WO2013075265A1

WO2013075265A1 - Method for extracting 3d surface profile of object

Info

Publication number: WO2013075265A1
Application number: PCT/CN2011/001949
Authority: WO
Inventors: 白净; 张舒; 陈颀潇
Original assignee: 清华大学
Priority date: 2011-11-23
Filing date: 2011-11-23
Publication date: 2013-05-30
Also published as: CN102869950B; CN102869950A

Abstract

The present invention relates to a method for extracting a 3D surface profile of an object, which comprises the following steps: 1) setting an imaging system comprising a CCD, a lens, several light sources, a turntable for placing an object to be imaged and a computer, the turntable being disposed with a controller; 2) obtaining an object image from the imaging system, and extracting the boundary of the object in a selected area in the image to be processed by using a level-set algorithm; 3) obtaining each tomographic image of the object according to the size of the image to be processed, and extracting, in each tomographic image, a surface profile of the layer; and 4) integrating each tomographic surface profile of the object obtained in Step 3), to form a 3D surface profile of the object, which specifically means that the boundary points are sequentially arranged and connected by layers according to a search angle. The present invention can be widely applied to rapidly acquire a 3D surface profile of an object.

Description

Method for extracting three-dimensional surface contour of object

The present invention relates to a method of extracting three-dimensional data on an object surface, and more particularly to a method of extracting a three-dimensional surface contour of an object on a simple imaging system. Background technique

In the technical fields of quality control, machine vision, medical imaging, etc., many specific applications are based on extracting the three-dimensional surface contour of an object. The optical methods for extracting the three-dimensional surface contour of an object in the prior art are: a laser-based method, a moiré method, an interferometry method, a photogrammetry method, a structured light method, etc., although a high-precision surface contour can be obtained by the above method, but these The method requires the use of a projector or other specific device to extract the three-dimensional surface contour, which not only increases the complexity of the entire system, but also increases the cost.

In the prior art, it is also possible to extract the three-dimensional surface contour of the object accurately and at full angle by using the silhouette or the visible light image, and the two methods are not required to use an additional device such as a projector, which greatly simplifies the imaging system and reduces the cost. The method of obtaining the surface contour of the object by using the silhouette adopts the visual shape technology as the core algorithm, but the visual shape technology is more complicated and difficult to implement. The method of obtaining the surface contour of an object by using visible light image is mainly based on Radon Transform, which is easy to implement, but it will produce the following problems: 1. This algorithm requires all pixels to participate in the back projection operation. Therefore, the speed of extracting the three-dimensional surface contour of the object is very slow; 2. The parallel beam back projection method adopted by such an algorithm does not match the actual lens imaging. Moreover, in extracting the boundary of an object in a visible light image, the prior art uses a binarization algorithm, which is difficult to accurately find the boundary of the object, resulting in subsequent contour extraction errors. Summary of the invention

In view of the above problems, it is an object of the present invention to provide a method for extracting a three-dimensional surface contour of an object quickly, accurately, and all angles on a simple imaging system.

In order to achieve the above object, the present invention adopts the following technical solutions: A method for extracting a three-dimensional surface contour of an object, comprising the following steps: 1) setting a CCD, a lens, a plurality of light sources, a turntable for placing an object to be imaged, and a An imaging system of a computer, wherein the turntable is provided with Controller; 2) extracting the object boundary of the selected area in the image to be processed from the image of the object obtained from the imaging system; 3) obtaining each tomographic image of the object according to the size of the image to be processed, and in each fault Extracting the surface contour of the layer on the image, which includes the following steps:

1 Define the rotating coordinate system and the shooting coordinate system in the imaging system; 2 Select a layer of the image to be processed, calculate the distance of the object boundary extracted from the layer image from the center line of the image; 3 Perform coordinate transformation, that is, calculate the object in each The coordinates of the lens centroid in the rotating coordinate system when imaging the angle, and the coordinates of the object boundary in the rotating coordinate system in all the images to be processed; 4 The non-parallel beam back projection method is used to obtain the object boundary back projection image; 5 through step 4 Extracting the boundary point coordinates of the layer of the object based on the boundary back projection image of the layer; 6 smoothing the surface contour obtained in step 5 by using a filter; 7 repeating steps 1 to 6 for each layer of the image to be processed, Obtaining the surface contour of each fault of the object; 4) combining the surface contours of the various faults of the object obtained by step 3) to obtain the three-dimensional surface contour of the object, specifically connecting the boundary points in an orderly arrangement according to the search angle. .

