A gear crack detection method based on wavelet multi-layer decomposition

Abstract

The present invention discloses a gear crack detection method based on wavelet multi-layer decomposition. The method adopts an off-line detection method, that is, regularly collecting images of working gears, and then applying wavelet algorithm and image processing technology to the collected images to analyze the characteristics of gear cracks, focusing on the detection of local cracks in the gears. When performing wavelet decomposition on the image, in order to see the defects more clearly, the sub-graphs are fused, and finally the histogram threshold method is used for segmentation, and the defects of the gears are finally clearly obtained. The present invention ensures timely and accurate detection and control of the damage state of the gears, thereby reducing or avoiding the occurrence of faults in the processing process, better ensuring the processing quality, reducing the scrapping of workpieces, and improving production efficiency.

Description

Technical Field

[0001] The invention relates to the field of detection technology, and in particular to a gear crack detection method based on wavelet multi-layer decomposition. Background Art

[0002] As early as the mid-1970s, people launched a series of research on gear fault diagnosis. The fault diagnosis methods for gears can be roughly divided into two categories:

[0003] (1) Through the dynamic signals such as vibration and noise generated during the operation of gears, signal processing methods are used to complete fault analysis and diagnosis. This paper focuses on this type of diagnostic method.

[0004] (2) According to the friction and wear theory, the purpose of gear fault diagnosis is achieved by analyzing the lubricating oil. Early gear fault diagnosis was mainly limited to directly measuring some simple vibration parameters, such as vibration peak value K, mean square value, etc. In order to eliminate the influence of factors such as machine load, some dimensionless parameters such as peak coefficient were also introduced. The measurement of these parameters is relatively simple and easy, and only requires the use of relatively simple vibration measuring instruments. The data results displayed at the end can also be compared with each other, so they are often used for simple diagnosis. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a gear crack detection method based on wavelet multi-layer decomposition in view of the above-mentioned deficiencies in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A gear crack detection method based on wavelet multi-layer decomposition adopts an offline detection method, that is, regularly collecting images of the working gear, and then applying the wavelet algorithm and image processing technology to the collected images to analyze the characteristics of the gear cracks, focusing on the detection of local cracks in the gear. When the image is decomposed by wavelet, in order to see the defects more clearly, the sub-images are fused, and finally the histogram threshold method is used for segmentation, and finally the gear defects are clearly obtained.

[0008] As a further solution of the present invention: after the image acquisition of the working gear, gear image preprocessing is required; image preprocessing refers to the db wavelet decomposition of the input image, fusion based on the wavelet transform method, grayscale and noise removal before threshold segmentation in image analysis. The purpose of image preprocessing is to eliminate the influence of irrelevant information in the image, restore useful real information, enhance the detectability of relevant information and simplify the use of data to the maximum extent, so as to facilitate feature extraction, image segmentation, matching and recognition.

[0009] As a further solution of the present invention: the grayscale processing refers to grayscale the gear color image;

[0010] The weighted average method is used to further calculate the grayscale value after conversion based on the average method. That is, according to the specific processing requirements, different weights are selected for the grayscale of the three components of the color image to perform weighted averaging. According to the visual characteristics of the human eye, the grayscale image that meets the requirements is usually calculated according to the weights of the following formula, as shown in formula (1-1):

[0011]

[0012] Median filtering gear image denoising:

[0013] In two-dimensional form, median filtering is performed by taking a certain form of two-dimensional window, sorting the pixels in the window, and generating a monotonous two-bit data sequence; let represent the gray value of each point in the digital image, and the two-dimensional median filter output g(x, y) of the filter window is:

[0014] g(i,j)=Med{f ij}=Med{f (i+k),(j+l) (k,l)∈A(i,j)∈I 2

[0015] (1-2).

