Piston cleanliness intelligent evaluation method, device and system and storage medium

Abstract

The invention provides an intelligent evaluation method, device and system for the cleanliness of a piston and a storage medium. The intelligent evaluation method for the cleanliness of the piston comprises the steps of obtaining image data of the piston; preprocessing the image data of the piston to obtain sediment image data of the piston after preprocessing; and inputting the preprocessed sediment image data into a trained intelligent evaluation model based on a deep learning network, identifying sediment feature information in the sediment image data by the intelligent evaluation model, and calculating and outputting an identification result according to the sediment feature information. The sediment feature information in the piston image data is identified through the trained intelligent evaluation model based on the deep learning network, the identification effect on the sediment of the piston is improved, the identification result is more scientific and objective, the evaluation result of the piston can be further prevented from being influenced by subjective factors, and the reliability is high.

Description

Technical Field

[0001] This application generally relates to the technical field of piston detection devices, and more specifically to an intelligent evaluation method, device, system, and storage medium for piston cleanliness. Background Art

[0002] As an important performance evaluation index of lubricating oil, the evaluation of high-temperature deposits, viscosity growth, and ring sticking of engine lubricating oil is carried out using an engine test bench. Due to the influence of heat and metal catalysis, engine oil continuously undergoes chemical reactions such as oxidation, cracking, and polymerization, increasing the viscosity of the lubricating oil and generating paint films and carbon deposits on the piston surface, resulting in wear and damage of engine components. Naphthenes, aromatics, and heteroatom compounds containing sulfur and nitrogen in the engine oil base oil are the main sources of sludge and paint films. In the engine test bench, the quality of the engine oil cleanliness is mainly evaluated by the score of the deposits on the piston surface. Piston cleanliness is an important index for evaluating the quality of lubricating oil.

[0003] In the prior art, the piston is generally fixed on a special auxiliary tooling to observe the piston ring grooves and skirt, and professional scoring personnel are required to visually score and evaluate, which has a large subjectivity, and there are also differences in the scoring results of each professional, which is not conducive to the objectivity of piston scoring.

[0004] Therefore, it is necessary to provide an intelligent evaluation method, device, system, and storage medium for piston cleanliness to at least partially solve the above problems. Summary of the Invention

[0005] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further described in detail in the Detailed Description section. The Summary of the Invention section of this application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0006] To at least partially solve the above problems, a first aspect of this application provides an intelligent evaluation method for piston cleanliness, and the intelligent evaluation method for piston cleanliness includes:

[0007] Obtain image data of the piston;

[0008] Preprocess the image data of the piston to obtain the deposit image data of the piston after preprocessing;

[0009] Input the preprocessed deposit image data into a trained intelligent evaluation model based on a deep learning network, and the intelligent evaluation model identifies the deposit feature information in the deposit image data, and calculates and outputs an identification result according to the deposit feature information.

[0010] In one embodiment of the present application, the step of obtaining the image data of the piston includes:

[0011] Divide the area of the piston to obtain a plurality of sub-areas;

[0012] Obtain the image data within each of the sub-areas;

[0013] The plurality of sub-areas respectively correspond to the piston ring grooves, piston skirt, piston ring platforms and piston inner tops of the piston.

[0014] In one embodiment of the present application, the step of identifying the sediment characteristic information in the sediment image data by the intelligent evaluation model and calculating and outputting the identification result according to the sediment characteristic information includes:

[0015] Identify the sediment characteristic information within each of the sub-areas by the intelligent evaluation model;

[0016] Calculate the identification result according to the sediment characteristic information within all the sub-areas.

[0017] In one embodiment of the present application, the step of preprocessing the image data of the piston includes:

[0018] Perform at least one of grayscale processing, image enhancement processing, filtering processing and binarization processing on the image data of the piston.

[0019] In one embodiment of the present application, the sediment characteristic information includes a sediment severity coefficient, a sediment property probability and a sediment area coefficient.

[0020] In one embodiment of the present application, the identification result includes the cleanliness score or sediment score of the piston.

[0021] In one embodiment of the present application, when the obtained image data of the piston includes video data, the step of preprocessing the image data of the piston further includes:

[0022] Decompose the video data to obtain picture data for inputting into the intelligent evaluation model.

