CN109727267B - Standard virtual sine linear vibration measurement method - Google Patents

Abstract

The invention discloses a standard virtual sinusoidal linear vibration measurement method, which includes calculating the grayscale gradient amplitude of an image, using the distribution of the grayscale gradient amplitude to find the maximum amplitude value in the interval, and according to the number of the maximum value Determine whether there is a superposition of objects with different moving positions in the region of interest in the image; improve the accuracy of subsequent image processing for different superposition situations of the region of interest; use a sub-pixel edge detection method to extract the feature edges of moving objects; introduce For the target vibration model, the least squares method is used for model fitting; finally, displacement error compensation is performed on the target vibration model to obtain the time-displacement measurement curve of virtual sinusoidal linear vibration. The invention can avoid the non-ideal sinusoidal vibration situation of the vibration table caused by factors such as mechanical manufacturing and motion control, output standard virtual sinusoidal linear vibration through Matlab programming, and provide a virtual traceability approach for plane motion measurement based on machine vision.

Description

A Standard Virtual Sinusoidal Linear Vibration Measurement Method

technical field

The invention belongs to the field of image processing and machine vision detection, and particularly relates to a standard virtual sinusoidal linear vibration measurement method.

Background technique

Vibration sensors and measuring instruments are widely used in real-time monitoring and vibration parameter measurement in many fields, such as bridge construction, earthquake, automobile, aerospace, etc. In order to ensure the reliability and accuracy of the measurement results obtained by the vibration sensor and measuring instrument, it is necessary to calibrate the vibration sensor and measuring instrument regularly.

Typical calibration methods for vibration sensors and measuring instruments include laser interferometric absolute calibration and comparison calibration. These calibration methods use different measurement methods to obtain the vibration graph of the shaker, and the calibration of the vibration sensor is achieved by comparing these measurement results. However, due to the influence of errors in mechanical processing and motion control of the vibrating table, the vibration form of the vibrating table is not a standard sinusoidal linear vibration, resulting in large errors in the existing calibration methods.

The method of the invention generates standard virtual sinusoidal linear vibration through Matlab programming, realizes the measurement of the standard virtual sinusoidal linear vibration through machine vision, avoids errors introduced by the vibration table, and further improves the vibration measurement accuracy. Besides, the method of the present invention provides a virtual traceability approach for the plane vibration measurement method.

SUMMARY OF THE INVENTION

In order to avoid errors introduced by external factors such as mechanical processing and motion control, so that the vibration form of the shaking table is non-standard sinusoidal linear vibration, resulting in low vibration measurement accuracy, the present invention proposes a standard virtual sinusoidal linear vibration measurement method, including :

Judgment of whether the image has a target overlapping area: it is used to judge whether there is a target overlapping area in the read sequence image, including: the selection of the region of interest of the moving target in the image, and the judgment of whether the area of interest has a target overlapping area;

Selection of image restoration method: used to improve the accuracy of subsequent image processing, including: judging the direction of movement of the target, and selecting an image restoration method based on the direction of movement;

High-precision detection of moving target feature edges: sub-pixel detection for extracting moving target feature edges to improve feature edge accuracy and ensure the reliability of motion measurement results;

Fitting the target vibration model: introduce the target vibration model and use the least squares method to fit the model;

Correction of the target vibration model: Compensate the displacement error of the fitted vibration model to obtain the optimal curve of displacement-time for virtual sinusoidal linear vibration measurement.

A standard virtual sinusoidal linear vibration measurement method, including the following steps:

S1: Calculate the grayscale gradient amplitude of the image, use the distribution of the grayscale gradient amplitude to find the peaks in the area, and judge whether there is a moving object superposition in the area of interest of the image according to the number of peaks;

S2: For images with overlapping moving objects in the area, the optical flow gradient method is used to determine the moving direction of the moving objects, and different image restoration methods are used to restore the images to improve the accuracy of subsequent image processing;

S3: Use the sub-pixel edge detection method to complete the extraction of moving target feature edges;

S4: Introduce the target vibration model and use the least squares method for model fitting;

S5: Perform displacement error compensation on the target vibration model to obtain a time-displacement measurement curve of virtual sinusoidal linear vibration.

The step S1 specifically includes:

(1) Determination of the region of interest of the moving target

Use the inter-frame difference method to determine the moving area of the moving target; in order to reduce the influence of noise, perform a connected domain analysis on the obtained moving area; calculate the area of each connected domain, and take the two areas with the largest area as the moving target. area;

(2) Judgment of the superimposed area of the moving target

In the region of interest of the moving object, calculate the gray gradient magnitude:

Among them, f(x,y) is the gray value at the pixel (x,y); _Ix and Iy are the grayscale gradients in the x and _y directions, respectively; ▽f(x,y) is the pixel (x) ,y) gradient vector; mag(▽f) is the amplitude corresponding to the gradient vector; when there is no moving target in the area, there are only two local maximum points in the gradient amplitude of the traversed area; when there is motion in the area When the targets are superimposed, the number of local maximum points is greater than two.

