Method and system for detecting carbon deposition on inner wall of combustion chamber based on machine vision

Abstract

The present application belongs to the technical field of image anomaly detection, and particularly relates to a combustion chamber inner wall carbon deposition detection method and system based on machine vision, to solve the technical problem that the existing method is prone to misjudgment due to interference because it ignores the texture directionality and lacks adaptability. The detection method comprises: S1, calculating the kurtosis coefficient of the global brightness histogram of the RGB image to obtain a first chroma component and a second chroma component; S2, obtaining high-frequency subbands of horizontal, vertical and diagonal directions under at least two scales; calculating a directionality index representing the anisotropy degree of chroma noise; S3, determining a determination threshold of the directionality index; S4, weighting and summing the chroma noise energy obtained under each scale in combination with the corresponding center frequency of each scale to obtain a comprehensive noise index representing the carbon deposition degree. The comprehensive noise index obtained by multi-scale information fusion can accurately represent the carbon deposition degree.

Description

Technical Field

[0001] This invention belongs to the technical field of image anomaly detection, specifically relating to a method and system for detecting carbon deposits on the inner wall of a combustion chamber based on machine vision. Background Technology

[0002] Carbon deposits in the engine combustion chamber are a significant factor affecting engine power, fuel economy, and emissions. Long-term accumulation can lead to a series of problems, including engine knocking, unstable idling, and increased fuel consumption. Traditional visual inspection methods rely primarily on the color and brightness information of images. However, the internal working environment of the combustion chamber is harsh, with uneven lighting conditions and various interfering factors such as strong metallic reflections, oil stains, and mechanical scratches on the inner wall surface. These interfering factors affect the stability of color and brightness characteristics, easily misjudging oil stains or shadows as carbon deposits, or missing carbon deposits in reflective areas.

[0003] Carbon deposits typically exhibit a rough, irregular texture at the microscopic level. Therefore, some methods employ techniques such as gray-level co-occurrence matrix (GLCM), local binary mode, or wavelet transform to extract texture features from images in an attempt to distinguish carbon deposits from smooth metal inner walls. However, the texture features of carbon deposits vary at different scales and are quite similar to the texture features of defects such as oil stains and corrosion spots. Most existing texture analysis methods focus on the energy or statistical characteristics of the texture, neglecting directional information. Scratches and processing marks commonly found on the inner walls of combustion chambers exhibit obvious directional textures. These structural textures can easily interfere with the random texture of carbon deposits, leading to misjudgments. Furthermore, when the overall brightness and contrast of the captured image change due to variations in operating conditions or equipment, detection performance often degrades, lacking the ability to adaptively adjust to the global state of the image. Summary of the Invention

[0004] This invention provides a machine vision-based method and system for detecting carbon deposits on the inner wall of a combustion chamber, in order to solve the technical problem that existing methods are prone to misjudgment due to neglecting texture directionality and lacking adaptability.

[0005] In a first aspect, the present invention provides a machine vision-based method for detecting carbon deposits on the inner walls of a combustion chamber, comprising the following steps:

[0006] S1. Obtain the RGB image of the inner wall of the combustion chamber, calculate the kurtosis coefficient of the global brightness histogram of the RGB image, and perform nonlinear function weighting on the R, G, and B components of each pixel based on the ratio of the neighborhood brightness mean of each pixel in the RGB image to the preset brightness benchmark to obtain the first chromaticity component and the second chromaticity component.

[0007] S2, perform two-dimensional discrete wavelet transform on the first chromaticity component and the second chromaticity component respectively to obtain high-frequency sub-bands in the horizontal, vertical and diagonal directions at at least two scales; at the same scale, summarize the high-frequency sub-band energy of the first chromaticity component and the second chromaticity component in the horizontal, vertical and diagonal directions to obtain the energy distribution in the three directions at the corresponding scale, and calculate the directional index representing the degree of anisotropy of chromaticity noise based on the energy distribution.

