GH4169 bar mixed crystal multi-dimensional fusion identification method based on digital image processing

Abstract

The invention relates to the technical field of material nondestructive testing, and provides a digital image processing-based GH4169 bar mixed crystal multi-dimensional fusion identification method, which comprises the following steps of: performing signal acquisition on a GH4169 bar containing a mixed crystal structure by adopting an ultrasonic water immersion system to obtain an ultrasonic C scanning graph containing an abnormal signal; performing image feature boundary identification on the obtained ultrasonic C scanning image containing the abnormal signal to obtain an area proportion corresponding to the abnormal signal; carrying out image characteristic numerical value parameter analysis and classification on the obtained ultrasonic C scanning image containing the abnormal signal to obtain a classification result corresponding to the abnormal signal; according to results obtained in the above steps, when the area proportion corresponding to the abnormal signal in the ultrasonic C scanning image is larger than or equal to 5%, the signal-to-noise ratio of the abnormal signal is smaller than or equal to 5, and the partition noise variation coefficient value is larger than or equal to 0.2, it is judged that the defect corresponding to the abnormal signal in the GH4169 bar is mixed crystal. By means of the method, the mixed crystal reflection signals can be effectively judged, and a rapid and reliable qualitative method is provided for the GH4169 bar ultrasonic detection technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of non-destructive testing of materials, and particularly relates to a multi-dimensional fusion recognition method for mixed crystals of GH4169 bars based on digital image processing. Background Art

[0002] Since non-destructive testing plays an increasingly important role in product quality control, non-destructive testing is rapidly developing towards automation, digitization, and imaging. In particular, ultrasonic testing technology has introduced a higher-precision digital imaging non-destructive testing system. The unique ultrasonic digital signal processing and image analysis technologies have higher detection sensitivity and reliability, and can detect defects of smaller sizes, thereby improving the quality control of products. Detecting mixed crystals means that there are grains of different sizes inside the bar, both larger-sized grains and smaller-sized grains, rather than a uniform grain structure. Under normal circumstances, the grain size of a bar after appropriate processing and treatment should be relatively uniform to ensure the stability and consistency of material properties. The accurate judgment of mixed crystal signals can improve the quality control of bars.

[0003] However, in the existing ultrasonic automatic immersion testing applications, there is still no clear detection method to determine the type of abnormal signals generated during the detection of GH4169 bars, nor can it be determined whether the abnormal signals detected in GH4169 bars are mixed crystal signals. Therefore, the currently used ultrasonic automatic immersion testing method cannot achieve the purpose of improving the quality control of bars.

[0004] In view of this, the present invention is specifically proposed. Summary of the Invention

[0005] In order to solve the problems in the prior art, a multi-dimensional fusion recognition method for mixed crystals of GH4169 bars based on digital image processing proposed by the present invention solves the problem of ineffective qualitative analysis of abnormal signals in ultrasonic C-scan testing of GH4169 bars.

[0006] A multi-dimensional fusion recognition method for mixed crystals of GH4169 bars based on digital image processing of the present invention includes the following steps:

[0007] S1. Use an ultrasonic immersion system to collect signals from a GH4169 bar with a mixed crystal structure to obtain an ultrasonic C-scan image containing abnormal signals;

[0008] S2. Perform image feature boundary recognition on the ultrasonic C-scan image containing abnormal signals obtained in S1 to obtain the area ratio corresponding to the abnormal signals in the ultrasonic C-scan image;

[0009] S3. Analyze and classify the numerical parameters of the image features of the ultrasonic C-scan image containing abnormal signals obtained in S1 to obtain the classification results corresponding to the abnormal signals in the ultrasonic C-scan image;

[0010] S4. According to the results obtained in S2 and S3, when the area ratio of the abnormal signals corresponding in the ultrasonic C-scan image is ≥ 5%, the signal-to-noise ratio of the abnormal signals is ≤ 5, and the partition noise coefficient of variation value is ≥ 0.2, it is determined that the defect corresponding to the abnormal signal in the GH4169 bar is mixed crystal.

[0011] Furthermore, the specific steps of using the ultrasonic immersion system to collect signals from the GH4169 bar with mixed crystal structure in S1 are as follows:

[0012] First, calibrate the sensitivity of the ultrasonic immersion system using a reference block, then, automatically inspect the GH4169 bar, and finally, obtain the ultrasonic C-scan image containing abnormal signals.

