CN114526694A - Method for measuring metal surface roughness - Google Patents

Abstract

A method of metal surface roughness measurement comprising the steps of: scanning the surface of the metal to be detected by adopting a laser structured light 3D vision system; when the scanning plane of the 3D vision system is not parallel to the surface of the measured metal, the 3D vision system levels the surface of the measured metal and obtains a leveling image; the 3D vision system scans and images the surface of the metal to be detected to obtain the 3D appearance of the surface of the metal to be detected; and measuring the surface roughness of the metal to be measured by utilizing a plurality of image processing algorithms. The method adopts the laser structured light 3D vision system to scan and image the metal surface to obtain the 3D morphology of the metal surface, and then utilizes an image processing algorithm to automatically measure the roughness of the metal surface, so that the method has higher real-time performance, increases the measurement reliability, and can realize the full automation of large-scale construction surface roughness detection.

Description

A kind of measuring method of metal surface roughness

technical field

The invention relates to a cross technology of mechanical engineering and machine vision, in particular to a method for measuring metal surface roughness by using laser structured light vision.

Background technique

In the production process of large-scale equipment such as ships, nuclear industry, heavy machinery, and vehicles, metal materials are a commonly used structural material. Among them, ferrous metals, especially steel, are used in a large amount. During the service process of these components, it is inevitable Problems such as corrosion will occur, which will lead to the reduction of the service life of the components, the reduction of the reliability of use and the reduction of the external aesthetics. Spraying the surface is a common method to solve the corrosion problem of ferrous metal surfaces. In order to improve the bonding between the sprayed material and the metal surface, the metal surface is often shot and sandblasted to improve the roughness of the metal surface and remove surface oxidation. impurities and other impurities.

After the metal surface is shot and sandblasted, the measurement of surface roughness is an important factor in determining the quality of spraying. Therefore, the measurement of surface roughness is an important measure of the quality of shot and sandblasting. At present, the main methods used for surface roughness measurement include visual inspection of comparative samples, microscope focusing method, stylus method and replication tape method. , there are problems such as low measurement efficiency, poor measurement objectivity, and low measurement accuracy.

In addition, the shot blasting and sand blasting environments are relatively harsh environments, and operators also have various unhealthy and dangerous factors during the inspection process. Therefore, it is an urgent problem to realize the automation of roughness measurement. In recent years, machine vision technology has made great progress, and the application of machine vision technology to automatic roughness measurement will help to improve the degree of automation and accuracy of this operation.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a method for measuring metal surface roughness using laser structured light vision for the above-mentioned defects of the prior art.

In order to achieve the above object, the present invention provides a method for measuring metal surface roughness, which includes the following steps:

S100, using a laser structured light 3D vision system to scan the surface of the metal to be tested;

S200. When the scanning plane of the 3D vision system is not parallel to the surface of the metal to be measured, the 3D vision system levels the surface of the metal to be measured and obtains a leveling image;

S300, the 3D vision system scans and images the surface of the metal to be tested, and obtains the surface 3D topography of the metal to be tested; and

S400, using a variety of image processing algorithms to measure the surface roughness of the metal to be tested.

In the above method for measuring metal surface roughness, in step S200, the leveling image is obtained by using a leveling algorithm, and the leveling algorithm includes the following steps:

S201, smooth the original image _Isr obtained by scanning the measured metal surface by the 3D vision system to obtain a smoothed image _Ism ;

S202, subtracting the original image _Isr and the smoothed image _Ism to obtain a subtracted image _Isu ; and

S203. Add the average value of the subtracted image I _su and the original image _{Is sr} to obtain the leveled image I _sl .

In the above method for measuring metal surface roughness, the various image processing algorithms in step S400 include:

S410, a depth image missing information compensation algorithm;

S420, a depth image noise removal algorithm;

S430. Specify the line segment and area roughness measurement algorithm;

S440. Improve the anti-jamming watershed segmentation algorithm; and

S450. Statistical algorithm of sand and shot blasting bunker parameters.