[0008] S3, based on the arctangent function of the difference between the kurtosis coefficient and the preset reference coefficient, determine the judgment threshold of the directional index; when the directional index is less than the judgment threshold, the sum of the high-frequency subband energies in the three directions at the corresponding scale is taken as the chromaticity noise energy; otherwise, the chromaticity noise energy is set to a preset low value.

[0009] S4, the chromatic noise energy obtained at each scale is weighted and summed with the center frequency corresponding to each scale to obtain the comprehensive noise index representing the degree of carbon deposition.

[0010] Furthermore, in S1, for any pixel in the RGB image, a 5×5 neighborhood window is constructed, and the arithmetic mean of the brightness of all pixels within this neighborhood window is calculated as the neighborhood brightness mean. ;

[0011] Calculate the weighting coefficients ,in, The preset brightness reference;

[0012] The R, G, and B components of the pixel are multiplied by the weighting coefficients respectively. The weighted components are obtained. , , ;

[0013] The first chromaticity component is calculated based on the weighted components. Second chromaticity component .

[0014] Furthermore, in S2, the Daubechies4 wavelet basis is used to perform three-level two-dimensional discrete wavelet decomposition on the first and second chromaticity components respectively to obtain the horizontal high-frequency sub-band, vertical high-frequency sub-band and diagonal high-frequency sub-band at the first, second and third scales.

[0015] Furthermore, in S2, for any scale and direction Calculate the first chromaticity component in the direction The average of the absolute values ​​of all wavelet coefficients in the high-frequency subband is used to obtain the high-frequency subband energy of the first chromaticity component. ;

[0016] Calculate the second chromaticity component in the direction The average of the absolute values ​​of all wavelet coefficients in the high-frequency subband is used to obtain the high-frequency subband energy of the second chromaticity component. ;

[0017] The high-frequency subband energy of the first chromaticity component High-frequency subband energy of the second chromaticity component By adding them together, we obtain the total high-frequency subband energy at that scale and in that direction. ,satisfy .

[0018] Furthermore, in S2, for any scale Calculate the high-frequency subband energy in the horizontal, vertical, and diagonal directions respectively. , , ;

[0019] Calculate the arithmetic mean of the high-frequency subband energies in three directions. and standard deviation ;

[0020] Calculate the corresponding scale Directional indicators below Let be the ratio of the standard deviation to the arithmetic mean, satisfying . .

[0021] Furthermore, in S3, the decision threshold T is calculated using the following formula: ,in, The kurtosis coefficient is the kurtosis coefficient of the global luminance histogram of the RGB image. As a preset baseline coefficient, This is the scaling factor. This is the bias coefficient.

[0022] Furthermore, the preset low value is 0.

[0023] Furthermore, in S4, the chromaticity noise energy obtained at the 1st, 2nd, and 3rd scales is denoted as follows: , , ;

[0024] Set weighting coefficients for each scale. , , The weighting coefficients are positively correlated with the center frequency of the corresponding scale;

[0025] Calculate the overall noise index ,satisfy .

[0026] Furthermore, , , .

[0027] Secondly, the present invention provides a machine vision-based combustion chamber wall carbon deposit detection system, including a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned machine vision-based combustion chamber wall carbon deposit detection method is implemented.

[0028] The beneficial effects are as follows: By applying a nonlinear function weighting to image pixels based on the ratio of the neighborhood brightness mean to a preset brightness benchmark, interference caused by uneven lighting and metallic reflection on the combustion chamber wall is suppressed, enhancing the stability of the first and second chromaticity components. This invention obtains high-frequency subbands in at least two scales (horizontal, vertical, and diagonal) by performing two-dimensional discrete wavelet transforms on the two chromaticity components. Based on the energy distribution of these three high-frequency subbands at each scale, a directional index representing the degree of anisotropy of chromaticity noise is constructed. This distinguishes between the irregular, isotropic textures of carbon deposits and structural textures with obvious directions, such as scratches and processing marks, thereby eliminating the energy of interfering textures and reducing the false positive rate. Furthermore, by correlating the judgment threshold of the directional index with the kurtosis coefficient of the image's global brightness histogram, the detection standard can be adjusted according to the overall contrast of the image, improving the consistency and reliability of the detection results under different shooting conditions. The comprehensive noise index obtained through multi-scale information fusion can accurately represent the degree of carbon deposits. Attached Figure Description

