Hand-held target visual coordinate measurement method and system based on phase characteristics, and medium

Abstract

The invention discloses a handheld target visual coordinate measurement method and system based on phase characteristics and a medium, and the measurement method comprises the steps: obtaining an RGB three-channel image through a binocular camera based on target characteristic points of a three-frequency superposition phase to obtain a three-step phase shift image, and achieving the three-dimensional reconstruction of a circular / elliptical cone with a single area array characteristic through the combination of a calibrated phase-height model, due to the fact that feature imaging is in a point-to-point one-to-one mapping relation, it can be effectively avoided that an image feature set is extracted in the perspective imaging process of a binocular camera, and the invention further comprises a system for executing the method and a medium for storing the method. According to the method, the measurement efficiency can be improved, the feature point extraction error is reduced, the measurement view field is enlarged, the influence of imaging system defocus on target surface feature point extraction is solved, and online high-precision measurement of the handheld target system is realized.

Description

Technical Field

[0001] The invention belongs to the technical field of visual measurement, and in particular relates to a method, system and medium for measuring visual coordinates of a handheld target based on phase characteristics. Background Art

[0002] With the increasing demand for various surface inspections in industrial sites, portable 3D coordinate measurement technology based on visual measurement is increasingly being used. Handheld target visual measurement systems usually only require one or several binocular cameras, calibration plates, handheld targets and special probes to complete in-situ / online measurement of large and complex parts. They are less affected by the on-site environment and have high reliability, especially for the measurement of complex cavities, deep and long holes, etc.

[0003] The handheld target is an important measuring device in the measurement system. No matter what form it presents, it consists of the following parts: visual features, probes, support carriers connecting visual features and other control and adjustment auxiliary equipment. Visual features are artificially constructed auxiliary features to solve the measurement of weak textures, invisible and other weak feature surface measurements. The 3D coordinates of the probe can be located through the rigid transformation relationship between the easy-to-capture and locate visual feature coordinates and the probe coordinates to be measured on the target. During measurement, the probe is manually contacted with the coordinate point to be measured and photographed to achieve spatial positioning of the point to be measured. At present, visual features are mainly constructed in two ways according to the form of light emission: passive and active. The passive type is composed of reflective materials such as glass beads to form a certain shape. The feature itself does not emit light and is formed by reflection from an external light source. The active type itself is a light source, such as an LED, which actively emits light to form feature points. Generally, the active type can control the brightness of the feature point to adapt to the on-site measurement light field environment, and thus is more widely used.

[0004] In order to accurately locate the coordinates of the feature points on the target surface, on the one hand, the "feature point" LED used needs to be small, such as a surface-mount LED (SMD-LED), so that the image plane feature imaging area is small, which is convenient for coordinate extraction. On the other hand, the feature point cannot be too small. The less obvious the visual contrast is at low resolution, the greater the coordinate extraction error. At the same time, pinhole camera imaging is a perspective imaging process, and there is a perspective distortion error. It needs to be corrected according to the feature point size and the target imaging posture to improve the accuracy of image plane coordinate extraction, and the algorithm is complex.

