An adaptive method for extracting texture features of machined surfaces

Abstract

The present invention discloses a method for adaptively extracting texture features of a machining surface, which is performed in the following steps: S1: using an image acquisition device to obtain a machining surface texture image; S2: converting the obtained machining surface texture image into a grayscale image; S3: setting the maximum decomposition layer number of adaptive extraction of empirical mode decomposition according to the required task requirements; S4: sequentially performing empirical mode decomposition on the grayscale image under different decomposition layer numbers; S5: calculating the correlation between the characteristic signals of each layer of the empirical mode decomposition of the grayscale image under different decomposition layer numbers and the original grayscale image; S6: selecting the characteristic signal with the maximum correlation as the extraction result of the machining surface texture. The present invention can adaptively extract machining texture features under the condition of a given surface image, and has the characteristics of strong anti-characteristic mode aliasing ability, rotation invariance, etc.

Description

Technical Field

[0001] The invention relates to the field of processed surface image processing, in particular to a method for adaptively extracting texture features of processed surfaces. Background Art

[0002] During the machining process, detailed information of the machining process will inevitably be left on the surface of the machined workpiece, and changes in its conditions will inevitably be left on the morphology of the machined surface. Therefore, the machined surface can effectively reflect the state of the machining process to a certain extent and leave mapping information on the machined surface. How to extract the quality information of the machining process from the analysis of the machined surface texture and finally extract the key feature information reflecting the machined surface is a key problem that needs to be solved urgently. With the development of surface morphology detection and image processing technology, some researchers have tried to obtain machining process information (such as tool wear, etc.) based on the machined surface morphology and surface texture images using statistical analysis and other methods, and have achieved good results. However, the machined surface texture is a typical two-dimensional feature information, which is easily affected by noise, etc. Therefore, it is extremely important to accurately extract its machining texture features, which determines the acquisition of machining process information. Summary of the invention

[0003] The purpose of the present invention is to provide a method for adaptively extracting texture features of a machined surface. The present invention can adaptively extract the features of machined textures under the condition of a given surface image, and has the characteristics of strong anti-feature mode aliasing capability and rotation invariance.

[0004] The technical solution of the present invention is a method for adaptively extracting texture features of a processed surface, characterized by: performing the following steps:

[0005] S1: Acquire the texture image of the processed surface using an image acquisition device;

[0006] S2: converting the acquired machined surface texture image into a grayscale image;

[0007] S3: Set the maximum number of decomposition layers for adaptive extraction of empirical mode decomposition according to the required task requirements;

[0008] S4: Perform empirical mode decomposition on the grayscale image under different decomposition layer conditions in turn;

[0009] S5: Calculate the correlation between the characteristic signals of each layer of the empirical mode decomposition of the grayscale image and the original grayscale image under different decomposition layer conditions; the calculation method of the correlation is:

[0010] Rel=λ×Similarity+(1-λ)×Smooth;

[0011] Among them, Rel is the calculated correlation, λ is the harmonic coefficient of similarity and smoothness, Similarity is the similarity between the feature signal and the original grayscale image, and Smooth is the smoothness of the feature signal;

[0012] S6: Select the feature signal with the maximum correlation as the extraction result of the machined surface texture.

[0013] In the aforementioned method for adaptively extracting texture features of machined surfaces, the similarity evaluation method in S5 is:

[0014]

[0015] Among them, I is the original grayscale image, μ is the symbol for average value calculation, I * It is the characteristic signal of the current decomposition layer of the grayscale image.

[0016] In the aforementioned method for adaptively extracting machined surface texture features, the evaluation method for the smoothness of the feature signal in S5 is:

[0017] Smooth = abs(skew(I * ));

[0018] Among them, I * is the feature of the current decomposition layer of the grayscale image, abs(·) is the symbol for absolute value calculation, and skew(·) is the skewness calculation.

[0019] In the aforementioned method for adaptively extracting machined surface texture features, the magnitude of the harmonic coefficient λ is adjusted to adjust the similarity and smoothness tendency during the extraction process.

[0020] In the aforementioned method for adaptively extracting texture features of a machined surface, the maximum number of decomposition layers is 8-20 layers.

[0021] In the aforementioned method for adaptively extracting texture features of a machined surface, the harmonic coefficient λ is 0.1-0.5.