[0016] As a further solution of the present invention: in order to achieve a more ideal effect on gear crack detection, the wavelet used for gear crack image decomposition must have high vanishing moment, good compact support, and fast decay; a DB2 wavelet with high vanishing moment and orthogonal compact support is constructed, and the relationship between the support length N and the vanishing moment M satisfies N=2M-1, and has a certain approximate symmetry; the DB wavelet is selected

[0017] As a further solution of the present invention: The purpose of gear image decomposition is to decompose the gear surface characteristics in different levels, and different detail sub-graphs contain different gear surface information. When the required gear crack information has been decomposed into detail sub-graphs at different levels, it is not necessary to decompose the approximate sub-graphs again; therefore, the three-level decomposition of DB2 wavelet is adopted, which are the approximate horizontal detail sub-graph, the approximate vertical detail sub-graph, and the approximate diagonal detail sub-graph of the image.

[0018] As a further solution of the present invention: the fusion based on the wavelet transform method is to first perform wavelet decomposition on each source image to form a multi-scale decomposition of the image, and then perform fusion processing on the corresponding wavelet decomposition layer to obtain a fused wavelet multi-scale image, and finally perform an inverse wavelet transform to obtain a fused image, so that a multi-layer fused sub-image can be obtained.

[0019] As a further solution of the present invention: the threshold segmentation refers to performing histogram threshold segmentation on the fused sub-image to obtain a histogram threshold segmentation image; the selection of the histogram threshold method shows that most of the pixel grayscale of the image is concentrated in two places, that is, it contains two "peaks". This type of histogram is called a bimodal histogram, which indicates that the contrast between the object and the background in the corresponding image is obvious; conversely, if the contrast between the object and the background in the image is obvious, and the grayscale consistency between the objects is good, then the corresponding histogram Figure 1 It must be a bimodal histogram; due to the influence of noise, the peaks and valleys on the histogram designed for a bimodal histogram will not be perfect, and the "peak" is often composed of several narrow peaks, but the grayscale of the highest point can represent the typical grayscale value of the interior of the object or the background point; the threshold can be set at a fixed position relative to the two peaks.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: the gear crack detection scheme proposed by the present invention uses advanced computer image and information technology to transform and upgrade traditional gear detection. From the perspective of the development direction of numerical control technology in the equipment manufacturing industry, the research project conforms to the development strategy advocated by the country. From the perspective of actual production and processing, the research scheme not only helps to improve the scientific and technological content of gear detection, ensure timely and accurate detection and control of gear damage status, thereby reducing or avoiding the occurrence of faults during the processing process, better guarantee processing quality, reduce workpiece scrapping, and improve production efficiency, but also can improve the competitiveness of equipment manufacturing products, and will also have a certain demonstration and guiding role in related industries, which is of great significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a schematic diagram of a detection system based on image processing in the present invention;

[0022] Figure 2 It is a local diagram of a longitudinal gear crack in the present invention;

[0023] Figure 3 It is a local crack image after median filtering and denoising in the present invention;

[0024] Figure 4 It is the wavelet decomposition diagram of the longitudinal gear crack in the present invention;

[0025] Figure 5 It is the fusion graph of the longitudinal gear crack sub-graphs in the present invention;

[0026] Figure 6 This is the final segmentation graph in the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] This embodiment provides a gear crack detection method based on wavelet multi-layer decomposition, including:

[0029] Gear image preprocessing, in the process of acquiring images by imaging devices, will be affected by many factors and must be preprocessed. Image preprocessing refers to the processing of input images including db2 wavelet decomposition, fusion based on wavelet transform method, graying and noise removal before threshold segmentation in image analysis. The main purpose of image preprocessing is to eliminate the influence of irrelevant information in the image, restore useful real information, enhance the detectability of relevant information and simplify the use of data to the maximum extent, so as to facilitate feature extraction, image segmentation, matching and recognition.