[0005] Another key issue in the measurement process is defocus. On the one hand, the handheld target measurement system usually measures large-sized objects in-situ and online, and the measurement site environment is relatively complex. Once the handheld target imaging is out of focus, it is difficult to accurately determine the image plane coordinates of the feature points, resulting in extraction errors that seriously affect the system accuracy. Therefore, a visual imaging system with a larger depth of field is required. On the other hand, considering the actual situation of industrial field measurements, the general imaging system uses an industrial camera with a telephoto fixed-focus lens to meet the measurement requirements of high precision and high resolution. Considering the comprehensive influence of focal length and aperture, in order to obtain a larger depth of field, the aperture diaphragm of the visual measurement system needs to be adjusted to the minimum, but at this time, the imaging shape of the LED feature point will be significantly deformed by the modulation of the small aperture diaphragm, which seriously affects the image processing accuracy of the center / center of gravity coordinates extracted based on the shape of the LED feature point image. For example, the patent application with publication number CN105678709B discloses an algorithm for correcting the optical center deviation of a handheld target of LED. By analyzing the causes of the optical center deviation of the imaging with LED as the spatial feature point of the handheld target, the correction direction of each feature point on the target surface of the handheld target is found, and a pair of target imaging surfaces are obtained using a binocular stereo vision system. The epipolar geometric constraint relationship of the feature point pair is established using the basic matrix F, and the optimal correction position of the feature point is selected through the process of coarse extraction, noise reduction, coarse positioning and fine positioning. However, this patent application has the above two problems of accurate extraction of feature points and large depth of field defocus imaging. A new technology is urgently needed to overcome the limitations of the existing feature point accurate extraction and defocus imaging on the depth of field of the measurement system. Summary of the invention

[0006] In order to overcome the defects of the above-mentioned prior art, the purpose of the present invention is to propose a handheld target visual coordinate measurement method, system and medium based on phase features. The target feature "point" is a surface structure array with phase distribution. The three-step phase shift image is obtained by separating the RGB three-channel image. Combined with the set "phase-height" model, the three-dimensional reconstruction of the circular / elliptical cone of a single surface array feature is realized. Since the feature imaging is a one-to-one mapping relationship from point to point, it can effectively avoid extracting the image feature set during the perspective imaging process of the binocular camera. At the same time, the phase information is insensitive to the influence of the system defocus imaging, reduces the influence of the small aperture aperture shape modulation, increases the measurement depth of field, and expands the measurement range of the visual system.

[0007] In order to achieve the above object, the present invention adopts the following technical scheme:

[0008] The handheld target visual coordinate measurement method based on phase characteristics includes the following steps:

[0009] Step 1: Build a handheld target visual coordinate measurement system, which includes a signal-connected binocular camera 1, a handheld target 2, and a computer 3; the probe of the handheld target 2 contacts the surface of the position to be measured, and the binocular camera 1 collects images of the target surface features and transmits them to the computer 3; before measurement, the binocular camera 1 is calibrated to obtain the system internal parameter matrices A1, A2, ..., A n and the external parameter matrices [R1|t1], [R2|t2], ..., [R n-1 |t n-1 ], where n is the number of binocular cameras 1, and the external parameter matrix [R n-1 |t n-1 ] represents the rotation matrix R from the n-1th binocular camera 1 to the nth binocular camera 1 n-1 With the translation matrix t n-1 The rigid body transformation position relationship;

[0010] Step 2: Calibrate the probe of the handheld target 2 in the handheld target visual coordinate measurement system to establish the target surface feature coordinate system O f With the probe center coordinate system O p The relative position relationship between them is represented by the rotation matrix R fp , translation matrix t fp It is expressed as:

[0011] M p =R fp *M f +t fp

[0012] Among them, M p is the probe center coordinate system O p The three-dimensional coordinates of a certain point, M f is the probe center coordinate system O f The three-dimensional coordinates of a certain point;

[0013] Step 3: Use the handheld target 2 to measure the surface to be measured. During the measurement, the target probe contacts the surface of the position to be measured, and the binocular camera 1 is used to collect images of the target surface features to obtain images;

[0014] Step 4: Extract features from the collected target surface image information, that is, demodulate the color fringe image of a single surface feature to obtain the phase feature, construct a circular / elliptical cone in the feature area according to the "phase-height model", use an iterative algorithm to accurately extract the sub-pixel coordinates of the circular / elliptical cone vertex, and establish the target surface feature coordinate system according to the sub-image plane coordinates corresponding to each circular / elliptical cone vertex in the array. f , the system internal parameter matrix A calibrated in step 1 i , external parameter matrix [R i-1 |t i-1] and the rotation matrix R of the probe calibration in step 2 fp , translation matrix t fp Calculate the spatial coordinates of the probe center position;

[0015] Step 5: Based on the spatial coordinates of the probe center position in step 4 and the target surface feature coordinate system O established in step 2 f With the probe center coordinate system O p The relative position relationship between them can be used to calculate the spatial coordinates of the probe during measurement, and ultimately achieve the spatial positioning of the position to be measured.