The step 2) comprises the following steps: 1 sequentially reading an object image from the imaging system and storing it as an image sequence, and reducing each image in the image sequence to the original according to the accuracy requirement of extracting the surface contour of the object After the image is obtained, the image to be processed is obtained, where 0 _{< s} ≤ l;

2 arbitrarily select an image from the image to be processed, manually select the region where the three-dimensional contour of the object needs to be extracted; 3 extract the boundary of the selected region of the object using the level set algorithm in the selected region for each image to be processed.

Performing the coordinate transformation in step 3) includes the following steps: 1 calculating the coordinates of the object boundary in the shooting coordinate system of the lens optical center and the image to be processed; 2 converting the coordinates of the object boundary in the shooting coordinate system by the coordinate transformation formula to Rotate in the coordinate system.

The specific process of obtaining the back-projection image of the object boundary by using the non-parallel beam back projection method in step 3) is: determining an anti-projection line by using the object boundary coordinates and the lens optical center coordinate, and all the boundaries of all the images to be processed are After all the back projection lines are determined, a back projection image of the object at a boundary of the layer is obtained, and the back projection line is opposite to the actual propagation path of the light.

In the step 3), the boundary point of the selected area of the layer of the extracted object is a method of first extracting the maximum connected area and searching by angle.

The boundary point of the selected area of the layer of the extraction object described in step 3) adopts a direct angle-by-angle search method.

The present invention has the following advantages due to the above technical solution: 1. The present invention adopts a method based on object boundary back projection, and only backprojects pixel points located on the boundary of the object. The amount of data involved in the back projection operation is greatly reduced, thus speeding up the contour extraction exponentially. 2. The invention adopts the level set method instead of the binarization method to realize the extraction of the object boundary, and can automatically recognize the object boundary when the parameter selection is appropriate, even in the case of uneven illumination, the object can be better The boundary is accurately extracted, which effectively reduces the error in extracting the contour of the surface of the object. 3. The invention adopts the method of non-parallel beam back projection, which is more in line with the physical law of lens imaging. 4. The present invention reduces the image obtained from the imaging system. If the accuracy of the surface contour of the object to be extracted is not high, the image can be reduced to reduce the calculation time, and the speed of extracting the surface contour can be further accelerated. The invention can be widely applied to quickly acquire a three-dimensional surface contour of an object. DRAWINGS

1 is a schematic structural view of an imaging system of the present invention;

2 is a schematic flow chart of extracting an object's three-dimensional surface contour according to the present invention;

Figure 3 is a schematic diagram of a coordinate system of the present invention;

4 is a schematic diagram of acquiring a tomographic image by using a non-parallel beam back projection;

Figure 5 is a diagram showing the state of the boundary back projection image of the present invention;

Figure 6 is a schematic view showing the surface profile extraction of the present invention;

Figure 7 is a schematic illustration of the three-dimensional results of the surface of the extracted object of the present invention. detailed description

The invention will now be described in detail in conjunction with the drawings and embodiments.

As shown in FIG. 1, the three-dimensional surface simple imaging system of the present invention comprises a CCD (Charge Coupled Device) 1 packaged in a case, and a lens 2 is arranged on the front end of the case with the CCD 1 , and a lens is disposed on both sides of the lens 2 The two light sources 3, 4 for illumination, the two light sources 3, 4 can be white or monochromatic light according to actual needs, and a turntable 5 for placing an object to be imaged is disposed directly in front of the lens 2, and a controller is arranged in the turntable 5 (in the figure) Not shown), the computer 6 sends a signal to control the object to be imaged to achieve 360 degree rotation and upper and lower translation of the object to be imaged, and selects the rotation angle interval of the object to be imaged according to actual needs during shooting, that is, determines that the CCD1 is to be photographed. The position of the imaged object, the CCD 1 sends the image of the captured object to the computer 6 to extract the three-dimensional contour information of the surface of the object.