[0013] Furthermore, the sensitivity used for inspection is Φ0.8mm, the circumferential scanning point is 0.5mm, and the axial scanning interval is 2mm.

[0014] Furthermore, the specific steps of identifying the image feature boundaries of the ultrasonic C-scan image containing abnormal signals in S2 are as follows:

[0015] First, with the maximum amplitude value corresponding to the abnormal signal in the ultrasonic C-scan image as the center, select a 100*100 area, perform median filtering on the selected ultrasonic C-scan image to smooth the image and re-image it. Then, based on CNN image enhancement and boundary bilinear interpolation, perform binary processing on the image, identify the boundary by combining the CANNY operator, automatically frame the abnormal signal area in the image using gradient threshold segmentation, and finally, obtain the area ratio of the abnormal signals corresponding.

[0016] Furthermore, the color presented by the abnormal signals after binary processing in the image is black.

[0017] Furthermore, the larger the area ratio of the abnormal signals corresponding to the black color, the larger the distribution range of the abnormal signals in the ultrasonic C-scan image. When the area ratio of the abnormal signals corresponding to the black color is ≥ 5%, the black area corresponding to the abnormal signals is marked as the mixed crystal area.

[0018] Furthermore, the specific steps of analyzing and classifying the numerical parameters of the image features of the ultrasonic C-scan image containing abnormal signals in S3 are as follows:

[0019] First, centered on the maximum value of the amplitude corresponding to the abnormal signal in the ultrasonic C-scan image, a 100*100 area is selected, and the signal-to-noise ratio of the abnormal signal in the selected ultrasonic C-scan image and the noise variation coefficient value within the partition are calculated. Then, the K-means clustering algorithm is used to classify the calculated signal-to-noise ratio and the noise variation coefficient value within the partition. Finally, two classification results corresponding to the abnormal signal in the ultrasonic C-scan image are obtained.

[0020] Furthermore, the two classification results are respectively:

[0021] When the calculated signal-to-noise ratio of the abnormal signal ≤ 5 and the partition noise variation coefficient value ≥ 0.2, then the abnormal signal is determined as a mixed crystal signal;

[0022] When the calculated signal-to-noise ratio of the abnormal signal > 5 and the partition noise variation coefficient value < 0.2, then the abnormal signal is determined as other reflection signals.

[0023] Furthermore, the other reflection signals include: forging crack signals of GH4169 bars, hole signals of GH4169 bars, and alloy dirty white spot defect signals of GH4169 bars.

[0024] Compared with the prior art, the method for multi-dimensional fusion recognition of mixed crystals in GH4169 bars based on digital image processing provided by the present invention, through clear process operations, combines digital image processing technology for automatic recognition and classification of image and numerical analysis fusion, has good operability, and the analysis results can be presented quickly and intuitively; moreover, it can quickly conduct a preliminary qualitative analysis of abnormal signals without damaging the GH4169 bars. Without special materials, it can effectively determine the mixed crystal reflection signals, providing a fast and reliable qualitative method for the ultrasonic detection technology of GH4169 bars. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a schematic flowchart of the method for multi-dimensional fusion recognition of mixed crystals in GH4169 bars based on digital image processing of the present invention;

[0026] Figure 2 is the ultrasonic C-scan image of abnormal signal 1 in the first embodiment of the present invention;

[0027] Figure 3 is the ultrasonic C-scan image of abnormal signal 1 after binarization processing in the first embodiment of the present invention;

[0028] Figure 4 is the automatic frame-taking statistical chart of the area of abnormal signal 1 in the first embodiment of the present invention;

[0029] Figure 5 is the metallographic dissection image corresponding to abnormal signal 1 in the first embodiment of the present invention;

[0030] Figure 6 It is the ultrasonic C-scan image of abnormal signal 2 in the second embodiment of the present invention;

[0031] Figure 7 It is the ultrasonic C-scan image of abnormal signal 2 after binarization processing in the second embodiment of the present invention;

[0032] Figure 8 It is the automatic frame-taking statistical chart of the area of abnormal signal 2 in the second embodiment of the present invention;

[0033] Figure 9 It is the metallographic dissection image corresponding to abnormal signal 2 in the second embodiment of the present invention;

[0034] Figure 10 It is the ultrasonic C-scan image of abnormal signal 3 in the third embodiment of the present invention;

[0035] Figure 11 It is the ultrasonic C-scan image of abnormal signal 3 after binarization processing in the third embodiment of the present invention;

[0036] Figure 12 It is the automatic frame-taking statistical chart of the area of abnormal signal 3 in the third embodiment of the present invention;

[0037] Figure 13 It is the metallographic dissection image corresponding to abnormal signal 3 in the third embodiment of the present invention. Detailed implementation manners

[0038] The present invention will be further explained and illustrated below in conjunction with the accompanying drawings of the specification and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0039] According to the embodiments of the present invention, before implementing the recognition method of the present invention, it is necessary to prepare in advance an ultrasonic water immersion zone automatic detection system and a GH4169 bar with a mixed crystal structure forged by a specific process.