The step S2 specifically includes:

(1) Determination of the moving direction of the target

The optical flow gradient method is used to calculate the optical flow of the region of interest of the moving target. The algorithm assumes that the brightness of the corresponding pixels of two adjacent frames of images is constant. The calculation formula is as follows:

I _x u+I _y v+I _t =0 (2)

Among them, I _t is the grayscale gradient of the image at time t; the optical flow u in the x direction and the optical flow v in the y direction are used to determine the moving direction of the target;

(2) Image restoration method

For images with strong gray contrast between the background and the moving target, the moving target is a solid rectangle; it is assumed that the moving target feature edge is always its right edge, and when the x-coordinate of the moving target feature edge in adjacent frame images shows an increasing trend, it is considered that the target The movement direction is the positive direction; the gray area of the target feature edge area is divided; the gray level of the rightmost area is enhanced for the region of interest with the positive direction of movement, and the gray level of the remaining areas is weakened; for the area of interest with the negative movement direction The gray level of the leftmost region is enhanced, and the gray level of the remaining regions is weakened.

The step S3 specifically includes:

The sub-pixel edge detection is realized by the Zernike moment method, and the n-order m-th order Zernike moment of the image f(x, y) is defined as:

in, is the orthogonal n-th order m-th order Zernike polynomial in the unit circle in the polar coordinate system; * represents the complex conjugate; A' _nm is the n-th order m-th order Zernike moment after the image is rotated by φ angle;

The sub-pixel edge detection is realized through three different orders of Zernike moments, namely A ₀₀ , A ₁₁ , and A ₂₀ , and the corresponding integral kernel functions are:

According to formulas (3) and (4), we get:

The sub-pixel positions of the edge of the image moving target feature are:

The step S4 specifically includes:

Introduce the target vibration model:

Model fitting was performed using the least squares method.

The step S5 specifically includes:

The image size is in pixels, so there is a quantization error in the virtual sinusoidal linear vibration sequence image obtained by the camera; the display output position s _d (t _i ) and the real position s (t _i ) satisfy the following formula:

s _d (t _i )=s(t _i )+Δs(t _i ) (8)

In order to improve the measurement accuracy, it is necessary to compensate the displacement error of the target vibration model. The formula is as follows:

Δs _c (t _j )=int[s _pp sin(ω _v (N _T t _j )-π/2)]-s _pp sin(ω _v (N _T t _j )-π/2) (9)

Among them, s _pp sin(ω _v (N _T t _j )-π/2) is expressed as: the displacement value corresponding to the vibration model at time t _j ;

According to formulas (8) and (9), the correction value of s _d (t _i ) at any position can be obtained:

s' _d (t _i )=s _d (t _i )-Δs _c (t _j ) (10)

After correcting the target vibration model, the measurement curve of virtual sinusoidal linear vibration with respect to time-displacement is obtained.

The virtual sinusoidal linear vibration measurement method of the present invention has the following advantages:

(1) The present invention improves the accuracy of vibration measurement through image restoration and model correction methods.

(2) The present invention adopts standard virtual sinusoidal linear vibration, which can avoid errors introduced by the shaking table.

(3) The method of the present invention provides a virtual traceability approach for the plane vibration measurement method.

Description of drawings

Fig. 1 is the schematic diagram of the method device of the present invention;

Fig. 2 is a kind of standard virtual sinusoidal linear vibration measurement method flow chart;

Fig. 3 is the flow chart of judging whether the image has the target superposition area;

Fig. 4 is the image restoration method selection flow chart based on the target movement direction;

FIG. 5 is a schematic diagram of the edge position of the moving target feature under different moving directions and different grayscale distributions;

Fig. 6 is the standard virtual sinusoidal linear vibration sequence image collected by the camera;

Fig. 7 is the time-displacement result diagram (frequency is 1Hz) that the method of the present invention measures standard virtual sinusoidal linear vibration;

8-11 are the displacement and phase mean absolute value histogram and the mean square deviation histogram obtained by measuring the standard virtual sinusoidal linear vibration at different frequencies by the method of the present invention.

Detailed ways

In order to avoid the error introduced by the shaking table, the present invention provides a standard virtual sinusoidal linear vibration measurement method. The method of the present invention has high precision for the measurement of plane motion. Detailed Description.

1 is a schematic diagram of an apparatus of the method of the present invention. The apparatus mainly includes: a light source 1 , a camera 2 , and a processing and display device 3 . The light source 1 provides illumination for the camera 2; the camera 2 is used to collect the sequence images of the standard virtual sinusoidal linear vibration; the processing and display device 3 is used to present the standard virtual sinusoidal linear vibration generated by Matlab and realize the processing of the sequence images.