The above-mentioned method for measuring metal surface roughness, wherein, the depth image missing information compensation algorithm includes the following steps:

S411, perform a closing operation on the leveling image I _sl with a morphological operator to obtain a closing operation image I _sc ;

S412, find the missing point of the original image I _sr of the tested metal, the value of the missing point is zero, and obtain the binary image I _zb of the missing area;

S413, perform point multiplication with the binary image I _zb of the missing area and the closed operation image I _sc to obtain the replacement image I _zr of the missing area; and

S414. Add the complemented image I _zr of the missing area to the original image _Isr to obtain the complemented image I _re .

The above method for measuring metal surface roughness, wherein, the depth image noise removal algorithm includes the following steps:

S421, carry out Laplacian filtering to the described supplementary image I _re , obtain high-frequency image I _lp ;

S422, seek absolute value to described high frequency image _Ilp , obtain absolute high frequency image _Ia1 ;

_S423 , carry out threshold segmentation to the described absolute high frequency image I a1, obtain high frequency binary image I _hb and the complementary low frequency binary image I _1b of the high frequency binary image I _hb ;

S424, carry out median filtering to the described supplementary image I _re , obtain the median image I _me ;

S425, perform point multiplication with the high-frequency binary image I _hb and the median image I _me to obtain a high-frequency smoothing image I _hs ;

S426, using the low-frequency binary image I _1b and the described complementary image I _re point multiplication, obtains the low-frequency complementary image I _1s ; And

S427. Add the high-frequency smoothed image I _hs and the low-frequency compensated image I _ls to obtain a denoised image I _dn .

The above-mentioned method for measuring metal surface roughness, wherein, the specified line segment and area roughness measuring algorithm includes:

S431, an area roughness measurement algorithm; and

S432, a measurement algorithm for line roughness.

The above-mentioned method for measuring metal surface roughness, wherein, the measuring algorithm of the area roughness includes:

S4311. Set a measurement area for roughness measurement to measure regional roughness; and

S4312. Find the highest point and the lowest point in the area of the measurement area, and obtain the difference between the two values of the highest point and the lowest point.

The above-mentioned method for measuring metal surface roughness, wherein the measurement algorithm of the line roughness includes:

S4321. Set a measurement line for roughness measurement to measure line roughness;

S4322, obtain the height value on each pixel of the measurement line;

S4323. Divide the line segment into n segments according to the roughness measurement requirements;

S4324, find the highest value and the lowest value in each of the n segments, and find the difference between the highest value and the lowest value; and

S4325. Calculate the average value of the n-stage difference to obtain the line roughness.

The above-mentioned method for measuring metal surface roughness, wherein, the improved anti-interference watershed segmentation algorithm is used to realize the segmentation of the surface pits of the metal to be measured, including the following steps:

S441, smoothing the denoised image I _dn after denoising to obtain a smoothed image I _sm ;

S442, subtracting the denoised image I _dn and the smoothed image I _sm to obtain a boundary image I _ed ;

S443, adding the denoised image I _dn and the boundary image I _ed to obtain an enhanced image I _eh ;

S444, perform a closing operation on the enhanced image I _eh to obtain a closed image I _cl ; and

S445, performing a watershed algorithm on the closed image I _cl to obtain a pit segmentation image.

The above-mentioned method for measuring the roughness of metal surfaces, wherein, the statistical algorithm for the parameters of the sand blasting sand pit includes the following steps:

S451, obtain the pixel marking map of each pit, count the number of pixels of the pit on the pixel marking map, and obtain the pit area according to the proportional relationship; and

S452: Obtain the boundary of the pit on the pixel marker map, and obtain the average height of the boundary of the pit, the deepest value of the pit, and the difference between the deepest value of the pit and the average height of the boundary.

The technical effect of the present invention is:

The invention adopts a laser structured light 3D vision system to scan and image the metal surface to obtain the 3D topography of the metal surface, and then uses an image processing algorithm to automatically measure the roughness of the metal surface, which at least has the following advantages:

1) Using the laser mechanism optical 3D vision to obtain the roughness information of the metal surface can provide reliable information for subsequent roughness evaluation and analysis, and at the same time provide the possibility of long-term preservation of information. The 3D vision mechanism cooperates with the mobile robot to further Realize full automation of surface roughness inspection of large-scale construction;

