A device and method for monitoring expansion of pressure-bearing components of a power station based on machine vision

Through the expansion monitoring device of the power station pressure-bearing component based on machine vision, combined with wireless graph transmission and EPnP weighted iterative algorithm, the problems of low boiler expansion monitoring accuracy and inconvenient data transmission are solved, and three-dimensional online monitoring and real-time data transmission of boiler expansion are realized. Improves the safety of boiler operation.

CN113959345BActive Publication Date: 2024-11-19ZHEJIANG ZHENENG TECHN RES INST CO LTD +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202111375051.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-11-17
Publication Date
2024-11-19
Estimated Expiration
2041-11-17

AI Technical Summary

Technical Problem

In the prior art, the boiler expansion monitoring accuracy is low and data transmission is inconvenient, resulting in safety risks of boiler operation, and the mechanical expansion indicator is difficult to meet the real-time data needs of smart power plants.

Method used

The expansion monitoring device for the pressure-bearing components of the power station is adopted based on machine vision. Video signals are collected through the camera device, combined with wireless image transmission and EPnP weighted iterative algorithm, online monitoring and remote transmission of three-dimensional expansion is realized, reducing the labor intensity of patrol personnel.

Benefits of technology

It realizes three-dimensional online monitoring of boiler expansion, improves measurement accuracy and data transmission convenience, provides real-time expansion information, and enhances the safety of unit operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113959345B_ABST
    Figure CN113959345B_ABST
Patent Text Reader

Abstract

The present invention relates to a device for monitoring the expansion of pressure-bearing components of a power station based on machine vision, comprising: a pressure-bearing component, an angle steel, a connector, and a camera device; wherein the fixed end of the angle steel is fixed on the pressure-bearing component, the fixed end of the angle steel is rigidly connected to the pressure-bearing component, and the three-dimensional expansion of the pressure-bearing component will be reflected as a three-dimensional displacement of the free end of the angle steel. The beneficial effects of the present invention are: the interior of the visual light source controllable device of the present invention is an LED with optimized distribution arrangement, high brightness, and a thin and light appearance; the visual light source controllable device is made of aerospace-grade aluminum alloy, which has strong corrosion resistance and tensile strength, so it can adapt to complex working environments; a surface light source and a sign board are also provided; the surface light source and the sign board are both installed within the field of view of the camera device, and the surface light source and the sign board are fixed in the same plane as a reference object, and the surface light source serves as the backlight of the sign board, thereby improving the adaptability of the visual measurement system to the environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of thermal power generation, relates to three-dimensional monitoring of expansion of pressure-bearing components, and in particular to a visual measurement device and method based on an EPnP weighted iterative algorithm. Background Art

[0002] Power plant boilers are the core equipment of thermal power plants. Boilers generally adopt a fully suspended structure, and each component of the pressure-bearing parts has sufficient expansion space when heated. The expansion amount and expansion direction of each point of the boiler are affected by many factors, such as unit load, burner organization, air distribution, load change rate, etc. Large boilers contain various complex pipe systems and steel structures. Due to the different materials of each component, the expansion value and direction of the boiler after heating are also different. Abnormal deformation of the heating surface can also be reflected by the expansion amount. The large-scale change of the load of the power plant boiler will cause the expansion characteristics of the boiler to become complicated. Once the expansion exceeds the allowable range, it will affect the normal operation of the boiler, and even cause major accidents such as damage and fracture of the boiler heating surface. It can be seen that real-time monitoring of the expansion amount of key points of the boiler and fault alarm are very important.

[0003] In recent years, with the expansion of the scale of new energy access, thermal power units have participated in deep peak regulation, and the load change rate has further increased; on the other hand, the State Grid has become more and more stringent in its AGC assessment indicators for thermal power generating units. These two reasons have led to a gradual increase in the load change rate of the unit, making the boiler expansion more significant and the expansion safety area further reduced. In order to ensure the reliability of boiler operation, it is necessary to obtain the expansion characteristics of the key points of the boiler more accurately and in real time. At present, large power station boilers are generally equipped with dozens of expansion indicators at key points for expansion monitoring, but most of the ones currently used are mechanical expansion indicators. Such devices have low measurement accuracy and large data deviation, and the expansion measurement of pressure-bearing components is inaccurate, resulting in hidden dangers in boiler operation safety. The indicators are installed on-site and scattered in multiple horizontal layers of the boiler. Mechanical expansion indicators can only be read on-site. Each boiler is equipped with dozens of expansion indicators scattered in various places of the boiler. There is no remote data transmission function. Inspection personnel must climb to various places of the boiler to complete readings and data recording, which is time-consuming and labor-intensive, and there are safety risks. In addition, on-site readings make it impossible for operating personnel to obtain the expansion changes of the boiler in the first place, resulting in the omission of parameter information, which does not meet the development trend of smart power plants at all. Inspection personnel generally pay close attention to the expansion volume only during the start-up and shutdown of the unit. During the normal operation of the unit, due to the difficulty of using the expansion indicator, it is generally impossible to conduct continuous expansion inspections. It can be seen that this mechanical indicator has low detection accuracy, is time-consuming and labor-intensive, and is not timely, and cannot provide sufficient data support for unit operation.

[0004] Patent CN106441105A and Patent CN204902781U proposed a boiler expansion monitoring method based on laser ranging. This method designed three reference plates and used three groups of lasers to measure the three-dimensional expansion of a point respectively. Patent CN1290833A used a multi-turn potentiometer to design a three-dimensional displacement indicator to vividly reflect the collected signal to the CRT in the central control room. The paper "Design of Boiler Expansion Detector Based on Micro-displacement Positioning Technology" designed a three-dimensional positioning mechanism based on an absolute encoder and a microcontroller for boiler expansion monitoring. The above solutions focus on remote monitoring of three-dimensional boiler expansion, but all have problems such as complex systems, high costs, and low applicability in boiler environments. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a device and method for monitoring the expansion of pressure-bearing components of a power station based on machine vision.

[0006] The device for monitoring the expansion of pressure-bearing components of a power station based on machine vision comprises: a pressure-bearing component, an angle steel, a connector, and a camera device; wherein the fixed end of the angle steel is fixed to the pressure-bearing component, and the fixed end of the angle steel is rigidly connected to the pressure-bearing component. The three-dimensional expansion of the pressure-bearing component will be reflected as a three-dimensional displacement of the free end of the angle steel, that is, the optical center O of the camera. c The three-dimensional relative displacement between the angle plate and the sign plate, the free end of the angle steel is equipped with a camera through a connector, and the camera is connected to the rigid body of the measured point.

