Self-adaptive detection device and detection method for highlight blade

By using the combination of adaptive exposure and GWO-ICP algorithm in the detection device, the problem of incomplete blade detection images in the high-light environment is solved, and high-precision blade morphology detection is achieved, ensuring the accuracy of measurement.

CN118936355APending Publication Date: 2024-11-12HARBIN UNIV OF SCI & TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410983862.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing detection devices are difficult to ensure the integrity of the blade morphology detection images in a high-light environment, resulting in low measurement accuracy.

Method used

Adaptive detection device of high-light blades, including No. 1 and No. 2 industrial camera units connected by fixed angles through mechanical structures, three motors are used to adaptively adjust the exposure parameters, and point cloud fusion splicing is performed in combination with GWO-ICP algorithm to achieve Three-dimensional reconstruction.

Benefits of technology

Effectively suppress specular reflection, improve image quality and measurement accuracy, ensure the integrity of the morphological detection image of the blade in a high-light environment, and improve the accuracy of the detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118936355A_ABST
    Figure CN118936355A_ABST
Patent Text Reader

Abstract

The invention discloses a self-adaptive detection device and method for a highlight blade, and relates to the field of mechanical structure light detection. The objective of the invention is to solve the problem that an existing detection device is incomplete in morphology detection image of a blade in a highlight environment. A high-quality image is obtained through adaptive exposure adjustment of two cameras based on a gray value, a high-quality fringe pattern is obtained through an improved multi-exposure fusion method, and then phase calculation is performed according to the obtained fringe pattern; a point cloud resolved by one camera and a point cloud resolved by another camera are fused and spliced through an algorithm combining an improved grey wolf algorithm and a nearest point iterative algorithm, the shape and position precision of the point cloud are both higher than those of a single monocular point cloud, and the precision of the final point cloud is higher than that of a single camera; the dynamic error caused by the movement of the mechanical arm is reduced, meanwhile, the error caused by manual adjustment is avoided, the completeness of the morphology detection image of the blade in the highlight environment is ensured, and the measurement accuracy is improved. The method is suitable for self-adaptive detection of the blade.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of mechanical structured light detection, and in particular to an adaptive detection device and a detection method for a high-light blade. Background Art

[0002] As the core component of the engine, the blade has a complex manufacturing process and high technological content. There are many problems encountered in the research and development process. It can be said that the performance of the engine depends to a large extent on the design and manufacturing level of the blade profile. If the shape and size of the blade cannot be guaranteed, it is very dangerous when the engine is running at high speed. Therefore, it is also very important to detect the profile and geometric dimensions of the blade. However, in the monocular structured light measurement of small and medium-sized metal blades, due to the influence of ambient light, projection stripes and shooting angle, it is very easy to cause the problem of high reflectivity of the aeroengine blades, resulting in incomplete captured images, which makes the point cloud reconstruction incomplete and fails to achieve measurement accuracy.

[0003] The main existing methods to solve the problem of high-reflection detection are: First, add a polarizer; by using a polarizer to adjust the polarized light beam, the mirror reflection is filtered out, but the overall contrast is weakened and the adjustment cannot be adaptive. Second, image processing; after the image is taken, the image processing method is used to repair the overexposed area. This method has high requirements for the image processing algorithm and requires good robustness, so that the overexposed area can be properly processed. Third, multiple exposures; by adjusting the exposure of the camera or projector, take multiple stripe images at different exposures, and then use image fusion to obtain a higher quality stripe image and increase the detection accuracy. With the in-depth study of high-reflective object detection, multi-exposure has achieved good success in high dynamic range detection. Therefore, the good results achieved by multi-exposure technology can be used to detect the high-reflection problem of small and medium-sized aircraft engine blades.

[0004] In summary, the existing detection device solves the problem of incomplete image detection of leaf shape in a high light environment. Summary of the invention

[0005] In order to solve the problem that the existing detection device has incomplete image detection of blade shape in a high-light environment, the present invention proposes an adaptive detection device and a detection method for high-light blades.

[0006] The present invention provides an adaptive detection device for high-gloss blades, which comprises a projector 1, a first industrial camera unit 11, a housing 12 and a second industrial camera unit 13;

[0007] The first industrial camera unit 11, the projector 1 and the second industrial camera unit 13 are connected at a fixed angle through the housing 12 to capture fringe information. The middle of the back of the housing 12 is connected to the output end of the robot arm, and the first industrial camera unit 11 and the second industrial camera unit 13 have the same mechanical structure;

[0008] Further, the industrial camera unit No. 1 11 includes a No. 1 external meshing gear 2, a No. 1 control component 3, a No. 1 motor 4, a No. 2 external meshing gear 5, a No. 2 control component 6, a No. 2 motor 7, a No. 3 external meshing gear 8, a No. 3 control component 9, a No. 3 motor 10 and a camera 15;