[0016] The step 4 is specifically as follows:

[0017] (4.1) For the collected target surface image, the feature points of its color image are in the order from inside to outside along the radial direction:

[0018] Red, I r =A+Bcos[2πr(u,v) / T-2π / 3]

[0019] Green, I g =A+Bcos[2πr(u,v) / T]

[0020] Blue, I b =A+Bcos[2πr(u,v) / T+2π / 3]

[0021] Color,I rgb =(I r , I g , I b )

[0022] Among them, I r ,I g ,I b Represents color image I rgb The phase difference of the three components of the feature point is 2π / 3.

[0023] (4.2) In a color image, the green pixel size is P G (u g ,v g ), blue shows pixel size is P B (u b ,v b ), red shows pixel size P R (u r ,v r ), and there is:

[0024] u r =u b =2u g

[0025] v r =v b =v g =3u g

[0026] The size of a single color pixel block is P RGB (u rgb ,v rgb )for:

[0027] u rgb =6u g

[0028] v rgb =v r =v b =v g =3u g

[0029] u rgb / v rgb =2:1

[0030] Set the period of a single feature point to k, 1.5≤k≤3, so the display size m*n of a single feature point is:

[0031] m=P*k*u rgb

[0032] n=P*2k*v rgb ;

[0033] Among them, P>2 8 pixels; P is the number of pixels of a periodic stripe width;

[0034] Calculate the interval range of the feature point matrix quantity parameter M*N by the target surface area of ​​the handheld target 2 and the display size m*n of a single feature point;

[0035] (4.3) In the range of M*N in step (4.2), the color grating feature points are captured by binocular camera 1, and the color image I is converted into rgb Information separation, resulting in three color images I r ,I g ,I b ;

[0036]

[0037] The unwrapped phase is:

[0038]

[0039] (4.4) Based on the “phase-height” model, the surface shape is reconstructed by unfolding the phase of the feature points of the color image in step (4.3), and the reconstructed surface shape is set to a circular / elliptical cone. Therefore, the set cone height h and the physical size radius R of the feature point have the following relationship:

[0040] h=R

[0041] Since the “phase-height” model is a linear model, the proportionality coefficient q is:

[0042] q=P*u rgb / 2π

[0043] Where P is the number of pixels of a periodic stripe width, u rgb is a color pixel block P RGB Horizontal physical dimensions;

[0044] (4.5) Expand the reconstructed surface to obtain a circular / elliptical conical surface;

[0045] (4.6) Since there is overlap between the 2π / 3 phase differences of the color feature points, the sub-pixel center coordinates of the feature point cone angle are obtained by iterative calculation.

[0046] The iterative calculation of step (4.6) is specifically as follows:

[0047] (4.6.1) Set the iteration threshold ε;

[0048] (4.6.2) Randomly extract the coordinates of color feature points that are not less than 50% of the number of pixels occupied, and fit the circular / elliptical cone surface:

[0049]

[0050] According to the fitting results, calculate the average distance ε' between the coordinates of the unextracted points and the plane passing through the center axis of the cone and the generatrix where they are located;

[0051] (4.6.3) Determine the size of ε' and ε. If ε'<ε, randomly select data points that are not less than 10% of the total number of color feature points from the unselected data points, fit the circular / elliptical cone, and calculate ε' to determine the size of ε' and ε; if ε'<ε is still satisfied after extracting 100% of the data points, adjust ε to ε / (2-3) and calculate from step (4.6.1);

[0052] If ε'≥ε, randomly select the coordinates of a number of points not less than 50% of the number of pixels occupied by the color feature points, and fit the circular / elliptical cone surface until the ε' of the sample fitted circular / elliptical cone surface is less than ε;

[0053] (4.6.4) Based on the circular / elliptical cone surface determined by the data points that finally meet the threshold requirements, extract the XOY plane sub-pixel coordinates (x1,1 ,y 1,1 );

[0054] (4.6.5) Calculate the number of all feature point array parameters M*N of the color feature point cone vertex sub-pixel coordinates (x m,n ,y m,n ) to realize the extraction of target surface feature points.