[0007] Preferably, a surface light source and a sign board are further included; the surface light source and the sign board are both installed within the field of view of the camera device, and the surface light source and the sign board are fixed in the same plane as reference objects, and the surface light source serves as the backlight of the sign board, thereby improving the adaptability of the visual measurement system to the environment.

[0008] Preferably, the angle steel is fixed to a key position on the pressure-bearing component where expansion monitoring is to be performed.

[0009] Preferably, the imaging device is a video camera.

[0010] The working method of the expansion monitoring device for pressure-bearing components of a power station based on machine vision specifically comprises the following steps:

[0011] Step 1, the camera device collects video signals. The horizontal elevation of the expansion monitoring point of the power station boiler is relatively high. It is inconvenient to wire on the boiler steel frame and the construction cost is high. In order to facilitate system installation and wiring, the present invention uses wireless image transmission in the image acquisition and transmission system to wirelessly transmit video signals, and the video signals are transmitted to the processing host through wireless image transmission; the wireless transmission of frequency signals adopts 5.8G wireless image transmission system, and 1W power is selected to cooperate with mushroom antenna. The image transmission distance can reach 15km, achieving the effect of contactless measurement and no manual reading is required; in terms of visual light source, a controllable visual light source device is selected as the surface light source, combined with the feature that the camera exposure time can be manually set, so that the image features are more significant, which is conducive to subsequent image processing; Step 2, the processing program in the processing host processes the video signal, and solves the three-dimensional coordinates of the camera device in the world coordinate system with the world coordinate system of the fixed marker in the field of view of the camera device as a reference;

[0012] Step 3. Calculate the rotation matrix R and translation matrix T corresponding to each frame of the image. Solving R and T is the key to obtaining the expansion: when the corresponding relationship between the world coordinate system coordinates P and the camera coordinate system coordinates p of n pairs of points is known, the rotation matrix R and translation matrix T are solved by the PnP algorithm; the three-dimensional displacement of the measured point is obtained, and the three-dimensional expansion of the pressure-bearing component is obtained.

[0013] Preferably, step 2 specifically comprises the following steps:

[0014] Step 2.1: According to the pinhole imaging model of visual measurement, the camera coordinate system is expressed as c X c Y c Z c , the world coordinate system is represented as O w X w Y w Z w ; There is a rigid body transformation relationship between the camera coordinate system and the world coordinate system, that is, rotation change and translation change; Let the 3×3 matrix R be the rotation matrix of the world coordinate system relative to the camera coordinate system, and each row and column of the matrix R satisfy the orthogonal relationship; Let the 3×1 matrix T be the three-dimensional coordinates of the origin of the world coordinate system in the camera coordinate system, Pw(X w , Y w , Z w ) is any point in the world coordinate system, and its corresponding expression in the camera coordinate system is Pc(X c , Y c , Z c ); the coordinate relationship between Pw and Pc is expressed as:

[0015]

[0016] In the above formula, X c , Yc , Z c are any point Pw(X w , Y w , Z w ) is the three-dimensional coordinate in the camera coordinate system, X w , Y w , Z w For any point Pw(X w , Y w , Z w ) three-dimensional coordinates; the matrix R is the rotation matrix of the world coordinate system relative to the camera coordinate system, and the matrix T is the three-dimensional coordinate translation matrix of the origin of the world coordinate system in the camera coordinate system;

[0017] For any point Pw(X w , Y w , Z w ), given the camera coordinate system coordinates Pc(X c , Y c , Z c ), rotation matrix R and translation matrix T, the world coordinate system coordinates Pw(X w , Y w , Z w );

[0018] Step 2.2: Camera optical center O in camera coordinate system c The coordinates of the camera are (0,0,0), then the camera optical center O c The coordinates in the world coordinate system are expressed as:

[0019]

[0020] In the above formula, the matrix R is the rotation matrix of the world coordinate system relative to the camera coordinate system, and the matrix T is the three-dimensional coordinate of the origin of the world coordinate system in the camera coordinate system; the coordinate (X c-Oc , Y c-Oc , Z c-Oc ) is the camera optical center O c Three-dimensional coordinates in the camera coordinate system; coordinates (X w-Oc , Y w-Oc , Z w-Oc ) is the camera optical center O c The three-dimensional coordinates in the world coordinate system; O c The change in coordinates in each direction is the expansion of the key point;

[0021] Step 2.3: Calculate the camera optical center O by formula (2): cIn the three-dimensional coordinates of the world coordinate system, since the world coordinate system is a stationary reference, the scale of its coordinate axis has a real length meaning, and the camera and the object to be measured are connected by a rigid body, therefore, the optical center of the camera O can be used. c The three-dimensional coordinate change in the world coordinate system is used to represent the three-dimensional displacement of the measured point:

[0022]

[0023] In the above formula, D x , Dy, D z Respectively represent the displacement of the measured point in the three directions of x, y, and z axes of the world coordinate system, ΔX w-Oc , ΔY w-Oc , ΔZ w-Oc Represent the camera optical center O c The coordinate transformation amount in the three directions of x, y, and z axes of the world coordinate system.

[0024] Preferably, step 3 specifically comprises the following steps:

[0025] Step 3.1, PnP algorithm solves the transformation matrix R and translation matrix T. It needs to match the pixel coordinates of multiple pairs of points with the world coordinates. Select multiple circles (6) of different areas as fixed markers, and use the center of mass of the circle as the marker point. The world coordinates of each marker point are obtained by offline measurement, and the image of the sign board is captured in real time by a camera. Since the image will be distorted during shooting, and the center of gravity position will not change, the ellipse recognition algorithm is used. Based on geometric features, the pixel coordinates of the center of gravity of the circular fixed mark are obtained through edge detection and curve fitting, and the world coordinates and pixel coordinates are matched according to the pixel area size. The algorithm is compared with other PnP algorithms through simulation tests. The results show that the proposed algorithm has more advantages than other algorithms.