2) Using the leveling algorithm based on digital images, it is difficult to ensure the parallel relationship between the trajectory of the scanning movement of the scanning mechanism and the scanned plane in the process of laser structured light 3D vision in the prior art, which seriously affects the follow-up. The problem of roughness evaluation;

3) The missing signal compensation algorithm has wide scale and shape adaptability, which solves the problem of local laser structured light depth image caused by reflection of metal surface and occlusion of convex parts in the laser structured light vision in the prior art. The signal is missing, and the size and shape of the missing signal area are indeterminate. The missing signal brings serious interference to the subsequent processing, and at the same time solves the problem of the reliability of the signal replenishment;

4) The noise removal algorithm based on the combination of Laplacian operator and median filter is insensitive to noise scale and shape, and can reasonably estimate the original information of noise parts; Good operability, with parallel computing hardware, high real-time performance;

5) The roughness measurement on the basis of the depth image can avoid the accidental interference of human operation in methods such as the visual inspection method of the comparative sample, the microscope focusing method, the stylus method and the replica belt method, etc., and the roughness measurement can be easily determined. The target area of algorithms and statistical algorithms; at the same time, it can effectively increase statistical points, thereby increasing the reliability of statistics;

6) The improved watershed algorithm is adopted, which effectively avoids the over-segmentation problem caused by the tiny depression in the prior art, and can accurately segment each bunker;

7) Statistical analysis of sand pits on digital image technology can obtain information such as shot peening, sand blasting pit area, area distribution, pit depth and depth distribution, pit shape, etc. The manufacturability of sandblasting provides an evaluation basis, and also provides more abundant information for the correlation research between surface treatment quality and spraying quality.

The present invention is described in detail below with reference to the accompanying drawings and specific embodiments, but is not intended to limit the present invention.

Description of drawings

FIG. 1 is a flow chart of a method according to an embodiment of the present invention;

2A-2D are renderings of image processing effects of metal scale roughness according to an embodiment of the present invention;

3A-3B are effect diagrams of over-segmented images by an improved watershed algorithm according to an embodiment of the present invention;

FIG. 4 is a histogram of the pit area of the sand blasting and shot blasting sand pit parameter statistics according to an embodiment of the present invention;

FIG. 5 is a result diagram of a roughness measurement algorithm according to an embodiment of the present invention.

Detailed ways

Below in conjunction with accompanying drawing, structure principle and working principle of the present invention are described in detail:

In the machine vision technology of the prior art, there are many ways of 3D vision, but relatively speaking, laser structured light vision is a 3D vision method with high precision and good reliability. Based on this background, the present invention adopts laser structured light. It scans the metal surface visually, and realizes the measurement of metal surface roughness through the algorithm of many key links from the depth signal to the roughness measurement.

The method for measuring metal surface roughness of the present invention adopts a laser structured light 3D vision system. First, the vision system is used to level the measured plane according to a leveling algorithm, and then the metal surface is scanned and imaged to obtain the 3D shape of the metal surface. Then, the roughness of the metal surface is automatically measured using a variety of image processing algorithms.

Referring to FIG. 1, FIG. 1 is a flowchart of a method according to an embodiment of the present invention. The method for measuring the metal surface roughness of the present embodiment includes the following steps:

Step S100, using a laser structured light 3D vision system to scan the surface of the metal to be tested;

Step S200, when the scanning plane of the 3D vision system is not parallel to the surface of the metal to be measured, the 3D vision system levels the surface of the metal to be measured and obtains a leveling image;

Step S300, the 3D vision system scans and images the surface of the metal to be tested, and obtains the surface 3D topography of the metal to be tested; and

Step S400, using a variety of image processing algorithms to measure the surface roughness of the metal to be measured.

In step S200, the leveling image is obtained by a leveling algorithm, which has a leveling function when the scanning plane is not parallel to the measured plane, and the leveling algorithm includes the following steps:

Step S201, smoothing the original image _Isr obtained by scanning the measured metal surface by the 3D vision system to obtain a smoothed image _Ism ;

Step S202, subtracting the original image _Isr and the smoothed image _Ism to obtain a subtracted image _Isu ; and

Step S203 , adding the average value of the subtracted image I _su and the original image _{Is sr} to obtain the leveled image I _sl .