[0009] The third motor 10, the second motor 7 and the first motor 4 are arranged in sequence from left to right. The aperture adjustment part of the camera 15 is sleeved with a first control part 3, and the output end of the first motor 4 is provided with a first external meshing gear 2, and the first external meshing gear 2 is meshed and connected with the first control part 3;

[0010] The shutter time adjustment part of the camera 15 is sleeved with a second control member 6, and the output end of the second motor 7 is provided with a second external meshing gear 5, and the second external meshing gear 5 is meshedly connected with the second control member 6;

[0011] The gain adjustment part of the camera 15 is sleeved with a No. 3 control member 9, and the output end of the No. 3 motor 10 is provided with a No. 3 external meshing gear 8, and the No. 3 external meshing gear 8 is meshedly connected with the No. 3 control member 9;

[0012] Furthermore, the first industrial camera unit 11 and the second industrial camera unit 13 are respectively fixed on the inner wall of the housing 12 through the camera holder 14;

[0013] Furthermore, through holes are respectively processed at both ends and the middle of the front side of the shell 12, and the through hole at one end of the front side of the shell 12 is embedded with the camera 15 in the No. 1 industrial camera unit 11, and the through hole at the other end of the front side of the shell 12 is embedded with the camera 15 in the No. 2 industrial camera unit 13, and the through hole in the middle of the front side of the shell 12 is arranged corresponding to the output end of the projector 1;

[0014] Furthermore, the specific method of the adaptive detection method of a high-gloss blade described in the present invention is as follows:

[0015] Step 1: Adjust the shooting posture of the device and calibrate the system;

[0016] Step 2: determine the automatic exposure camera adjustment parameters according to the average gray value of the leaf and set the parameters through the three motors in the first industrial camera unit 11 and the second industrial camera unit 13 to capture the image of the leaf to be inspected;

[0017] Step 3, performing phase calculation on the fringes in the improved image by using the phase shift method;

[0018] Step 4: Use the improved GWO-ICP algorithm to dephase the two phase images to obtain a point cloud image and fuse and stitch them;

[0019] Step 5: 3D reconstruction;

[0020] Furthermore, the specific process of setting the parameters in step 2 is as follows: according to the obtained grayscale image of the aircraft engine blade, a functional mapping relationship between the pixel grayscale value and the camera exposure is established, and the optimal exposure is found by the grayscale value of the image; the average grayscale value of the blade image is obtained and calculated as follows:

[0021]

[0022] Wherein, B is the average gray value of the image, D is the gray value of the pixel, and n is the number of pixels; the gray range of the optimal exposure interval is mainly determined by the reflectivity of the surface of the object to be measured. In this embodiment, an aluminum alloy aeroengine blade is used, and the optimal exposure interval of the aluminum alloy is 75-86, so the optimal exposure gray value interval is set to 70-90;

[0023] Furthermore, in step 2, it is considered that the same exposure value can be obtained by combining different shutter speeds and aperture sizes, so the relationship between the image grayscale and the exposure is:

[0024] B=k×L×G×t×(F) -2 (2)

[0025] In the formula, k is a constant reflecting the lighting conditions of the scene, L represents the brightness of the surrounding environment, G and t represent the gain and exposure time of the camera respectively, F is the aperture number of the camera lens, and B is the grayscale of the current image. For a certain scene, when the time interval between the previous and next frames is relatively short, L and G remain unchanged, and the exposure time is fixed at 0.001s.

[0026] First, use aperture priority automatic exposure, and adjust the aperture to the ideal grayscale of 80 according to the average grayscale value. Determine whether the image grayscale value is in the optimal exposure range of 70 to 90. If not, adjust the shutter time for fine tuning. The calculation is as follows:

[0027]

[0028] Where B0 is the ideal grayscale, B1 is the image grayscale after the aperture is roughly adjusted, t1 is the exposure time, and t2 is the exposure time of the next frame;

[0029] After adjusting the exposure time to t2, calculate the current image grayscale B2. If B2 is within the optimal exposure range, the adaptive exposure adjustment is completed. If it is outside the optimal exposure range, repeat the previous step until the collected image grayscale is within the optimal exposure range.

[0030] Furthermore, in step 3, the multi-exposure fusion image is improved, and multiple images under optimal exposure are obtained by adaptive exposure control and threshold preprocessing is performed, and a mask image is produced. The projector projects an image onto the object using the optimal exposure obtained by the adaptive exposure method, and the images taken by the camera are sequentially stored in the sequence Is. Similarly, four grating fringe images with a phase shift of π / 2 are collected according to the optimal exposure, and the captured images are stored in the sequence J. s The specific operation is to normalize the grayscale image, select a suitable threshold, and perform the following operations on the sequence Is and sequence J. s The images in the sequence Is are preprocessed by threshold. In this implementation, the threshold is 250. When the grayscale value of the image pixel is less than 250, it is retained and stored in the mask image sequence. Otherwise, it is removed. The images in the sequence Is are made into mask images, and the mask image sequence is defined as M. S , the calculation formula is as follows:

