A Diesel Engine Abrasive Color Recognition Method Based on Adaptive Clustering and Dual Mapping Standard

Abstract

This invention provides a diesel engine abrasive particle color recognition method based on adaptive clustering and a dual-mapping standard, comprising the following steps: automatic extraction of diesel engine abrasive particle color through adaptive clustering; and automatic recognition of diesel engine abrasive particle color based on the extracted abrasive particle color using a dual-mapping standard. The beneficial effects of this invention are: automatically extracting abrasive particle color and associating it with abrasive particle material and harsh wear process, satisfying the two major work objectives of diesel engine abrasive particle material identification and harsh wear process identification, and achieving automatic recognition of diesel engine abrasive particle color using a dual-mapping standard; introducing a standard color system, the extracted color features are first compared with the standard color system to determine the color name, and then compared with an abrasive particle color library based on the standard color system for identification, which can effectively reduce the impact of uncertain contaminants such as dust and fibers on the identification process and reduce identification errors.

Description

Technical Field

[0001] This invention belongs to the field of diesel engine abrasive particle recognition technology, and in particular relates to a diesel engine abrasive particle color recognition method based on adaptive clustering and dual mapping standard. Background Technology

[0002] Diesel engines are among the most widely used internal combustion engines, playing a vital role in generator sets, construction machinery, and military equipment, and will continue to play an irreplaceable role for a considerable period. Currently, diesel engines are developing towards lightweight and high-strength designs, leading to increasingly harsh working conditions for their friction pairs, pushing them closer to their mechanical limits than ever before, thus creating a stronger demand for friction and wear monitoring. Abrasive analysis is one of the main methods for friction and wear monitoring and is widely used in the research and development and application stages of enhanced diesel engines. Abrasive analysis typically involves six dimensions: color, contour, edge details, surface problems, size, and thickness. Color is the primary feature considered in diesel engine abrasive analysis, often used to identify and initially determine the abrasive material and whether it has undergone severe wear such as high temperature or oxidation during its formation. Other features are identified after color recognition to determine whether further identification or auxiliary identification is necessary. Therefore, color recognition plays a crucial and irreplaceable role in diesel engine abrasive analysis.

[0003] At present, visual observation and experience-based judgment are the main methods for identifying the color of diesel engine abrasive particles. However, this method relies heavily on the professional knowledge and experience of the analysts, making it difficult to form a unified quantitative evaluation standard. The evaluation results are greatly affected by subjective factors, and the consistency of identification is difficult to guarantee.

[0004] Problems with existing technology:

[0005] (1) The K-means algorithm's hyperparameter K needs to be preset by human intervention, making it impossible to achieve adaptive automatic calculation and subject to subjective factors. (2) The first principal component color is used as the first criterion for judgment, but diesel engine abrasive particles often contain not only the original abrasive color, but also abrasive oxidation color and tempering color features, and the area ratio of each color is random. The first principal component color of abrasive particles of the same material is uncertain. Therefore, there is a large possibility of misjudgment in specific application scenarios of diesel engines. (3) The standard color library only contains material colors, while the ferrographic image of diesel engine lubricating oil often contains a variety of interference colors such as air dust, gravel, fibers, and mica flakes. Using the minimum Euclidean distance from the standard material color as the evaluation criterion will force various pollution colors to be identified as abrasive particles, causing misjudgment. (4) Color recognition is carried out in the RGB space, but the RGB space is based on the linear response design of hardware devices. Colors with similar numerical Euclidean distances may have great differences in human perception, thus causing misjudgment. (5) Focusing only on abrasive material identification, in diesel engine abrasive color identification, assessing whether the abrasive has experienced severe wear conditions such as high temperature and oxidation during the abrasive formation process is also an important working goal of abrasive color identification.

[0006] In summary, those skilled in the art often hold a common technical bias when facing the challenge of automatic color recognition of diesel engine abrasive particles: they believe that simply applying general image color recognition techniques (such as K-means with a fixed K value, calculated in RGB space) directly to abrasive particle images and establishing a simple material color library will achieve effective recognition. This bias stems from neglecting the unique characteristics of diesel engine abrasive particle colors, namely, the coexistence of multiple colors in a single abrasive particle, random color proportions, complex interference colors, and a strong correlation between color and the wear process. Therefore, existing technical solutions, in this specific application scenario, consistently fail to overcome the fundamental defects of high false positive rates and inability to identify the wear process. This invention is proposed precisely to overcome this technical bias. Summary of the Invention

[0007] In view of this, the present invention aims to propose a diesel engine abrasive particle color recognition method based on adaptive clustering and dual mapping standard, so as to solve at least one of the problems of the above-mentioned prior art.

