CN112037126A

CN112037126A - Image synthesis method for detecting object surface crack defects based on laser scanning method

Info

Publication number: CN112037126A
Application number: CN202010632157.1A
Authority: CN
Inventors: 张凯; 梁荣; 陈力
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2020-07-02
Filing date: 2020-07-02
Publication date: 2020-12-04
Anticipated expiration: 2040-07-02
Also published as: CN112037126B

Abstract

The invention relates to an image synthesis method for detecting a structural member with a complex surface appearance by adopting a laser scanning thermal imaging nondestructive testing technology. In the invention, the surface of the structural member is uniformly thermally excited by adopting line laser scanning, and the surface change temperature of the structural member in the laser scanning process is recorded in real time by a thermal imager. Aiming at the phenomena of gradual change of a laser curve, step sudden change and over-temperature points outside the laser curve in an image caused by the change of the surface topography of an object, the three phenomena are distinguished by a distance-setting squat and peak value judging method, and a segment fitting method is adopted to obtain the correct laser curve position. And extracting and interpolating according to the required cooling time after excitation to obtain the gray value of the curve with the fixed interval with the laser curve, and splicing into new images in sequence. The invention not only can compensate the excitation time delay difference of each position of the object surface in a single frame image collected by the thermal imager, but also realizes the correct synthesis of the image under the condition that the object surface has complex morphology.

Description

Image synthesis method for detecting object surface crack defects based on laser scanning method

Technical Field

The invention relates to an image processing method in a laser scanning thermal imaging detection process, which is used for synthesizing a crack defect detection image on the surface of a workpiece with a complex morphology and belongs to the technical field of infrared nondestructive detection.

Background

Taking line laser scanning detection as an example, the detection process is line laser in generalThe light scans from one end of the surface of the object to the other end, the thermal imager continuously collects the surface temperature of the object in the scanning process, and the thermal imager collects corresponding sequence images after the scanning is finished. As shown in fig. 2, if the laser scanning speed is constant, and thus the collected images are browsed frame by frame, it can be seen that the laser line gradually moves from one end of the object surface to the other end at fixed intervals, and the interval Δ x of each spatial movement is the product of the time Δ t of collecting one frame of image by the thermal imager and the laser scanning speed Δ v. Taking any frame of image from the sequence image, setting the position of the laser line as x, and assuming that the object surface area corresponding to the position x in the image is in the excitation of the laser line, the current moment is t₀. Then from the above analysis it can be seen that the laser line moves by a distance of deltax, i.e. t, in the next image frame₀The laser line should be located at x + Δ x at time + Δ t. This represents t₀The object surface region corresponding to the x + Δ x position at the + Δ t time is under laser excitation, and the region corresponding to the x position at this time is the time cooled by Δ t after laser excitation. Therefore, in a single image, different positions on the surface of the object have different laser scanning delay times, and the state of the surface at the same laser scanning delay time cannot be simultaneously represented. If the same laser scanning delay time is required at each position of the surface of the object in the single-frame image, as in the case of adopting the excitation of the whole frame, the pixels in each frame of the image of the sequence image need to be recombined according to the delay difference of the laser scanning, so that each pixel point in the recombined image has the same laser scanning delay. The common method is that pixel point rows with the same time delay are extracted from each frame of image frame by frame, and then recombined in sequence to form a new image, and each point in the image after recombination has the same laser scanning time delay.

If the surface of the object is flat and the laser is parallel to the thermal imager, the laser line in the image collected by the thermal imager is a vertical straight line (the laser line is assumed to be vertical to the thermal imager). However, in the actual detection process, the surface of the object usually has a complex morphology, and the image acquired by the thermal imager may have gradual change in the form of laser curve inclination and bending, abrupt change in the form of step jump deformation, and the phenomenon of an over-temperature point on the surface of the object outside the laser curve. The step jump deformation type change and the over-temperature point phenomenon are both expressed as position mutation when the maximum gray value is searched in a traversing manner. Fig. 3 is a schematic diagram of laser curve variation under different morphologies of the object surface, wherein (a) represents the field of view of the laser vertical thermal imager, (b) represents the inclination and bending of the laser curve, (c) represents the step jump of the laser curve, and (d) represents the existence of an over-temperature point on the object surface.

The thermal imager is used as a focal plane detector, the structure of the thermal imager is a two-dimensional area array formed by a large number of detection units, and the output of the thermal imager is a planar image with pixel points as units, so that the minimum resolution of the thermal imager is the size of an area corresponding to a single pixel point. When the laser curve changes, the situation that the laser curve stretches across two pixel points appears in an image collected by the thermal imager, and the gray value of the middle position of the two pixel points cannot be extracted only by taking the pixel points as the minimum unit in image extraction. At this time, the gray value of the extracted pixel is probably not the true maximum value, and the horizontal stripes also appear in the synthesized image.