[0031]

[0032] In the formula, is the mask image sequence M S The gray value of the i-th image at the coordinate (x, y); D i Sequence I s The gray value of the i-th image at (x, y);

[0033] The mask image sequence M S With the fringe diagram sequence J s Multiply them one by one and define G s is the synthetic fringe pattern sequence, we can get:

[0034]

[0035] In the formula, is the synthetic fringe pattern sequence G s The gray value of the i-th image at the coordinate (x, y); is the fringe pattern sequence J s The gray value of the i-th image at the coordinate (x, y);

[0036] I represents the fused raster image, and the gray value I(x,y) at the coordinate point (x,y) is calculated as follows:

[0037]

[0038] The GLAHE algorithm is used on the fused image I to improve the contrast and clarity of the fringe pattern;

[0039] Furthermore, the GWO-ICP algorithm of step 4 comprises the following steps:

[0040] Step 41: Given the point cloud obtained by the first industrial camera as the reference point cloud P = {pi, i = 1, 2, ...}, the point cloud obtained by the second industrial camera is Q = {qj, j = 1, 2, ...};

[0041] Step 42: Use GWO algorithm for rough registration;

[0042] Step 43: According to the initial position provided by the GWO algorithm, the point sets P and Q1 are finely aligned by ICP;

[0043] Furthermore, the specific steps of performing rough registration using the GWO algorithm in step 42 are as follows:

[0044] Step 421: Initialize R of GWO algorithm x , R y , R z , T x , T y , T z , μ and other parameters, among which R x , R y , R z is the element of the rotation matrix R, T x , T y , T z is the element of the translation matrix T, μ is the scaling parameter, the initial rotation matrix R0=I, I is the unit matrix, and the translation matrix T=

[000] T ;

[0045] Step 422: Calculate a new transformation point set Q1=R·Q+T;

[0046] Step 423: Get fitness through objective function:

[0047]

[0048] Where N is the number of paired point clouds; p xi 、p yi and p zi They are the X, Y, and Z coordinates of the point cloud obtained by the first industrial camera; p xj 、p yj and p zj They are the X, Y, and Z coordinates of the point cloud obtained by the second industrial camera;

[0049] Step 424: Update the position of the gray wolf according to the standard gray wolf algorithm update formula, and calculate the fitness value of the gray wolf after the update as f s ; Gray wolf position update formula:

[0050] D=|C·X P (t)-X(t)| (8)

[0051] X(t+1)=X P (t)-A·D (9)

[0052] Where D represents the current distance between the wolf and its prey; t is the current iteration number; A and C are coefficient vectors; X p (t) is the position vector of the prey; X(t) is the current position vector of the gray wolf; X(t+1) is the update of the gray wolf's position;

[0053] Formula for coefficient vectors C and A:

[0054] C=2r1 (10)

[0055] A=2a·r2-a (11)

[0056] Among them, r1 and r2 are random numbers between [0, 1]; α is the convergence factor, which is a key parameter to balance the exploration and development capabilities of GWO. During the entire iteration process, α decreases from 2 to 0;

[0057] Step 425: Determine whether the current number of iterations t=t+1 satisfies the iteration termination condition f s <ε, if not satisfied, return to step 421, if satisfied, use the new point cloud dataset Q1 obtained after registration as the initial position of ICP fine registration;

[0058] Furthermore, in step 43, the specific steps of performing ICP fine registration of point sets P and Q1 according to the initial position provided by the GWO algorithm are as follows:

[0059] Step 431: Initialize the initial position of the point cloud P0=p, Q1=[1000000] T , the number of iterations k = 0;

[0060] Step 432: Search for the nearest neighbor point set Y k =c(P k , Q);

[0061] Step 433, obtain the space coordinate transformation matrix R and T and the error d;

[0062] Step 434: Update the data point set P using the transformation matrices R and T k+1 =R K ·P K +T k ;

[0063] Step 4, 3, 5: The difference between two adjacent iteration errors is less than the threshold Right now The iteration ends. If it is not satisfied, return to step 433;

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] The present invention overcomes the shortcomings of the prior art and utilizes three motors in the No. 1 industrial camera unit and the No. 2 industrial camera unit to realize adaptive adjustment of the camera exposure parameters. When the robot arm is moving, the camera and the projector can always maintain a better shooting angle, so that the device has a better effect on highlight detection; the projection structured light angle and the camera observation angle are always controlled within [0°, 15°], which can effectively suppress the mirror reflection effect;