[0008] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0009] A diesel engine abrasive grain color recognition method based on adaptive clustering and dual mapping standard includes the following steps:

[0010] S1. Automatic extraction of diesel engine abrasive particle color through adaptive clustering;

[0011] S2. Based on the extracted diesel engine abrasive color, perform automatic identification of diesel engine abrasive color using a dual-mapping standard;

[0012] In step S1, the diesel engine abrasive particle color adaptive clustering automatic extraction includes:

[0013] S11. Input abrasive particle image;

[0014] S12. Based on the abrasive grain image, extract the abrasive grain contour boundary and pixel information within the contour.

[0015] S13. Convert the pixel information within the outline from RGB space to Lab space;

[0016] S14. Based on the pixel information within the contour in Lab space, determine the optimal K value for the K-means algorithm;

[0017] S15. Based on the optimal K value and the pixel information within the contour in Lab space, perform K-means clustering to obtain the color of the abrasive grain color class;

[0018] In step S2, based on the extracted diesel engine abrasive grain color, the diesel engine abrasive grain color is automatically identified using a dual-mapping standard, including:

[0019] S21. Select a standard color system;

[0020] S22. Based on the selected standard color system, construct the correspondence between the standard color and the abrasive material and the abrasive formation process;

[0021] S23. Abrasive Grain Color Recognition: Based on the color of the abrasive grain color class, match the corresponding standard color in the standard color system;

[0022] S24. Based on the abrasive color identified in step S23 and the corresponding relationship in step S22, identify the abrasive material and the abrasive formation process.

[0023] Furthermore, in step S12, based on the abrasive grain image, the abrasive grain contour boundary and pixel information within the contour are extracted, including:

[0024] S121. Extract the abrasive grain contour boundary based on the abrasive grain image;

[0025] S122. Set the abrasive area threshold, filter out small abrasive particles based on the abrasive area threshold, set the remaining abrasive particles as the abrasive particles of interest, and number the abrasive particles of interest according to their area size.

[0026] S123. In the abrasive image, draw the outline of the abrasive grain of interest, and display the abrasive grain number at the center of the abrasive grain of interest.

[0027] S124. Extract pixel information within the contour of the abrasive grains of interest.

[0028] Furthermore, in step S13, the pixel information within the contour is converted from RGB space to Lab space, including:

[0029] S131. Convert the pixel information within the outline from RGB space to Lab space;

[0030] S132, for pixels , , The vector is normalized.

[0031] Furthermore, in step S14, determining the optimal K value for the K-means algorithm includes:

[0032] S141. Set the initial optimization range for the K value;

[0033] S142. Based on the initial optimization range of K value, traverse the K values ​​in the Lab space, perform K-means clustering for each K value, and calculate the SSE value;

[0034] S143. Set the absolute value of the difference between a certain K value and the SSE value of the previous K value as the absolute difference of the certain K value. Based on the absolute difference of each K value Calculate the mean absolute difference Find the last one that is greater than of The corresponding K value is used as the initial optimal K value;

[0035] S144. Set the range for the second-order optimization of the K value;

[0036] S145. Based on the quadratic optimization range of K values, traverse the K values ​​in the Lab space, perform K-means clustering for each K value, and calculate the silhouette coefficient;

[0037] S146. Find the K value corresponding to the maximum profile coefficient and set it as the optimal K value.

[0038] Furthermore, in step S15, based on the optimal K value and the pixel information within the contour in Lab space, K-means clustering is performed to obtain the color of the abrasive grain color class, including:

[0039] S151. Based on the optimal K value and the pixel information within the contour of Lab space, the K-means algorithm is used to cluster the pixels of the abrasive grains to form abrasive grain color class.

[0040] S152. Obtain the color of the abrasive color class: Record the L, a, and b values ​​of the cluster center points as the color of the abrasive color class, and sort the clusters based on the number of pixels.

[0041] Furthermore, in step S22, the correspondence between the standard color and the abrasive material and abrasive formation process is established, including:

[0042] S221. Organize the color representation of abrasive grains of various materials and map them to the standard color system;

[0043] S222. Organize the color expression of abrasive particles of various materials after oxidation and high temperature during diesel engine operation, and correspond them to the standard color system.

[0044] Furthermore, in step S23, based on the color of the abrasive grain color class extracted in step S15, abrasive grain color identification is performed in the standard color system, including:

[0045] Based on the standard color system, the Euclidean distance between the color of the abrasive color class and each standard color is calculated, and the standard color with the smallest Euclidean distance is identified as the abrasive color.

[0046] Compared with existing technologies, the diesel engine abrasive particle color recognition method based on adaptive clustering and dual mapping standard described in this invention has the following advantages:

[0047] (1) Based on image processing technology, this invention automatically extracts the color of abrasive particles and associates the abrasive material with the harsh wear process, which meets the two major work objectives of diesel engine abrasive material identification and harsh wear process identification, and realizes the automatic identification of diesel engine abrasive color dual mapping standard.