Referring to FIG. 2, it is a flow chart of a standard virtual sinusoidal linear vibration measurement method. The virtual sinusoidal linear vibration measurement method of the present invention mainly comprises the following steps:

Step S160: generate standard virtual sinusoidal linear vibration through Matlab programming;

Step S180: obtaining a standard virtual sinusoidal linear vibration sequence image through a camera;

Step S200: read in the virtual sinusoidal linear vibration sequence image;

Step S220: judging whether the image has a target overlapping area, which includes: detecting a region of interest of a moving object in the image, and judging whether the region of interest has a target overlapping area;

Step S240: Selection of an image restoration method, which includes: judging the movement direction of the moving target, and selecting an image restoration method based on the movement direction;

Step S260: sub-pixel edge detection of the moving target feature edge of the sequence image;

Step S280: Fitting and correcting the target vibration model, which includes: fitting the target vibration model by the least squares method, and compensating the displacement error of the vibration model.

Referring to FIG. 3 , it is a flow chart of judging whether the region of interest has a target overlapping region. The judgment of whether there is a target overlapping area in the region of interest of the present invention includes the following steps:

Step S221: use the inter-frame difference method to obtain the motion area of the target;

Step S221: Perform a connected domain analysis on the motion area to determine the area of interest of the motion target;

Step S223: in the region of interest, perform gradient magnitude calculation on the image;

Step S224: Find the local maximum value of the gradient amplitude;

Step S225: Determine the number of local maxima. When the number of local maxima is greater than 2, it is considered that there is a target overlapping area in the area of interest; otherwise, it is considered that there is no target overlapping area in the area of interest.

Referring to FIG. 4 , it is a flowchart for selecting an image restoration method based on the moving direction of the target. The image restoration method selection based on the target movement direction of the present invention comprises the following steps:

Step S241: use the optical flow method to determine the moving direction of the target in the image: when the moving direction is positive, go to step S242; when the moving direction is negative, go to step S246;

Step S242: Judging the grayscale distribution of the target overlapping area in the image: when the grayscale distribution is three grayscale segments, skip to step S243; when the grayscale distribution is four grayscale segments, skip to step S245;

Step S243: Judging the size of the area in the three-gray-scale segmented area: when the area of the left area is smaller than the area of the right area, skip to step S244; otherwise, skip to step S245;

Step S244: Swap the area of the right area and the area of the left area, so that in the three-gray-level segmented area, the area of the left area is always larger than the area of the right gray-scale area;

Step S245: the gray level of the right area is enhanced, and the gray level of the remaining areas is weakened;

Step S246: Judging the grayscale distribution of the target overlapping area in the image: when the grayscale distribution is three grayscale segments, skip to step S247; when the grayscale distribution is four grayscale segments, skip to step S249;

Step S247: Judging the area size of the three-gray-scale segmented area: when the area of the left area is larger than the area of the right area, skip to step S248; otherwise, skip to step S249;

Step S248: Swap the area of the right area with the area of the left area, so that in the three-gray-level segmented area, the area of the right area is always larger than the area of the left gray-scale area;

Step S249: Enhance the gray level of the left area, and weaken the gray level of the other areas.

Referring to FIG. 5, it is a standard virtual sinusoidal linear vibration sequence image collected by the camera according to the method of the present invention. The specific parameters of the method and device of the invention are: a German AVTManta G-125B industrial camera with a resolution of 1292×964 and a frame rate of 30fps, a lens focal length of 8mm, and a 60W incandescent lamp as the light source.

Referring to FIG. 6 , the time-displacement curve (frequency is 1 Hz) obtained by measuring the standard virtual sinusoidal linear vibration by the method of the present invention. It can be seen from the time-displacement curve that the present invention has high detection accuracy for the moving target feature edge of the standard virtual sinusoidal linear vibration sequence image, and the error between it and the target vibration model obtained by least squares fitting is small.

Referring to FIGS. 7-10 , the displacement and phase mean absolute value histograms and the mean square deviation histograms obtained by measuring the standard virtual sinusoidal linear vibration at different frequencies by the method of the present invention. It can be seen from the bar graph that the present invention has high measurement accuracy for standard virtual sinusoidal linear vibrations of different frequencies.

The above description is a detailed introduction of the process of the present invention, and is not intended to limit the present invention in any form. Those skilled in the art can make a series of improvements, optimizations and modifications on the basis of the present invention. The scope of protection of the present invention should therefore be defined by the appended claims.