[0029] Figure 1 This is a flowchart of a machine vision-based method for detecting carbon deposits on the inner walls of a combustion chamber.

[0030] Figure 2 A flowchart for determining chromaticity noise energy based on directional indicators;

[0031] Figure 3 This is a schematic diagram of a machine vision-based combustion chamber wall carbon deposit detection system. Detailed Implementation

[0032] An embodiment of the machine vision-based method for detecting carbon deposits on the inner walls of a combustion chamber provided by this invention:

[0033] like Figure 1 As shown, the machine vision-based method for detecting carbon deposits on the inner walls of a combustion chamber includes the following steps:

[0034] S1. Obtain the RGB image of the combustion chamber wall, calculate the kurtosis coefficient of the global brightness histogram of the RGB image, and perform nonlinear function weighting on the R, G, and B components of each pixel based on the ratio of the neighborhood brightness mean of each pixel in the RGB image to the preset brightness benchmark to obtain the first chromaticity component and the second chromaticity component.

[0035] An M×N pixel RGB image of the combustion chamber wall is obtained by imaging with an endoscope. This RGB image is then converted to a grayscale image. The luminance component L of the grayscale image is calculated, and the number of pixels at each luminance level from 0 to 255 is counted to obtain a luminance histogram H. The kurtosis coefficient K is calculated based on the luminance histogram H, which is the ratio of the fourth-order central moment of the image to the fourth power of the standard deviation.

[0036] In an optional embodiment, in S1, for any pixel in the RGB image, a 5×5 neighborhood window is constructed, and the arithmetic mean of the brightness of all pixels within the neighborhood window is calculated as the neighborhood brightness mean. ;

[0037] Calculate the weighting coefficients ,in, The preset brightness reference;

[0038] The R, G, and B components of the pixel are multiplied by the weighting coefficients respectively. The weighted components are obtained. , , ;

[0039] The first chromaticity component is calculated based on the weighted components. Second chromaticity component .

[0040] For example, consider a pixel at coordinates (100, 150) in an RGB image. The pixel's original R, G, B values ​​are 160, 120, and 80, respectively. A 5×5 pixel window is constructed centered on this pixel, extending from coordinates (98, 148) to (102, 152). The brightness of all 25 pixels within this window is calculated, and the arithmetic mean of the brightness is obtained. Assume the average brightness of this neighborhood is... The value is 110.

[0041] Set the preset brightness reference The value is 128. Substitute this value into the formula to calculate the weighting coefficient. Weighting coefficients Approximately 0.933. The weighting coefficients... Multiply by the R, G, and B components of the original pixel respectively to obtain the weighted new component, i.e. It is 149.28. It is 111.96. The value is 74.64. The chromaticity components are calculated based on the weighted components; the first chromaticity component... =37.32; Second chromaticity component =55.98.

[0042] Among them, the preset brightness reference The following steps can be used to make the selection:

[0043] (1) Select a set of representative combustion chamber wall images as a reference set. These combustion chamber wall images should cover different lighting conditions, combustion states and carbon deposition levels to ensure the generalization ability of the model.

[0044] (2) Perform global brightness histogram analysis on each image to obtain the pixel frequency distribution of brightness values ​​from 0 to 255; calculate the cumulative probability density function of each brightness level, and determine the brightness value corresponding to a certain percentage of cumulative probability as a candidate benchmark value, such as 50%.