[0048] (2) Based on the K-means algorithm, this invention realizes automatic extraction of abrasive grain color in Lab space and proposes an adaptive method for determining the hyperparameter of K value of K-means clustering algorithm, which reduces the subjective role of human in the process.

[0049] (3) In this invention, a standard color system is introduced. The extracted color features are compared with the standard color system to determine the color name. Then, they are compared with the abrasive color library based on the standard color system for identification. This can effectively reduce the impact of uncertain pollutants such as dust and fibers on the identification process and reduce identification errors.

[0050] (4) The diesel engine abrasive color recognition method based on adaptive clustering and dual mapping standard provided by the present invention can effectively improve work efficiency and recognition consistency, and promote the application of automated abrasive analysis in this field. Attached Figure Description

[0051] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0052] Figure 1 This is a schematic diagram of the automatic extraction process for adaptive clustering of diesel engine abrasive particle color according to an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram of the automatic identification process for diesel engine abrasive grain color using the dual-mapping standard as described in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the input abrasive grain image as described in an embodiment of the present invention;

[0055] Figure 4 This is a schematic diagram showing the outline and numbering of the abrasive particles of interest according to an embodiment of the present invention. Detailed Implementation

[0056] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0057] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0058] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0059] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0060] like Figure 1 As shown, an automatic extraction method for adaptive clustering of diesel engine abrasive grain color is as follows:

[0061] A1: Input abrasive image. Input a JPG format abrasive image taken by equipment such as ferrography, including the foreground color of the abrasive and the background color of the oil sample.

[0062] A2: Extract the abrasive grain contour boundary and pixel information within the contour, as detailed below:

[0063] A21: Extract abrasive grain contour boundaries;

[0064] A22: Set the abrasive grain area threshold (number of pixels), filter out small abrasive grains that are not of interest, and number the abrasive grains of interest in order of area size;

[0065] A23: Draw the outline of each abrasive grain of interest in the original abrasive grain image and display the abrasive grain number at the center of the abrasive grain;

[0066] A24: Extract pixel information within the contour of each abrasive grain of interest in sequence.

[0067] A3: Convert the pixel information within the outline from RGB space to Lab space, as follows:

[0068] A31: Convert pixel information within the outline from RGB space to Lab space;

[0069] A32: For each pixel , , The vectors are normalized using the following formula:

[0070] ;

[0071] ;

[0072] ;

[0073] In the formula, , For normalized data, , , This is the data before normalization.

[0074] A4: Determine the optimal K value for the K-means algorithm, as follows:

[0075] A41: Settings Preliminary Optimization Range ;

[0076] A42: Traverse within the Lab space. Value range For each Perform K-means clustering and calculate the clustering error SSE value;

[0077] A43: Calculate each The SSE value of the previous absolute value of the difference And calculate the mean absolute difference. Find the last one that is greater than of Corresponding Value, as the initial optimal value ,in, In the k-means algorithm The initial optimization value of the value, where, and Calculate according to the following formulas respectively;

[0078] ;

[0079] ;

[0080] A44: Confirm Value of the second optimization range , ;

[0081] A45: Traverse within the Lab space. Value range For each The values ​​are subjected to K-means clustering, and the silhouette coefficients are calculated.

[0082] A46: Find the point corresponding to the maximum profile coefficient. Value, determine the optimal value ,in, In the k-means algorithm The final optimal value.

[0083] A5: Extract the color contained in the abrasive grains, based on the optimal result obtained in step A4. value In the Lab space, the K-means algorithm is used to perform clustering calculations on each pixel of the abrasive grain, record the L, a, b values ​​of each cluster center point, and sort them according to the number of pixels in each category to extract the colors contained in the abrasive grain (Note: the L, a, b values ​​of each cluster center point are the L, a, b values ​​of each color contained in the abrasive grain).

[0084] like Figure 2 As shown, a method for automatic identification of diesel engine abrasive grain color using a dual-mapping standard is as follows:

[0085] B1: Choose a suitable standard color system, such as the CSS Color Module Level 4 standard color system, the RAL color chart standard color system, the Pantone color chart standard color system, etc.

[0086] B2: Establish the correspondence between standard colors and abrasive materials and abrasive formation processes, as detailed below:

[0087] The color representation of abrasive particles from various materials in diesel engines, as observed in ferrography and other equipment, is compiled and mapped to the standard colors described in step B1. Simultaneously, the color representation of abrasive particles formed after oxidation and high temperatures during diesel engine operation is also compiled and mapped to the standard colors described in B1. See Table 1.