[0023] In one embodiment of the present application, the training of the intelligent evaluation model includes the following training steps:

[0024] Obtain a data set, and the data set includes a training set;

[0025] Input the sample images in the training set into the intelligent evaluation model to be trained, and the intelligent evaluation model to be trained performs the following operations:

[0026] Feature extraction is performed on the sample image to obtain sediment image features;

[0027] The sediment image features are converted into sediment feature information;

[0028] The sediment feature information is recognized, and the obtained result is compared with the standard result of the sample image, and model weight parameters are output based on the comparison result;

[0029] Iterate the above training steps until the intelligent evaluation model to be trained converges, and obtain the trained intelligent evaluation model based on the deep learning network.

[0030] In an embodiment of the present application, before inputting the sample image in the training set into the intelligent evaluation model to be trained, it further includes:

[0031] Image processing operations are performed on the sample image in the training set through image processing operations, so as to input the sample image after image processing into the intelligent evaluation model to be trained;

[0032] Among them, the image processing operations include at least one of grayscale processing, image enhancement processing, filtering processing, and binarization processing.

[0033] In an embodiment of the present application, the data set further includes a validation set, and the training steps of the intelligent evaluation model further include:

[0034] The intelligent evaluation model after multiple trainings is verified through the validation set to adjust the parameters of the intelligent evaluation model.

[0035] In an embodiment of the present application, the data set further includes a test set, and the test set is used to test the trained intelligent evaluation model based on the deep learning network to verify the recognition accuracy of the intelligent evaluation model.

[0036] A second aspect of the present application provides a piston cleanliness intelligent evaluation device, and the piston cleanliness intelligent evaluation device includes:

[0037] An image acquisition module for acquiring image data of a piston;

[0038] A preprocessing module for preprocessing the image data of the piston to obtain the sediment image data of the piston after preprocessing; and

[0039] An image recognition module, configured to input the preprocessed sediment image data into a trained intelligent evaluation model based on a deep learning network, and the intelligent evaluation model identifies sediment feature information in the sediment image data, calculates and outputs an identification result according to the sediment feature information.

[0040] A third aspect of the present application provides another intelligent piston cleanliness evaluation device, including a memory and a processor, wherein a computer program run by the processor is stored on the memory, and when the computer program is run by the processor, the processor executes the piston cleanliness intelligent evaluation method described in the first aspect of the present application.

[0041] A fourth aspect of the present application provides an intelligent piston cleanliness evaluation system, and the system includes:

[0042] An acquisition device, configured to acquire image data of a piston; and

[0043] The intelligent piston cleanliness evaluation device according to the second or third aspect of the present application, and the intelligent piston cleanliness evaluation device is electrically connected to the acquisition device to obtain the image data of the piston acquired by the acquisition device.

[0044] In an embodiment of the present application, the acquisition device includes a detection table and a positioning boss, the positioning boss is detachably connected to the detection table and is used for installing the piston to be evaluated, and the piston is configured to be rotatably connected to the positioning boss around a rotation axis.

[0045] In an embodiment of the present application, the acquisition device further includes:

[0046] A support frame, the support frame is connected to the detection table, the support frame includes a cross beam and a longitudinal beam, the longitudinal beam extends in the vertical direction, and the cross beam is used to be arranged above the piston in the vertical direction; and

[0047] A camera, the camera is arranged on the support frame, and the camera is configured to be movable along the extension direction of the longitudinal beam, and the camera is also configured to be movable along the extension direction of the cross beam.

[0048] In an embodiment of the present application, the acquisition device further includes a light source device, and the light source device is arranged around the detection table to illuminate the piston on the detection table.

[0049] In an embodiment of the present application, a power supply unit is integrated on the intelligent piston cleanliness evaluation device; and / or

[0050] The intelligent piston cleanliness evaluation device obtains electric energy from the outside.

[0051] The fifth aspect of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the processor is caused to execute the piston cleanliness intelligent evaluation method according to the first aspect of the present application.