[0055] A handheld target visual coordinate measurement system based on phase characteristics is installed in a computer 3, and includes a processing instruction module for running each step of the handheld target visual coordinate measurement method based on phase characteristics.

[0056] A computer-readable storage medium stores a calculation program of a handheld target visual coordinate method based on phase features. When the computer 3 executes the calculation program, the handheld target visual coordinate method based on phase features can be implemented.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] (1) In step 4 of the present invention, based on the target feature points of the three-frequency superposition phase, a binocular camera is used to obtain an RGB three-channel image to obtain a three-step phase-shifted image, and a calibrated phase-height model is combined to realize the three-dimensional reconstruction of a circular / elliptical cone of a single array feature. Since the feature imaging is a one-to-one mapping relationship between points, it can effectively avoid extracting the image feature set during the perspective imaging process of the binocular camera. At the same time, the phase information is insensitive to the influence of the system defocus imaging, which reduces the influence of the small aperture diaphragm shape modulation, increases the measurement depth of field, and expands the measurement range of the visual system.

[0059] (2) Step (4.3) of the present invention is based on the principle of spatial phase unwrapping and has no restriction on the motion state of the measured object, so real-time dynamic measurement can be achieved. At the same time, the measurement accuracy can be maintained even when a large-sized object is out of focus due to a small depth of field in a fixed-focus measurement system.

[0060] In summary, the present invention aims at the bottleneck of accurate extraction of luminous feature point coordinates in the measurement process of a handheld target measurement system; a planar array target feature structure and an accurate extraction method are proposed, which can improve measurement efficiency, reduce feature point extraction errors, increase the measurement field of view, solve the influence of imaging system defocus on target surface feature point extraction, and realize online high-precision measurement of a handheld target system. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of a handheld target visual coordinate measurement system constructed by the present invention.

[0062] Figure 2 It is a schematic diagram of the handheld target 2 and the characteristic points of the target surface.

[0063] Figure 3 It is a flow chart of the iterative algorithm of step (4.6) of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] This method is a measurement method for measuring a three-dimensional object in space using a binocular stereoscopic vision system with a handheld target 2 formed by a characteristic phase. The method aims to combine phase shift technology with RGB projection and three-channel filtering methods based on the uniqueness of phase characteristics in photographic space and Euclidean space transformation, construct the spatial transformation invariance of characteristic points on the characteristic target surface, and realize the rapid and accurate determination of the spatial points of the object to be measured. The specific implementation operation exists in the measurement process of the binocular stereoscopic vision system. The present invention is further described in detail below in conjunction with a certain measurement process.

[0066] The handheld target visual coordinate measurement method based on phase characteristics includes the following steps:

[0067] Step 1: Reference Figure 1 , a handheld target visual coordinate measurement system is built, which includes a signal-connected binocular camera 1, a handheld target 2 and a computer 3; the probe of the handheld target 2 contacts the surface of the position to be measured, and the binocular camera 1 collects images of the target surface features and transmits them to the computer 3; before measurement, the BPP method (Calibration of stereo rigs based on the Backward Projection Process [J]. Measurement Science and Technology, 2016, 27 (8): 085007) proposed by Gu, FF is used to calibrate the binocular camera 1 first, and the system internal parameter matrices A1, A2, ..., A n and the external parameter matrices [R1|t1], [R2|t2], ..., [R n-1 |t n-1 ], where n is the number of binocular cameras 1, and the external parameter matrix [R n-1 |t n-1 ] represents the rotation matrix R from the n-1th binocular camera 1 to the nth binocular camera 1 n-1 With the translation matrix t n-1 The rigid body transformation position relationship;