In the above embodiment, the number of light sources may be selected according to actual experimental conditions as long as the illumination conditions required for the object at the time of shooting can be satisfied. As shown in FIG. 1 to FIG. 2, when the image of the object to be imaged is collected by the embodiment of the present invention, the objects to be imaged on the turntable 5 are irradiated by the two white light sources 3 and 4, and the object to be imaged is selected along the turntable 5 every 5°. The direction rotation (counterclockwise or clockwise) means that the CCD 1 takes an image of the object every time the object to be imaged is rotated by 5 °, and the CCD 1 sequentially transmits the captured 72 object images to the computer 6, and the computer 6 performs the acquired object image. The method for extracting surface three-dimensional contour information includes the following steps:

1. Extract the object boundary of the selected area in the image to be processed by using the level set algorithm on the obtained object image.

1) The computer 6 sequentially reads 72 object images and stores them into image sequences. According to the accuracy requirements of the surface contours to be extracted, each image in the image sequence is reduced to the original image to obtain a to-be-processed image. /, where 0 _{< S} ≤ 1. For example, the original image size is 512x512, which means that the image is composed of 512 rows and 512 columns of pixels, and the image is reduced to 0.5 times of the original image, and the size of the reduced image is 256x256.

The above image reduction may be performed by an image reduction method based on wavelet transform, discrete cosine transform, and downsampling (separated point sampling) in the prior art. In this embodiment, an image reduction method based on downsampling is adopted, which is specifically: For the 512 pixels, the method of downsampling is used to reduce the number of pixels by 1, 5, 5, 7, ..., 511, and 256 pixels are discarded, and the pixels with the even number are discarded; For an image composed of two-dimensional 512x512 pixels, this method is used to reduce the number of pixels by 0. 5 times, which is an odd number of pixels, such as (1,1), (1,3), etc., a total of 256x256 Pixels, which get an image with a size of 256x256.

2) Select an image from the image to be processed obtained in the above step 1), and manually select the region where the three-dimensional contour of the object needs to be extracted.

Selecting any image in the processed image is displayed on the computer 6, and the region W on which the three-dimensional contour needs to be extracted is selected on the image by using the mouse, and the coordinates of the selected region are recorded at this time, since the sizes of all the images to be processed are the same. For example, 256X256, so according to the coordinates of the selected area, the position of the area on the other images can be obtained.

3) Extracting the selected area of the object using the level set algorithm for each image to be processed/in the selected area?

In the above formula, ^ is a level set function, which is time (the level set function ^ is a function of time, Is the evolution time of the level set function), Δ is the Laplacian operator, V is the gradient, div is the divergence, 0) is the Dirac function, g = is the boundary indication function,

Medium is a Gaussian kernel function with a standard deviation of σ, / is the image to be processed, is the internal energy weight, is the weight of the horizontal set curve length, and V is the convergence driving force weight. The values of μ, JL, and v should be determined according to actual needs. The algorithm uses iterative method of level set evolution. In the prior art, there are various iteration stop conditions, such as satisfying ^^, etc., where is the level set function. Increment, A is the number of iterations, f > 0 is the iteration stop threshold, and the appropriate iterative stop condition and iterative stop threshold can be selected according to actual needs. In this implementation, it is selected as the iterative stop condition, and the relevant parameter selected is ^ = 0.2. = 8, v = 8, At = \ , _f = 0.01. 2. Obtaining each tomographic image of the object according to the size of the image to be processed, and extracting the surface contour of the layer on each tomographic image, specifically including the following steps:

1) Define a rotating coordinate system and a shooting coordinate system in the imaging system.

As shown in FIG. 3, the rotating coordinate system X _F O _F defined in the imaging system is located at the rotation axis of the turntable 5; the defined shooting coordinate system is such that its origin is located at the optical center of the lens 2.

2) Select the first layer of the image to be processed, and calculate the distance between the boundary of the object extracted by the layer image from the center line of the image, that is, the size of the image to be processed is 256x 256, then the image to be processed is 256 layers, and the first layer of the image is stored. For an image of size 256 X 256, the center line of the image is 128. 5 columns.

3) Perform coordinate transformation to calculate the coordinates of the optical center of the lens 2 in the rotating coordinate system at each imaging angle, and the coordinates of the object boundary in the rotating coordinate system in all the images to be processed, including the following steps:

1 Calculate the coordinates of the optical center of the lens 2 and the boundary of the object in the image to be processed in the shooting coordinate system.