[0045] (3) Based on experimental or empirical data, select a brightness value that can balance the performance under different lighting conditions as the final preset brightness benchmark. For example, you can choose the mean or median of all candidate benchmark values.

[0046] (4) Use the selected preset brightness reference Multiple test images are processed, and the results are evaluated to determine if they meet expectations. If significant deviations are found, the preset brightness benchmark needs to be re-evaluated and adjusted. The value of .

[0047] S2, perform two-dimensional discrete wavelet transform on the first chromaticity component and the second chromaticity component respectively to obtain high-frequency sub-bands in the horizontal, vertical and diagonal directions at at least two scales; at the same scale, summarize the high-frequency sub-band energy of the first chromaticity component and the second chromaticity component in the horizontal, vertical and diagonal directions to obtain the energy distribution in the three directions at the corresponding scale, and calculate the directional index representing the degree of anisotropy of chromaticity noise based on the energy distribution.

[0048] In an optional embodiment, in S2, the Daubechies4 wavelet basis is used to perform three-level two-dimensional discrete wavelet decomposition on the first chromaticity component and the second chromaticity component respectively to obtain the horizontal high-frequency sub-band, vertical high-frequency sub-band and diagonal high-frequency sub-band at the first, second and third scales respectively.

[0049] Specifically, for the first chromaticity component calculated from all pixels... The constructed two-dimensional data matrix is ​​subjected to the first-level two-dimensional discrete wavelet decomposition using the Daubechies 4 wavelet basis. The original two-dimensional data matrix is ​​decomposed into a low-frequency approximate subband. And three high-frequency sub-bands, the three high-frequency sub-bands are in the horizontal direction Vertical direction and diagonal direction The three high-frequency subbands constitute the detailed information at the first scale.

[0050] The low-frequency approximate subband obtained from the first-level decomposition Perform the same two-dimensional discrete wavelet decomposition again to generate a new low-frequency subband. And the three high-frequency sub-bands at the second scale, namely the horizontal ones. Vertical direction and diagonal direction The decomposition process continues to repeat, and the low-frequency approximate subbands obtained from the second-level decomposition are... Further decomposition yields the horizontal high-frequency subband at the third scale. Vertical high-frequency subband and diagonal high-frequency subband After completing the three-layer decomposition, the first chromaticity component was obtained. Nine high-frequency subbands at three different scales. The same complete three-layer decomposition process is also applied to the second chromaticity component. The constructed two-dimensional data matrix yields the second chromaticity component. Nine additional high-frequency subbands across three scales.

[0051] In an optional embodiment, in S2, for any scale and direction Calculate the first chromaticity component in the direction The average of the absolute values ​​of all wavelet coefficients in the high-frequency subband is used to obtain the high-frequency subband energy of the first chromaticity component. ;

[0052] Calculate the second chromaticity component in the direction The average of the absolute values ​​of all wavelet coefficients in the high-frequency subband is used to obtain the high-frequency subband energy of the second chromaticity component. ;

[0053] The high-frequency subband energy of the first chromaticity component High-frequency subband energy of the second chromaticity component By adding them together, we obtain the total high-frequency subband energy at that scale and in that direction. ,satisfy .

[0054] Specifically, taking the calculation of the total high-frequency subband energy in the horizontal direction of the second scale as an example, it is necessary to locate the high-frequency subband in the horizontal direction of the second scale, that is, the first chromaticity component obtained above. Quantity Subband and Second Chromaticity Component Quantity Sub-band. Assume of The subband is a matrix containing 4096 wavelet coefficients. During calculation, the absolute value of each of the 4096 wavelet coefficients is calculated, all absolute values ​​are summed, and the sum is then divided by the total number of coefficients, 4096.

