: a method for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace is constructed, using a dual camera, comprising:

[0008] The dual cameras simultaneously collect the left and right spot plane images during the crystal pulling process;

[0009] The image identifies the left and right image points corresponding to the same point on the outer edge of the crystal rod in the left and right spot plane images, and then calculates the actual spatial position coordinates of the same point through optical geometric relationship according to the positions of the left and right image points;

[0010] Obtaining the actual spatial position coordinates of the same point and converting them into world coordinates with the liquid surface as the Z=0 plane and the center of the light spot as the coordinate origin;

[0011] Substitute the x,y coordinates of the world coordinates into the equation , calculate a, that is, the diameter of the crystal rod; where: x, y, z are the horizontal, vertical and vertical coordinates of the world coordinates respectively, and a and b are the coefficients to be determined.

[0012] According to the method for detecting the diameter of the crystal rod in the CZ silicon single crystal furnace provided by the present invention, the matrix conversion has a variety of specific forms of implementation methods, which can be implemented step by step, such as: first, the liquid surface is taken as the Z=0 plane and the upper left corner of the image is taken as the coordinate origin for conversion.

[0013] According to the method for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention, the identification includes:

[0014] Perform edge detection on the left and right spot plane images, and calculate the edge area and edge center of gravity of all closed edges;

[0015] Each spot plane image selects the closed edge with the largest area and the center of gravity at the center of the image as its own characteristic area to generate the corresponding left and right measurement windows;

[0016] Divide the left and right measurement windows equally in the vertical height, and calculate the points with the largest edge gradients at each divided height by scanning from left to right or from right to left as the left and right feature image point sets to be selected;

[0017] Multiple identical points are matched according to the corresponding identical average height and scanning order, and each identical point corresponds to only one left candidate feature image point set element and one right candidate feature image point set element.

[0018] According to the method for detecting the diameter of the crystal rod in a CZ silicon single crystal furnace provided by the present invention, the equivalent value includes but is not limited to 3, 5 or 7.

[0019] According to the method for detecting the diameter of the crystal rod of the CZ silicon single crystal furnace provided by the present invention, the generation includes calculating the measurement window size of the area to be detected, the width is: 2*(X1-X0) and the height is (Y3-Y2) / 2; wherein: X0, X1, Y2, and Y3 are respectively the abscissa of the center of gravity position of the area to be detected, the abscissa of the left edge limit position, the ordinate of the upper edge limit position, and the ordinate of the lower edge limit position.

[0020] According to the method for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention, the light spot is a light spot generated by the crystal pulling process itself without the need for an external light source.

[0021] According to the method for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention, dual cameras collect data through two or the same observation windows on the CZ silicon single crystal furnace.

[0022] Another technical problem of the present invention is solved as follows: a device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace is constructed, characterized in that it comprises:

[0023] Binocular image acquisition module with two cameras: used to simultaneously collect the left and right spot plane images during the crystal pulling process in the furnace through the observation window on the CZ silicon single crystal furnace;

[0024] The input end is electrically connected to the recognition module of the binocular image acquisition module: used for image recognition of the left and right image points corresponding to the same point on the outer edge of the crystal rod in the left and right spot plane images;

[0025] The input end is electrically connected to the detection module of the recognition module: used to calculate the spatial position coordinates of the corresponding same point according to the positions of the left and right image points, and then calculate the spatial position coordinates of the same point according to the fitting equation Calculate a, that is, the diameter of the crystal rod; where: x, y, z are the horizontal, vertical and vertical coordinates of the world coordinate system with the liquid surface as the Z=0 plane and the center of the light spot as the coordinate origin, and a and b are the coefficients to be calculated.

[0026] According to the device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention, the detection module includes a spatial position measuring unit for calculating the spatial position coordinates corresponding to the same point according to the positions of the left and right image points.

[0027] According to the device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention, the detection module further comprises a space size measuring unit for fitting an ellipse or circle equation according to the spatial positions of a plurality of identical points.

[0028] According to the device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention, the camera includes but is not limited to a charge-coupled device (CCD) or a complementary metal oxide semiconductor (CMOS).

[0029] The method and device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace provided by the present invention adopts spot recognition based on automatic detection and direct spatial coordinate measurement, and has at least the following advantages compared with the prior art:

[0030] 1. Avoid the adverse detection effect caused by the change of the position of the crystal rod spot as the crystal rod shakes during the seeding process;

[0031] 2. Avoid the adverse detection effect caused by part of the light spot being blocked when the diameter is reduced during the seeding process;

[0032] 3. Avoid detection errors caused by liquid level fluctuations during the crystal pulling process. The diameter of the crystal rod can be directly calculated to avoid system errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Figure 1 This is a schematic diagram of the functional module structure of a preferred CZ silicon single crystal furnace crystal rod diameter detection device of the present invention;

[0035] Figure 2 yes Figure 1 Schematic diagram of the detection program control flow of the device shown.