[0066] The present invention obtains high-quality images through adaptive exposure adjustment of two cameras based on grayscale values, obtains high-quality fringe patterns through an improved multi-exposure fusion method, and then performs phase solution based on the obtained fringe patterns; the point cloud accuracy obtained by existing monocular cameras is relatively poor, and the present invention fuses and splices the point cloud solved by one camera with the point cloud solved by another camera through an algorithm combining an improved grey wolf algorithm with a nearest point iteration algorithm, and the shape and position accuracy are higher than those of a single monocular point cloud, and the accuracy of the final point cloud is also higher than that of a single camera; the equipment of the present invention is simple, low in cost, and easy to operate, and the two cameras are integrated with a projector, which reduces the dynamic error caused by the movement of the mechanical arm and avoids the error of manual adjustment, thereby ensuring the integrity of the blade morphology detection image in a high-light environment and improving the accuracy of the measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a schematic diagram of the three-dimensional structure of an adaptive detection device for high-gloss blades according to the present invention in a use state;

[0068] Figure 2 It is a schematic diagram of the three-dimensional structure of an adaptive detection device for high-gloss blades according to the present invention;

[0069] Figure 3 It is a front view of an industrial camera unit of an adaptive detection device for high-gloss blades according to the present invention;

[0070] Figure 4 It is a flow chart of an adaptive detection method of a high-gloss blade according to the present invention;

[0071] Figure 5 It is a flow chart of adaptively adjusting exposure in an adaptive detection method for highlight blades described in the present invention. DETAILED DESCRIPTION

[0072] Specific implementation method 1: Combination Figures 1 to 3The present embodiment is described. The present embodiment describes an adaptive detection device for high-gloss blades, which comprises a projector 1, a first industrial camera unit 11, a housing 12 and a second industrial camera unit 13;

[0073] Industrial camera unit No. 1 11, projector 1 and industrial camera unit No. 2 13 are connected at a fixed angle through a shell 12 for capturing fringe information. The middle of the back side of the shell 12 is connected to the output end of the robot arm, and the mechanical structures of industrial camera unit No. 1 11 and industrial camera unit No. 2 13 are the same.

[0074] Specific implementation method 2: Combination Figures 1 to 3 This embodiment is described. This embodiment is a further limitation of the detection device described in the specific embodiment 1. In this embodiment, an adaptive detection device for high-gloss blades is described. The first industrial camera unit 11 includes a first external meshing gear 2, a first control component 3, a first motor 4, a second external meshing gear 5, a second control component 6, a second motor 7, a third external meshing gear 8, a third control component 9, a third motor 10 and a camera 15.

[0075] The third motor 10, the second motor 7 and the first motor 4 are arranged in sequence from left to right. The aperture adjustment part of the camera 15 is sleeved with a first control part 3, and the output end of the first motor 4 is provided with a first external meshing gear 2, and the first external meshing gear 2 is meshed and connected with the first control part 3;

[0076] The shutter time adjustment part of the camera 15 is sleeved with a second control member 6, and the output end of the second motor 7 is provided with a second external meshing gear 5, and the second external meshing gear 5 is meshedly connected with the second control member 6;

[0077] A third control component 9 is sleeved on the gain adjustment portion of the camera 15 , and a third external meshing gear 8 is provided on the output end of the third motor 10 , and the third external meshing gear 8 is meshedly connected with the third control component 9 .

[0078] Specific implementation method three: Combination Figures 1 to 3 To explain this embodiment, this embodiment is a further limitation of the detection device described in the specific embodiment 1. This embodiment describes an adaptive detection device for high-gloss blades, and the industrial camera unit No. 1 11 and the industrial camera unit No. 2 13 are respectively fixed on the inner wall of the shell 12 through the camera holder 14.

[0079] Specific implementation method four: Combination Figures 1 to 3The present embodiment is described as a further limitation of the detection device described in the first specific embodiment. The present embodiment describes an adaptive detection device for high-gloss blades, wherein the two ends and the middle of the front side of the shell 12 are respectively processed with through holes, and the through hole at one end of the front side of the shell 12 is embedded with the camera 15 in the No. 1 industrial camera unit 11, and the through hole at the other end of the front side of the shell 12 is embedded with the camera 15 in the No. 2 industrial camera unit 13, and the through hole in the middle of the front side of the shell 12 is arranged corresponding to the output end of the projector 1.

[0080] Specific implementation method five: Combination Figures 1 to 5 The present embodiment describes an adaptive detection method for high-gloss blades, and the specific method is as follows:

[0081] Step 1: Adjust the shooting posture of the device and calibrate the system;

[0082] Step 2: determine the automatic exposure camera adjustment parameters according to the average gray value of the leaf and set the parameters through the three motors in the first industrial camera unit 11 and the second industrial camera unit 13 to capture the image of the leaf to be inspected;

[0083] Step 3, performing phase calculation on the fringes in the improved image by using the phase shift method;

[0084] Step 4: Use the improved GWO-ICP algorithm to dephase the two phase images to obtain a point cloud image and fuse and stitch them;

[0085] Step 5: 3D reconstruction.