[0068] Step 2: Use the probe calibration method proposed by Ma, YY (Handheld target probe tip centerposition calibration for target-based vision measurement system [J]. Measurement Science and Technology, Meas. Sci. Technol. 30 (2019) 065013) to calibrate the handheld target 2 in the handheld target vision coordinate measurement system and establish the target surface feature coordinate system O f With the probe center coordinate system O p The relative position relationship between them is expressed by the rotation matrix R fp , translation matrix t fp It is expressed as:

[0069] M p =R fp *M f +t fp

[0070] Among them, M p is the probe center coordinate system O p The three-dimensional coordinates of a certain point, M f is the probe center coordinate system O f The three-dimensional coordinates of a certain point;

[0071] Step 3: Use the handheld target 2 to measure the surface to be measured, such as a deep hole, an inner cavity, or an obstructed area. During the measurement, the target probe contacts the surface of the position to be measured, and the binocular camera 1 is used to capture the target surface features to obtain an image;

[0072] Step 4: Extract features from the collected target image information, that is, demodulate the color fringe image of a single surface feature to obtain the phase feature, and construct a circular / elliptical cone of the feature area according to the "phase-height model". In order to eliminate the problem of inaccurate coordinates of the top of the circular / elliptical cone caused by crosstalk between different frequency bands, an iterative algorithm is used here to accurately extract the sub-pixel coordinates of the top of the circular / elliptical cone. Figure 2 , according to the sub-image coordinates corresponding to each circular / elliptical cone vertex in the array, the target surface feature coordinate system O is established f , the system internal parameter matrix A calibrated in step 1 i , external parameter matrix [R i-1 |t i-1 ] and the rotation matrix R of the probe calibration in step 2 fp , translation matrix t fp Calculate the spatial coordinates of the probe center position.

[0073] The step 4 is specifically as follows:

[0074] (4.1) For the collected target surface image, the feature points of its color image are circular, and the order from inside to outside along the radial direction is:

[0075] Red, I r =A+Bcos[2πr(u,v) / T-2π / 3]

[0076] Green, I g =A+Bcos[2πr(u,v) / T]

[0077] Blue, I b =A+Bcos[2πr(u,v) / T+2π / 3]

[0078] Color,I rgb =(I r , I g , I b )

[0079] Among them, I r ,I g ,I b Represents color image I rgb The phase difference of the three components of the feature point is 2π / 3.

[0080] (4.2) In a color image, the green pixel size is P G (u g ,v g ), blue shows pixel size is P B (u b ,v b ), red shows pixel size P R (u r ,v r ), and there is:

[0081] u r =u b =2u g

[0082] v r =v b =v g =3u g

[0083] The size of a single color pixel block is P RGB (u rgb ,v rgb )for:

[0084] u rgb =6u g

[0085] v rgb =v r =v b =v g =3u g

[0086] u rgb / v rgb =2:1

[0087] Set the period of a single feature point to k, 1.5≤k≤3, so the lattice parameter of a single feature point, that is, the physical size m*n is:

[0088] m=P*k*u rgb

[0089] n=P*2k*v rgb ;

[0090] In order to obtain a high-precision feature point phase map, the number of pixels in a periodic fringe width needs to satisfy the following: the number of pixels in a periodic fringe width P>2 8 pixels, set the period of a single feature point to k, 1.5≤k≤3, so the dot matrix parameter of a single feature point, that is, the physical size m*n is:

[0091] m=P*k*u rgb

[0092] n=P*2k*v rgb

[0093] Among them, P>2 8 pixels; P is the number of pixels of a periodic stripe width;