[0055] For example, if of The sum of the absolute values ​​of all coefficients in the subband is 81920, so the energy of the high-frequency subband is... It's 20. Similarly, for... of The same calculation is performed on the sub-bands, assuming the calculated high-frequency sub-band energy... The value is 16. Adding the two energy values ​​yields the total high-frequency subband energy in the second-scale horizontal direction. The value is 36. This calculation process needs to be repeated for all three scales and all three directions, namely horizontal, vertical, and diagonal, for a total of nine pairs of subbands to obtain nine total high-frequency subband energies.

[0056] In an optional embodiment, in S2, for any scale Calculate the high-frequency subband energy in the horizontal, vertical, and diagonal directions respectively. , , ;

[0057] Calculate the arithmetic mean of the high-frequency subband energies in three directions. and standard deviation ;

[0058] Calculate the corresponding scale Directional indicators below Let be the ratio of the standard deviation to the arithmetic mean, satisfying . .

[0059] Specifically, the directionality index is used to determine whether chromatic noise at a specific scale is uniformly distributed or exhibits significant directionality. Continuing with the example from the previous step, let's focus on the second scale and assume that the total high-frequency sub-band energies in the three directions at this scale have been calculated: horizontal direction... It is 36, in the vertical direction. It is 38, diagonal direction The value is 34. Calculate the arithmetic mean of the energies of the three high-frequency subbands. , which is 36.

[0060] Calculate the energy and arithmetic mean of the three high-frequency subbands. The squared differences are 0, 4, and 4 respectively. The average of these three squared differences (i.e., the variance) is approximately 2.67. Taking the square root gives the standard deviation. The value is approximately 1.63. The directionality index D(2) at the second scale is calculated by dividing the standard deviation by the arithmetic mean, and it is approximately 0.045. The low directionality index indicates that the noise energy is relatively uniformly distributed in the three directions at the second scale, and the noise is isotropic. Similarly, the above calculations are performed on the first and third scales, respectively, to obtain the corresponding directionality indices D(1) and D(3).

[0061] It should be noted that, before calculation, the arithmetic mean was... Make a judgment; if its value is less than a very small positive number (such as...) ), then directly Set to 0.

[0062] S3. Based on the arctangent function of the difference between the kurtosis coefficient and the preset reference coefficient, determine the judgment threshold of the directional index; when the directional index is less than the judgment threshold, the sum of the high-frequency subband energies in the three directions at the corresponding scale is taken as the chromaticity noise energy; otherwise, the chromaticity noise energy is set to a preset low value.

[0063] At the k-th scale, the calculated directionality index D(k) is compared with the judgment threshold T: if D(k) is less than T, it indicates that the noise energy at this scale is relatively uniformly distributed in all directions, and is judged to be noise caused by carbon deposits. The chromatic noise energy at this scale is then... This equals the sum of the high-frequency subband energies in the three directions at that scale. If D(k) ≥ T, it indicates that the noise energy has a clear directionality, and is judged as interference such as scratches. Therefore, the chromatic noise energy at that scale... Set it to a very small positive number, such as Figure 2 .

[0064] In an optional embodiment, in S3, the determination threshold T is calculated using the following formula: ,in, The kurtosis coefficient is the kurtosis coefficient of the global luminance histogram of the RGB image. As a preset baseline coefficient, This is the scaling factor. This is the bias coefficient.

[0065] Specifically, it is necessary to calculate the kurtosis coefficient K of the global brightness histogram of the input image. Kurtosis represents the sharpness of the data distribution. Assuming that for an input image, by analyzing the brightness value distribution of all pixels, the calculated kurtosis coefficient K is 4.2, a preset baseline coefficient is set. The scaling factor is 3.0. The bias coefficient is 0.15. The value is 0.25. Substituting this into the formula, the decision threshold T is calculated to be 0.381. The decision threshold T is then compared with the calculated directional index D(k) for each scale to make a judgment.