[0086] Specific implementation method six: Combination Figures 1 to 5 This embodiment is described. This embodiment is a further limitation of the detection method described in the specific embodiment 5. The specific process of setting parameters in the step 2 of the adaptive detection method of a high-gloss blade described in this embodiment is as follows: according to the obtained grayscale image of the aircraft engine blade, a function mapping relationship between the pixel grayscale value and the camera exposure is established, and the optimal exposure is found by the grayscale value of the image; the average grayscale value of the blade image is obtained, and the calculation is as follows:

[0087]

[0088] Wherein, B is the average gray value of the image, D is the gray value of the pixel, and n is the number of pixels; the gray range of the optimal exposure interval is mainly determined by the reflectivity of the surface of the object to be measured. In this embodiment, an aluminum alloy aero-engine blade is used, and the optimal exposure interval of the aluminum alloy is 75-86, so the optimal exposure gray value interval is set to 70-90.

[0089] Specific implementation method seven: Combination Figures 1 to 5This embodiment is described. This embodiment is a further limitation of the detection method described in the sixth embodiment. In the adaptive detection method for highlight blades described in this embodiment, in step 2, it is considered that the same exposure value can be obtained by a combination of different shutter speeds and aperture sizes, so the relationship between the image grayscale and the exposure is:

[0090] B=k×L×G×t×(F) -2 (2)

[0091] In the formula, k is a constant reflecting the lighting conditions of the scene, L represents the brightness of the surrounding environment, G and t represent the gain and exposure time of the camera respectively, F is the aperture number of the camera lens, and B is the grayscale of the current image. For a certain scene, when the time interval between the previous and next frames is relatively short, L and G remain unchanged, and the exposure time is fixed at 0.001s.

[0092] First, use aperture priority automatic exposure, and adjust the aperture to the ideal grayscale of 80 according to the average grayscale value. Determine whether the image grayscale value is in the optimal exposure range of 70 to 90. If not, adjust the shutter time for fine tuning. The calculation is as follows:

[0093]

[0094] Where B0 is the ideal grayscale, B1 is the image grayscale after the aperture is roughly adjusted, t1 is the exposure time, and t2 is the exposure time of the next frame;

[0095] After adjusting the exposure time to t2, calculate the current image grayscale B2. If B2 is within the optimal exposure range, the adaptive exposure adjustment is completed. If it is outside the optimal exposure range, repeat the previous step until the collected image grayscale is within the optimal exposure range.

[0096] Specific implementation method eight: Combination Figures 1 to 5 This embodiment is described. This embodiment is a further limitation of the detection method described in the specific embodiment 5. In the adaptive detection method of a high-gloss blade described in this embodiment, the multi-exposure fusion image is improved in the step 3, and multiple images under the optimal exposure are obtained by adaptive exposure control and threshold preprocessing is performed, and a mask image is made. The optimal exposure obtained by the adaptive exposure method is used, and the projector projects an image onto the object, and the images taken by the camera are sequentially stored in the sequence Is. Similarly, 4 grating stripe images with a phase shift of π / 2 are collected according to the optimal exposure, and the captured images are stored in the sequence J. s The specific operation is to normalize the grayscale image, select a suitable threshold, and perform the following operations on the sequence Is and sequence J. sThe images in the sequence Is are preprocessed by threshold. In this implementation, the threshold is 250. When the grayscale value of the image pixel is less than 250, it is retained and stored in the mask image sequence. Otherwise, it is removed. The images in the sequence Is are made into mask images, and the mask image sequence is defined as M. S , the calculation formula is as follows:

[0097]

[0098] In the formula, is the mask image sequence M S The gray value of the i-th image at the coordinate (x, y); D i Sequence I s The gray value of the i-th image at (x, y);

[0099] The mask image sequence M S With the fringe diagram sequence J s Multiply them one by one and define G s is the synthetic fringe pattern sequence, we can get:

[0100]

[0101] In the formula, is the synthetic fringe pattern sequence G s The gray value of the i-th image at the coordinate (x, y); is the fringe pattern sequence J s The gray value of the i-th image at the coordinate (x, y);

[0102] I represents the fused raster image, and the gray value I(x,y) at the coordinate point (x,y) is calculated as follows:

[0103]

[0104] The GLAHE algorithm is used on the fused image I to improve the contrast and clarity of the fringe pattern;

[0105] In this specific implementation, the fused images obtained by two industrial cameras are phase resolved. This implementation adopts a four-step phase shift method to solve the main phase value. Four fringe images need to be projected, and the grayscale distribution of each fringe image is:

[0106]

[0107] The phase difference between two adjacent grating fringe images in the four-step phase shift is π / 2. The phase principal value extraction formula at this time is:

[0108]

[0109] Through the above method, high-precision three-dimensional point cloud data of highly reflective aircraft engine blades can be obtained, and then the point clouds obtained by the two cameras are fused and spliced, and the point clouds obtained from the two perspectives are gathered into a unified coordinate system to form a complete data point cloud.

[0110] As an implementable method, the specific method is to extract the neighborhood geometric features between the point clouds solved by the two cameras, quickly determine the corresponding relationship between the point pairs between the two through the geometric features, and then calculate this relationship to obtain the transformation matrix.