As shown in fig. 4, for the case of vertical and gradual change of the laser line, fitting and interpolation methods are usually used to obtain the actual laser curve position and its gray value. The method comprises the following specific steps:

1. and taking out single-frame images in the sequence images, traversing the images row by row to respectively take out the maximum gray value position points of the row, and taking out the maximum gray value position point rows corresponding to the image rows after traversing. The curve obtained by fitting with the curve as a parameter can be represented as an actual laser curve.

2. And extracting a position point row with a distance x from the position of the laser line in the image, and calculating the actual gray value of the curve position by using an interpolation algorithm, namely, the gray value at the moment when t is x/delta v after laser excitation.

3. And extracting pixel point rows which are far away from the laser line x frame by frame in the same step, and sequentially arranging and splicing the pixel point rows into an image which can represent a temperature distribution image with the same time delay t at each position on the surface of the object.

When the surface of an object has a step mutation condition, the method is difficult to accurately fit the position of the laser peak curve. Due to the smoothing effect of the fitting, errors can occur between the fitted curve and the actual laser position at the jump of the laser curve. One possible approach is to segment the abrupt laser curve segment from the main curve segment, each segment performing the fitting separately. The method not only avoids the error generated by fitting at the jump position, but also realizes the simultaneous delay display of two end faces with sudden change on the surface of the object.

When the over-temperature point exists on the surface of the object, the existence of the over-temperature point can cause the confusion of the judgment of the maximum value position. For example, a small piece of black paint is sprayed on an iron block with a smooth surface. Because the black paint absorption rate is greater than the iron block absorption rate, in a few frames of images of the black paint area just scanned by laser, the gray value of the black paint area is greater than that of the laser curve, and thus an over-temperature point appears. Due to the existence of the over-temperature points, two gray value peak values can appear in the pixels of the single row, wherein one gray value peak value is the position of the laser curve, the other gray value peak value is the position of the over-temperature point, and the gray value of the position of the over-temperature point is larger than the gray value of the position of. The maximum value taken after traversal is the gray value of the over-temperature point and not the gray value of the laser curve. One possible method is to perform line-by-line pixel fitting in the lateral direction in a small-range fitting manner, compare the line-by-line pixel fitting with the peak position of the previous line, and extract the peak point with a small distance, which is the laser curve position.

In actual detection, due to uncertainty of the surface topography of an object, one or more phenomena described above may occur simultaneously in an image acquired by a thermal imager, which results in diversity of fitting error forms. However, any one of the fitting methods is not suitable for different situations to fit a correct laser curve, so that the laser scanning infrared detection method is greatly limited by the surface topography of the object to be detected.

Disclosure of Invention

Aiming at the defects of the image synthesis technology, the invention provides an image synthesis technology for detecting objects with complex surface morphology by laser scanning infrared imaging. The invention can actively distinguish the phenomena of gradual change, jump change and over-temperature point of the laser curve, and split the complex situation, thus simplifying the process. And aiming at different phenomena, different fitting methods are adopted, so that the correct laser curve position can be fitted.

And distinguishing gradual change and abrupt change of the position point of the maximum gray value in the image by setting a distance criterion delta p, and judging whether a laser curve in the image has step jump or an over-temperature point. The magnitude of the distance criterion Δ p varies according to the change in the topography of the object surface, and the more severe the object surface is tilted or curved, the greater the magnitude of the criterion should be.

And traversing each row of pixel points line by line from top to bottom in the image to find the position point of the maximum gray value, and comparing the position point obtained by traversing with the position point of the previous row from the second row. And if the distance is larger than the distance criterion delta p, judging that the position of the maximum gray value is suddenly changed, and recording the row position of the suddenly changed pixel. Fitting the mutation pixel line, and if only one peak point exists in a fitting curve, indicating that the mutation is caused by the adjustment and hopping of the laser curve; if two or more peak points exist in the fitting curve, the over-temperature point exists at the position. The laser curve in the image can be distinguished from the over-temperature point according to the method. Using different fitting methods for different phenomena

In actual detection, the method can be combined to effectively compensate laser curve fitting errors caused by the surface appearance change of the object, so that temperature distribution images with the same time delay at all positions of the surface of the object are correctly spliced.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

FIG. 1 is a schematic structural diagram of a laser scanning thermal imaging detection system;

FIG. 2 is a graph of the temperature distribution of the surface of a laser scanning excited object;

FIG. 3 is a schematic diagram of a laser scanning peak curve under a complex topography;

FIG. 4 is a schematic diagram of the steps for executing image reconstruction under the conditions of vertical and gradual change of laser lines;

fig. 5 is a schematic diagram illustrating an implementation procedure of the image reconstruction method according to the present patent.