[0052] Details of one or more embodiments of the present application are set forth in the following drawings and description. Other features, objects, and advantages of the present application will become apparent from the specification, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The following drawings of the embodiments of the present application are hereby incorporated as part of the present application for understanding the present application. The embodiments of the present application shown in the drawings and their descriptions are used to explain the principles of the present application. In the drawings,

[0054] Figure 1 is a three-dimensional structural schematic diagram of an acquisition device of a piston cleanliness intelligent evaluation system according to a preferred embodiment of the present application;

[0055] Figure 2 is Figure 1 a front structural schematic diagram of the acquisition device in;

[0056] Figure 3 is a schematic flowchart of a piston cleanliness intelligent evaluation method according to an embodiment of the present application;

[0057] Figure 4 is a schematic structural block diagram of a piston cleanliness intelligent evaluation system according to an embodiment of the present application;

[0058] Figure 5 is a schematic structural block diagram of a piston cleanliness intelligent evaluation device according to an embodiment of the present application;

[0059] Figure 6 is a schematic structural block diagram of another piston cleanliness intelligent evaluation device according to an embodiment of the present application; and

[0060] Figure 7 is another schematic flowchart of a piston cleanliness intelligent evaluation method according to an embodiment of the present application.

[0061] DESCRIPTION OF REFERENCE NUMERALS:

[0062] 10: Piston cleanliness intelligent evaluation system

[0063] 11: Acquisition device

[0064] 101: Detection table

[0065] 101a: Positioning boss

[0066] 102: Support frame

[0067] 102a: Longitudinal beam

[0068] 102b: Cross beam

[0069] 103: Camera

[0070] 104: Light source device

[0071] 104a: Support plate

[0072] 104b: Light emitting strip

[0073] 12: Intelligent piston cleanliness evaluation device

[0074] 13: Piston to be evaluated

[0075] 100: Intelligent piston cleanliness evaluation method

[0076] 200: Intelligent piston cleanliness evaluation device

[0077] 210: Image acquisition module

[0078] 220: Preprocessing module

[0079] 230: Image recognition module

[0080] 300: Intelligent piston cleanliness evaluation device

[0081] 310: Memory

[0082] 320: Processor

[0083] D1: Vertical direction Detailed implementation manners

[0084] In the following description, numerous specific details are given to provide a more thorough understanding of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without one or more of these details. In other instances, in order to avoid confusion with the embodiments of the present application, some well-known technical features are not described.

[0085] In order to thoroughly understand the embodiments of the present application, detailed structures will be presented in the following description. Obviously, the implementation of the embodiments of the present application is not limited to the specific details familiar to those skilled in the art. The preferred embodiments of the present application are described in detail below. However, in addition to these detailed descriptions, the present application can also have other embodiments and should not be construed as limited to the embodiments presented here.

[0086] It should be understood that the purpose of the terms used herein is only to describe specific embodiments and is not a limitation of the present application. The singular forms "a", "an" and "the" are also intended to include the plural forms unless the context clearly indicates otherwise. When the terms "comprising" and / or "including" are used in this specification, they specify the presence of the stated features, wholes, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations. The terms "upper", "lower", "front", "rear", "left", "right" and similar expressions used in this application are for illustrative purposes only and are only used to represent the relative positional relationship between relevant parts, rather than to limit the absolute positions of these relevant parts.

[0087] The ordinal numbers such as "first" and "second" cited in this application are only identifiers and do not have any other meanings, such as a specific order, etc. Moreover, for example, the term "first component" does not imply the existence of a "second component" by itself, and the term "second component" does not imply the existence of a "first component" by itself.

[0088] In this article, "equal", "same", etc. are not strict mathematical and / or geometric limitations, and also include the allowable errors that can be understood by those skilled in the art and are allowed in manufacturing or using, etc.

[0089] Unless otherwise specified, the numerical ranges in this article include not only the entire range within its two endpoints, but also several sub-ranges included therein.

[0090] Hereinafter, specific embodiments of the present application will be described in more detail with reference to the accompanying drawings. These drawings show representative embodiments of the present application and do not limit the present application.

[0091] First, refer to Figure 3 to describe the intelligent evaluation method 100 for piston cleanliness for implementing the embodiments of the present invention. As Figure 3 shown, the intelligent evaluation method 100 for piston cleanliness may include the following steps:

[0092] In step S110, obtain the image data of the piston.

[0093] In step S120, preprocess the image data of the piston to obtain the sediment image data of the piston after preprocessing.

[0094] In step S130, input the preprocessed sediment image data into a trained intelligent evaluation model based on a deep learning network. The intelligent evaluation model identifies the sediment feature information in the sediment image data, and calculates and outputs the recognition result according to the sediment feature information.