[0111] The edge features are extracted from the point cloud of the first industrial camera, and the edge features are also extracted from the point cloud of the second industrial camera, and the rotation matrix and translation matrix of the corresponding points are calculated. The point clouds of the first industrial camera and the second industrial camera are spliced ​​according to the corresponding rigid body transformation matrix.

[0112] The closest point iteration algorithm has the advantages of simple operation, strong robustness, and high registration accuracy, but it has strict requirements on the initial position and is prone to fall into local optimality, resulting in registration failure. At the same time, too many iterations increase the time complexity and reduce the search efficiency. Therefore, the present invention proposes an algorithm (GWO-ICP) that combines the gray wolf algorithm and the closest point iteration algorithm to give full play to the respective advantages of the GWO and ICP algorithms. By combining the gray wolf algorithm with the ICP algorithm, the convergence speed and registration accuracy of the ICP algorithm can be improved. The gray wolf algorithm can further optimize the rigid transformation matrix by searching a wider solution space to better align the source point cloud and the target point cloud.

[0113] Specific implementation method nine: Combination Figures 1 to 5 This embodiment is described. This embodiment is a further limitation of the detection method described in the fifth embodiment. The adaptive detection method for high-gloss leaves described in this embodiment, the GWO-ICP algorithm in step 4, includes the following steps:

[0114] Step 41: Given the point cloud obtained by the first industrial camera as the reference point cloud P = {pi, i = 1, 2, ...}, the point cloud obtained by the second industrial camera is Q = {qj, j = 1, 2, ...};

[0115] Step 42: Use GWO algorithm for rough registration;

[0116] Step 43: According to the initial position provided by the GWO algorithm, the point sets P and Q1 are finely aligned by ICP;

[0117] Furthermore, the specific steps of performing rough registration using the GWO algorithm in step 42 are as follows:

[0118] Step 421: Initialize R of GWO algorithm x , Ry , R z , T x , T y , T z , μ and other parameters, among which R x , R y , R z is the element of the rotation matrix R, T x , T y , T z is the element of the translation matrix T, μ is the scaling parameter, the initial rotation matrix R0=I, I is the unit matrix, and the translation matrix T=

[000] T ;

[0119] Step 422: Calculate a new transformation point set Q1=R·Q+T;

[0120] Step 423: Get fitness through objective function:

[0121]

[0122] Where N is the number of paired point clouds; p xi 、p yi and p zi They are the X, Y, and Z coordinates of the point cloud obtained by the first industrial camera; p xj 、p yj and p zj They are the X, Y, and Z coordinates of the point cloud obtained by the second industrial camera;

[0123] Step 424: Update the position of the gray wolf according to the standard gray wolf algorithm update formula, and calculate the fitness value of the gray wolf after the update as f s ; Gray wolf position update formula:

[0124] D=|C·X P (t)-X(t)| (8)

[0125] X(t+1)=X P (t)-A·D (9)

[0126] Where D represents the current distance between the wolf and its prey; t is the current iteration number; A and C are coefficient vectors; X p (t) is the position vector of the prey; X(t) is the current position vector of the gray wolf; X(t+1) is the update of the gray wolf's position;

[0127] Formula for coefficient vectors C and A:

[0128] C=2r1 (10)

[0129] A=2a·r2-a (11)

[0130] Among them, r1 and r2 are random numbers between [0, 1]; α is the convergence factor, which is a key parameter to balance the exploration and development capabilities of GWO. During the entire iteration process, α decreases from 2 to 0;

[0131] Step 425: Determine whether the current number of iterations t=t+1 satisfies the iteration termination condition f s <ε, if not satisfied, return to step 421, if satisfied, the new point cloud dataset Q1 obtained after registration is used as the initial position of ICP fine registration.

[0132] Specific implementation method ten: Combination Figures 1 to 5 This embodiment is described. This embodiment is a further limitation of the detection method described in the specific embodiment 5. In the adaptive detection method of a highlight blade described in this embodiment, the specific steps of performing ICP fine registration of the point sets P and Q1 according to the initial position provided by the GWO algorithm in step 43 are as follows:

[0133] Step 431: Initialize the initial position of the point cloud P0=p, Q1=[1000000] T , the number of iterations k = 0;

[0134] Step 432: Search for the nearest neighbor point set Y k =c(P k , Q);

[0135] Step 433, obtain the space coordinate transformation matrix R and T and the error d;

[0136] Step 434: Update the data point set P using the transformation matrices R and T k+1 =R K ·P K +T k ;

[0137] Step 4, 3, 5: The difference between two adjacent iteration errors is less than the threshold Right now The iteration ends. If it is not satisfied, return to step 433.

[0138] Specific implementation method eleven: Combination Figures 1 to 5 This embodiment is described. This embodiment further limits the detection method described in the specific embodiment five. This embodiment describes an adaptive detection method for high-gloss blades. The three-dimensional reconstruction in step five also provides a computer device, including: a memory for storing a program for implementing the adaptive exposure and point cloud fusion of the blades; and a processor for executing the program stored in the memory.