[0026] Step 3.2, reduce the coupling between landmarks: Linearly represent any landmark by the coordinates of four non-coplanar virtual control points, which can reduce the coupling between landmarks, and the world coordinates of the four virtual control points can be set arbitrarily. First, solve the coordinates of the four virtual control points in the camera coordinate system, and then solve the classic 3D-3D absolute positioning problem to obtain the camera pose parameters R and T;

[0027] Step 3.2.1: Let the world coordinates of the n marking points on the signboard be P w i (i=1,2,3...n), the corresponding camera coordinate system coordinate is P c i (i=1,2,3...n); and let the world coordinates of the four virtual control points be C w j(j=1,2,3,4), the coordinates of the four virtual control points corresponding to the camera coordinate system are C c j (j=1,2,3,4), then:

[0028]

[0029] In the above formula, α ij (j=1, 2, 3, 4) are the 4 weighting coefficients corresponding to each landmark point, and the sum of the 4 weighting coefficients is 1;

[0030] Let P w i The corresponding pixel coordinates are (u i , v i ), the camera coordinate system coordinates of the jth (j=1,2,3,4) virtual control point are (x c j ,y c j , z c j ), according to formula (1), the corresponding relationship between each pair of calibration points and pixel points is obtained:

[0031]

[0032] In the above formula, f x , f y ,u x , v y is the internal parameter of the camera device, obtained by calibration of the camera device, (u i , v i ) is the pixel coordinate corresponding to the i-th marker point. Assuming there are n marker points, the world coordinate of the i-th marker point is P w i (i=1,2,3...n), P w i The corresponding pixel coordinates are (u i , v i );

[0033] When there are n pairs of spatial reference points corresponding to pixel coordinate points, 2n equations are obtained, which are recorded in matrix form:

[0034] M*x=0 (6)

[0035] In the above formula, the vector is a 12×1 vector containing the non-homogeneous coordinates of the four virtual control points in the camera coordinate system; the solution x of equation (6) belongs to ker(M T M),ker(M T M) represents the matrix M TThe null space of M, x belongs to ker(M T M) means x is an element of this set; M is a 2n×12 matrix, M is expressed as:

[0036]

[0037] In the above formula, α ij is the weight coefficient corresponding to each landmark point, (u i , v i ) is the pixel coordinate corresponding to the i-th landmark point; M T The zero eigenvalue dimension of M depends on the characteristics of the spatial reference points (such as distribution, number, etc.) and the focal length of the camera, and usually 1 to 4 groups of eigenvectors corresponding to zero eigenvalues ​​are selected to solve equation (7); according to the distance-preserving property of rigid body transformation, the distance constraint condition is added to obtain the coordinates of the four control points in the camera coordinate system;

[0038] Step 3.2.2: Solve the transformation relationship between the two coordinate systems based on the three-dimensional coordinates of multiple points in the camera coordinate system and the world coordinate system; if the coordinates P of n marker points in the world coordinate system are known i w (i=1,2,3...n) and the coordinate P in the camera coordinate system i c (i=1,2,3...n), definition These are the centroid coordinates of n landmark points in the world coordinate system and the camera coordinate system:

[0039]

[0040] In the above formula, n is the number of landmark points, P i w (i=1,2,3...n) are the coordinates of n marker points in the world coordinate system, P i c (i=1,2,3...n) are the coordinates of n marker points in the camera coordinate system;

[0041] The structural construction matrix F is:

[0042]

[0043] In the above formula, P n w is the coordinate of the landmark point n in the world coordinate system, P n c is the coordinate of the marker point n in the camera coordinate system; are the centroid coordinates of n landmark points in the world coordinate system and the camera coordinate system respectively; perform SVD decomposition on the matrix F:

[0044] F=UΣV T (10)

[0045] In the above formula, U, ∑ and V are matrices obtained by SVD decomposition of F;

[0046] Step 3.3, calculate the final rotation matrix R * , the final translation vector T * , that is, the camera pose parameters are obtained:

[0047]

[0048] In the above formula, are the centroid coordinates of n marker points in the world coordinate system and the camera coordinate system respectively; U and V are matrices obtained by SVD decomposition of F; R is the rotation matrix;

[0049] Step 3.4, set weight coefficients for marker points; when actually measuring three-dimensional displacement, glare, motion, aero-optical effects, etc. will cause the marker point image to be blurred, resulting in varying degrees of error in the positioning of the marker points, which in turn causes abnormal coordinate data to affect the results. In response to the above problems, the present invention sets weights for the errors of each marker point to reduce the impact of the marker point distribution (depth change) on the results, and also suppresses the impact of abnormal data and image noise on the pose solution through iteration;

[0050] Define the error of each landmark point as r i , the mean error of all landmark points is r; the weight ω of each landmark point is constructed according to the Huber function i :

[0051]

[0052] In the above formula, if the reprojection error of the i-th landmark point r i When the absolute value of is less than the mean, the weight coefficient is 1; when the absolute value of the reprojection error is greater than the mean, the weight coefficient is less than 1; and the larger the error value, the smaller the weight coefficient, and the smaller its contribution to the overall error. Since the solution of formula (6) can be transformed into a least squares problem, after incorporating the weight coefficient into the objective function of formula (6), the solution of formula (6) is converted to:

[0053]

[0054] In the above formula, ω i The landmark point P i w The weight of the projection error is determined by formula (12); is the projection matrix, The expression is:

[0055]

[0056] In the above formula, Assume that the focal length of the imaging plane is f = 1, (u i ,v i ,1) T is the pixel coordinate of the landmark point in the world coordinate system projected onto the normalized imaging plane;

[0057] Step 3.5, use iterative calculation to solve equation (13) to obtain the rotation matrix R and the translation matrix T; according to the extreme value condition, the partial derivative of equation (13) with respect to the rotation matrix R must be 0, and the translation matrix T at the kth iteration is obtained k (R k ) About the rotation matrix R k The function expression is:

[0058]

[0059] After substituting formula (15) into formula (13), formula (13) is expressed as:

[0060]

[0061] In the above formula, R k+1 is the optimal value of the rotation matrix, R is the rotation matrix, T is the translation matrix, P i w is the coordinate of the i-th marker point in the world coordinate system; the initial value R of the rotation matrix R 0 It is obtained by the PnP algorithm, and then the weight of each landmark point is calculated by formula (13), and the new weight ω i Substitute into equation (16) to solve the optimal problem; then update the weight ω i , repeatedly iterate the above calculation process; set a threshold ε (usually set to 2mm-5mm) for the objective function, and stop the iteration when the objective function value is less than the threshold ε (usually set to 2mm-5mm).