[0040] As Figure 1 shown, the present invention provides a multi-dimensional fusion recognition method for mixed crystals of GH4169 bars based on digital image processing, including the following steps:

[0041] S1. Use an ultrasonic immersion system to collect signals from GH4169 bars with ambiguous crystal structures, and obtain an ultrasonic C-scan image containing abnormal signals. Specifically: First, calibrate the sensitivity of the ultrasonic immersion system using a reference block. Then, automatically inspect the GH4169 bars. Finally, obtain an ultrasonic C-scan image containing abnormal signals. The sensitivity used for inspection is Φ0.8mm, the circumferential scanning point is 0.5mm, and the axial scanning interval is 2mm.

[0042] S2. Identify the boundary of the image features of the ultrasonic C-scan image containing abnormal signals obtained in S1, and obtain the area ratio corresponding to the abnormal signals in the ultrasonic C-scan image. Specifically: First, take the maximum amplitude value corresponding to the abnormal signals in the ultrasonic C-scan image of the GH4169 bar containing mixed crystals as the center, select a 100*100 area, perform median filtering on the selected ultrasonic C-scan image to smooth the image and re-image it. Then, based on CNN image enhancement and bilinear interpolation of the boundary, perform binary processing on the image, identify the boundary using the CANNY operator, and automatically frame the black abnormal signal area in the image using gradient threshold segmentation. Finally, obtain the area ratio corresponding to the black abnormal signals. The larger the area ratio corresponding to the black abnormal signals, the larger the distribution range of the abnormal signals in the ultrasonic C-scan image. When the area ratio corresponding to the black abnormal signals ≥ 5%, the black area corresponding to the abnormal signals is marked as the mixed crystal area.

[0043] According to the embodiments of the present invention, the present invention only performs median filtering on the abnormal signal area. After median filtering, the abnormal signals appear black, while the non-abnormal signals appear white.

[0044] S3. At the same time, perform image feature numerical parameter analysis and classification on the ultrasonic C-scan image containing abnormal signals obtained in S1, and obtain the classification results corresponding to the abnormal signals in the ultrasonic C-scan image. Specifically: First, take the maximum amplitude value corresponding to the abnormal signals in the ultrasonic C-scan image as the center, select a 100*100 area, calculate the signal-to-noise ratio of the abnormal signals and the noise variation coefficient value within the partition in the selected ultrasonic C-scan image. Then, use the K-means clustering algorithm to classify the calculated signal-to-noise ratio and noise variation coefficient value within the partition. Finally, obtain two classification results corresponding to the abnormal signals in the ultrasonic C-scan image. The two classification results are: When the calculated signal-to-noise ratio of the abnormal signals ≤ 5 and the partition noise variation coefficient value ≥ 0.2, it proves that the reflection signal of the abnormal signal has a small difference in acoustic impedance from the matrix, and the abnormal signal is determined as a mixed crystal signal; when the calculated signal-to-noise ratio of the abnormal signals > 5 and the partition noise variation coefficient value < 0.2, it proves that the reflection signal of the abnormal signal has a large difference in acoustic impedance from the matrix, and the abnormal signal is determined as other reflection signals.

[0045] According to an embodiment of the present invention, other reflected signals include: GH4169 bar forging crack signals, GH4169 bar hole signals, GH4169 bar alloy dirty white spot defect signals, etc.

[0046] S4. According to the results obtained in S2 and S3, when the area ratio of the abnormal signal corresponding in the ultrasonic C-scan image ≥ 5%, the signal-to-noise ratio of the abnormal signal ≤ 5, and the partition noise coefficient of variation value ≥ 0.2, it is determined that the defect corresponding to the abnormal signal in the ultrasonic testing process of the GH4169 bar is mixed crystal.