Claims

1. An adaptive detection device for high-gloss blades, characterized in that: It comprises a projector (1), a first industrial camera unit (11), a housing (12) and a second industrial camera unit (13); The first industrial camera unit (11), the projector (1) and the second industrial camera unit (13) are connected at a fixed angle through a housing (12) for capturing fringe information. The middle of the back of the housing (12) is connected to the output end of the mechanical arm, and the first industrial camera unit (11) and the second industrial camera unit (13) have the same mechanical structure.

2. The self-adaptive detection device for high-gloss blades according to claim 1, characterized in that: The industrial camera unit No. 1 (11) comprises a No. 1 external meshing gear (2), a No. 1 control component (3), a No. 1 motor (4), a No. 2 external meshing gear (5), a No. 2 control component (6), a No. 2 motor (7), a No. 3 external meshing gear (8), a No. 3 control component (9), a No. 3 motor (10) and a camera (15); The third motor (10), the second motor (7) and the first motor (4) are arranged in sequence from left to right, the aperture adjustment part of the camera (15) is sleeved with a first control component (3), and the output end of the first motor (4) is provided with a first external meshing gear (2), and the first external meshing gear (2) is meshedly connected with the first control component (3); A second control member (6) is sleeved on the shutter time adjustment portion of the camera (15), and a second external meshing gear (5) is provided on the output end of the second motor (7), and the second external meshing gear (5) is meshedly connected with the second control member (6); A third control component (9) is sleeved on the gain adjustment portion of the camera (15), and a third external meshing gear (8) is provided on the output end of the third motor (10), and the third external meshing gear (8) is meshingly connected with the third control component (9).

3. The self-adaptive detection device for high-gloss blades according to claim 1, characterized in that: The first industrial camera unit (11) and the second industrial camera unit (13) are respectively fixed on the inner wall of the housing (12) via camera seats (14).

4. The self-adaptive detection device for high-gloss blades according to claim 1, characterized in that: The shell (12) is respectively provided with through holes at both ends and in the middle of the front side, and the through hole at one end of the front side of the shell (12) is embedded with a camera (15) in the No. 1 industrial camera unit (11), and the through hole at the other end of the front side of the shell (12) is embedded with a camera (15) in the No. 2 industrial camera unit (13), and the through hole in the middle of the front side of the shell (12) is arranged corresponding to the output end of the projector (1).

5. A detection method using the self-adaptive detection device for high-gloss blades according to any one of claims 1 to 4, characterized in that: The specific method is as follows: Step 1: Adjust the shooting posture of the device and calibrate the system; Step 2: determining the automatic exposure camera adjustment parameters according to the average gray value of the leaf and setting the parameters through three motors in the first industrial camera unit (11) and the second industrial camera unit (13) to capture an image of the leaf to be inspected; Step 3: Perform phase calculation on the fringes in the image by using the phase shift method; Step 4: Use the GWO-ICP algorithm to dephase the two phase images to obtain a point cloud image and fuse and stitch them; Step 5: 3D reconstruction.

6. The method for adaptively detecting a high-gloss blade according to claim 5, characterized in that: The specific process of setting parameters in step 2 is as follows: according to the grayscale image of the aircraft engine blade, a function mapping relationship between the pixel grayscale value and the camera exposure is established, and the optimal exposure is found by the grayscale value of the image; the average grayscale value of the blade image is obtained and calculated as follows: In the formula, B is the average gray value of the image, D is the gray value of the pixel, and n is the number of pixels; The grayscale range of the optimal exposure interval is determined by the reflective ability of the surface of the object to be tested. For aluminum alloy aircraft engine blades, the optimal exposure range of aluminum alloy is 75-86, so the optimal exposure grayscale value range is set to 70-90.

7. The method for adaptively detecting a high-gloss blade according to claim 6, characterized in that: In the step 2, it is considered that the same exposure value is obtained by combining different shutter speeds and aperture sizes, so the relationship between the image grayscale and the exposure is: B=k×L×G×t×(F) -2 (2) In the formula, k is a constant reflecting the lighting conditions of the scene, L represents the brightness of the surrounding environment, G and t represent the gain and exposure time of the camera respectively, F is the aperture number of the camera lens, and B is the grayscale of the current image. For a certain scene, when the time interval between the previous and next frames is short, L and G remain unchanged, and the exposure time is fixed at 0.001s. First, use aperture priority automatic exposure, and adjust the aperture to the ideal grayscale of 80 according to the average grayscale value. Determine whether the image grayscale value is in the optimal exposure range of 70 to 90. If not, adjust the shutter time for fine tuning. The calculation is as follows: Where B0 is the ideal grayscale, B1 is the image grayscale after the aperture is roughly adjusted, t1 is the exposure time, and t2 is the exposure time of the next frame; After adjusting the exposure time to t2, calculate the current image grayscale B2. If B2 is within the optimal exposure range, the adaptive exposure adjustment is completed. If it is outside the optimal exposure range, repeat the previous step until the collected image grayscale is within the optimal exposure range.