[0030] 1. Grayscale color image:

[0031] The weighted average method makes further calculations on the basis of the average method in selecting the grayscale value after conversion. That is, according to the specific processing requirements, different weights are selected for the grayscale of the three components of the color image for weighted averaging. According to the visual characteristics of the human eye, the grayscale image that meets the requirements is usually calculated by the weights of the following formula, as shown in formula (1-1):

[0032]

[0033] Median filtering gear image denoising:

[0034] In two-dimensional form, median filtering takes a certain form of two-dimensional window, sorts the pixels in the window, and generates a monotonous two-bit data sequence. Let represent the gray value of each point in the digital image, and the two-dimensional median filter output g(x,y) of the filter window is:

[0035] g(i,j)=Med{f ij}=Med{f (i+k),(j+l) (k,l)∈A(i,j)∈I 2

[0036] (1-2)

[0037] 2. Wavelet decomposition of gear crack image:

[0038] In order to achieve a more ideal effect on gear crack detection, the wavelet used for gear crack image decomposition must have high vanishing moment, good compact support, and fast decay. Daubechies constructed a DB2 wavelet with high vanishing moment and orthogonal compact support from the perspective of the filter, and the relationship between the support length N and the vanishing moment M satisfies N = 2M-1, and has a certain approximate symmetry. Select DB wavelet

[0039] The purpose of gear image decomposition is to decompose the gear surface characteristics at different levels. Different detail sub-graphs contain different gear surface information. When the required gear crack information has been decomposed into detail sub-graphs at different levels, there is no need to decompose the approximate sub-graphs again. It is easy to know that the more layers of wavelet decomposition, the more obvious the defects of gear cracks are. In order to observe the wear condition more clearly, the (3-5) layer decomposition of db2 wavelet can better process the image. Finally, we choose the three-layer decomposition of db2 wavelet, which are the approximate horizontal detail sub-graph, the approximate vertical detail sub-graph, and the approximate diagonal detail sub-graph of the image.

[0040] 3. Gear crack image fusion:

[0041] The usual fusion algorithm is to perform simple weighted merging and select the maximum wavelet coefficient for fusion. Since the human retina processes images in different channels, the fusion method based on wavelet transform can also obtain a fusion effect that is more similar to human visual characteristics. According to the characteristics of wavelet transform itself, it can be seen that the image fusion method based on wavelet transform has greater advantages. The image fusion method based on wavelet transform is to first perform wavelet decomposition on each source image to form a multi-scale decomposition of the image, and then perform fusion processing on the corresponding wavelet decomposition layer to obtain the fused wavelet multi-scale image, and finally perform inverse wavelet transform to obtain the fused image. A three-layer fusion sub-image can be obtained.

[0042] 4. Gear crack image segmentation:

[0043] The histogram threshold method shows that most of the pixel grayscale of the image is concentrated in two places, that is, it contains two "peaks". This type of histogram is called a bimodal histogram, which indicates that the contrast between the object and the background in the corresponding image is obvious; conversely, if the contrast between the object and the background in the image is obvious, and the grayscale consistency between the objects is good, then the corresponding histogram is Figure 1It must be a bimodal histogram. The automatic threshold algorithm in this paper is designed for bimodal histograms. Due to the influence of noise, the peaks and valleys on the histogram will not be perfect. The "peak" is often composed of several narrow peaks. However, the grayscale of the highest point can generally represent the typical grayscale value of the object or background point. Set the threshold at a fixed position relative to the two peaks, such as the middle position. Generally speaking, this result will be more reliable than directly looking for the least common grayscale value, that is, estimating the position of the valley of the histogram. Perform histogram threshold segmentation on the final fused sub-image to obtain the histogram threshold segmentation image.

[0044] This embodiment determines the use of an offline detection method. The main idea is to regularly collect images of the working gears, and then apply the wavelet algorithm and image processing technology to the collected images to analyze the characteristics of the gear cracks. This article mainly focuses on the detection of local cracks in gears and studies the application of wavelet analysis in gear crack detection. When performing wavelet decomposition on the image, the more suitable db2 wavelet is used to generate approximate sub-images, horizontal detail sub-images, vertical detail sub-images, and diagonal detail sub-images. In order to see the defects more clearly, the sub-images are fused, and finally the histogram threshold method is used for segmentation, and the gear defects are finally clearly seen.