[0062] As a preference, the iterative calculation in step 3.5 is specifically as follows: In order to ensure global convergence, E(R k+1 )≤E(R k ), where E(R k ) is the kth iteration error, E(R k+1 ) k+1th iteration error; define the i-th landmark point P i w The camera coordinate system coordinate is q i , i = 1, 2, 3...n, after the kth iteration q i The expression is:

[0063]

[0064] In the above formula, k is the number of iterations, R k is the rotation matrix, T k is the translation matrix, P i w is the coordinate of the i-th marker point in the world coordinate system; then the k+1 iteration error E(R k+1 )for:

[0065]

[0066] In the above formula, ω i The landmark point P i w The weight of the projection error, q i is the i-th landmark point P i w The camera coordinate system coordinates of is the pixel coordinate of the landmark point in the world coordinate system projected onto the normalized imaging plane;

[0067] From equations (16) and (17), we know that at the (k+1)th iteration, i k Compared to q i k+1 Leave The distance is closer, so the following inequality (19) holds:

[0068]

[0069] Subtract both ends of equation (19) have to:

[0070]

[0071] In formula (20), the second term on the right side of the inequality is always non-zero when the objective function has not reached the minimum value. Therefore, E(R k+1 )≤E(R k ) condition is met, and the PnP algorithm meets the global convergence property.

[0072] Preferably, the visual light source controllable device in step 1 has LEDs with optimized distribution arrangement inside, high brightness and thin appearance; the visual light source controllable device is made of aerospace-grade aluminum alloy, which has strong corrosion resistance and tensile strength, and can therefore adapt to complex working environments; the light intensity of the visual light source controllable device is controlled by a flicker-free LED dimming controller.

[0073] The beneficial effects of the present invention are as follows: the present invention will use a method based on visual measurement to realize three-dimensional online monitoring of the expansion amount of the pressure-bearing components of the power station when the boiler expands, simplify the system wiring through wireless communication, and improve the availability of the system. The present invention reduces the labor intensity of the inspection personnel through remote signal transmission, and provides direct video signals to enhance the intuitive experience. Wireless communication is used to improve the convenience of the system, and the output information of the system provides a basis for fault alarm; the present invention aims to provide complete boiler expansion information for the unit operating personnel and improve the safety of the unit operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a schematic diagram of a boiler three-dimensional expansion measurement device based on computer vision;

[0075] Figure 2 It is a schematic diagram of the camera coordinate system and the world coordinate system model;

[0076] Figure 3 It is a schematic diagram of the communication model of the thermal power station expansion monitoring system;

[0077] Figure 4 This is a schematic diagram of the video signal routing;

[0078] Figure 5 This is a schematic diagram of a flicker-free dimming controller;

[0079] Figure 6 Schematic diagram of the measurement image.

[0080] Description of reference numerals:

[0081] Pressure-bearing component 1, angle steel 2, connector 3, camera device 4, surface light source 5, sign board 6, industrial control computer 7, video acquisition card 8, wireless image transmission system 9, expansion processing instrument 10. DETAILED DESCRIPTION

[0082] The present invention is further described below in conjunction with embodiments. The description of the following embodiments is only used to help understand the present invention. It should be noted that for ordinary persons in the art, without departing from the principle of the present invention, the present invention can also be modified in some ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

[0083] Aiming at the demand for three-dimensional expansion monitoring of boilers, the present invention provides a device and method for expansion monitoring of pressure-bearing components of power plants based on computer vision.

[0084] Embodiment 1:

[0085] Vision-based three-dimensional expansion measurement device for boilers Figure 1As shown. The angle steel is fixed on the key position of the pressure-bearing component that needs to be monitored for expansion. The angle steel is rigidly connected to the pressure-bearing component, so the three-dimensional expansion of the pressure-bearing component will be reflected as the three-dimensional displacement of the free end of the angle steel. An industrial camera is installed at the free end of the angle steel through a connector. In order to improve the adaptability of the visual measurement system to the environment, a surface light source is configured as the backlight of the sign board. The surface light source and the sign board are in the same plane and are both within the field of view of the camera. The video signal is transmitted to the processing host through the wireless communication system, and the three-dimensional displacement measurement of the free end of the angle steel is realized in conjunction with the processing program on the host, that is, the three-dimensional expansion of the pressure-bearing component is obtained.

[0086] (1) Visual signal transmission

[0087] The horizontal elevation of the expansion monitoring point of the power station boiler is relatively high. It is inconvenient to wire on the boiler steel frame and the construction cost is high. In order to facilitate system installation and wiring, this embodiment adopts wireless image transmission to realize wireless transmission of video signals. Figure 3 As shown. The AV video signal output by the industrial camera is connected to the transmitting end of the wireless image transmission system, and the receiving end is connected to the video acquisition card, and the video acquisition card is connected to the expansion processing instrument via USB. In order to facilitate installation and ensure high reliability of operation, the present invention uses an embedded industrial computer as the core equipment for expansion processing.

[0088] The video transmission sampling 5.8G image transmission system of the expansion monitoring system uses 1W power and mushroom antenna. The ideal image transmission distance can reach 15km, which can also meet the application requirements of the present invention under the interference of obstacles. Figure 4 As shown in the figure. Each power plant boiler needs to monitor the expansion at dozens of points, so multiple expansion processing instruments are configured to coordinate the expansion monitoring of a boiler. In general, the expansion processing instruments are installed diagonally, such as Figure 4 As shown, the video signal in front of the boiler and on the right side enters the 2# angle expansion processor, while the video signal in the back and on the left side enters the 1# angle expansion processor. Figure 4 The arrangement shown can ensure that there are no large-section obstacles between the sending and receiving ends of the image transmission system. The obstacles are only steel frame equipment, which has little impact on the image transmission system. Therefore, the selection of a wireless image transmission system can meet the requirements of real-time image transmission.

[0089] (2) Since the detection system of this embodiment needs to work continuously and in real time for a long time, it is necessary to consider the exposure problem of the camera in dim light conditions (at night). In addition, the cluttered background on the boiler steel frame will cause certain troubles to the later processing of the picture. In response to the above problems, this embodiment combines the characteristics of the camera exposure time that can be manually set, and selects a set of visual light source controllable devices, in which the light source board is a high-brightness backlight source, the appearance design is very light and thin, the internal structure is an optimized arrangement of LED distribution, the light intensity distribution is uniform, and the material is aerospace-grade aluminum alloy, which has strong corrosion resistance and tensile strength, so it can adapt to complex working environments.

[0090] The light intensity of the light source panel is realized by a flicker-free LED dimming controller. Figure 5 As shown, it adopts constant current power supply, the light source intensity can be manually adjusted, and the light source level is divided very finely, with a total of 255 levels, which can meet the needs of various scenes. The most suitable brightness level can be selected in the on-site test, which can meet the measurement needs while achieving the purpose of energy saving.