[0095] Piston cleanliness is an important indicator for evaluating the quality of lubricating oil. By observing the image characteristics of deposits at various positions of the piston by evaluators, the cleanliness of the piston is scored. Usually, to ensure the accuracy of the evaluation, it is necessary to observe various positions in the circumferential direction of the piston multiple times during the evaluation. Even so, the evaluation results will still be affected by the subjective factors of the evaluators due to different evaluators. In the embodiment of the present application, the intelligent evaluation method 100 for piston cleanliness provides a method for realizing the intelligent evaluation of the cleanliness of the piston. Specifically, the intelligent evaluation method 100 for piston cleanliness uses a trained intelligent evaluation model based on a deep learning network to identify the deposit feature information in the image data of the piston 13 to be evaluated, improving the recognition effect of the deposits on the piston, and the recognition result is more scientific and objective, which can avoid the influence of the subjective factors of the evaluators on the evaluation result of the piston, and the reliability of the evaluation result is high.

[0096] In the embodiment of the present application, the step of obtaining the image data of the piston may include: dividing the area of the piston (the piston 13 to be evaluated) to obtain multiple sub-areas; obtaining the image data within each sub-area. Exemplarily, the multiple sub-areas may respectively correspond to the piston ring grooves, piston skirt, piston ring platform, and inner top of the piston. The sub-areas of the piston can be flexibly divided according to the actual situation of the piston.

[0097] Specifically, an image acquisition device 11 (such as a camera 103, a video camera) can be used to collect images of each sub-area of the piston, so as to obtain the image data of each sub-area. Then, the image data within each sub-area can be preprocessed respectively to obtain the deposit image data within each preprocessed sub-area.

[0098] In the embodiment of the present application, the step of preprocessing the image data of the piston to be evaluated includes: performing at least one of grayscale processing, image enhancement processing, filtering processing, and binarization processing on the image data of the piston. The quality and accuracy of the image data are improved through image preprocessing, and the reliability of the recognition result is indeed improved.

[0099] The steps of identifying the deposit feature information in the deposit image data by the intelligent evaluation model and calculating and outputting the recognition result according to the deposit feature information include: identifying the deposit image data within each sub-area by the intelligent evaluation model; summarizing and calculating the deposit feature information within all sub-areas to obtain the recognition result and output it.

[0100] In the embodiment of the present application, the recognition result includes the cleanliness score or deposit score of the piston. After identifying the deposit feature information in the deposit image data by the intelligent evaluation model, the cleanliness score or deposit score of the piston can be further calculated according to the deposit feature information.

[0101] In an embodiment of the present application, the sediment characteristic information includes a sediment severity coefficient, a sediment property probability, and a sediment area coefficient. Exemplarily, the sediment severity coefficient is a scoring coefficient obtained by dividing the sediment of the piston according to the sediment property, and the corresponding sediment severity coefficient is 0 when there is no sediment. The sediment area coefficient may refer to the area ratio of sediments of a certain property in a certain sub-area, which can be identified and output by an intelligent evaluation model. The sediment property probability is a quantitative result of the degree of conformity between the input image given by the intelligent evaluation model and the corresponding sediment property. For example, when the input image is the true color of the piston, the intelligent evaluation model outputs a sediment severity coefficient of 0 and a sediment property probability of 1; when the input image data shows that the piston has a slight sedimentation and tends to be true color, the intelligent evaluation model outputs a sediment severity coefficient of 0 and a sediment property probability of a parameter coefficient greater than 0.5 and less than 1.

[0102] In the embodiment of the present application, after obtaining the sediment image data in each sub-region after pre-processing, the sub-region can be further divided (can be divided equally by area), and then the sediment properties and sediment areas in each divided region are identified by the intelligent evaluation model, and the sediment severity coefficient, sediment property probability and sediment area coefficient of various sediments in the sub-region can be obtained by summarizing and calculating. Then, the score (sediment score or cleanliness score) in the sub-region can be determined, and the piston score can be calculated based on the scores of each sub-region.

[0103] Exemplarily, calculating the score of the piston 13 to be evaluated based on the deposit characteristic information includes two different calculation methods.