Claims (6)

1. a standard virtual sinusoidal linear vibration measurement method, is characterized in that, comprises the following steps: S1: Calculate the grayscale gradient amplitude of the image, use the distribution of the grayscale gradient amplitude to find the peaks in the area, and judge whether there is a moving object superposition in the area of interest of the image according to the number of peaks; S2: For images with overlapping moving objects in the area, the optical flow gradient method is used to determine the moving direction of the moving objects, and different image restoration methods are used to restore the images to improve the accuracy of subsequent image processing; S3: Use the sub-pixel edge detection method to complete the extraction of moving target feature edges; S4: Introduce the target vibration model and use the least squares method for model fitting; S5: Perform displacement error compensation on the target vibration model to obtain a time-displacement measurement curve of virtual sinusoidal linear vibration. 2. a kind of standard virtual sinusoidal line vibration measurement method according to claim 1, is characterized in that: The step S1 specifically includes: (1) Determination of the region of interest of the moving target Use the inter-frame difference method to determine the moving area of the moving target; in order to reduce the influence of noise, perform a connected domain analysis on the obtained moving area; calculate the area of each connected domain, and take the two areas with the largest area as the moving target. area; (2) Judgment of the superimposed area of the moving target In the region of interest of the moving object, calculate the gray gradient magnitude:

Among them, f(x, y) is the gray value at the pixel (x, y); I _x and I _y are the gray gradients in the x and y directions, respectively;

is the gradient vector at the pixel (x, y); is the amplitude corresponding to the gradient vector; when there are no moving objects superimposed in the area, there are only two local maximum points in the gradient amplitude of the traversed area; when there are moving objects superimposed in the area, the number of local maximum points is greater than two . 3. a kind of standard virtual sinusoidal line vibration measurement method according to claim 1, is characterized in that: The step S2 specifically includes: (1) Determination of the moving direction of the target The optical flow gradient method is used to calculate the optical flow of the region of interest of the moving target. The algorithm assumes that the brightness of the corresponding pixels of two adjacent frames of images is constant. The calculation formula is as follows: I _x u+I _y v+I _t =0 (2) Among them, I _t is the grayscale gradient of the image at time t; the optical flow u in the x direction and the optical flow v in the y direction are used to determine the moving direction of the target; (2) Image restoration method For images with strong gray contrast between the background and the moving target, the moving target is a solid rectangle; it is assumed that the moving target feature edge is always its right edge, and when the x-coordinate of the moving target feature edge in adjacent frame images shows an increasing trend, it is considered that the target The movement direction is the positive direction; the gray area is divided into the edge area of the target feature; the gray level of the rightmost area is enhanced for the region of interest with the positive movement direction, and the gray level of the remaining areas is weakened; for the area of interest with the negative movement direction The gray level of the leftmost region is enhanced, and the gray level of the remaining regions is weakened. 4. a kind of standard virtual sinusoidal line vibration measurement method according to claim 1, is characterized in that: The step S3 specifically includes: The sub-pixel edge detection is realized by the Zernike moment method, and the n-order m-th Zernike moment of the image f(x, y) is defined as:

in, is the orthogonal n-th order m-th order Zernike polynomial in the unit circle in the polar coordinate system; * represents the complex conjugate; A' _nm is the n-th order m-th order Zernike moment after the image is rotated by φ angle; The sub-pixel edge detection is realized through three different orders of Zernike moments, namely A ₀₀ , A ₁₁ , and A ₂₀ , and the corresponding integral kernel functions are: According to formulas (3) and (4), we get:

The sub-pixel positions of the edge of the image moving target feature are: 5. a kind of standard virtual sinusoidal line vibration measurement method according to claim 1, is characterized in that: The step S4 specifically includes: Introduce the target vibration model:

Model fitting was performed using the least squares method. 6. a kind of standard virtual sinusoidal line vibration measurement method according to claim 1 is characterized in that: The step S5 specifically includes: The image size is in pixels, so there is a quantization error in the virtual sinusoidal linear vibration sequence image obtained by the camera; the display output position s _d (t _i ) and the real position s (t _i ) satisfy the following formula: s _d (t _i )=s(t _i )+Δs(t _i ) (8) In order to improve the measurement accuracy, it is necessary to compensate the displacement error of the target vibration model. The formula is as follows: Δs _c (t _j )=int[s _pp sin(ω _v (N _T t _j )-π/2)]-s _pp sin(ω _v (N _T t _j )-π/2) (9) Among them, s _pp sin(ω _v (N _T t _j )-π/2) is expressed as: the displacement value corresponding to the vibration model at time t _j ; According to formulas (8) and (9), the correction value of s _d (t _i ) at any position can be obtained: s' _d (t _i )=s _d (t _i )-Δs _c (t _j ) (10) After correcting the target vibration model, the measurement curve of virtual sinusoidal linear vibration with respect to time-displacement is obtained.

Images