Detailed Description

The invention is further illustrated by the following specific figures and examples.

Fig. 1 is a schematic diagram of a laser scanning thermal imaging detection system, which includes an upper computer 1, a thermal imager 2, a light source 3, a galvanometer 4, a platform 5 and an object 6 to be detected. The upper computer 1 has the following functions: controlling the on-off of the light source 3 and the intensity of the output light; controlling the galvanometer 4 to realize laser uniform scanning; the thermal imager 3 is controlled and receives image data, and at the same time, post image processing can be performed. The image synthesis technology is applied to the post-processing process of the upper computer image. The light source 3 part provides a line laser light source for the detection system, and the length of the line laser is assumed to be larger than the field of view of the thermal imager, and the width of the line laser is extremely small and not more than the field of view of a single pixel of the thermal imager. The thermal imager 2 is used for recording the surface change temperature of an object in the laser excitation process, the resolution of the thermal imager is X multiplied by Y, the acquisition frame frequency is f, and the time delta t for acquiring a single-frame image is 1/f. And assuming that the single detection time length is T, the thermal imager can acquire N-T/delta T images after detection. The upper computer controls the variable-speed rotation of the galvanometer 4 to realize the uniform scanning of the linear laser on the surface of the sample, and if the scanning speed is v, the laser moves to X within T time_lThe unit time Δ t is moved by the Δ x distance. And after single scanning excitation, the thermal imager acquires a sequence image and performs image synthesis on the sequence image. Fig. 5 is a block diagram of a flow of image synthesis, which includes the following specific steps:

1. and selecting a frame image in the sequence image as a starting end of image synthesis. On the premise of ensuring that the synthesized image can contain the region of interest on the surface of the object, any frame image in the sequence image can be selected as a starting end. The same is true of the end. The image synthesis calculation amount can be reduced by properly reducing the synthesis frame number, and the burden of an upper computer is reduced. The image synthesis is carried out in the whole sequence image range by taking the first frame image as a starting end and the last frame image as an ending end in the synthesis.

2. And selecting the interested area on the surface of the object by selecting a rectangular frame with a proper size, and traversing and searching the position point with the maximum gray value line by line in the frame. Similarly, the range selection can reduce the traversal data volume and shorten the image synthesis time. And assuming that the framed selection range is the whole picture, acquiring a row of position point number rows with the number equal to the number Y of the rows of the thermal imager after traversing.

3. And setting a distance criterion delta p. And sequentially calculating the distance difference between two adjacent position points in the maximum gray value position point sequence. If the distance difference is greater than the distance criterion Δ p, it indicates that there is a position mutation at the position point. Recording and setting the position point as a₀And then in turn is a₁，a₂… are provided. According to the difference of the flatness of the surface topography of the object, the distance criterion delta p is properly adjusted. If the flatness of the surface topography of the object is poor, for example, the surface is severely inclined or the curvature of the curve is large, the value of Δ p should be increased to prevent misjudgment. Conversely, the value of Δ p should be decreased to improve accuracy. Generally, Δ p should be no less than 2. When a laser line crosses the field of view of two adjacent pixels of the thermal imager in the laser scanning process, the position of the maximum value may fluctuate between the two pixels, so as to avoid misjudgment that Δ p is greater than the position difference, namely 1, between the two adjacent pixels. This step is used to distinguish between site point gradual changes and abrupt changes.

4. Aiming at the position mutation point a recorded in the previous step₀，a₁，a₂…, taking out the gray value set of the pixels in the row corresponding to the position point to perform fitting, and calculating the number of peak points of the fitted curve. If the number of the peak points is 1, the peak points are the laser line position points and no over-temperature points exist. If the number of peak points is greater than 1, it indicates that there may be over-temperature points in the row in addition to the laser line. The step can be used for preliminarily screening the step jump and over-temperature point phenomena of the laser curve.

5. And traversing the sequence images frame by frame, and repeatedly executing the steps 2-4 until the end images are finished. And finally extracting N columns of maximum gray value position point columns.

6. And (4) aiming at the maximum gray value position catastrophe points with the number of the peak points larger than 1 after fitting in the step (4), taking images in a multi-frame range before and after the peak points, and observing the position moving condition of the peak points. If the position in which the peak point exists does not move in the previous and subsequent frames of images, the position is indicated as an over-temperature point. From the above analysis, it can be seen that the laser curve position in the sequence image moves frame by frame at fixed intervals Δ x, and the over-temperature point position does not change. Combining step 4 and step 6, we can distinguish the laser curve step jump from the over-temperature point.