According to the following optical imaging formula and the parameter ratio relationship of the selected CCD1, the lens 2 is obtained. The coordinates of the optical center in the shooting coordinate system are (0,0), the coordinates of the layer; t boundaries in the shooting coordinate system (v, o, where the optical imaging formula is:

丄―丄+丄

f ^uv (2)

The proportional relationship of the parameters of CCD1:

u _ FOV

(3) In the above formula, / is the focal length of the lens 2, "is the object distance, V is the image distance, w _CCD is the width of the actual size of the selected CCD1, and ror is the field of view of the selected CCD1.

2 The coordinates in the shooting coordinate system obtained above are converted into a rotating coordinate system by a coordinate transformation formula.

As shown in FIG. 3, in the embodiment, the turntable 5 drives the object to be imaged to rotate clockwise, and the relative shooting coordinate system rotates counterclockwise with respect to the rotating coordinate system. When the turning angle of the turntable 5 is , the shooting coordinate system becomes xO, and the coordinate transformation formula of the coordinates of the shooting coordinate system to the rotating coordinate system is:

x _r = x _c cos^-y _c sin^+wcos^

y _r =x _c sin^+ '_(; cos^+Msin^ (4) That is, the coordinates of the optical center of the lens 2 in the rotating coordinate system are: ( _MC0S ^ _MS in^ _; the layer is the first; t boundaries are rotating The coordinates in the coordinate system are: ( vcos<p— d _k sin^ + wcos^vsin^+^ cos^ + wsin^).

4) Using the non-parallel beam back projection method to obtain the object boundary back projection image.

As shown in FIG. 4, an inverse projection line can be determined from the obtained object boundary coordinates and the optical center coordinates of the lens 2. This back projection line is opposite to the actual propagation path of the light, and the back projection lines of all the boundaries of the 72 images are After all calculations are completed, the object is back projected from the boundary of the layer (as shown in Figure 5). In the boundary back projection image of the embodiment, the back projection line is represented by a dot having a pixel value of 1 (as shown in the white portion of FIG. 4 and FIG. 5), and the remaining portion is a dot having a pixel value of 0 (FIG. 4, FIG. 5). In the black part).

5) Obtain the boundary point coordinates of the selected area of the object based on the boundary back projection image of the first layer by the above step 4).

There are two methods for extracting the surface contour from the back projection image, one is to extract the largest connected region first, then the angle-by-angle search method, and the other is the direct angle-by-angle search method.

This embodiment adopts a method of first extracting the maximum connected area and searching by angle. Through the opposite side The boundary back projection image (shown in Figure 5) extracts the maximum connected region. First, the first tomographic image of the object is obtained. The maximum connected region is represented by a pixel with a pixel value of 1 (as in the white part of Figure 6), and the rest is pixelized. A point with a value of 0 indicates (as in the black part of Figure 6), then the center point of the largest connected area is used as the search starting point, and the counterclockwise direction is the search direction, and the boundary is searched within the 360° range, that is, when the search angle is "When you search for the boundary point from the search starting point, once the boundary point is found, the search angle at this time" and the length ^ (the length from the search start point to the boundary point) are recorded, that is, in polar coordinates. Stored boundary point coordinates. The search starting point of this method is the center point of the largest connected area, and there are a total of pixels in the maximum connected area. The coordinates of each pixel in the rotating coordinate system are P ( ^X "W, then the center point of the largest connected area is

N N

N N , in this embodiment, the search angle increment Δ" = 1°.

The direct angle-by-angle search method is applied to the case where each plane of the object contains a rotating shaft, and the contour search starting point is fixed, which may be the center point of the back projection image, that is, the origin of the rotating coordinate system.

6) The surface contour obtained by the above step 5), that is, the boundary point, is smoothed by a finite impulse response digital low-pass filter.

7) Repeat steps 1) ~6) for each layer of the image to be processed to obtain the surface contour of each fault of the object.

3. Combine the surface contours of the various faults of the objects obtained through the above steps to obtain the three-dimensional surface contour of the imaged object. Specifically, the boundary points are arranged in an orderly arrangement according to the search angle in the order of the layers (as shown in FIG. 7).