[0104] The first calculation method is to evaluate the cleanliness of the piston (cleanliness score):

[0105] The calculation formula includes:

[0106] Cleanliness score of piston = score coefficient * Σ(cleanliness score of sub-region) m ;

[0107] Cleanliness score of sub-area = W-Σ(sediment property probability) j *(Sediment severity coefficient) i *(Sediment area coefficient) i ;

[0108] In the above formula, m represents any sub-region of the piston; i represents a certain type of deposit in the sub-region of the piston; W represents the severity coefficient corresponding to severe carbon deposition, and the severity coefficient of deposits corresponding to no deposits is 0.

[0109] The second calculation method is to calculate and evaluate the deduction items for the cleanliness of the piston, that is, to evaluate the deposits on the piston (sediment score):

[0110] The calculation formula includes:

[0111] Piston deposit score = Σ(sub-region deposit weighted score) m ;

[0112] Sub-area Sediment Weighted Score = (Sub-area Sediment Score) m *(Weighting coefficient) m ;

[0113] Sediment score of sub-area = Σ(sediment property probability) j (Sediment severity coefficient) j *(Sediment area coefficient) j ;

[0114] In the above formula, m represents any sub-region; j represents a sediment of a certain nature in the sub-region of the piston.

[0115] Specifically, the deposits of the piston can be classified according to the properties of the deposits. For example, from severe deposits to no deposits, the deposit properties can be classified into six categories: black paint film, dark brown paint film, brown paint film, light brown paint film, light yellow paint film, and natural color; it can be understood that the natural color is the original color of the piston, and the corresponding deposit severity coefficient is 0, and the black paint film usually includes carbon deposits. This degree of deposit properties indicates severe deposits, and the corresponding deposit severity coefficient is W.

[0116] The probability of sediment properties is the quantitative result of the "degree of conformity" between the input image data and the corresponding sediment properties given by the intelligent evaluation model. For example, when the input image is the original color of the piston, the intelligent evaluation model outputs a sediment severity coefficient of 0 and a sediment property probability of 1; when the input image is between the light yellow paint film and the original color, but tends to be the original color, the intelligent evaluation model outputs a sediment severity coefficient of 0 and a sediment property probability (a parameter less than 1 and greater than 0.5, such as 0.9 or 0.8, determined according to actual conditions).

[0117] From severe deposition to no deposition, the properties of the deposits can be further divided into nine categories: severe carbon deposits, moderate carbon deposits, mild carbon deposits, black paint film, dark brown paint film, brown paint film, light brown paint film, light yellow paint film, and natural color. It can be understood that this classification has more categories, and the corresponding cleanliness evaluation results will be more reliable. The deposit severity coefficient corresponding to severe carbon deposits is W, and the deposit severity coefficient corresponding to the natural color is 0.

[0118] To further improve the objectivity and reliability of the evaluation results, the sediment properties can be further divided into more detailed and numerous classification levels.

[0119] In the embodiments of the present application, the image data collected by the image acquisition device 11 can be video data and / or picture data that record each sub-region of the piston to be evaluated. When the image data includes video data, the operation steps for preprocessing the image data further include: decomposing the video data to obtain picture data for inputting into the intelligent evaluation model. Specifically, the video data can be decomposed into picture data by software for converting video to pictures at a specified frame rate, such as every 1 frame, every 5 frames, etc., or the specified frames of the video data, such as the image data corresponding to each sub-region respectively, can be decomposed into picture data. The software for converting video to pictures can be opencv, Adobe Premiere Pro, etc., and is not limited thereto.

[0120] In the embodiments of the present application, before the intelligent evaluation model identifies the sediment image data, the intelligent evaluation model needs to be trained. Referring to Figure 7 , the training of the intelligent evaluation model can include the following steps: obtaining a data set, where the data set includes a training set; inputting the sample images in the training set into the intelligent evaluation model to be trained, and the intelligent evaluation model to be trained performs the following operations:

[0121] extracting features of the sample images to obtain sediment image features; converting the sediment image features into sediment feature information; identifying the sediment feature information, comparing the obtained result with the standard result of the sample images, and outputting model weight parameters based on the comparison result;

[0122] iterating the above training steps until the intelligent evaluation model to be trained converges to obtain a trained intelligent evaluation model based on a deep learning network.

[0123] In the above process, through continuous iteration, the model weight parameters are optimized multiple times until the optimized model weight parameters meet the corresponding requirements, thereby obtaining a trained intelligent evaluation model based on a deep learning network. In addition, generally speaking, as the training set increases, the accuracy of the trained intelligent evaluation model also increases.