[0094] The interval of the feature point matrix quantity parameter M*N is calculated by the target surface area of ​​the handheld target 2 and the display size m*n of a single feature point;

[0095] (4.3) In the M*N interval of step (4.2), the color grating feature points are captured by the binocular camera 1, and the color image I is converted into rgb Information separation, resulting in three color images I r ,I g ,I b , based on step (4.1), since the phase difference of the feature point units of the three color images is 2π / 3, calculate their wrapped phase and unfolded phase:

[0096] The wrapping phase is:

[0097]

[0098] The unwrapped phase is:

[0099]

[0100] (4.4) In order to obtain more accurate sub-pixel center coordinates X(x, y), a "phase-height" model is required to reconstruct the surface shape of the unfolded phase in step (4.3). In order to maintain the consistency of the horizontal and vertical positioning accuracy of the target surface, the reconstructed surface shape is set to a circular / elliptical cone. Therefore, it is necessary to set the cone height h and the physical size radius R of the feature point, which have the following relationship:

[0101] h=R

[0102] Therefore, the "phase-height" model is a linear model with a proportionality coefficient q of:

[0103] q=P*u rgb / 2π

[0104] Where P is the number of pixels of a periodic stripe width, u rgb is a color pixel block P RGB Horizontal physical dimensions.

[0105] (4.5) Since the set surface shape is a circular / elliptical cone, the phase of any generatrix is ​​continuous and linear, so the reconstructed surface shape is linearly expanded to obtain a circular / elliptical cone surface shape;

[0106] (4.6) Considering the overlap between the 2π / 3 phase differences of the color feature points, which leads to the crosstalk problem between the color channels of the binocular camera 1, there will be reconstruction phase errors in the overlapping areas of different colors, which will cause water ripples in the corresponding reconstructed surface area. In order to obtain the accurate sub-pixel center coordinates of the cone angle, the following iterative algorithm is used for calculation:

[0107] (4.6.1) Set the iteration threshold ε. Based on the resolution R of the binocular camera 1, the effective size of the target surface of the handheld target 2, the number parameter M*N of the feature point array, and the maximum value of the phase deviation caused by the crosstalk between the color channels, the threshold ε is selected based on experience.

[0108] (4.6.2) Randomly extract the coordinates of 50% of the number of pixels occupied by the color feature points, and fit the circular / elliptical cone surface:

[0109]

[0110] According to the fitting results, calculate the average distance ε' between the coordinates of the un-extracted points and the plane passing through the center axis of the cone and the generatrix where they are located;

[0111] (4.6.3) Determine the size of ε' and ε. If ε'<ε, randomly select 10% of the total number of color feature points from the unselected data points, fit the circular / elliptical cone, and calculate ε' to determine the size of ε' and ε; if ε'<ε is still satisfied after extracting 100% of the data points, adjust ε to ε / 2 and calculate from step (4.6.1);

[0112] If ε'≥ε, randomly extract the coordinates of 50% of the number of pixels occupied by the color feature points, and fit the circular / elliptical cone surface until the ε' of the sample fitted circular / elliptical cone surface is less than ε;

[0113] (4.6.4) Based on the circular / elliptical cone surface determined by the data points that finally meet the threshold requirements, extract the XOY plane sub-pixel coordinates (x 1,1 ,y 1,1 );

[0114] (4.6.5) Calculate the number of all feature point array parameters M*N of the color feature point cone vertex sub-pixel coordinates (x m,n ,y m,n ) to realize the extraction of target surface feature points.

[0115] Step 5: Based on the spatial coordinates of the probe center position in step 4 and the target surface feature coordinate system O established in step 2 f With the probe center coordinate system O p The relative position relationship between them can be used to calculate the spatial coordinates of the probe during measurement, and ultimately achieve the spatial positioning of the position to be measured.