[0022] Compared with the prior art, the present invention first obtains the processing surface texture image through an image acquisition device and converts it into a grayscale image, then sets the maximum decomposition layer number for adaptive extraction, and successively performs empirical mode decomposition on the grayscale image under different decomposition layer numbers, then calculates the correlation between the features of each layer of the empirical mode decomposition of the grayscale image under different decomposition layer numbers and the original grayscale image, and finally selects the feature signal with the maximum correlation as the texture extraction result. The proposed method of the present invention has the characteristics of adaptive feature extraction, strong anti-feature mode aliasing ability, rotation invariance, etc. It can adaptively extract processing texture features under the condition of a given surface image, and has the advantage of accurate results. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the present invention;

[0024] Figure 2 is the grayscale image of the processed surface in Example 1;

[0025] Figure 3 is a partial hierarchical feature component of the grayscale image in Example 1;

[0026] Figure 4 is a correlation distribution diagram of the grayscale image of the processed surface in Example 1;

[0027] Figure 5 is the final extracted machined surface texture feature in Example 1;

[0028] Figure 6 is the grayscale image of the processed surface in Example 2;

[0029] Figure 7 It is the partial hierarchical feature component of the grayscale image in Example 2;

[0030] Figure 8 is a correlation distribution diagram of the grayscale image of the processed surface in Example 2;

[0031] Fig. 9 It is the processing surface texture feature finally extracted in Example 2. DETAILED DESCRIPTION

[0032] The present invention is further described below in conjunction with the accompanying drawings and embodiments, but they are not intended to limit the present invention.

[0033] Embodiment 1: A method for adaptively extracting texture features of a processed surface, such as Figure 1 As shown, the following steps are included:

[0034] S1: Acquire a texture image of a machined surface by using an image acquisition device, wherein the image acquisition device may be a camera, and acquire the texture image of the machined surface by photographing with the camera;

[0035] S2: Convert the acquired surface texture image into a grayscale image; the grayscale image conversion can be performed using software such as PS. Figure 2 As shown;

[0036] S3: Set the maximum number of decomposition layers according to the required task requirements. In this embodiment, the maximum number of decomposition layers of the adaptive extraction of the empirical mode decomposition is set; the maximum number of decomposition layers set in this embodiment is 12 layers; Empirical Mode Decomposition (EMD) is a new type of adaptive signal time-frequency processing method creatively proposed by NE Huang and others at NASA in 1998, which is suitable for analyzing nonlinear and non-stationary signal sequences and has a high signal-to-noise ratio;

[0037] S4: Perform empirical mode decomposition on the grayscale image under different decomposition levels, and the decomposed hierarchical feature components are as follows: Figure 3 As shown;

[0038] S5: Calculate the correlation between the feature signals of each layer of the empirical mode decomposition of the grayscale image under different decomposition layers and the original grayscale image;

[0039] In this step, the calculation method of the correlation degree is as follows:

[0040] Rel=λ×Similarity+(1-λ)×Smooth;

[0041] Wherein, Rel is the calculated correlation degree, λ is the harmonic coefficient of similarity and smoothness, in this embodiment, λ is set to 0.2, Similarity is the similarity between the feature signal and the original grayscale image, and Smooth is the smoothness of the feature signal.

[0042] The similarity evaluation method is:

[0043]

[0044] Among them, I is the original grayscale image, μ is the symbol for average calculation, and I* is the feature of the current decomposition layer of the image.

[0045] The smoothness evaluation method is:

[0046] Smooth = abs(skew(I * ))

[0047] Among them, I* is the feature of the current decomposition layer of the image, the symbol abs(·) is the symbol of absolute value calculation, and skew(·) is the skewness calculation.

[0048] S6: The grayscale image correlation distribution diagram corresponding to this embodiment is as follows: Figure 4As shown in the figure, the maximum value of the correlation occurs in the 8th feature component of the 8th layer, and its maximum correlation is about 0.5. The feature signal with the maximum correlation is selected as the texture extraction result. In this embodiment, the feature signal with the maximum correlation is as follows Figure 5 As shown in the figure, this figure is the final texture extraction result.