[0124] Referring to Figure 7 , in the embodiments of the present application, obtaining the data set can include the following steps:

[0125] obtaining sample images; marking the sample images and converting the marked images into a target format file; dividing the target format file into a training set, a validation set, and a test set according to a set ratio.

[0126] Among them, the sample images can be sourced from existing piston image datasets, and the corresponding sediment harshness coefficients and piston scores for the piston image data therein are obtained. The piston score is a score related to piston deposits and can be a piston cleanliness score or a sediment score. It should be noted that the types and sizes of the pistons used as sample images can be different.

[0127] To improve the accuracy of the intelligent evaluation model, before inputting the sample images in the training set into the intelligent evaluation model to be trained, the following steps can also be included:

[0128] Perform image processing operations on the sample images in the training set through image processing operations, so as to input the sample images after image processing into the intelligent evaluation model to be trained;

[0129] Among them, the image processing operations include at least one of grayscale processing, image enhancement processing, filtering processing, and binarization processing.

[0130] In the embodiments of the present application, the steps of preprocessing the image data of the piston to be evaluated and the steps of image processing operations on the sample images specifically include the following detailed steps:

[0131] 1) Image preprocessing: Perform grayscale, image enhancement, filtering, binarization, etc. on the image to overcome image interference and improve the quality and accuracy of the image;

[0132] 2) Image segmentation: Adopt common image segmentation methods such as threshold segmentation, edge detection, region growing, clustering, etc. to divide the preprocessed image into multiple sub-regions, and the multiple sub-regions can respectively correspond to the first ring groove, the second ring groove, the inner top of the piston, the piston skirt, etc. of the piston.

[0133] 3) Image post-processing: Eliminate the noise in the sediment images of each sub-region, enhance the contrast of the sediment images, adjust the brightness of the sediment images, etc.

[0134] In the embodiments of the present application, the target recognition model of the intelligent evaluation model based on the deep learning network can be a convolutional neural network model, such as deep CNN models like ResNet (Deep Residual Neural Network), GoogleNet, etc. There is no limitation on this.

[0135] Exemplarily, the sample images can be labeled through an image annotation tool. Among them, the image annotation tool can be a LabelImg tool, a roLabelImg tool, a labelme tool, a Vott tool, a CVAT tool, etc. There is no limitation on this.

[0136] Taking the LabelImg tool as an example, the sediment images in the image data can be labeled using the LabelImg tool. The output format can be selected as PascalVOC, thereby generating a file in the VOC (xml) format of the sediment image target detection coordinates. The file in the VOC (xml) format of the sediment image target detection coordinates can be placed in the datasets folder at the same level as the code. Then, the file in the VOC (xml) format of the sediment image target detection coordinates can be converted into a dataset suitable for the intelligent evaluation model based on the deep learning network. Subsequently, the dataset can be divided into a training set, a validation set, and a test set according to set ratios such as 6:2:2, 8:1:1, 7:2:1, etc.

[0137] Exemplarily, the steps of extracting the sediment image features by performing feature extraction on the sample image may include:

[0138] Using feature extraction methods such as edge detection, corner detection, and color space conversion to extract the features related to the sediment of the piston from the sediment image;

[0139] Using feature selection methods such as information gain, correlation analysis, and principal component analysis to select the extracted features to obtain the sediment image features.

[0140] Through the above steps, information loss and computational complexity can be minimized to the greatest extent, while increasing the diversity of difficult samples and data.

[0141] In the embodiments of the present application, the training of the intelligent evaluation model may further include:

[0142] Whenever the operation (training the intelligent evaluation model with the input training set) is executed multiple times, the intelligent evaluation model after multiple trainings is verified through the validation set to adjust the parameters of the intelligent evaluation model. It should be noted that the parameters here do not specifically refer to the model weight parameters in the above text, but can also be other parameters of the intelligent evaluation model, such as the learning rate, the number of layers, the number of neurons in each layer, etc.

[0143] Reference Figure 7 , in the embodiments of the present application, the test set can be used to test the trained intelligent evaluation model based on the deep learning network to verify the recognition accuracy of the intelligent evaluation model. It should be noted that if the recognition accuracy of the intelligent evaluation model does not meet the requirements, the intelligent evaluation model can be iteratively trained again until the recognition accuracy of the intelligent evaluation model meets the requirements. In addition to accuracy, indicators such as recall rate and F1 value can also be used to evaluate the performance of the classification model and evaluate the trained classification model.