[0036] Among them, the figure marks are: 11-binocular image acquisition unit, 12-recognition module, 13-detection module. DETAILED DESCRIPTION

[0037] First, the principle and core of the present invention are described:

[0038] principle

[0039] During the single crystal silicon pulling process, it is necessary to monitor the pulling situation inside the single crystal furnace in real time. During the crystal pulling process, a bright spot will form at the solid-liquid interface between the crystal rod and the molten silicon liquid. The CCD industrial camera is used to collect the image of the pulling process, and the image is first corrected to eliminate the image distortion. The grayscale feature of the corrected image is first extracted to determine the grayscale value of the spot image part, so that the image spot part can be distinguished from the molten silicon solution through threshold segmentation, and then the image features of the spot can be extracted for subsequent detection.

[0040] core

[0041] The present invention installs two area array CCD cameras that have been calibrated on the cover of a single crystal furnace, collects images during the crystal pulling process through the two CCD cameras, first pre-processes the collected images, then performs limit correction on the pre-processed images, analyzes the images after limit correction, analyzes the imaging characteristics of the bright spot image at the solid-liquid interface between the crystal rod and the molten silicon solution, calculates the coordinates of the corresponding feature points in the two edge images in the camera coordinate system by finding the maximum value of the bright spot edge gradient at the solid-liquid interface between the crystal rod and the molten silicon solution, calculates the parallax of the same point in the image of the two industrial cameras, calculates the actual three-dimensional space coordinates of the feature points in the camera coordinate system by the triangulation method, calculates the conversion matrix of the camera coordinate system relative to the world coordinate system, and brings the world coordinates of the feature points calculated by the triangulation method into the calculation formula. The world coordinates are a coordinate system with the liquid surface as the Z=0 plane and the center of the light spot as the coordinate origin.

[0042] Method 1: Let the equation of the ellipse be , calculate the ellipse major axis coefficient a, where a is the crystal rod diameter.

[0043] Method 2: Near the intersection of the major axis and the ellipse, the distance between the centroid point O and the feature point can be approximately equal, so the diameter of the circle can be calculated using the formula for calculating the radius using the chord length of the circle.

[0044] The present invention collects images of the bright spot part, and the liquid level fluctuation within a small range has little effect on the bright spot part and has no effect on the detection. The present invention can avoid the detection error caused by the liquid level fluctuation during the crystal pulling process. The diameter of the crystal rod can be directly calculated.

[0045] Another core of the present invention is the light spot recognition based on automatic detection.

[0046] Second, briefly describe the device for detecting the diameter of the crystal rod of a CZ silicon single crystal furnace of the present invention:

[0047] like Figure 1 As shown, the device comprises:

[0048] The binocular image acquisition module 11 is composed of respective encoders and two independent cameras. The two independent cameras preferably use CCD image sensors to acquire binocular images through the observation window on the single crystal furnace: left and right spot plane images;

[0049] The recognition module 12 is used for image recognition of the left and right image points corresponding to the same point on the outer edge of the crystal rod (outer edge of the light spot) in the left and right light spot plane images;

[0050] The detection module 13 is mainly composed of a spatial position calculation unit and a spatial size calculation unit. The spatial position calculation unit calculates the spatial position coordinates of the same point according to the positions of the left and right image points through triangular geometric optical relationships; the spatial size calculation unit performs ellipse or circle fitting according to the spatial position coordinates of several same points to solve the corresponding size parameters of the corresponding equation.

[0051] Third, the present invention is further described in detail in conjunction with the best embodiment of the present invention:

[0052] The present invention preferably uses a control program in a device for detecting the diameter of a crystal rod in a CZ silicon single crystal furnace based on machine vision, specifically as follows Figure 2 As shown, the following steps are included:

[0053] 201) Start;

[0054] 202) Collect pictures during the crystal pulling process, and use the camera mapping matrix obtained from the first camera CCD1 and the second camera CCD2 to perform extreme correction on the collected pictures to obtain extreme corrected pictures. The extreme corrected picture can ensure that in the corrected image, in the actual crystal pulling process, the actual physical target point of the crystal rod is located at the same image height in the left camera corrected picture and the right camera corrected picture. The extreme corrected picture calculates the spot position in the left and right corrected pictures through the spot recognition algorithm, sets the feature point calculation window of the same height within the field of view, divides the measurement window equally along the Y direction, and at each equally divided height, calculates the maximum point of the edge gradient from left to right as the feature point. The pixel coordinates of the feature point are recorded to calculate the parallax formed by the target point on the left and right views. The target point is also called the same point;

[0055] 203) Analyze the position of the spot area in the image after the extreme correction, obtain the pixel coordinates of the edge points of the spot area, mark the image point coordinates of CCD1 as set A, and the image point coordinates of CCD2 as set B;