[0116] The present invention also includes a handheld target visual coordinate measurement system and medium based on phase characteristics, specifically:

[0117] A handheld target visual coordinate measurement system based on phase characteristics is installed in a computer 3, and includes a processing instruction module for running each step of the handheld target visual coordinate measurement method based on phase characteristics.

[0118] A computer-readable storage medium stores a calculation program of a handheld target visual coordinate method based on phase features. When the computer 3 executes the calculation program, the handheld target visual coordinate method based on phase features can be implemented.

[0119] The present invention is not directed to the display handheld target 2 features of a certain type of display device, whether it is an LCD screen, LED, microLED, OLED, or whether the measurement system is monocular, binocular, or multi-eye; nor is it directed to the arrangement of the array of phase feature points in a plane array, such as a triangle, square, ring, or other special shapes, etc., and any array composed of the feature points. That is, the use of the feature structure and measurement method in the handheld target 2 visual measurement system can be considered as an expansion and deformation of the method as the core.

Claims

1. A handheld target visual coordinate measurement method based on phase characteristics, characterized in that: The following steps are involved: Step 1: Build a handheld target visual coordinate measurement system, which includes a signal-connected binocular camera (1), a handheld target (2) and a computer (3); the probe of the handheld target (2) contacts the surface of the position to be measured, and the binocular camera 1 collects images of the target surface features and transmits them to the computer (3); before measurement, the binocular camera (1) is calibrated to obtain the system internal parameter matrices A1, A2, ..., A n and the external parameter matrices [R1|t1], [R2|t2], ..., [R n-1 |t n-1 ], where n is the number of binocular cameras (1), and the external parameter matrix [R n-1 |t n-1 ] represents the rotation matrix R from the n-1th binocular camera (1) to the nth binocular camera (1) n-1 With the translation matrix t n-1 The rigid body transformation position relationship; Step 2: Calibrate the probe of the handheld target (2) in the handheld target visual coordinate measurement system to establish the target surface feature coordinate system O f With the probe center coordinate system O p The relative position relationship between them is represented by the rotation matrix R fp , translation matrix t fp It is expressed as: M p =R fp *M f +t fp Among them, M p is the probe center coordinate system O p The three-dimensional coordinates of a certain point, M f is the probe center coordinate system O f The three-dimensional coordinates of a certain point; Step 3: Use the handheld target (2) to measure the surface to be measured. During the measurement, the target probe contacts the surface of the position to be measured, and the binocular camera (1) is used to capture the image of the target surface features to obtain an image; Step 4: Extract features from the collected target surface image information, that is, demodulate the color fringe image of a single surface feature to obtain the phase feature, construct a circular / elliptical cone of the feature area according to the "phase-height model", use an iterative algorithm to accurately extract the sub-pixel coordinates of the circular / elliptical cone vertex, and establish the target surface feature coordinate system according to the sub-image plane coordinates corresponding to each circular / elliptical cone vertex in the array. f , the system internal parameter matrix A calibrated in step 1 i , external parameter matrix [R i-1 |t i-1 ] and the rotation matrix R of the probe calibration in step 2 fp , translation matrix t fp Calculate the spatial coordinates of the probe center position; Step 5: Based on the spatial coordinates of the probe center position in step 4 and the target surface feature coordinate system O established in step 2 f With the probe center coordinate system O p The relative position relationship between them can be used to calculate the spatial coordinates of the probe during measurement, and ultimately achieve the spatial positioning of the position to be measured.