[0045] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0046] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A gear crack detection method based on wavelet multi-layer decomposition, characterized in that: This method adopts an offline detection method, that is, regularly collect images of working gears, and then apply wavelet algorithm and image processing technology to the collected images to analyze the characteristics of gear cracks, focusing on the detection of local cracks in gears. When performing wavelet decomposition on the image, in order to see the defects more clearly, the sub-images are fused, and finally the histogram threshold method is used for segmentation, and finally the gear defects are clearly obtained; After the working gear image is acquired, gear image preprocessing is required; Image preprocessing refers to the process of decomposing the input image by db wavelet, fusion based on wavelet transform method, graying and noise removal before threshold segmentation in image analysis. The purpose of image preprocessing is to eliminate the influence of irrelevant information in the image, restore useful real information, enhance the detectability of relevant information and simplify the use of data to the maximum extent, so as to facilitate feature extraction, image segmentation, matching and recognition; In order to achieve a more ideal effect on gear crack detection, the wavelet used for gear crack image decomposition must have high vanishing moment, good compact support, and fast decay. The DB2 wavelet with high vanishing moment and orthogonal compact support is constructed, and the relationship between the support length N and the vanishing moment M satisfies N=2M-1, and has a certain approximate symmetry. The DB wavelet is selected. The purpose of gear image decomposition is to decompose the gear surface characteristics at different levels. Different detail sub-graphs contain different gear surface information. When the required gear crack information has been decomposed into detail sub-graphs at different levels, there is no need to decompose the approximate sub-graphs again. Therefore, the three-level decomposition of db2 wavelet is adopted, which are the approximate horizontal detail sub-graph, the approximate vertical detail sub-graph, and the approximate diagonal detail sub-graph of the image. The fusion based on the wavelet transform method is to first perform wavelet decomposition on each source image to form a multi-scale decomposition of the image, then perform fusion processing on the corresponding wavelet decomposition layer to obtain a fused wavelet multi-scale image, and finally perform inverse wavelet transform to obtain a fused image, so as to obtain a multi-layer fusion sub-image; The threshold segmentation refers to performing histogram threshold segmentation on the fused sub-image to obtain a histogram threshold segmentation image; the selection of the histogram threshold method shows that most of the pixel grayscales of the image are concentrated in two places, that is, it contains two "peaks". This type of histogram is called a bimodal histogram, which indicates that the object and the background in the corresponding image have a clear contrast; Conversely, if the contrast between the objects and the background in the image is obvious, and the grayscale consistency between the objects is good, then the corresponding histogram must be a bimodal histogram; for the bimodal histogram designed due to the influence of noise, the peaks and valleys on the histogram will not be perfect, the "peak" is composed of several narrow peaks, but the grayscale of the highest point can represent the typical grayscale value of the object's interior or background points; the threshold can be set at a fixed position relative to the two peaks.

2. The gear crack detection method based on wavelet multi-layer decomposition according to claim 1 is characterized in that: The grayscale processing refers to grayscale the gear color image; The weighted average method is used to further calculate the grayscale value after conversion based on the average method. That is, according to the specific processing requirements, different weights are selected for the grayscale of the three components of the color image for weighted averaging. According to the visual characteristics of the human eye, the grayscale image that meets the requirements can be obtained by calculating the weights according to the following formula, as shown in formula (1-1): Median filtering gear image denoising: In two-dimensional form, median filtering is performed by taking a certain form of two-dimensional window, sorting the pixels in the window, and generating a monotonous two-bit data sequence; let represent the gray value of each point in the digital image, and the two-dimensional median filter output g(x, y) of the filter window is: g ( i,j ) =With { f ij } =With{f i+k),(j+l) (k,l)∈A(i,j)∈I 2 (1-2)。

Images