[0056] 204) Since the viewing angles of the two cameras CCD1 and CCD2 are different, the edge points are different. The left and right image point coordinates of the same point in the CCD1 and CCD2 extreme correction images are obtained from the set A and the set B in step 203);

[0057] 205) For the multiple groups of left and right image point coordinates of the same point obtained in step 204), the parallax of the same point in the images of the two industrial cameras is calculated, and the spatial position coordinates of the same point are calculated by triangulation method;

[0058] 206) The spatial position coordinates are converted into world coordinates with the liquid surface as the Z=0 plane and the center of the light spot as the coordinate origin through the relationship matrix, and then the x and y coordinates of the world coordinates are substituted into the fitting equation , calculate a, i.e. the diameter of the crystal rod; wherein: x, y, z are the horizontal coordinate, vertical coordinate and vertical coordinate of the world coordinate respectively, and a and b are the coefficients to be calculated;

[0059] 208)End.

[0060] Wherein: the spot recognition algorithm in step 202) uses automatic detection of the spot position and image information, which can accurately extract accurate spot information on the original image, facilitate the follow-up movement of the detection window, and improve the detection accuracy. It does not adopt the method of setting the feature area fixed at the image position, and then calculating the spot position and other detection information in the feature area, so as to avoid the shaking of the crystal rod during the seeding process, the change of the position of the crystal rod spot, and the poor detection effect caused by the shrinkage of the diameter during the equal diameter process and the blocking of part of the spot.

[0061] The spot recognition algorithm is as follows:

[0062] 301) Obtaining the gray value V_l of the liquid surface, the gray value mean V_mean of the entire image, and the gray value maximum V_max;

[0063] 302) Perform threshold segmentation on the image and calculate the threshold V_thre = V_1 + (V_max - V_mean) / 2;

[0064] 303) For the image after threshold segmentation, edge detection is performed, and the edge area and edge center of gravity of all closed edges are calculated;

[0065] 304) Set the area screening threshold to screen the candidate areas whose area exceeds a certain range, and the spot with the center of gravity on the left side of the image field of view is the candidate area;

[0066] 305) According to the centroid coordinates and edge point coordinates of the selected area, the centroid coordinates P(X, Y), the left edge limit position P1(X1, Y1), the upper edge coordinate point P2(X2, Y2), and the lower edge point coordinates are: P3(X3, Y3);

[0067] 306) Calculate the measurement window size of the area to be detected, the width is: 2*(X1 - X) and the height is (Y3 - Y2) / 2;

[0068] 307) In the original image, the point with the largest edge gradient is calculated from left to right in the measurement window as the feature point to be selected.

[0069] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A method for detecting the diameter of a CZ silicon single crystal furnace crystal rod, characterized in that: It uses dual cameras, including: The dual cameras simultaneously collect the left and right spot plane images during the crystal pulling process; The image identifies the left and right image points corresponding to the same point on the outer edge of the crystal rod in the left and right spot plane images, and then calculates the actual spatial position coordinates of the same point through optical geometric relationship according to the positions of the left and right image points; Obtaining the actual spatial position coordinates of the same point and converting them into world coordinates with the liquid surface as the z=0 plane and the center of the light spot as the coordinate origin; Substitute the x,y coordinates of the world coordinates into the equation , calculate a, i.e. the radius of the crystal rod; where: x, y, z are the horizontal, vertical and vertical coordinates of the world coordinates, respectively, and a and b are the coefficients to be calculated The identification includes: Performing edge detection on the left and right light spot plane images, and calculating edge areas and edge centroids for all closed edges; Each spot plane image selects the closed edge with the largest area and the center of gravity at the center of the image as its own characteristic area to generate the corresponding left and right measurement windows; The left and right measurement windows are equally divided in the longitudinal direction of the windows, and the points with the largest edge gradients are calculated at each equally divided height by horizontal scanning from left to right or from right to left as the left and right feature image point sets to be selected; Multiple identical points are matched according to the corresponding identical average height and scanning order, and each identical point corresponds to only one element of the left candidate feature image point set and one element of the right candidate feature image point set.

2. The diameter detection method according to claim 1, characterized in that: The equivalent value is 3, 5 or 7.

3. The diameter detection method according to claim 1, characterized in that: The generation includes calculating the measurement window size of the area to be detected, the width is: 2*(X1-X0) and the height is (Y3-Y2) / 2; wherein: X0, X1, Y2, Y3 are respectively the horizontal coordinate of the center of gravity position of the area to be detected, the horizontal coordinate of the left edge limit position, the vertical coordinate of the upper edge limit position and the vertical coordinate of the lower edge limit position.

4. The diameter detection method according to claim 1, characterized in that: The light spot is a light spot generated by the crystal pulling process itself without the need for an external light source.

5. The diameter detection method according to any one of claims 1 to 4, characterized in that: The dual cameras collect data through two or the same observation windows on the CZ silicon single crystal furnace.

Images