The multiple image processing algorithms in step S400 of this embodiment include five image processing algorithms, specifically including:

Step S410, a depth image missing information compensation algorithm;

Step S420, a depth image noise removal algorithm;

Step S430, specifying a line segment and a region roughness measurement algorithm;

Step S440, improving the anti-interference watershed segmentation algorithm; and

Step S450, a statistical algorithm for sand blasting and shot blasting bunker parameters.

Wherein, in step S410, the depth image missing information compensation algorithm is a missing information compensation algorithm for laser structured light vision, including the following steps:

Step S411, performing a closing operation on the leveling image I _sl with a morphological operator to obtain a closing operation image I _sc ;

Step S412, searching for the missing point of the original image I _sr of the tested metal, the value of the missing point is zero, and obtaining the binary image I _zb of the missing area;

Step S413, performing dot multiplication with the binary image I _zb of the missing area and the closed operation image I _sc to obtain the replacement image I _zr of the missing area; and

Step S414 , adding the complemented image I _zr of the missing area to the original image _Isr to obtain the complemented image I _re .

In step S420, the depth image noise removal algorithm includes the following steps:

Step S421, performing Laplacian filtering on the complementary image I _re to obtain a high-frequency image I _lp ;

Step S422, the absolute value of the high-frequency image _Ilp is obtained to obtain the absolute high-frequency image _Ial ;

Step _S423 , performing threshold segmentation on the absolute high-frequency image I a1 , obtaining the high-frequency binary image I _hb and the complementary low-frequency binary image I _lb of the high-frequency binary image I _hb ;

Step S424, performing median filtering on the complemented image I _re to obtain the median image I _me ;

Step S425, performing point multiplication with the high-frequency binary image I _hb and the median image I _me to obtain a high-frequency smoothing image I _hs ;

Step S426, using the low-frequency binary image I _1b and the complementary image I _re point multiplication, obtains the low-frequency complementary image I _s ; And

Step S427, adding the high-frequency smoothed image I _hs and the low-frequency complementary image I _ls to obtain a denoised image I _dn .

In step S430, the specified line segment and region roughness measurement algorithm includes two kinds of roughness measurement algorithms, which are respectively a region roughness measurement algorithm and a line roughness measurement algorithm, namely:

Step S431, area rough measurement algorithm; and

Step S432, a measurement algorithm for line roughness.

The area roughness measurement algorithm in step S431 includes:

In step S4311, a measurement area for roughness measurement can be set by itself, so as to measure the regional roughness; and

Step S4312, find the highest point and the lowest point in the area of the measurement area, and find the difference between the two values of the highest point and the lowest point.

The line roughness measurement algorithm in step S432 includes:

In step S4321, a measurement line for roughness measurement can be set by itself, so as to measure the line roughness;

Step S4322, obtaining the height value on each pixel on the measurement line;

Step S4323, dividing the line segment into n segments according to the roughness measurement requirements;

Step S4324, finding the highest value and the lowest value in each of the n segments, and finding the difference between the highest value and the lowest value; and

Step S4325: Calculate the average value of the n-stage difference values to obtain the line roughness.

In step S440, the improved anti-interference watershed segmentation algorithm is used to realize the segmentation of the surface pits of the tested metal, including the following steps:

Step S441, smoothing the denoised image I _dn after denoising to obtain a smoothed image I _sm ;

Step S442, subtracting the denoised image I _dn and the smoothed image I _sm to obtain a boundary image I _ed ;

Step S443, adding the denoised image I _dn and the boundary image I _ed to obtain an enhanced image I _eh ;

Step S444, performing a closing operation on the enhanced image I _eh to obtain a closed image I _cl ; and

Step S445, performing a watershed algorithm on the closed image I _cl to obtain a pit segmentation image.

In step S450, the statistical algorithm of the parameters of the sand blasting and shot blasting can realize the statistical analysis of the pit parameters on the basis of the pit segmentation, and the statistical parameters include the area of the pit, the depth of the pit, etc., including the following steps :

Step S451, obtain the pixel marking map of each pit, count the number of pixels of the pit on the pixel marking map, and obtain the pit area according to the proportional relationship; and

Step S452, obtain the boundary of the pit on the pixel marking map, and obtain the average value of the boundary height of the pit, the deepest value of the pit and the difference between the deepest value of the pit and the average value of the boundary height .