[0066] Among them, the preset benchmark coefficient The following method can be used to determine the baseline coefficient: acquire multiple images of the combustion chamber wall under conditions of no or light carbon buildup, calculate the kurtosis coefficient of the global brightness histogram of each image, and take the statistical median as the preset baseline coefficient. .

[0067] In an optional embodiment, the preset low value is 0.

[0068] Specifically, in the preceding steps, a directional index D(k) was calculated for each scale k, and a global decision threshold T was calculated. D(k) was compared with T, and if D(k) < T, the noise at that scale was considered to be isotropic, i.e., uniform and unstructured noise, and the noise energy was regarded as chromatic noise energy.

[0069] Assume the directionality index D(2) of the second scale is less than the judgment threshold T. The noise at the second scale is determined to be chromatic noise. In this case, the chromatic noise energy corresponding to this scale is set to a preset low value. The arithmetic mean originally calculated at the second scale... Forced zeroing. The purpose of this operation is to ignore or reduce the contribution of chromatic noise identified as uniform and non-directional in noise assessment, because this type of noise usually has little impact on visual quality or is unrelated to image texture.

[0070] In other embodiments, the preset low value can be set to a very small positive number close to zero, for example, the preset low value can be... .

[0071] S4, the chromatic noise energy obtained at each scale is weighted and summed with the center frequency corresponding to each scale to obtain the comprehensive noise index representing the degree of carbon deposition.

[0072] Specifically, the chromaticity noise energy calculated at the three scales , , Weighted summation is performed, and since carbon deposit texture is more pronounced in the mid-to-high frequency range, while the low frequency range is mostly background light or smooth areas, the weights of each scale are adjusted accordingly. It shows a positive correlation with the center frequency of the corresponding scale. Let the center frequencies of the 1st, 2nd, and 3rd scales be respectively... The weights then decrease sequentially from high to low frequency. The overall noise index I is calculated using the formula... Calculations show that the higher the value of the comprehensive noise index I, the more severe the carbon buildup on the inner wall of the combustion chamber.

[0073] In an optional embodiment, in S4, the chromaticity noise energy obtained at the first, second, and third scales is acquired, and denoted as follows: , , ;

[0074] Weighting coefficients are assigned to the corresponding scales based on the order of their center frequencies from high to low. , , ;

[0075] Calculate the overall noise index ,satisfy .

[0076] Specifically, the chromaticity noise energy is collected after judgment and processing at all scales: assuming that after comparison with the judgment threshold, the noise at scale 1 is judged to be anisotropic, the chromaticity noise energy... The average energy value of the original calculation is retained as 50; the noise at the second scale is determined to be isotropic, and the chromaticity noise energy... Set to 0; the noise at the third scale is also determined to be anisotropic, and the chromaticity noise energy... Retain the original value of 15.

[0077] The calculation is performed based on preset weighting coefficients, which reflect the differences in the impact of noise of different frequencies on human vision. High-frequency noise is usually more easily perceived and therefore has the highest weight. Weighting of the first scale. The weight of the second scale is 0.5. The weight of the third scale is 0.3. The value is 0.2. Multiplying the chromaticity noise energy at each scale by the corresponding weighting coefficient and summing all the results, the overall noise index I is 28.

[0078] An embodiment of the machine vision-based combustion chamber wall carbon deposit detection system provided by the present invention:

[0079] like Figure 3As shown, the machine vision-based combustion chamber wall carbon deposit detection system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned machine vision-based combustion chamber wall carbon deposit detection method.