[0047] According to an embodiment of the present invention, the ultrasonic C-scan adopted in the present invention is a two-dimensional imaging technology, which is used in non-destructive testing, especially in the detection and evaluation of internal defects of materials. The ultrasonic C-scan moves the probe on the surface of the object to be detected and records the ultrasonic reflection signals at each position. It generates a two-dimensional image to visually display the internal structure and defects of the material, and has the advantages of high resolution, comprehensive coverage, quantitative analysis, etc. For the ultrasonic C-scan abnormal signal image, an image feature boundary recognition and quantification classification method is adopted, combined with an image feature numerical parameter analysis classification method, and through multi-dimensional fusion recognition of the abnormal signals in the ultrasonic C-scan image, the type of the abnormal signal is effectively determined.

[0048] According to an embodiment of the present invention, the median filtering adopted in the present invention is a non-linear digital image processing technology, which is often used to remove noise in images. The median filtering achieves the denoising effect by replacing the value of each pixel with the median of all pixel values in its neighborhood. This method can effectively retain the edge information of the image while removing noise to smooth the image.

[0049] According to an embodiment of the present invention, the CANNY operator adopted in the present invention is a widely used edge detection algorithm, which can be used to accurately extract the edge information in the image, which is crucial for image analysis. Through edge detection, further operations such as image segmentation, feature extraction and recognition can be carried out and it helps to enhance the edge information in the image, etc. Therefore, it has many applications in image analysis and feature extraction. In image classification technology, the convolutional neural network technology (CNN) is often used. Through its unique convolution and pooling operations, it can effectively extract features and classify grid structure data such as images, so it has been widely used in fields such as computer vision and natural language processing.

[0050] According to an embodiment of the present invention, the K-means clustering algorithm adopted in the present invention is a widely used clustering analysis method, mainly used to divide a data set into K clusters, and the center of each cluster is determined through an iterative process to minimize the sum of the squares of the distances from each data point to its cluster center. The K-means clustering algorithm is widely applied in various fields such as data analysis, signal processing, machine learning, etc. The K-means clustering algorithm is fast and has efficient data processing capabilities.

[0051] The signal-to-noise ratio (SNR) of ultrasonic detection is generally used to judge whether a signal is acceptable and is defined as follows:

[0052]

[0053] Where S is the maximum amplitude of the suspicious indication of ultrasonic detection, the M value is the average value of the noise in the surrounding or adjacent area, and the N value is the highest amplitude value of the noise in the surrounding or adjacent area except for the electrical noise. Generally, when the signal-to-noise ratio ≥ 2.5, there is a risk of an abnormal signal.

[0054] The coefficient of variation is a parameter used to measure the dispersion or variability of data, which can be used to analyze the fluctuation degree of a sample relative to its average value and is used to evaluate the quality consistency of products. The lower the coefficient of variation, the higher the product quality stability. For the automatic ultrasonic immersion zone detection of superalloy bars, the coefficient of variation is generally between 0.1 - 0.2.

[0055] To better demonstrate the technical effects of the present invention, the following specific embodiments will be further described.

[0056] Embodiment 1

[0057] S1. For a GH4169 bar with a specification of Φ250mm, the sensitivity of Φ0.8mm is used for inspection. The circumferential scanning point for the automatic ultrasonic immersion zone detection of the bar is 0.5mm, and the axial scanning spacing is 2mm. After the detection is completed, the ultrasonic C-scan image of abnormal signal 1 is obtained, as Figure 2 shown.

[0058] S2. Taking the maximum value of the amplitude of abnormal signal 1 as the center, a 100*100 area is selected from the ultrasonic C-scan image of the Φ250mm GH4169 bar. The selected ultrasonic C-scan image is subjected to median filtering to smooth the image and re-image it. Then, based on CNN image enhancement and boundary bilinear interpolation, the image is binarized, as Figure 3 shown. Then, the boundary is identified by combining the CANNY operator, and the gradient threshold segmentation is applied to automatically frame the abnormal signal area, as Figure 4 shown. Finally, it is obtained that the proportion of the area of the black abnormal signal data is 9.3%, indicating that the distribution range of abnormal signal 1 is relatively large.

[0059] S3. Meanwhile, taking the maximum value of the amplitude of the abnormal signal 1 as the center, select a 100*100 area from the ultrasonic C-scan image of the Φ250mm GH4169 bar, calculate the signal-to-noise ratio of the abnormal signal 1 and the noise coefficient of variation within the partition in the selected ultrasonic C-scan image, and obtain a signal-to-noise ratio of 3.5 for the abnormal signal 1 and a noise coefficient of variation within the partition of 0.21.