The invention adopts a laser structured light 3D vision system, firstly, the vision system is used to level the measured plane according to the leveling algorithm, and then the metal surface is scanned and imaged to obtain the 3D topography of the metal surface, and then a variety of image processing algorithms are used. , which automatically measures the roughness of metal surfaces. Its working principle is:

As shown in Figures 2A-2D, Figure 2A is the original image. It can be seen that the upper half of the original image is darker, while the lower half is brighter. Figure 2B is the original image after digital leveling. You can see It can be seen that the brightness of the upper and lower half of the image has been relatively balanced, indicating that the surface has no large-scale inclination; Figure 2C is the depth map of the metal surface restored by the missing signal compensation algorithm, and the signal missing area in Figure 2B ( Black area), which has been completely filled in in Figure 2C, and the remaining black areas are noise points; Figure 2D is the image obtained by the denoising algorithm, and the black noise areas have been completely removed.

Figures 3A-3B are the pit images segmented by the watershed algorithm, wherein Figure 3A is the pit image segmented by the general watershed algorithm, and Figure 3B is the pit image segmented by the improved watershed algorithm. It can be seen that Figure 3A produces excessive segmentation, and the effect of the segmentation in Figure 3B is much better than that of the segmentation in Figure 3A.

Fig. 4 is the result after the histogram analysis is performed on the basis of the pit extraction, and the present invention can easily obtain its statistical distribution. In the figure, the horizontal axis is the area of the pit, where the unit of the area is the number of pixels occupied by a single pit, every 2000 is an interval, and the vertical axis is the number of pits that appear in the area.

Figure 5 shows the roughness measurements obtained according to the standard. The horizontal axis in the figure is the pixel sequence on the measurement line segment, which is divided into 5 segments, each segment corresponds to an actual length of 2.5mm, separated by a vertical dotted line, and the vertical axis is the height of each measurement point on the line segment, representing the metal surface The height of , the unit is μm, the highest point and the lowest point are obtained in each segment, the highest point is represented by *, and the lowest point is represented by o. The final roughness is calculated by first finding the difference between the highest point and the lowest point of each segment, and then finding the average of five differences.

The invention uses laser structured light vision to measure the metal surface roughness, and uses a laser structured light 3D vision system to scan and image the metal surface to obtain the 3D topography of the metal surface, and then use an image processing algorithm to measure the roughness of the metal surface. Take measurements. It also has a leveling function when the scanning plane is not parallel to the measured plane. The image processing algorithm includes roughness measurement algorithm in the case of setting area or line segment; compensation algorithm for missing signal of laser structured light depth signal; denoising algorithm for laser structured light depth signal; improved watershed algorithm, It can realize the segmentation of surface pits in the case of complex interference; and on the basis of pit segmentation, the statistical algorithm of pit parameters. Surface detection in the prior art mainly focuses on roughness. On the basis of roughness detection, the present invention adds the analysis of shot peening and sandblasted surface pits and the corresponding image processing algorithm for subsequent surface quality evaluation and process correlation analysis. Provides richer information.

Of course, the present invention can also have other various embodiments, without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding Changes and deformations should belong to the protection scope of the appended claims of the present invention.

Claims (10)