[0088] Table 1

[0089] ;

[0090] B3: Abrasive Grain Color Recognition: Calculate the Euclidean distance between the colors of each cluster center mentioned in step A5 (Note: the L, a, and b values ​​of each cluster center point are the L, a, and b values ​​of each color contained in the abrasive grain) and the standard colors in the standard color system mentioned in step B1, and match the color name with the smallest Euclidean distance. The calculation expression is as follows:

[0091] ;

[0092] In the formula, , , Normalized for each cluster center , , value, , , Normalized for each standard color , , value.

[0093] B4: Identification of abrasive material and formation process, as detailed below:

[0094] Based on the correspondence between the abrasive color recognition results in step B3 and the abrasive color, material, and abrasive formation process in step B2, the abrasive material and abrasive formation process are identified, and the recognition conclusion is output. The recognition rules mainly include the following:

[0095] (1) All colors in step B3 that do not exceed the range of Table 7 are used for abrasive material identification according to Table 7, and are comprehensively judged according to the corresponding relationship;

[0096] (2) When all colors in step B3 exceed the range described in Table 1, they are determined to be other unidentified abrasive particles;

[0097] (3) When there is a conflict between the colors pointing to the material, the color with more pixels should be used as the primary color.

[0098] (4) When the color in step B3 includes the tempering color or oxidation color described in step B2, it indicates that there is abnormal wear during the abrasive grain formation process, and a relevant reminder is issued.

[0099] (5) When the abrasive material determined by color is not unique, output all possible results.

[0100] Example 1:

[0101] This embodiment provides an automatic extraction method for adaptive clustering of diesel engine abrasive grain color, as detailed below:

[0102] C1: Input abrasive image. Input a JPG format abrasive image taken by equipment such as ferrography, including the foreground color of the abrasive particles and the background color of the oil sample, such as... Figure 3 As shown.

[0103] C2: Extract the abrasive grain contour boundary and pixel information within the contour, as detailed below:

[0104] C21: Extracting abrasive grain contour boundaries, implemented using programming languages ​​such as MATLAB, as follows: First, the input abrasive grain image is converted into a 3D RGB matrix with width pixels × height pixels × 3, and then converted into a 2D matrix with width pixels × height pixels through grayscale transformation. Second, adaptive histogram equalization and 2D median filtering are performed on the grayscale image. Based on the Otsu method, the global threshold of the image is calculated to convert the grayscale image into a black and white binary image. Finally, morphological denoising is performed on the binary image, and abrasive grain contour boundary information is extracted based on the watershed algorithm.

[0105] C22: Set the abrasive grain area threshold to 8000 pixels to filter out small abrasive grains that are not of interest, and number the abrasive grains of interest in order of their area size.

[0106] C23: Draw the outline of the abrasive grain of interest in the original abrasive grain image and display the abrasive grain number at the center of the grain, such as... Figure 4 As shown;

[0107] C24: Extract the pixel information within the contour of each abrasive grain of interest in sequence. The pixel information within the contour is shown in Table 2.

[0108] Table 2

[0109] ;

[0110] C3: Convert the pixel information within the contour from RGB space to Lab space. Convert the pixel information within the contour of each abrasive grain of interest from RGB space to Lab space. Use rgb2lab in MATLAB software to achieve the color space conversion. The converted pixel information within the contour is shown in Table 3.

[0111] Table 3

[0112] ;

[0113] C4: Determine the optimal K value for the K-means algorithm, as follows:

[0114] C41: Settings Preliminary Optimization Range ;

[0115] C42: Traverse within the Lab space. Value range For each Perform K-means clustering and calculate the clustering error SSE value;

[0116] C43: Calculate each The SSE value of the previous absolute value of the difference And calculate the mean absolute difference. Find the last one that is greater than of Corresponding Value, as the initial optimal value ,in, In the k-means algorithm Preliminary optimization value, and The calculation expression is as follows:

[0117] ;

[0118] ;

[0119] SSE corresponding to each K value and As shown in Table 4, the optimal abrasive particles were initially determined. value It is 3;

[0120] Table 4

[0121] ;

[0122] C44: Confirm Value of the second optimization range , abrasive grains for ;

[0123] C45: Traverse within the Lab space. Value range For each K-means clustering was performed on the values, and the silhouette coefficients were calculated. The silhouette coefficients corresponding to each K value are shown in Table 5.

[0124] Table 5

[0125] ;

[0126] C46: Find the point corresponding to the maximum profile coefficient. Value, determine the optimal value The optimal abrasive grains are determined value The value is 2, where, In the k-means algorithm The final optimal value.