The above embodiments are only for illustrating the present invention, wherein the structure of the imaging system is changeable, and the steps of the method of extracting the three-dimensional surface contour of the object and the sequence of the implementation may be varied, and the technical solution of the present invention Equivalent transformations and improvements made on the basis of the invention should not be excluded from the scope of protection of the present invention.

Claims

Rights request

What is claimed is: 1. A method of extracting a three-dimensional surface contour of an object, comprising the steps of:

1) providing an imaging system including a CCD, a lens, a plurality of light sources, a turntable for placing an object to be imaged, and a computer, wherein the turntable is provided with a controller;

2) extracting the object boundary of the selected area in the image to be processed by using the level set algorithm from the image of the object obtained from the imaging system;

3) obtaining respective tomographic images of the object according to the size of the image to be processed, and extracting the surface contour of the layer on each of the tomographic images, which comprises the following steps:

1 defining a rotating coordinate system and a shooting coordinate system in the imaging system;

2 selecting a layer of the image to be processed, and calculating the distance between the boundary of the object extracted on the image of the layer and the center line of the image;

3 performing coordinate transformation, that is, calculating the coordinates of the lens optical center in the rotating coordinate system when the object is at each imaging angle, and the coordinates of the object boundary in the rotating coordinate system in all the images to be processed;

4 using the non-parallel beam back projection method to obtain the object boundary back projection image;

5, by step 4, extracting the boundary point coordinates of the layer of the object based on the boundary back projection image of the layer;

6 smoothing the surface contour obtained in step 5 by using a filter;

7 Repeat steps 1 to 6 for each layer of the image to be processed to obtain the surface contour of each fault of the object;

4) Combine the surface contours of the respective faults of the object obtained by the step 3) to obtain the three-dimensional surface contour of the object, specifically, the boundary points are sequentially arranged in the order of the layers according to the search angle.

2. A method of extracting a three-dimensional surface contour of an object according to claim 1, wherein: said step 2) comprises the steps of:

1 sequentially reading an object image from the imaging system and storing it as an image sequence, and according to the accuracy requirement of extracting the surface contour of the object, each image in the image sequence is reduced to the image of the original image to obtain a to-be-processed image, Where 0 _{< s} ≤ l ;

2 arbitrarily select an image from the image to be processed, and manually select an area where the three-dimensional contour of the object needs to be extracted;

3 For each image to be processed, the level set algorithm is used to extract the boundary of the selected area of the object in the selected area.

3. A method for extracting a three-dimensional surface contour of an object according to claim 1, wherein: performing coordinate transformation in step 3) comprises the following steps:

1 calculating the coordinates of the object boundary in the shooting coordinate system in the lens optical center and the image to be processed;

2 Convert the coordinates of the object boundary in the shooting coordinate system to the rotating coordinate system by the coordinate transformation formula.

4. A method for extracting a three-dimensional surface contour of an object according to claim 2, wherein: performing coordinate transformation in step 3) comprises the following steps:

5. A method for extracting a three-dimensional surface contour of an object according to claim 1 or 2 or 3 or 4, wherein: in step 3), a non-parallel beam back projection method is used to obtain an object boundary back projection image. The specific process is: using an object boundary coordinate and a lens optical coordinate to determine an anti-projection line, all the back projection lines of all the boundaries of the image to be processed are determined, and then the boundary back projection image of the object at a certain layer is obtained, The back projection line is opposite to the actual propagation path of the light.

The method for extracting a three-dimensional surface contour of an object according to claim 1 or 2 or 3 or 4, wherein: the boundary point of the selected region of the layer is extracted from the object in step 3) The method of extracting the maximum connected area first and then searching by angle is adopted.

The method for extracting a three-dimensional surface contour of an object according to claim 5, wherein: in the step 3), the boundary point of the selected region of the layer is extracted first, and the maximum connected region is extracted first. , then the method of searching by angle.

The method for extracting a three-dimensional surface contour of an object according to claim 1 or 2 or 3 or 4, wherein: the boundary point of the selected region of the layer of the extracting object described in step 5) A direct angle-by-angle search method is employed.

The method for extracting a three-dimensional surface contour of an object according to claim 5, wherein: the boundary point of the selected region of the extracted object in step 3) is a direct angle-by-angle search method. .