[0060] S4. According to the results obtained in S2 and S3, and simultaneously satisfying the mixed crystal criterion, it is determined that the defect corresponding to the abnormal signal 1 is a mixed crystal.

[0061] S5. Dissect the area corresponding to the abnormal signal 1, and the specific morphology of the dissection is as Figure 5 shown, and it is concluded that the defect corresponds to a mixed crystal of grain size 2 + 8.

[0062] Example 2

[0063] S1. For the Φ250mm specification GH4169 bar, a sensitivity of Φ0.8mm is used for inspection, the circumferential scanning point for the automatic detection of the bar by immersion zoning is 0.5mm, and the axial scanning interval is 2mm. After the detection is completed, the ultrasonic C-scan image of the abnormal signal 2 is obtained, as Figure 6 shown.

[0064] S2. Taking the maximum value of the amplitude of the abnormal signal 2 as the center, select a 100*100 area from the ultrasonic C-scan image of the Φ250mm GH4169 bar, perform median filtering on the selected ultrasonic C-scan image to smooth the image and re-image it, and then perform binary processing on the image based on CNN image enhancement and boundary bilinear interpolation, as Figure 7 shown, and then combine the CANNY operator to identify the boundary, and automatically frame the abnormal signal area by applying the gradient threshold segmentation, as Figure 8 shown. Finally, it is obtained that the area ratio of the black abnormal signal data is 8.1%, indicating that the distribution range of the abnormal signal 2 is relatively large.

[0065] S3. Meanwhile, taking the maximum value of the amplitude of the abnormal signal 2 as the center, select a 100*100 area from the ultrasonic C-scan image of the Φ250mm GH4169 bar, calculate the signal-to-noise ratio of the abnormal signal 2 and the noise coefficient of variation within the partition in the selected ultrasonic C-scan image, and obtain a signal-to-noise ratio of 3.6 for the abnormal signal 1 and a noise coefficient of variation within the partition of 0.22.

[0066] S4. According to the results obtained in S2 and S3, and simultaneously satisfying the mixed crystal criterion, it is determined that the defect corresponding to the abnormal signal 2 is a mixed crystal.

[0067] S5. Dissect the area corresponding to the abnormal signal 2, and the specific morphology of the dissection is as Figure 9 shown, and it is concluded that the defect corresponds to a mixed crystal of grain size 2.5 + 9.

[0068] Example 3

[0069] S1. For the GH4169 bar with a Φ250mm specification, the inspection uses a sensitivity of Φ0.8mm. The circumferential scanning point for the automatic detection of the bar by water immersion zoning is 0.5mm, and the axial scanning interval is 2mm. After the detection is completed, the ultrasonic C-scan image of abnormal signal 3 is obtained, as Figure 10 shown.

[0070] S2. Take the ultrasonic C-scan image of the Φ250mm GH4169 bar and select a 100*100 area centered on the maximum amplitude value of abnormal signal 3. Perform median filtering on the selected ultrasonic C-scan image to smooth the image and re-image it. Then, based on CNN image enhancement and boundary bilinear interpolation, perform binary processing on the image, as Figure 11 shown. Then, combine the CANNY operator to identify the boundary, and apply gradient threshold segmentation to automatically frame the abnormal signal area, as Figure 12 shown. Finally, it is obtained that the area ratio of the black abnormal signal data is 0.9%, indicating that the distribution range of abnormal signal 3 is relatively large.

[0071] S3. At the same time, take the ultrasonic C-scan image of the Φ250mm GH4169 bar and select a 100*100 area centered on the maximum amplitude value of abnormal signal 3. Calculate the signal-to-noise ratio of abnormal signal 3 and the noise variation coefficient value within the partition in the selected ultrasonic C-scan image, and obtain that the signal-to-noise ratio of abnormal signal 3 is 4.5 and the partition noise variation coefficient value is 0.18.

[0072] S4. According to the results obtained in S2 and S3, the mixed crystal criterion cannot be satisfied simultaneously, and it is determined that the defect corresponding to abnormal signal 3 is not a mixed crystal.

[0073] S5. Dissect the area corresponding to abnormal signal 3, and the specific morphology of the dissection is as Figure 13 shown. It is obtained that the defect corresponds to an intergranular crack with an average size of about 3155μm.