1. a method for measuring metal surface roughness, is characterized in that, comprises the steps: S100, using a laser structured light 3D vision system to scan the surface of the metal to be tested; S200. When the scanning plane of the 3D vision system is not parallel to the surface of the metal to be measured, the 3D vision system levels the surface of the metal to be measured and obtains a leveling image; S300, the 3D vision system scans and images the surface of the metal to be tested, and obtains the surface 3D topography of the metal to be tested; and S400, using a variety of image processing algorithms to measure the surface roughness of the metal to be tested. 2 . The method for measuring metal surface roughness according to claim 1 , wherein, in step S200 , the leveling image is obtained by a leveling algorithm, and the leveling algorithm comprises the following steps: 3 . S201, smooth the original image _Isr obtained by scanning the measured metal surface by the 3D vision system to obtain a smoothed image _Ism ; S202, subtracting the original image _Isr and the smoothed image _Ism to obtain a subtracted image _Isu ; and S203. Add the average value of the subtracted image I _su and the original image _{Is sr} to obtain the leveled image I _sl . 3. The method for measuring metal surface roughness according to claim 1 or 2, wherein the multiple image processing algorithms in step S400 include: S410, a depth image missing information compensation algorithm; S420, a depth image noise removal algorithm; S430. Specify the line segment and area roughness measurement algorithm; S440. Improve the anti-jamming watershed segmentation algorithm; and S450. Statistical algorithm of sand and shot blasting bunker parameters. 4. The method for measuring metal surface roughness according to claim 3, wherein the algorithm for complementing missing information of the depth image comprises the following steps: S411, perform a closing operation on the leveling image I _sl with a morphological operator to obtain a closing operation image I _sc ; S412, find the missing point of the original image I _sr of the tested metal, the value of the missing point is zero, and obtain the binary image I _zb of the missing area; S413, perform point multiplication with the binary image I _zb of the missing area and the closed operation image I _sc to obtain the replacement image I _zr of the missing area; and S414. Add the complemented image I _zr of the missing area to the original image _Isr to obtain the complemented image I _re . 5. The method for measuring metal surface roughness according to claim 4, wherein the depth image noise removal algorithm comprises the following steps: S421, carry out Laplacian filtering to the described supplementary image I _re , obtain high-frequency image I _lp ; S422, seek absolute value to described high frequency image _Ilp , obtain absolute high frequency image _Ia1 ; _S423 , carry out threshold segmentation to the described absolute high frequency image I a1, obtain high frequency binary image I _hb and the complementary low frequency binary image I _1b of the high frequency binary image I _hb ; S424, carry out median filtering to the described supplementary image I _re , obtain the median image I _me ; S425, perform point multiplication with the high-frequency binary image I _hb and the median image I _me to obtain a high-frequency smoothing image I _hs ; S426, using the low-frequency binary image I _1b and the described complementary image I _re point multiplication, obtains the low-frequency complementary image I _1s ; And S427. Add the high-frequency smoothed image I _hs and the low-frequency compensated image I _ls to obtain a denoised image I _dn . 6. The method for measuring metal surface roughness according to claim 3, wherein the specified line segment and area roughness measuring algorithm comprises: S431, an area roughness measurement algorithm; and S432, a measurement algorithm for line roughness. 7. The method for measuring metal surface roughness according to claim 6, wherein the measurement algorithm for the area roughness comprises: S4311. Set a measurement area for roughness measurement to measure regional roughness; and S4312. Find the highest point and the lowest point in the area of the measurement area, and obtain the difference between the two values of the highest point and the lowest point. 8. The method for measuring metal surface roughness according to claim 6, wherein the measurement algorithm for the line roughness comprises: S4321. Set a measurement line for roughness measurement to measure line roughness; S4322, obtain the height value on each pixel of the measurement line; S4323. Divide the line segment into n segments according to the roughness measurement requirements; S4324, find the highest value and the lowest value in each of the n segments, and find the difference between the highest value and the lowest value; and S4325. Calculate the average value of the n-stage difference to obtain the line roughness. 9. The method for measuring metal surface roughness according to claim 5, wherein the improved anti-interference watershed segmentation algorithm is used to realize the segmentation of the surface pits of the measured metal, comprising the steps of: S441, smoothing the denoised image I _dn after denoising to obtain a smoothed image I _sm ; S442, subtracting the denoised image I _dn and the smoothed image I _sm to obtain a boundary image I _ed ; S443, adding the denoised image I _dn and the boundary image I _ed to obtain an enhanced image I _eh ; S444, perform a closing operation on the enhanced image I _eh to obtain a closed image I _cl ; and S445, performing a watershed algorithm on the closed image I _cl to obtain a pit segmentation image. 10. The method for measuring metal surface roughness according to claim 3, wherein the statistical algorithm for the parameters of the sand and shot blasting bunker comprises the following steps: S451, obtain the pixel marking map of each pit, count the number of pixels of the pit on the pixel marking map, and obtain the pit area according to the proportional relationship; and S452: Obtain the boundary of the pit on the pixel marker map, and obtain the average height of the boundary of the pit, the deepest value of the pit, and the difference between the deepest value of the pit and the average height of the boundary.

Images