[0091] In order to facilitate subsequent detection and matching, this embodiment uses multiple circular markers of different areas. After acquiring the sample image, the pixel coordinates of the center of each marker can be obtained through the circle detection method. The world coordinates of the center of the circle can be calibrated in advance through offline measurement. The pixel coordinates and world coordinates can be matched based on the pixel area size of the marker, and then the WIEPnP algorithm can be used to calculate the three-dimensional displacement corresponding to this image.

[0092] Embodiment 2:

[0093] The working method of the expansion monitoring device for pressure-bearing components of a power station based on machine vision specifically includes the following steps:

[0094] Step 1, the camera device 4 collects video signals. The horizontal elevation of the expansion monitoring point of the power station boiler is relatively high. It is inconvenient to wire on the boiler steel frame and the construction cost is high. In order to facilitate system installation and wiring, the present invention uses wireless image transmission in the image acquisition and transmission system to wirelessly transmit video signals, and the video signals are transmitted to the processing host through wireless image transmission; the wireless transmission of frequency signals adopts 5.8G wireless image transmission system, and 1W power is selected to cooperate with mushroom antenna. The image transmission distance can reach 15km, achieving the effect of contactless measurement and no manual reading is required; in terms of visual light source, a controllable visual light source device is selected as the surface light source 5, combined with the feature that the camera exposure time can be manually set, so that the image features are more significant, which is conducive to subsequent image processing; Step 2, the processing program in the processing host processes the video signal, and takes the world coordinate system of the fixed marker in the field of view of the camera device 4 as a reference to solve the three-dimensional coordinates of the camera device 4 in the world coordinate system;

[0095] Step 3, calculate the rotation matrix R and translation matrix T corresponding to each frame image. Solving R and T is the key to obtaining the expansion amount: when the correspondence between the world coordinate system coordinates P and the camera coordinate system coordinates p of n pairs of known points is known, the rotation matrix R and translation matrix T are solved by the PnP algorithm; the three-dimensional displacement of the measured point is obtained, and the three-dimensional expansion of the pressure-bearing component 1 is obtained.

[0096] Embodiment 3:

[0097] Based on Example 2, step 2 is further refined as follows:

[0098] According to the pinhole imaging model of visual measurement, the camera coordinate system and the world coordinate system are expressed as c X c Y c Z c and O w X w Y w Z w There is a rigid body transformation relationship between the two coordinate systems, that is, rotation and translation. Let the 3×3 matrix R be the rotation transformation matrix of the world coordinate system relative to the camera coordinate system, and each row and column of the matrix satisfies the orthogonal relationship; let the 3×1 matrix T represent the three-dimensional coordinates of the origin of the world coordinate system in the camera coordinate system. w (X w , Y w , Z w ) is any point in the world coordinate system, and its corresponding point in the camera coordinate system is represented by P c (X c , Y c , Z c ), the model of Pw and Pc in the coordinate system is as follows Figure 2 shown.

[0099] The coordinate relationship between Pw and Pc is expressed as:

[0100]

[0101] It can be seen from formula (1) that for any point, the world coordinate system coordinates of the point can be calculated under the premise that the camera coordinate system coordinates, rotation matrix R and translation matrix T of the point are known. Based on this, in this embodiment, the camera is fixed to one end of the angle steel, the other end of the angle steel is connected to the rigid body of the pressure-bearing component, and a fixed sign plate is set in the camera field of view, such as Figure 1 As shown in the figure, the three-dimensional expansion of the key points of the pressure-bearing component is expressed as the camera optical center O c The three-dimensional relative displacement between the camera and the signboard. The optical center of the camera in the camera coordinate system O c The coordinates of are (0,0,0), then its coordinates in the world coordinate system can be expressed as:

[0102]

[0103] In formula (2), the coordinate (X w-Oc , Y w-Oc , Z w-Oc ) is the three-dimensional coordinate of the camera optical center in the world coordinate system, and the coordinate (X c-Oc , Y c-Oc , Z c-Oc ) is the three-dimensional coordinate of the camera optical center in the camera coordinate system. c The change in coordinates in each direction is the expansion of the key point.

[0104] The camera collects images in real time, and calculates the rotation matrix R and translation matrix T corresponding to each frame of the image. According to formula (2), the three-dimensional expansion of the key points of the boiler relative to the sign plate can be obtained in real time. Therefore, solving R and T is the key to obtaining the expansion. When the corresponding relationship between the world coordinate system coordinates P and the camera coordinate system coordinates p of n pairs of points is known, R and T can be solved by the PnP algorithm, and then the three-dimensional coordinates of the camera optical center in the world coordinate system are calculated by formula (2). Since the world coordinate system is a stationary reference object, its coordinate axis scale has a real length meaning, and the camera and the object to be measured are connected by a rigid body, the three-dimensional coordinate change of the camera optical center in the world coordinate system can be used to represent the three-dimensional displacement of the measured point. The three-dimensional displacement expression is as follows:

[0105]

[0106] In the above formula, D x , Dy, D z Respectively represent the displacement of the measured point in the three directions of x, y, and z axes of the world coordinate system, ΔX w-Oc , ΔY w-Oc , ΔZ w-Oc Represent the camera optical center O c The coordinate transformation amount in the three directions of x, y, and z axes of the world coordinate system.

[0107] Embodiment 4:

[0108] On the basis of Example 2 and Example 3, a three-dimensional expansion displacement measurement algorithm WIEPnP is introduced:

[0109] It can be seen from the above-mentioned measuring device that the key point of solving the displacement measurement algorithm is mainly to solve the PnP problem, and solving PnP requires matching the pixel coordinates of multiple pairs of points with the world coordinates. The present invention uses several circles of different areas as markers, and uses their centroids as marker points. The world coordinates of each marker point can be obtained by offline measurement, and the image of the sign board is photographed in real time with an industrial camera. Since the image will be distorted during shooting, and the center of gravity position will not change, the pixel coordinates of the center of gravity of the circular marker can be obtained by using an ellipse recognition algorithm. According to the relationship between the pixel area of ​​the marker, the matching of the world coordinates and the pixel coordinates is achieved. The process of solving R and T is as follows:

[0110] (1) Reduce the coupling between markers

[0111] Using the coordinates of four non-coplanar virtual control points to linearly represent any landmark point can reduce the coupling between the landmark points, and the world coordinates of the four virtual control points can be set arbitrarily. First, the coordinates of the four virtual control points in the camera coordinate system are solved, and then the camera pose parameters R and T are obtained by solving the classic 3D-3D absolute positioning problem.