[0127] C5: Extract color from abrasive grains, based on the optimal result obtained in step C4. value In the Lab space, the K-means algorithm is used to perform clustering calculations on each pixel of the abrasive grain, and the L, a, b values ​​of each cluster center point are recorded. The clusters are then sorted according to the number of pixels in each category to extract the colors contained in the abrasive grain, as shown in Table 6.

[0128] Table 6

[0129] ;

[0130] This embodiment provides an automatic identification method for the color of diesel engine abrasive particles using a dual-mapping standard, as detailed below:

[0131] D1: Select a suitable standard color system. In this example, the CSS Color Module Level 4 standard color system developed by the W3C Alliance is selected. This system provides 143 standard colors, including white, silver, yellow blue, chrysanthemum, sand brown, and Peruvian.

[0132] D2: Construct a correspondence between standard colors and abrasive materials and formation processes. Taking iron, copper, and aluminum as examples, compile the color representation of abrasive particles of various diesel engine materials in equipment such as ferrography instruments, and map them to the standard colors described in step D1. Simultaneously, compile the color representation of abrasive particles formed by various materials after oxidation and high temperature during diesel engine operation in equipment such as ferrography instruments, and map them to the standard colors described in D1. As shown in Table 7.

[0133] Table 7

[0134] ;

[0135] D3: Abrasive Grain Color Recognition: Calculate the Euclidean distance between the colors of each cluster center in Table 6 and the standard colors in the CSS ColorModule Level 4 standard color system described in step D1. Match the color name with the smallest Euclidean distance. After calculation, the standard color corresponding to cluster center 1 (0.78 0.53 0.66) in Table 6 is sandy brown, and the standard color corresponding to cluster center 2 (0 0 0) is black. The calculation expression is as follows:

[0136] ;

[0137] In the formula, , , Normalized cluster centers , , value, , , Normalized for each standard color , , value.

[0138] D4: Identification of abrasive material and formation process, as detailed below:

[0139] Based on the abrasive color recognition results (sand brown, black) in step D3 and the correspondence between abrasive color, material, and abrasive formation process in step D2, the abrasive material and formation process are identified. The identification rules mainly include the following:

[0140] (1) All colors in step D3 that do not exceed the range of Table 7 are included in the abrasive material identification according to Table 7, and are comprehensively judged according to the corresponding relationship;

[0141] (2) When all colors in step D3 exceed the range described in Table 7, they are determined to be other unidentified abrasive particles;

[0142] (3) When there is a conflict between the colors pointing to the material, the color with more pixels should be used as the primary color.

[0143] (4) When the color in step D3 includes the tempering color or oxidation color described in step D2, it indicates that there is abnormal wear during the abrasive grain formation process, and a warning is issued.

[0144] (5) When the abrasive material determined by color is not unique, output all possible results.

[0145] According to rule (1), both sand brown and black are within the range of Table 7. Based on Table 7, sand brown corresponds to the original color of copper material, and black corresponds to the oxidation color of three materials such as iron, copper and aluminum. Therefore, it can be determined that the abrasive material is copper.

[0146] According to rule (4) of the identification rules, black is an oxidation color, which indicates that there may be abnormal wear during the abrasive grain formation process, and relevant reminders should be made.

[0147] Recognition conclusion: The abrasive grains in the input image are made of copper and show signs of oxidation, indicating that the engine may have abnormal wear.

[0148] Problems with existing technology:

[0149] (1) The K value hyperparameter of the K-means algorithm needs to be preset by human in advance, and cannot be automatically calculated adaptively.

[0150] (2) The color of the first principal component is used as the first criterion. However, the abrasive grains of diesel engines often contain not only the original color of the abrasive grains, but also the abrasive grain oxidation color and tempering color. Moreover, the area ratio of each color is random. The color of the first principal component of the abrasive grains of the same material is uncertain. Therefore, there is a large possibility of misjudgment in specific application scenarios of diesel engines.

[0151] (3) The standard color library only contains material colors, while the diesel engine lubricating oil ferrography image often contains a variety of interfering colors such as air dust, gravel, fibers, and mica flakes. Using the minimum Euclidean distance from the material standard color as the evaluation basis will force various pollution colors to be identified as abrasive particles, resulting in misjudgment.

[0152] (4) Color recognition is carried out in the RGB space. The RGB space is based on the linear response design of hardware devices. Colors with similar Euclidean distances may appear very different to the human eye, which may lead to misjudgment.

[0153] (5) Focusing only on abrasive material identification, in diesel engine abrasive color identification, assessing whether the abrasive has experienced severe wear conditions such as high temperature and oxidation during the abrasive formation process is also an important working goal of abrasive color identification.