[0074] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for multi-dimensional fusion recognition of mixed crystals in GH4169 bars based on digital image processing, characterized in that, It includes the following steps: S1. Use an ultrasonic immersion system to collect signals from GH4169 bars with ambiguous crystal structures, and obtain an ultrasonic C-scan image containing abnormal signals; S2. Identify the boundary of the image features of the ultrasonic C-scan image containing abnormal signals obtained in S1, and obtain the area ratio corresponding to the abnormal signals in the ultrasonic C-scan image; S3. Analyze and classify the numerical parameters of the image features of the ultrasonic C-scan image containing abnormal signals obtained in S1, and obtain the classification results corresponding to the abnormal signals in the ultrasonic C-scan image; S4. According to the results obtained in S2 and S3, when the area ratio corresponding to the abnormal signals in the ultrasonic C-scan image is ≥5%, the signal-to-noise ratio of the abnormal signals is ≤5, and the partition noise variation coefficient value is ≥0.2, it is determined that the defect corresponding to the abnormal signals in the GH4169 bars is ambiguous crystals.

2. The method for multi-dimensional fusion recognition of mixed crystals of GH4169 bars based on digital image processing according to claim 1, wherein, The specific steps of using the ultrasonic immersion system to collect signals from GH4169 bars with ambiguous crystal structures in S1 are as follows: First, calibrate the sensitivity of the ultrasonic immersion system using a reference block, then, automatically inspect the GH4169 bars, and finally, obtain an ultrasonic C-scan image containing abnormal signals.

3. The method for multi-dimensional fusion recognition of mixed crystals of GH4169 bars based on digital image processing according to claim 2, characterized in that, The sensitivity used for inspection is Φ0.8mm, the circumferential scanning point is 0.5mm, and the axial scanning interval is 2mm.

4. The method for multi-dimensional fusion recognition of mixed crystals of GH4169 bars based on digital image processing according to claim 1, wherein The specific steps of identifying the boundary of the image features of the ultrasonic C-scan image containing abnormal signals in S2 are as follows: First, take the maximum value of the amplitude corresponding to the abnormal signals in the ultrasonic C-scan image as the center, select a 100*100 area, perform median filtering on the selected ultrasonic C-scan image to smooth the image and re-image it, then, based on CNN image enhancement and bilinear interpolation of the boundary, perform binary processing on the image, identify the boundary using the CANNY operator, automatically frame the abnormal signal area in the image using gradient threshold segmentation, and finally, obtain the area ratio corresponding to the abnormal signals.

5. The method for multi-dimensional fusion recognition of mixed crystals of GH4169 bars based on digital image processing according to claim 4, wherein The color of the abnormal signals presented in the image after binary processing is black.

6. The method for multi-dimensional fusion recognition of mixed crystals in GH4169 bars based on digital image processing according to claim 5, characterized in that The larger the area ratio corresponding to the abnormal signals presented in black, the larger the distribution range of the abnormal signals in the ultrasonic C-scan image. When the area ratio corresponding to the abnormal signals presented in black is ≥5%, the black area corresponding to the abnormal signals is marked as the ambiguous crystal area.

7. The method for multi-dimensional fusion recognition of mixed crystals in GH4169 bars based on digital image processing according to claim 1, wherein The specific steps of analyzing and classifying the numerical parameters of the image features of the ultrasonic C-scan image containing abnormal signals in S3 are as follows: First, take the maximum value of the amplitude corresponding to the abnormal signals in the ultrasonic C-scan image as the center, select a 100*100 area, calculate the signal-to-noise ratio of the abnormal signals in the selected ultrasonic C-scan image and the partition noise variation coefficient value, then, use the K-means clustering algorithm to classify the calculated signal-to-noise ratio and partition noise variation coefficient value, and finally, obtain two classification results corresponding to the abnormal signals in the ultrasonic C-scan image.

8. The method for multi-dimensional fusion recognition of mixed crystals of GH4169 bars based on digital image processing according to claim 7, wherein The two classification results are respectively: When the calculated signal-to-noise ratio of the abnormal signals is ≤5 and the partition noise variation coefficient value is ≥0.2, the abnormal signals are determined to be ambiguous crystal signals; When the calculated signal-to-noise ratio of the abnormal signals is >5 and the partition noise variation coefficient value is <0.2, the abnormal signals are determined to be other reflection signals.

9. The method for multi-dimensional fusion recognition of mixed crystals of GH4169 bars based on digital image processing according to claim 8, wherein The other reflected signals include: the forging crack signal of GH4169 bar, the hole signal of GH4169 bar, and the alloy dirty white spot defect signal of GH4169 bar.

Images