8. The method for adaptively detecting a high-gloss blade according to claim 5, characterized in that: In the step 3, the multi-exposure fusion image is improved, and multiple images under optimal exposure are obtained by adaptive exposure control and threshold preprocessing is performed, and a mask image is prepared; the projector projects an image onto the object according to the optimal exposure obtained by the adaptive exposure method, and the images taken by the camera are sequentially stored in the sequence Is, and 4 grating fringe images with a phase shift of π / 2 are also collected according to the optimal exposure, and the captured images are stored in the sequence J s The specific operation is to normalize the grayscale image, select a threshold, and perform the following operations on the sequence Is and sequence J. s The images in the sequence Is are preprocessed by threshold, and the threshold is selected as 250. When the gray value of the image pixel is less than 250, it is retained and stored in the mask image sequence, otherwise it is removed; the images in the sequence Is are made into mask images, and the mask image sequence is defined as M S , the calculation formula is as follows: In the formula, is the mask image sequence M S The gray value of the i-th image at the coordinate (x, y); D i Sequence I s The gray value of the i-th image at (x, y); The mask image sequence M S With the fringe diagram sequence J s Multiply them one by one and define G s is the synthetic fringe pattern sequence, we can get: In the formula, is the synthetic fringe pattern sequence G s The gray value of the i-th image at the coordinate (x, y); is the fringe pattern sequence J s The gray value of the i-th image at the coordinate (x, y); I represents the fused raster image, and the gray value I(x,y) at the coordinate point (x,y) is calculated as follows: The GLAHE algorithm is used on the fused image I.

9. The method for adaptively detecting a high-gloss blade according to claim 5, characterized in that: The GWO-ICP algorithm of step 4 comprises the following steps: Step 41: Given the point cloud obtained by the first industrial camera as the reference point cloud P = {pi, i = 1, 2, ...}, the point cloud obtained by the second industrial camera is Q = {qj, j = 1, 2, ...}; Step 42: Use GWO algorithm for rough registration; Step 43: According to the initial position provided by the GWO algorithm, the point sets P and Q1 are finely aligned by ICP; The specific steps of the step 42 using the GWO algorithm for rough registration are as follows: Step 421: Initialize R of GWO algorithm x , R y , R z , T x , T y , T z , μ parameter, where R x , R y , R z is the element of the rotation matrix R, T x , T y , T z is the element of the translation matrix T, μ is the scaling parameter, the initial rotation matrix R0=I, I is the unit matrix, and the translation matrix T=[000] T ; Step 422: Calculate a new transformation point set Q1=R·Q+T; Step 423: Get fitness through objective function: Where N is the number of paired point clouds; p xi 、p yi and p zi They are the X, Y, and Z coordinates of the point cloud obtained by the first industrial camera; p xj 、p yj and p zj They are the X, Y, and Z coordinates of the point cloud obtained by the second industrial camera; Step 424: Update the position of the gray wolf according to the standard gray wolf algorithm update formula, and calculate the fitness value of the gray wolf after the update as f s ; Gray wolf position update formula: D=|C·X P (t)-X(t)| (8) X(t+1)+X P (t)-A·D (9) Where D represents the current distance between the wolf and its prey; t is the current iteration number; A and C are coefficient vectors; X p (t) is the position vector of the prey; X(t) is the current position vector of the gray wolf; X(t+1) is the update of the gray wolf's position; Formula for coefficient vectors C and A: C=2r1 (10) A=2a·r2-a (11) Among them, r1 and r2 are random numbers between [0, 1]; α is the convergence factor, which is a key parameter to balance the exploration and development capabilities of GWO. During the entire iteration process, α decreases from 2 to 0; Step 425: Determine whether the current number of iterations t=t+1 satisfies the iteration termination condition f s <ε, if not satisfied, return to step 421, if satisfied, the new point cloud dataset Q1 obtained after registration is used as the initial position of ICP fine registration.

10. The method for adaptively detecting a high-gloss blade according to claim 5, characterized in that: The specific steps of performing ICP fine registration of point sets P and Q1 according to the initial position provided by the GWO algorithm in step 43 are as follows: Step 431: Initialize the initial position of the point cloud P0=p, Q1=[1000000] T , the number of iterations k = 0; Step 432: Search for the nearest neighbor point set Y k =c(P k ,Q); Step 433, obtain the space coordinate transformation matrix R and T and the error d; Step 434: Update the data point set P using the transformation matrices R and T k+1 =R K ·P K +T k ; Step 4, 3, 5: The difference between two adjacent iteration errors is less than the threshold Right now The iteration ends. If it is not satisfied, return to step 433.

Citation Information

Cited By

  • Glass container surface defect detection method based on visual inspection

    CN120142307A