2. The handheld target visual coordinate measurement method based on phase feature according to claim 1 is characterized in that: The step 4 is specifically as follows: (4.1) For the collected target surface image, the feature points of its color image are in the order from inside to outside along the radial direction: Red, I r =A+Bcos[2πr(u,v) / T-2π / 3] Green, I g =A+Bcos[2πr(u,v) / T] Blue, I b =A+Bcos[2πr(u,v) / T+2π / 3] Coloring, I rgb = (I r , I g , I b ) Among them, I r ,I g ,I b Represents color image I rgb The phase difference of the three components of the feature point is 2π / 3. (4.2) In a color image, the green pixel size is P G (u g ,v g ), blue shows pixel size is P B (u b ,v b ), red shows pixel size P R (u r ,v r ), and there is: in r =in b =2u g v r =v b =v g =3u g The size of a single color pixel block is P RGB (u rgb ,v rgb )for: u rgb =6u g v rgb =v r =v b =v g =3u g u rgb / v rgb =2:1 Set the period of a single feature point to k, 1.5≤k≤3, so the lattice parameter of a single feature point, that is, the physical size m*n is: m=P*k*u rgb n=P*2k*v rgb ; Among them, P>2 8 pixels; P is the number of pixels of a periodic stripe width; Calculate the interval range of the feature point array quantity parameter M*N by using the target surface area of ​​the handheld target (2) and the display size m*n of a single feature point; (4.3) In the range of M*N in step (4.2), the color grating feature points are captured by the binocular camera (1), and the color image I is converted into rgb Information separation, resulting in three color images I r ,I g ,I b , based on step (4.1), since the phase difference of the feature point units of the three color images is 2π / 3, calculate their wrapped phase and unfolded phase: The wrapping phase is: The unwrapped phase is: (4.4) Based on the "phase-height" model, the surface shape is reconstructed by unfolding the phase of the feature points of the color image in step (4.3), and the reconstructed surface shape is set to a circular / elliptical cone. Therefore, the set cone height h and the physical size radius R of the feature point have the following relationship: h=R Since the "phase-height" model is a linear model, the proportionality coefficient q is: q=P*u rgb / 2π Where P is the number of pixels of a periodic stripe width, u rgb is a color pixel block P RGB Horizontal physical dimensions; (4.5) Expand the reconstructed surface to obtain a circular / elliptical conical surface; (4.6) Since there is overlap between the 2π / 3 phase differences of the color feature points, the sub-pixel center coordinates of the feature point cone angle are obtained by iterative calculation.

3. The handheld target visual coordinate measurement method based on phase feature according to claim 2 is characterized in that: The iterative calculation of step (4.6) comprises the following steps: (4.6.1) Set the iteration threshold ε; (4.6.2) Randomly extract the coordinates of color feature points that are not less than 50% of the number of pixels occupied, and fit the circular / elliptical cone surface: According to the fitting results, calculate the average distance ε' between the coordinates of the unextracted points and the plane passing through the center axis of the cone and the generatrix where they are located; (4.6.3) Determine the size of ε' and ε. If ε'<ε, randomly select data points that are not less than 10% of the total number of color feature points from the unselected data points, fit the circular / elliptical cone, and calculate ε' to determine the size of ε' and ε; if ε'<ε is still satisfied after extracting 100% of the data points, adjust ε to ε / (2-3) and calculate from step (4.6.1); If ε'≥ε, randomly select the coordinates of a number of points not less than 50% of the number of pixels occupied by the color feature points, and fit the circular / elliptical cone surface until the ε' of the sample fitted circular / elliptical cone surface is less than ε; (4.6.4) Based on the circular / elliptical cone surface determined by the data points that finally meet the threshold requirements, extract the XOY plane sub-pixel coordinates (x 1,1 ,y 1,1 ); (4.6.5) Calculate the number of all feature point array parameters M*N of the color feature point cone vertex sub-pixel coordinates (x m,n ,y m,n ) to realize the extraction of target surface feature points.

4. A handheld target visual coordinate measurement system based on phase characteristics, characterized in that: Installed in a computer (3), it includes a processing instruction module for running each step of the phase feature-based handheld target visual coordinate method described in any one of claims 1 to 3.

5. A computer-readable storage medium, characterized in that: A calculation program for the handheld target visual coordinate method based on phase features as described in any one of claims 1 to 3 is stored, and the computer (3) can implement the handheld target visual coordinate method based on phase features when executing the calculation program.

Images