[0144] Exemplarily, in combination with Figure 7, taking the sediment properties being 6 categories as an example, the training process of the intelligent evaluation model may include the following process:

[0145] First, collect the image data of the piston including 6 categories of sediments scored by professionals and the corresponding sediment severity coefficients. For example, the image data of the piston includes 1000 pictures.

[0146] Secondly, label and classify the 1000 pictures (images and corresponding sediment severity coefficients) into 6 categories, and divide the 1000 pictures into a training set (700 pictures), a test set (200 pictures) and a validation set (100 pictures) according to 7:2:1.

[0147] Establish the relationship between the image data and the sediment severity coefficients through a convolutional neural network for the training set to obtain the trained intelligent evaluation model, and verify it with the test set and the validation set.

[0148] Finally, collect the image data of the piston to be evaluated, input the image data of the piston to be evaluated into the trained intelligent evaluation model for recognition, and obtain the corresponding sediment characteristic information (sediment severity coefficient, sediment property probability and sediment area coefficient); calculate and output the cleanliness score (or sediment score) of the piston to be evaluated.

[0149] The above is only an example with 6 categories of sediments. If there are more classifications of sediment properties, the image data will be labeled and classified according to the new and more categories, and the network will be trained with more categories of image data, corresponding sediment severity coefficients and scores (train the convolutional neural network model).

[0150] Based on the above description, according to the piston cleanliness intelligent evaluation method 100 of the embodiment of the present application, the sediment characteristic information in the image data of the piston is identified through the trained intelligent evaluation model based on the deep learning network, and the cleanliness of the piston is scientifically and objectively evaluated by means of machine learning, avoiding being affected by subjective factors, improving the reliability of the evaluation result, and actually being able to greatly improve the efficiency of piston cleanliness evaluation.

[0151] The piston cleanliness intelligent evaluation method 100 according to the embodiment of the present application is described above by way of example. The following combines Figure 5 and Figure 6 , to describe the piston cleanliness intelligent evaluation device 200 provided by another aspect of the present application. As Figure 5As shown, the intelligent piston cleanliness evaluation device 200 includes an image acquisition module 210, a preprocessing module 220, and an image recognition module 230. Among them, the image acquisition module 210 is used to acquire image data of the piston; the preprocessing module 220 is used to preprocess the image data of the piston to obtain the sediment image data of the piston after preprocessing; the image recognition module 230 is used to input the sediment image data after preprocessing into a trained intelligent evaluation model based on a deep learning network, and the intelligent evaluation model identifies the sediment feature information in the sediment image data, and calculates and outputs the recognition result according to the sediment feature information.

[0152] Among them, the image acquisition module 210, the preprocessing module 220, and the image recognition module 230 can be implemented by a processor in an electronic device having the intelligent piston cleanliness evaluation device of the present application running program instructions stored in a memory, and can execute the corresponding steps in the piston cleanliness intelligent evaluation method 100 according to the foregoing embodiments of the present invention. Those skilled in the art can understand the specific operations of the intelligent piston cleanliness evaluation device 300 according to the embodiments of the present application in combination with the foregoing content. For the sake of brevity, the specific details are not described herein again.

[0153] As Figure 6 shown, the present application also provides another intelligent piston cleanliness evaluation device 300. Specifically, the intelligent piston cleanliness evaluation device includes a memory 310 and a processor 320, and a computer program is stored on the memory 310 and run by the processor 320. When the computer program is run by the processor 320, the processor 320 is caused to execute the piston cleanliness intelligent evaluation method 100 according to the embodiments of the present application described above. Those skilled in the art can understand the specific operations of the intelligent piston cleanliness evaluation device 300 according to the embodiments of the present application in combination with the foregoing content. For the sake of brevity, the specific details are not described herein again.

[0154] In addition, according to the embodiments of the present application, a computer-readable storage medium is also provided. A computer program is stored on the storage medium, and when the computer program is run by a computer or a processor, it is used to execute the corresponding steps of the piston cleanliness intelligent evaluation method 100 according to the embodiments of the present application. The storage medium may include, for example, a memory card of a smart phone, a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0155] In an embodiment of the present application, when the computer program is run by a computer or a processor, it can implement each functional module of the piston cleanliness intelligent evaluation device according to the embodiments of the present invention, and / or can execute the piston cleanliness intelligent evaluation method 100 according to the embodiments of the present invention.