7. The gradual change of the laser curve, the sudden change of the laser curve and the over-temperature point phenomenon can be separated in sequence through the steps. At this time, the error can be compensated by adopting corresponding methods respectively aiming at different phenomena. When the laser curve is gradually changed, a direct fitting method is adopted; when the laser curve jumps, a piecewise fitting method is adopted to divide the sudden-changed laser curve segment and the main curve segment, and each segment is independently fitted; and when the over-temperature point exists, abandoning the peak value of the over-temperature point, selecting the peak value point close to the position of the peak value point in the previous row as the position of the laser curve, eliminating the influence of the over-temperature point, and then performing sectional fitting.

The invention can separate the phenomena of gradual change of the laser curve, sudden change of the laser curve and over-temperature points in the image through the steps, and then respectively compensate the respective errors, thereby reducing the image synthesis complexity under the condition of complex surface appearance of the object.

Claims

1. An image processing method in a laser scanning thermal imaging detection process is used for image synthesis when a structural part with a complex surface appearance is detected, and is characterized by comprising the following steps:

s1: and setting a distance criterion delta p. The distance criterion delta p is adjusted according to the change of the flatness of the surface topography of the object. Then, S2 is executed.

S2: and selecting a single-frame image, and traversing line by line to search the position point of the maximum gray value of the pixel point of the line. And obtaining a data point column containing the position points of the maximum gray value of all the rows after traversing. Then, S3 is executed.

S3: and sequentially calculating the distance difference between two adjacent position points in the maximum gray value position point number sequence. If the distance difference is greater than the distance criterion Δ p, it indicates that there is a position mutation at the position point. Executing S4 if the position catastrophe point exists after the traversal calculation; if not, execution proceeds to S7.

S4: and taking out the gray value number columns of the row pixels corresponding to the catastrophe points at all positions, respectively fitting, and calculating the number of peak points of a fitting curve. If the number of the peak points is 1, the peak points are the laser line position points and no over-temperature points exist. If the number of peak points is greater than 1, it indicates that there may be over-temperature points in the row in addition to the laser line. If the position catastrophe points with the peak point number larger than 1 exist in the whole data, executing S5; otherwise, S6 is executed.

S5: and (4) taking the maximum gray value position point of the position catastrophe point with the number of the peak value points larger than 1, and comparing the maximum gray value position point with the maximum gray value position points at the position points of the previous and next frames of images. If the position of the peak point is not changed, the position is the over-temperature point. And if the overtemperature point exists, selecting a peak value point close to the position point of the maximum gray value in the previous row from the transversely fitted peak value points as a laser curve position point. Then, S6 is executed.

S6: dividing the whole column of data into a plurality of sections by taking the position catastrophe point with the peak point number of 1 as an interval, and executing S7 for each section respectively.

S7: and fitting the obtained maximum gray value position points to obtain an actual maximum gray value position curve. Then, S8 is executed.

S8: and setting the required laser excitation time delay t, and traversing and selecting the curve of the laser curve x after the distance fitting. Where the distance x is the product of the laser scanning speed v and the required time delay t. Then, S9 is executed.

S9: and traversing and interpolating to obtain the actual gray value of each point on the time delay curve. Then, S10 is executed.

S10: the above steps S2 to S8 are repeatedly executed by traversing the sequence image. And combining the multiple rows of gray value numbers obtained finally into a new image.

2. The image processing method as claimed in claim 1, wherein the light source for laser scanning is a line laser having a length greater than the width of the region of interest on the surface of the object to be inspected. Meanwhile, the width of the line laser should be smaller than the width of the field of view of a single pixel of the thermal imager. During actual detection, the laser scanning speed is constant, so that the excitation time of the surface of an object in a scanning plane is uniform.

3. An image processing method as claimed in claim 1, characterized in that the traversal ranges of the sequence of images, i.e. the start and end, are arbitrarily chosen while ensuring that the synthesized image completely contains the region of interest on the surface of the object. Reducing the number of the traversal images can reduce the calculation amount, but the size of the images is also reduced, and an appropriate traversal range is selected. Meanwhile, the size of the synthesized picture is also affected by the traversal mode of the sequence image. The synthesized image is the same as the original image during frame-by-frame traversal; and the image is compressed proportionally when every other multiframe traversal mode is carried out.

4. The image processing method according to claim 1, wherein the more difference Δ p in flatness of the surface topography of the object in step S1 is increased to prevent erroneous judgment. Conversely, the value of Δ p should be decreased to improve accuracy. In order to avoid the erroneous judgment caused by the normal fluctuation of the maximum gray value position due to the fact that laser scanning crosses two adjacent pixel points, the delta p is more than 2 usually.