[0080] The machine vision-based combustion chamber wall carbon deposit detection system also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0081] In addition, in the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

Claims

1. A method for detecting carbon deposition on the inner wall of a combustion chamber based on machine vision, characterized in that, The method comprises the following steps: S1, acquiring an RGB image of the inner wall of the combustion chamber, calculating a kurtosis coefficient of a global brightness histogram of the RGB image, based on a ratio of a neighborhood brightness average of each pixel point in the RGB image to a preset brightness reference, performing nonlinear function weighting on R, G and B components of each pixel point to obtain a first chroma component and a second chroma component, comprising: for any pixel point in the RGB image, constructing a 5x5 neighborhood window, calculating an arithmetic average of brightness of all pixels in the neighborhood window as a neighborhood brightness average ; Computing weighting coefficients wherein, is a preset luminance reference; The R, G, B components of the pixel point are multiplied by the weighting coefficients respectively , to obtain weighted components , , ; calculating a first chroma component based on the weighted components , a second chroma component ; S2, performing two-dimensional discrete wavelet transform on the first chroma component and the second chroma component respectively to obtain high-frequency subbands in three directions of horizontal, vertical and diagonal respectively at least two scales; at the same scale, the high-frequency subband energies of the first chroma component and the second chroma component in the three directions of horizontal, vertical and diagonal are summarized to obtain energy distribution in the three directions at the corresponding scale, and a directionality index representing the anisotropy degree of chroma noise is calculated based on the energy distribution; for any scale , the high-frequency subband energies in the three directions of horizontal, vertical and diagonal are calculated respectively 、 、 ; calculating an arithmetic mean of the high-frequency subband energies in the three directions and the standard deviation ; Computing the directional index under the corresponding scale is the ratio of the standard deviation to the arithmetic mean, satisfying ;​ S3, determining that the decision threshold of the directionality index satisfies: a kurtosis coefficient for a global luminance histogram of the RGB image, a preset reference coefficient, a scaling coefficient, a bias coefficient; When the directionality index is less than the decision threshold, the sum of the high-frequency sub-band energies of the three directions under the corresponding scale is taken as the chroma noise energy; otherwise, the chroma noise energy is set as a preset low value; S4, weighting and summing the chroma noise energies obtained under each scale in combination with the center frequencies corresponding to each scale to obtain a comprehensive noise index representing the degree of carbon deposition.

2. The machine vision-based combustion chamber inner wall carbon deposition detection method according to claim 1, characterized in that, In S2, the first chroma component and the second chroma component are respectively subjected to three-layer two-dimensional discrete wavelet decomposition by using a Daubechies4 wavelet base to obtain the horizontal high-frequency sub-band, the vertical high-frequency sub-band and the diagonal high-frequency sub-band under the 1st, 2nd and 3rd scales.

3. The machine vision-based combustion chamber inner wall carbon deposition detection method according to claim 2, characterized in that, In S2, for any scale and direction , the average of the absolute values of all wavelet coefficients of the first chroma component in the high frequency subband of direction is calculated to obtain the high frequency subband energy of the first chroma component; calculating an average of absolute values of all wavelet coefficients of the second chrominance component in a high frequency subband of the direction , obtaining a high frequency subband energy of the second chrominance component ; adding high frequency subband energy of the first chroma component to high frequency subband energy of the second chroma component to obtain total high frequency subband energy in the scale, the direction satisfying .

4. The machine vision-based combustion chamber inner wall carbon deposition detection method according to claim 1, characterized in that, The preset low value is 0.

5. The machine vision-based combustion chamber inner wall carbon deposition detection method according to any one of claims 1-4, characterized in that, In S4, the chrominance noise energy obtained under the 1st, 2nd and 3rd scales is acquired, respectively denoted as , , ; Set a weighting coefficient for each scale respectively , , The weighting coefficient is positively correlated with the center frequency of the corresponding scale. Computing an overall noise index , satisfying .

6. The machine vision-based combustion chamber inner wall carbon deposition detection method according to claim 5, wherein, 、 、 。 7. A machine vision-based system for detecting internal wall carbon deposition in a combustion chamber, characterized in that, The device comprises a memory and a processor, and the memory stores computer program instructions, which, when executed by the processor, implement the method for detecting carbon deposition on the inner wall of a combustion chamber based on machine vision according to any one of claims 1-6.

Images