[0112] Suppose the world coordinates of the n marking points on the signboard are P w i (i=1,2,3...n), the corresponding camera coordinate system coordinate is P c i (i=1,2,3...n); and let the world coordinates of the four virtual control points be C w j (j=1,2,3,4), the corresponding camera coordinate system coordinate is C c j (j=1,2,3,4), then the following relationship holds:

[0113]

[0114] In formula (4), each landmark point corresponds to 4 weighting coefficients α ij (j=1,2,3,4), and their sum is 1. Let P w i The corresponding pixel coordinates are (u i , v i ), the camera coordinate system coordinates of the jth (j=1,2,3,4) virtual control point are (x c j ,y c j , z c j ), combined with the transformation relationship between pixel coordinates and camera coordinates, the corresponding relationship between each pair of calibration points and pixel points can be obtained:

[0115]

[0116] In formula (4), f x , f y ,u x , v y is the internal parameter of the camera, obtained by camera calibration, (u i , v i ) is the pixel coordinate corresponding to the i-th marker point. When there are n pairs of spatial reference points corresponding to pixel coordinate points, 2n equations can be obtained, which are recorded in matrix form:

[0117] M*x=0 (6)

[0118] In formula (6), M is a 2n×12 matrix, which can be expressed as:

[0119]

[0120] vector is a 12×1 vector containing the non-homogeneous coordinates of the four virtual control points in the camera coordinate system. The solution x of equation (6) belongs to ker(M T M), M T The zero eigenvalue dimension of M depends on the characteristics of the spatial reference points (such as distribution, number, etc.) and the focal length of the camera, and usually 1 to 4 sets of eigenvectors corresponding to zero eigenvalues ​​are selected to solve the equation. According to the distance-preserving property of rigid body transformation, the distance constraint condition is added to obtain the coordinates of the four control points in the camera coordinate system. The coordinates of the marker points in the camera coordinate system can be calculated by equation (4), and then directly solved based on the absolute positioning problem.

[0121] The absolute value positioning problem solves the transformation relationship between two coordinate systems based on the three-dimensional coordinates of multiple points in two coordinate systems. If the world coordinates P of n landmark points are known wi (i=1,2,3...n) and the coordinate P in the camera coordinate system ci (i=1,2,3...n); definition are the centroid coordinates of the n landmark points in the two coordinate systems, namely:

[0122]

[0123] Then construct the matrix F according to the following structure:

[0124]

[0125] In the above formula, P n w is the coordinate of the marker point n in the world coordinate system, is the coordinate of the marker point n in the camera coordinate system; are the centroid coordinates of n landmark points in the world coordinate system and the camera coordinate system respectively; perform SVD decomposition on the matrix F:

[0126] F=UΣV T (10)

[0127] In the above formula, U, ∑ and V are matrices obtained by SVD decomposition of F;

[0128] Further calculate the final rotation matrix R * , the final translation vector T * As shown in formula (11), the camera pose parameters are obtained:

[0129]

[0130] (2) Setting weight coefficients for landmark points

[0131] When actually measuring three-dimensional displacement, glare, motion, aero-optical effects, etc. will cause the image of the marker point to be blurred, resulting in different degrees of error in the positioning of the marker point, which in turn causes abnormal coordinate data to affect the result. In view of the above problems, the present invention sets a weight for each marker point error to reduce the impact of the marker point distribution (depth change) on the result, and also suppresses the impact of abnormal data and image noise on the pose solution through iteration.

[0132] Define the error of each landmark point as r i , the error mean of all landmark points is r. The weight of each landmark point ω i Constructed according to the Huber function:

[0133]

[0134] In formula (12), if the reprojection error of the i-th landmark point r i When the absolute value of is less than the mean, the weight coefficient is 1; when the absolute value of the reprojection error is greater than the mean, the weight coefficient is less than 1; and the larger the error value, the smaller the weight coefficient, and the smaller its contribution to the overall error. Since the solution of formula (6) can be transformed into a least squares problem, after incorporating the weight coefficient into the objective function of the problem, the solution of formula (6) can be converted to:

[0135]

[0136] In formula (13), ω i is the weight of the reprojection error of the landmark point Piw, determined by formula (12). is the projection matrix, and its expression is shown in formula (14):

[0137]

[0138] In formula (14), It is the pixel coordinate of the marker point in the world coordinate system projected onto the normalized image plane (assuming the imaging plane with focal length f=1).

[0139] (3) Iterative solution

[0140] The present invention uses iterative calculation to solve equation (13) to obtain T and R. According to the extreme value condition, the partial derivative of equation (13) with respect to R must be 0, thus the translation matrix T at the kth iteration can be obtained: k (R k ) About the rotation matrix R k The function expression is:

[0141]

[0142] After substituting equation (15) into equation (13), equation (13) can be expressed as the optimization problem shown in equation (16):

[0143]

[0144] The initial value of the rotation matrix R 0 It is obtained by the EPnP algorithm, and then the weight of each landmark point is calculated by formula (13), and the new weight ω i Substitute into equation (16) to solve the optimal problem; then update the weight ω i , repeatedly iterate the above calculation process. Set a threshold ε (usually set to 2mm-5mm) for the objective function, and stop iterating when the objective function value is less than the threshold ε (usually set to 2mm-5mm).

[0145] WIEPnP is an iterative algorithm. To ensure global convergence, E(R k+1 )≤E(R k ), where E(R k ) is the kth iteration error. The following verifies the convergence condition of WIEPnP. Define the i-th (i=1,2,3...n) landmark point P iw The camera coordinate system coordinate is q i , after the kth iteration q i The expression is:

[0146]

[0147] Then the k+1 iteration error E(R k+1 )for:

[0148]

[0149] From equations (16) and (17), we can see that at the (k+1)th iteration, ik Compared to q i k+1 Leave is closer. Therefore, the following inequality holds:

[0150]

[0151] Subtract both ends of equation (19) have to:

[0152]

[0153] In formula (20), the second term on the right side of the inequality is always non-zero when the objective function has not reached the minimum value. Therefore, E(R k+1 )≤E(R k ) condition is established, the PnP algorithm meets the global convergence; the specific execution process of the weighted iterative EPnP algorithm is shown in Table 1 below;

[0154] Table 1 Specific execution flow chart of weighted iterative EPnP algorithm (WIEPnP)

[0155]

[0156] In order to verify the effect of this embodiment in actual application, a monocular vision three-dimensional expansion measurement prototype system was built in the laboratory. In the prototype system, the sign plate is fixed to the three-dimensional displacement platform, and the three-dimensional displacement of the platform simulates the expansion of the key points of the boiler. The experiment uses a HiconVision camera as the measuring element, the camera focal length is 12mm, and the camera internal parameters are obtained by chessboard calibration. The three-dimensional displacement platform has scales in three directions, which can read out the precise three-dimensional displacement to compare with the visual measurement results and evaluate the error performance of visual measurement.