[0049] Embodiment 2: A method for adaptively extracting texture features of a processed surface, such as Figure 1 As shown, the following steps are included:

[0050] S1: Acquire a texture image of a machined surface by using an image acquisition device, wherein the image acquisition device may be a camera, and acquire the texture image of the machined surface by photographing with the camera;

[0051] S2: Convert the acquired surface texture image into a grayscale image; the grayscale image conversion can be performed using software such as PS. Figure 6 As shown;

[0052] S3: Setting the maximum number of decomposition layers according to the required task requirements. In this embodiment, the maximum number of decomposition layers for adaptive extraction of empirical mode decomposition is set; the maximum number of decomposition layers set in this embodiment is 12 layers;

[0053] S4: Perform empirical mode decomposition on the grayscale image under different decomposition levels, and the decomposed hierarchical feature components are as follows: Figure 7 As shown;

[0054] S5: Calculate the correlation between the feature signals of each layer of the empirical mode decomposition of the grayscale image under different decomposition layers and the original grayscale image;

[0055] In this step, the calculation method of the correlation degree is as follows:

[0056] Rel=λ×Similarity+(1-λ)×Smooth;

[0057] Wherein, Rel is the calculated correlation degree, λ is the harmonic coefficient of similarity and smoothness, in this embodiment, λ is set to 0.2, Similarity is the similarity between the feature signal and the original grayscale image, and Smooth is the smoothness of the feature signal.

[0058] The similarity evaluation method is:

[0059]

[0060] Among them, I is the original grayscale image, μ is the symbol for average value calculation, I * is the current decomposition layer feature of the image.

[0061] The smoothness evaluation method is:

[0062] Smooth = abs(skew(I * ))

[0063] Among them, I * is the feature of the current decomposition layer of the image, abs(·) is the absolute value calculation symbol, and skew(·) is the skewness calculation.

[0064] S5: The grayscale image correlation distribution diagram corresponding to this embodiment is as follows: Figure 8 As shown in the figure, the maximum value of the correlation occurs in the 8th feature component of the 8th layer, and its maximum correlation is about 0.6; the feature signal with the maximum correlation is selected as the texture extraction result. In this embodiment, the feature signal with the maximum correlation is as follows Fig. 9 As shown in the figure, this figure is the final texture extraction result.

[0065] from Figure 5 and Fig. 9 It can be seen from the extraction results that the present invention can clearly show the characteristic information of the processed surface texture image and the results are accurate.

[0066] In summary, the present invention first obtains the processing surface texture image through the image acquisition device and converts it into a grayscale image, then sets the maximum decomposition layer number of adaptive extraction, and performs empirical mode decomposition on the grayscale image under different decomposition layer numbers in turn, and then calculates the correlation between the features of each layer of the empirical mode decomposition of the grayscale image under different decomposition layer numbers and the original grayscale image, and finally selects the feature signal with the maximum correlation as the texture extraction result. The proposed method of the present invention has the characteristics of adaptive feature extraction, strong anti-feature mode aliasing ability, rotation invariance, etc. It can adaptively extract processing texture features under the condition of a given surface image, and has the advantage of accurate results.

[0067] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for adaptively extracting texture features of a machined surface, characterized in that: Follow these steps: S1: Acquire the texture image of the processed surface using an image acquisition device; S2: converting the acquired machined surface texture image into a grayscale image; S3: Set the maximum number of decomposition levels according to the required task requirements; S4: Perform empirical mode decomposition on the grayscale image under different decomposition layer conditions in turn; S5: Calculate the correlation between the characteristic signals of each layer of the empirical mode decomposition of the grayscale image and the original grayscale image under different decomposition layer conditions; the calculation method of the correlation is: Among them, Rel is the calculated correlation, λ is the harmonic coefficient of similarity and smoothness, Similarity is the similarity between the feature signal and the original grayscale image, and Smooth is the smoothness of the feature signal; S6: Select the feature signal with the maximum correlation as the extraction result of the machined surface texture.

2. The method for adaptively extracting texture features of a machined surface according to claim 1, characterized in that: The evaluation method of the similarity in S5 is: Among them, I is the original grayscale image, μ is the symbol for average value calculation, I * It is the characteristic signal of the current decomposition layer of the grayscale image.

3. The method for adaptively extracting texture features of a machined surface according to claim 1, characterized in that: The evaluation method of the smoothness of the characteristic signal in S5 is: Smooth=abs(skew(I * )); Among them, I * is the feature of the current decomposition layer of the grayscale image, abs(·) is the symbol for absolute value calculation, and skew(·) is the skewness calculation.

4. The method for adaptively extracting texture features of a machined surface according to claim 1, characterized in that: The size of the harmonic coefficient λ is adjusted to adjust the similarity and smoothness tendency during the extraction process.

5. The method for adaptively extracting texture features of a machined surface according to claim 1, characterized in that: The maximum number of decomposition layers is 8-20 layers.

6. The method for adaptively extracting texture features of a machined surface according to claim 1, characterized in that: The harmonic coefficient λ is 0.1-0.5.

Images