[0154] The technical problem solved by this invention:

[0155] (1) This invention provides a diesel engine abrasive color recognition method based on adaptive clustering and dual mapping standard, including an automatic extraction method for diesel engine abrasive color by adaptive clustering and an automatic recognition method for diesel engine abrasive color by dual mapping standard. The computer recognition replaces the manual recognition, so as to solve the problems of high degree of human involvement, poor recognition consistency and easy misjudgment due to fatigue in the current stage, and improve the automation level of abrasive color recognition.

[0156] (2) The present invention provides an automatic extraction method for adaptive clustering of diesel engine abrasive color, which focuses on solving the problems of automatic determination of the number of colors contained in abrasive and automatic extraction of all main colors. It converts color information into accurate Lab color space values, realizes objective measurement of color features, eliminates human bias, and makes the extraction results repeatable and comparable. It extracts the abrasive contour boundary and pixel information within the contour, converts the color space from RGB to Lab, and then automatically optimizes the optimal K value of the K-means algorithm to accurately determine the number of colors contained in the abrasive. Based on K-means clustering, it extracts the L, a, and b values ​​of each color.

[0157] (3) The present invention provides an automatic identification method for diesel engine abrasive color based on dual mapping standard, which focuses on solving the color identification problem based on color extraction and the correlation mapping problem between color and abrasive material and wear process. By introducing a standard color system and using the minimum Euclidean distance, the abrasive color is accurately identified based on the clustering results. Based on the color performance of diesel engine abrasive of various materials under equipment such as ferrography, the corresponding standard color is matched, and its original color, oxidation color and standard color are statistically analyzed. The identified abrasive color is compared with the above statistical results to map the abrasive material and whether there is abnormal wear such as oxidation or high temperature during the abrasive formation process.

[0158] (4) This invention provides a diesel engine abrasive color recognition method based on adaptive clustering and dual mapping standard. From color space conversion to cluster K value optimization, to cluster center point color recognition, and then to associating abrasive material and abrasive formation process, it realizes the automation of abrasive color recognition in diesel engine friction and wear monitoring, solves the problem of excessive human intervention at the present stage, and provides a friction and wear monitoring method for strengthening diesel engines.

[0159] The inventiveness of this invention:

[0160] The inventiveness of this invention does not lie in the isolated use of individual technical features such as the K-means algorithm, Lab color space, or standard color system, but rather in proposing a complete set of interconnected and mutually supportive collaborative solutions for the specific, complex, and interference-ridden technical scenario of "diesel engine abrasive particle color analysis." This solution solves long-standing technical problems through a series of non-obvious improvements and combinations, specifically in the following three aspects:

[0161] (1) A “two-step adaptive K-value determination method” was created, which overcame the bottleneck of identifying the dynamic changes in the number of multi-color abrasive particles.

[0162] The technical challenge to be addressed: Diesel engine abrasive particles often simultaneously contain primary color, oxidation color, and tempering color, and their color variety and area proportion are dynamic, unknown, and random. The commonly used K-means algorithm requires a preset K value, which contradicts the needs of practical applications and is the core obstacle to automated identification.

[0163] Non-obviousness: This invention does not employ the simple Elbow Method or the contour coefficient method, but creatively combines SSE change rate analysis (preliminary optimization) with contour coefficient fine-tuning (secondary optimization). First, the steep decreasing trend of the SSE change rate quickly narrows down the approximate range of K values, effectively avoiding the subjectivity of the Elbow Method. Then, within this narrow range, the contour coefficient is used to find the optimal K value for the best clustering effect. This "coarse-fine" combined secondary optimization strategy is a customized solution for the high uncertainty of abrasive grain color clustering. Its parameter selection and judgment logic (such as "the last K value greater than the average absolute difference") are not common knowledge in the field, but rather effective methods verified through extensive experiments, ensuring accurate identification of the true number of abrasive grain color types even in complex backgrounds.

[0164] (2) A “color-material-process dual mapping model based on standard color system” was constructed to realize comprehensive diagnosis from color to wear status.

[0165] Technical challenges to be addressed: Existing technologies only map color to material, while seriously neglecting the more critical assessment of the "severity of wear process" (such as whether it has experienced high temperature or oxidation) in diesel engine condition monitoring.

[0166] Non-obviousness: This invention, for the first time within the framework of a standard color system, systematically and structurally establishes a dual mapping relationship between color and both "abrasive material" and "abrasive formation process" (see Tables 1 and 7 in the specification). This is not merely a simple database expansion, but a revolutionary diagnostic paradigm. By listing color features characterizing abnormal wear processes, such as "oxidation color" and "tempering color," alongside "original color" as identification targets, this method outputs not just a single material conclusion, but a comprehensive diagnostic report encompassing both "material identification" and "abnormal wear process warning." This dual mapping relationship cleverly utilizes the physicochemical origins of abrasive color under specific diesel engine operating conditions, transforming color information into a direct indicator of equipment health status, with technical effects far exceeding simple material identification.