[0157] In the experiment, the three-dimensional displacement platform was first adjusted to make the marker plate at a certain position, the scale reading at this time was recorded, and three images were taken for expansion measurement, and three measurement results were obtained, as shown in "Measurement value 1" to "Measurement value 3" in Table 3. Then, a certain direction (for example, x direction -10mm) was adjusted to another position, the displacement reading was recorded and three images were taken. The experiment took three positions in each of the three directions for verification, and a total of 27 images were obtained. Figure 6Only one image is listed for each position, and the size of each image is 3072 pixels by 2048 pixels. Six circles of different sizes are engraved on the signboard in the experiment as signs. The pixel coordinates of the centroid of each rectangle can be obtained through edge detection and curve fitting, and then the pixel coordinates are matched with the world coordinates according to the pixel area size relationship. The WIEPnP algorithm is then used to calculate the camera pose, and the three-dimensional expansion is calculated by formula (2). The experimental results are presented in Table 2 below. In order to prove that the WIEPnP algorithm meets the real-time requirements in actual measurement, this experiment counts the time taken for each image from preprocessing to obtaining the three-dimensional expansion, and the recorded results are presented in Table 3 below;

[0158] Table 2 Three-dimensional displacement measurement experimental results of prototype system

[0159]

[0160] It can be seen from Table 2 that in the prototype measurement system, the measurement values ​​in the x, y, and z directions are relatively accurate, with an average absolute error of less than 1 mm.

[0161] Table 3 Prototype system measurement time consumption experimental results

[0162]

[0163] It can be seen from Table 3 that in the prototype measurement system, the average time consumption for the three-dimensional displacement measurement of 27 images is 1.81 s, and the variance is 0.08. The fluctuation of the measurement time consumption of each image is small, and the measurement time consumption of a single image is less than 2 s. Therefore, the WIEPnP algorithm meets the real-time requirements in actual measurement.

Claims

1. A working method of a device for monitoring expansion of pressure-bearing components of a power station based on machine vision, It is characterized in that include: A pressure-bearing component (1), an angle steel (2), a connector (3), and a camera device (4); wherein the fixed end of the angle steel (2) is fixed to the pressure-bearing component (1), the fixed end of the angle steel (2) is rigidly connected to the pressure-bearing component (1), and the free end of the angle steel (2) is equipped with the camera device (4) via the connector (3); specifically comprising the following steps: Step 1, the camera device (4) collects video signals, and wireless image transmission is used on the image acquisition and transmission system to wirelessly transmit the video signals, and the video signals are transmitted to the processing host through the wireless image transmission; a visual light source controllable device is selected as a surface light source (5); Step 2: The processing program in the processing host processes the video signal, and uses the world coordinate system of the fixed marker in the field of view of the camera device (4) as a reference to solve the three-dimensional coordinates of the camera device (4) in the world coordinate system; Step 3, calculate the rotation matrix R and translation matrix T corresponding to each frame image: when the corresponding relationship between the world coordinate system coordinates P and the camera coordinate system coordinates p of n pairs of points is known, solve the rotation matrix R and translation matrix T by PnP algorithm; obtain the three-dimensional displacement of the measured point, and obtain the three-dimensional expansion of the pressure-bearing component (1); Step 2 specifically includes the following steps: Step 2.1: Express the camera coordinate system as O c X c Y c Z c , the world coordinate system is represented as O w X w Y w Z w ; Let the 3×3 matrix R be the rotation matrix of the world coordinate system relative to the camera coordinate system, and each row and column of the matrix R satisfy the orthogonal relationship; let the 3×1 matrix T be the three-dimensional coordinates of the origin of the world coordinate system in the camera coordinate system, Pw(X w , Y w , Z w ) is any point in the world coordinate system, and its corresponding expression in the camera coordinate system is Pc(X c , Y c , Z c ); the coordinate relationship between Pw and Pc is expressed as: In the above formula, X c , Y c , Z c are any point Pw(X w , Y w , Z w ) is the three-dimensional coordinate in the camera coordinate system, X w , Y w , Z w For any point Pw(X w , Y w , Z w ) three-dimensional coordinates; the matrix R is the rotation matrix of the world coordinate system relative to the camera coordinate system, and the matrix T is the three-dimensional coordinate translation matrix of the origin of the world coordinate system in the camera coordinate system; For any point Pw(X w , Y w , Z w ), given the camera coordinate system coordinates Pc(X c , Y c , Z c ), rotation matrix R and translation matrix T, the world coordinate system coordinates Pw(X w , Y w , Z w ); Step 2.2: Camera optical center O in camera coordinate system c The coordinates of the camera are (0,0,0), then the camera optical center O c The coordinates in the world coordinate system are expressed as: In the above formula, the matrix R is the rotation matrix of the world coordinate system relative to the camera coordinate system, and the matrix T is the three-dimensional coordinate of the origin of the world coordinate system in the camera coordinate system; the coordinate (X c-Oc , Y c-Oc , Z c-Oc ) is the camera optical center O c Three-dimensional coordinates in the camera coordinate system; coordinates (X w-Oc , Y w-Oc , Z w-Oc ) is the camera optical center O c Three-dimensional coordinates in the world coordinate system; Step 2.3: Calculate the camera optical center O by formula (2): c The three-dimensional coordinates in the world coordinate system are the camera optical center O c The three-dimensional coordinate change in the world coordinate system is used to represent the three-dimensional displacement of the measured point: In the above formula, D x , Dy, D z Respectively represent the displacement of the measured point in the three directions of x, y, and z axes of the world coordinate system, ΔX w-Oc、 ΔY w-Oc、 ΔZ w-Oc Represent the camera optical center O c The coordinate transformation amount in the three directions of x, y, and z axes of the world coordinate system; Step 3 specifically includes the following steps: Step 3.1, select a plurality of circles of different areas as fixed markers, and use the centroid of the circle as the marker point, obtain the world coordinates of each marker point through offline measurement, use a camera device (4) to take an image of the marker plate (6) in real time, use an ellipse recognition algorithm, based on geometric features, obtain the pixel coordinates of the centroid of the circular fixed marker through edge detection and curve fitting, and match the world coordinates with the pixel coordinates according to the size of the pixel area; Step 3.2, reduce the coupling between the marker points: linearly represent any marker point by the coordinates of the four non-coplanar virtual control points to solve the coordinates of the four virtual control points in the camera coordinate system; Step 3.2.1: Let the world coordinates of the n marking points on the signboard be P w i , i=1,2,3...n, the corresponding camera coordinate system coordinate is P c i i=1,2,3...n; and let the world coordinates of the four virtual control points be C w j , j = 1, 2, 3, 4, the coordinates of the four virtual control points corresponding to the camera coordinate system are C c j , j=1,2,3,4, then: In the above formula, α ij , j = 1, 2, 3, 4 are the 4 weighting coefficients corresponding to each landmark point, and the sum of the 4 weighting coefficients is 1; Let P w i The corresponding pixel coordinates are (u i, v i ), the camera coordinate system coordinates of the jth, j=1, 2, 3, 4 virtual control points are (x c j ,y c j , z c j ), according to formula (1), the corresponding relationship between each pair of calibration points and pixel points is obtained: In the above formula, f x , f y ,u x , v y is the internal parameter of the camera device, obtained by calibration of the camera device, (u i, v i ) is the pixel coordinate corresponding to the i-th marker point. Assuming there are n marker points, the world coordinate of the i-th marker point is P w i , i=1,2,3...n,P w i The corresponding pixel coordinates are (u i, v i ); When there are n pairs of spatial reference points corresponding to pixel coordinate points, 2n equations are obtained, which are recorded in matrix form: M*x=0 (6) In the above formula, the vector is a 12×1 vector containing the non-homogeneous coordinates of the four virtual control points in the camera coordinate system; the solution x of equation (6) belongs to ker(M T M),ker(M T M) represents the matrix M T The null space of M, x belongs to ker(M T M) means x is the matrix M T The elements of the null space set of M; M is a 2n×12 matrix, and M is represented as: In the above formula, α ij is the weight coefficient corresponding to each landmark point, (u i , v i ) is the pixel coordinate corresponding to the i-th landmark point; Step 3.2.2: Solve the transformation relationship between the two coordinate systems based on the three-dimensional coordinates of multiple points in the camera coordinate system and the world coordinate system; if the coordinates P of n marker points in the world coordinate system are known i w , i = 1, 2, 3...n and the coordinates P in the camera coordinate system i c ,i=1,2,3...n,definition These are the centroid coordinates of n landmark points in the world coordinate system and the camera coordinate system: In the above formula, n is the number of landmark points, P i w , i=1,2,3...n, are the coordinates of n marker points in the world coordinate system, P i c , i=1,2,3...n, are the coordinates of n marker points in the camera coordinate system; The structural construction matrix F is: In the above formula, is the coordinate of the marker point n in the world coordinate system, is the coordinate of the marker point n in the camera coordinate system; are the centroid coordinates of n landmark points in the world coordinate system and the camera coordinate system respectively; perform SVD decomposition on the matrix F: F=UΣV T (10) In the above formula, U, ∑ and V are matrices obtained by SVD decomposition of F; Step 3.3, calculate the final rotation matrix R * , the final translation vector T * : In the above formula, are the centroid coordinates of n marker points in the world coordinate system and the camera coordinate system respectively; U and V are matrices obtained by SVD decomposition of F; R is the rotation matrix; Step 3.4, set weight coefficients for landmark points; Define the error of each landmark point as r i , the mean error of all landmark points is r; the weight ω of each landmark point is constructed according to the Huber function i : In the above formula, if the reprojection error of the i-th landmark point r i When the absolute value of is less than the mean, the weight coefficient is 1; when the absolute value of the reprojection error is greater than the mean, the weight coefficient is less than 1; after incorporating the weight coefficient into the objective function of formula (6), the solution of formula (6) is converted to: In the above formula, ω i The landmark point P i w The weight of the projection error is determined by formula (12); is the projection matrix, The expression is: In the above formula, Assume that the focal length of the imaging plane is f = 1, (u i ,v i ,1) T is the pixel coordinate of the landmark point in the world coordinate system projected onto the normalized imaging plane; Step 3.5, use iterative calculation to solve equation (13) to obtain the rotation matrix R and the translation matrix T; according to the extreme value condition, the partial derivative of equation (13) with respect to the rotation matrix R must be 0, and the translation matrix T at the kth iteration is obtained k (R k ) About the rotation matrix R k The function expression is: After substituting formula (15) into formula (13), formula (13) is expressed as: In the above formula, R k+1 is the optimal value of the rotation matrix, R is the rotation matrix, T is the translation matrix, P i w is the coordinate of the i-th marker point in the world coordinate system; the initial value R of the rotation matrix R 0 It is obtained by the PnP algorithm, and then the weight of each landmark point is calculated by formula (13), and the new weight ω i Substitute into equation (16) to solve the optimal problem; then update the weight ω i , repeatedly iterate the above calculation process; set a threshold ε for the objective function, and stop the iteration when the objective function value is less than the threshold ε.

2. According to the working method of the machine vision-based expansion monitoring device for pressure-bearing components of a power station according to claim 1, It is characterized in that The specific iterative calculation in step 3.5 is: satisfy E(R k+1 )≤E(R k ), where E(R k ) is the kth iteration error, E(R k+1 ) k+1th iteration error; define the i-th landmark point P i w The camera coordinate system coordinate is q i , i = 1, 2, 3...n, after the kth iteration q i The expression is: q i k =R k P i w +T k (R k ) (17) In the above formula, k is the number of iterations, R k is the rotation matrix, T k is the translation matrix, P i w is the coordinate of the i-th marker point in the world coordinate system; then the k+1 iteration error E(R k+1 )for: In the above formula, ω i The landmark point P i w The weight of the projection error, q i is the i-th landmark point P i w The camera coordinate system coordinates of is the pixel coordinate of the landmark point in the world coordinate system projected onto the normalized imaging plane; From equations (16) and (17), we know that at the (k+1)th iteration, compared to Leave The distance is closer, so the following inequality (19) holds: Subtract both ends of equation (19) have to: In formula (20), the second term on the right side of the inequality is always non-zero when the objective function has not reached the minimum value. Therefore, E(R k+1 )≤E(R k ) condition is met, and the PnP algorithm meets the global convergence property.

3. According to the working method of the machine vision-based expansion monitoring device for pressure-bearing components of a power station as claimed in claim 1, Features: In step 1, the visual light source controllable device has LEDs with optimized distribution arrangement inside; the visual light source controllable device is made of aerospace-grade aluminum alloy; and the light intensity of the visual light source controllable device is controlled by a flicker-free LED dimming controller.

Citation Information

Cited By

  • Visual dilation measurement system image quality control method capable of reducing light pollution

    CN120935471A