[0167] (3) A “multi-color parallel recognition and decision-making mechanism in Lab space” was designed, which effectively improved the robustness in complex scenarios.

[0168] The technical challenge to be solved: The RGB space is based on the linear response design of hardware devices. Colors that are numerically close in Euclidean distance may appear very different to the human eye, leading to misjudgment.

[0169] Non-obviousness: This invention organically combines three steps: color space conversion to Lab, extraction of all primary colors (not just the first primary color), and introduction of a standard color system as a "color dictionary," forming a powerful anti-interference recognition chain.

[0170] Clustering and distance calculations are performed in a Lab space that aligns with human visual perception, fundamentally reducing misjudgments caused by perceptual differences.

[0171] By identifying all clustered colors and comparing them with a standard color system, "interference colors" that do not belong to any abrasive material / process can be effectively identified and eliminated in the decision-making process.

[0172] The final decision rules (such as "all colors within the range are included in the identification", "in case of color conflict, the color with more pixels is prioritized", and "an alarm is triggered if oxidation / tempering colors are included") are expert system logic tailored for diesel engine abrasive analysis. This mechanism ensures that the system can make comprehensive and robust judgments even in complex samples with interference.

[0173] In the adaptive determination of hyperparameters for K-means clustering algorithms, existing techniques require pre-setting of K values, which is subject to subjective influence. This invention, however, achieves adaptive automatic calculation of K values ​​by setting initial and secondary optimization ranges for K values ​​and combining the calculation of SSE values ​​and contour coefficients. Specifically, in the initial optimization stage, the SSE value corresponding to each K value is calculated by traversing the range of K values, and the absolute difference and average absolute difference between adjacent K values' SSE values ​​are calculated. The last K value greater than the average absolute difference is found as the initial optimal K value. In the secondary optimization stage, a secondary optimization range is set near the initial optimal K value. This range is traversed, and the contour coefficient corresponding to each K value is calculated. The K value corresponding to the largest contour coefficient is found as the optimal K value. This method is based on in-depth analysis and research of the color distribution characteristics of abrasive grain images, comprehensively considering clustering error and clustering effect, and can automatically and accurately determine K values, avoiding the subjectivity and uncertainty of manually pre-setting K values. Compared with existing technologies, the adaptive K-value determination method of the present invention is a completely new technical approach, which requires a deep understanding and mastery of clustering algorithms and abrasive image features, and is not a simple improvement of existing technologies.

[0174] The introduction of a standard color system is also a key innovation of this invention. Existing standard color libraries only contain material colors and use the minimum Euclidean distance from the material's standard color as the evaluation criterion, making them susceptible to interference colors. This invention introduces a standard color system, where extracted color features are first compared with the standard color system to determine the color name, and then compared with an abrasive color library based on the standard color system for identification. In practical applications, the standard color system provides a unified color standard, making abrasive color identification more accurate and objective. For example, when faced with interference colors in the ferrographic image of diesel engine lubricating oil, comparison with the standard color system can distinguish the interference colors from the abrasive colors, effectively reducing the impact of uncertain contaminants such as dust and fibers on the identification process. This method of introducing a standard color system requires in-depth research into color theory and abrasive analysis; it is a cross-disciplinary technological innovation, not a simple patchwork of existing technologies.

[0175] (4) The various technical features of this invention are not simply superimposed, but rather, based on a deep understanding of the special characteristics of diesel engine abrasive particles, they are coupled and work synergistically to form a complete technical solution specifically designed to solve a particular technical problem in this field. This solution produces multiple synergistic and beneficial effects, such as "improving identification accuracy, realizing wear process diagnosis, and enhancing anti-interference capabilities." These effects are difficult for those skilled in the art to anticipate using conventional logic based on existing technology.

[0176] Advantages and benefits of this invention:

[0177] (1) Based on image processing technology, this invention automatically extracts the color of abrasive particles and associates the abrasive material with the harsh wear process, thus meeting the two major work objectives of diesel engine abrasive material identification and harsh wear process identification, and realizing the automatic identification of diesel engine abrasive color.

[0178] (2) Based on the K-means algorithm, this invention realizes automatic extraction of abrasive grain color in Lab space and proposes an adaptive method for determining the hyperparameter of K value of K-means clustering algorithm, which reduces the subjective role of human in the process.

[0179] (3) In this invention, a standard color system is introduced. The extracted color features are compared with the standard color system to determine the color name. Then, they are compared with the abrasive color library based on the standard color system for identification. This can effectively reduce the impact of uncertain pollutants such as dust and fibers on the identification process and reduce identification errors.

[0180] (4) The automatic method for the color of diesel engine abrasive particles provided by the present invention can effectively improve work efficiency and recognition consistency, and promote the application of automated abrasive particle analysis in this field.

[0181] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A diesel engine abrasive particle color recognition method based on adaptive clustering and double mapping criteria, characterized in that: Includes the following steps: S1. Automatic extraction of diesel engine abrasive particle color through adaptive clustering; S2. Based on the extracted diesel engine abrasive color, perform automatic identification of diesel engine abrasive color using a dual-mapping standard; In step S1, the diesel engine abrasive particle color adaptive clustering automatic extraction includes: S11. Input abrasive particle image; S12. Based on the abrasive grain image, extract the abrasive grain contour boundary and pixel information within the contour, including: S121. Extract the abrasive grain contour boundary based on the abrasive grain image; S122. Set the abrasive area threshold, filter out small abrasive particles based on the abrasive area threshold, set the remaining abrasive particles as the abrasive particles of interest, and number the abrasive particles of interest according to their area size. S123. In the abrasive image, draw the outline of the abrasive grain of interest, and display the abrasive grain number at the center of the abrasive grain of interest. S124. Extract pixel information within the contour of the abrasive grains of interest; S13. Convert the pixel information within the contour from RGB space to Lab space, and normalize the L, a, and b components of the pixel. The converted Lab space is used for subsequent cluster analysis. S14. Based on the pixel information within the contour in Lab space, a two-step adaptive method is used to determine the optimal K value of the K-means algorithm, including: S141. Traverse the K values ​​within the preset initial optimization range, perform K-means clustering for each K value and calculate the corresponding SSE; S142, calculate the absolute value of the SSE difference between each K value and the previous K value and based on all calculate the average absolute difference ; S143, select the last satisfied item within the initial optimization range. The K value was determined to be the preliminary optimal value. ; S144, with Define a quadratic optimization range around the center and iterate through the K values ​​within that range. For each K value, K-means clustering is performed and the silhouette coefficient is calculated; S145. The K value with the largest contour coefficient is determined as the optimal K value; S15. Based on the optimal K value and the pixel information within the contour in Lab space, perform K-means clustering to obtain the colors of all color classes of the abrasive grains, i.e., all principal component colors. In step S2, based on the extracted diesel engine abrasive grain color, the diesel engine abrasive grain color is automatically identified using a dual-mapping standard, including: S21. Select the standard color system. The standard color system selected is the CSS Color Module Level 4 standard color system formulated by the W3C Alliance. This standard color system provides 143 standard colors, including white, silver, yellow blue, chrysanthemum, sand brown, and Peruvian. S22. Based on the selected standard color system, construct the correspondence between the standard color and the abrasive material and the abrasive formation process; S23. Abrasive color recognition: Calculate the distance between all abrasive color categories obtained in step S15 and the standard colors in the standard color system, match the standard color with the smallest distance and output the corresponding standard color name; S24. Abrasive material and formation process identification: Based on the abrasive color identified in step S23 and the corresponding relationship in step S22, the abrasive material and abrasive formation process are identified, and warnings are given for the identified severe wear processes. In step S24, the identification rules are as follows: In step S23, all colors that do not exceed the range of the corresponding relationship library in step S22 are used for abrasive material identification based on the corresponding relationship in step S22, and a comprehensive judgment is made based on the corresponding relationship. When all colors in step S23 exceed the range described in the corresponding relational library in step S22, they are determined to be other unidentified abrasive particles. When there is a conflict between the colors pointing to different materials, the color with more pixels should be prioritized. When the color in step S23 includes the tempering color or oxidation color described in step S22, it indicates that there is abnormal wear during the abrasive grain formation process, and a warning is issued. When the abrasive material determined by color is not unique, output all possible results.

2. The diesel engine abrasive particle color recognition method based on adaptive clustering and dual mapping standard according to claim 1, characterized in that: In step S15, based on the optimal K value and the pixel information within the contour in Lab space, K-means clustering is performed to obtain the color of the abrasive grain color class, including: S151. Based on the optimal K value and the pixel information within the contour of Lab space, the K-means algorithm is used to cluster the pixels of the abrasive grains to form abrasive grain color class. S152. Obtain the color of the abrasive color class: Record the L, a, and b values ​​of the cluster center points as the color of the abrasive color class, and sort the clusters based on the number of pixels.

3. The diesel engine abrasive particle color recognition method based on adaptive clustering and dual mapping standard according to claim 1, characterized in that: In step S22, the correspondence between standard colors and abrasive materials and abrasive formation processes is established, including: S221. Organize the color representation of abrasive grains of various materials and map them to the standard color system; S222. Organize the color expression of abrasive particles of various materials after oxidation and high temperature during diesel engine operation, and correspond them to the standard color system.

Images