[0156] In an embodiment of the present application, when a computer program is run by a computer or a processor, it causes the computer or the processor to perform the following steps: acquiring image data of a piston; preprocessing the image data of the piston to obtain sediment image data of the piston after preprocessing; inputting the sediment image data after preprocessing into a trained intelligent evaluation model based on a deep learning network, and the intelligent evaluation model identifies sediment feature information in the sediment image data and outputs an identification result.

[0157] Reference Figure 4 , in another embodiment of the present application, there is provided a piston cleanliness intelligent evaluation system 10. The piston cleanliness intelligent evaluation system 10 includes a collection device 11 and the piston cleanliness intelligent evaluation device according to the embodiments of the present application described above. The collection device 11 is used to collect image data of a piston; the piston cleanliness intelligent evaluation device is electrically connected to the collection device 11 to obtain the image data of the piston 13 to be evaluated collected by the collection device 11.

[0158] As Figure 1 and Figure 2 shown, in an embodiment of the present application, the collection device 11 includes a detection table 101, a support frame 102, and a camera 103. The detection table 101 is used to mount the piston 13 to be evaluated, and the piston 13 to be evaluated is configured to be able to rotate relative to the detection table 101 about a rotation axis.

[0159] The support frame 102 is connected to the detection table 101. The support frame 102 includes a longitudinal beam 102a extending in the vertical direction D1 and a cross beam 102b extending in the horizontal direction. Among them, the longitudinal beam 102a is located on the side of the piston 13 to be evaluated, and the cross beam 102b is located vertically above the piston 13 to be evaluated. Refer to Figure 1 , the two longitudinal beams 102a are respectively located on opposite sides of the piston 13 to be evaluated, and the cross beam 102b is connected to the tops of the two longitudinal beams 102a.

[0160] The camera 103 is set on the support frame 102, and the camera 103 is configured to be movable in the extending direction of the longitudinal beam 102a, so that the position (height position) of the camera 103 in the vertical direction D1 can be adjusted according to the different sizes of the piston 13 to be evaluated; the camera 103 is also configured to be movable in the extending direction of the cross beam 102b, so that it can be moved above the piston 13 to be evaluated vertically to collect images of the top of the piston 13 to be evaluated. Thus, image data of the piston 13 to be evaluated can be obtained. It can be understood that the image data of the piston includes complete images of various positions including piston ring grooves, piston skirts, piston ring platforms, and piston inner tops.

[0161] Exemplarily, a moving device can be provided on the support frame 102 and the camera 103 to facilitate the camera 103 to move in the vertical direction D1 (and / or horizontal direction) and collect image data of the piston 13 to be evaluated.

[0162] In an embodiment of the present application, the acquisition device 11 further includes a light source device 104, and the light source device 104 is arranged around the detection table 101 to illuminate all directions of the piston on the detection table 101. As Figure 1 shown, a light source device 104 is respectively arranged in the circumferential direction of the piston. The light source device 104 includes a support plate 104a and a light-emitting band 104b. The support plate 104a can be installed on the detection table 101, and the light-emitting band 104b extends in the vertical direction D1.

[0163] See Figure 2 , in this embodiment, a positioning boss 101a is arranged at a substantially central position of the detection table 101. The size of the positioning boss 101a matches the size of the piston 13 to be evaluated. The piston 13 to be evaluated can be installed on the positioning boss 101a and rotated relative to the positioning boss 101a around the rotation axis to switch the image acquisition position of the camera 103 for the piston 13 to be evaluated. For example, the main and secondary thrust sides and non-thrust sides of the piston 13 to be evaluated can be photographed respectively to obtain images of the entire piston surface. Figure 1 and Figure 2 In, the rotation axis of the piston is its own central axis and extends in the vertical direction D1.

[0164] Preferably, the positioning boss 101a is detachably connected to the detection table 101 to facilitate replacing the corresponding positioning boss 101a according to the size of the piston 13 to be evaluated.

[0165] In an embodiment of the present application, a power supply unit is also integrated on the piston cleanliness intelligent evaluation device, and the camera 103, the processor, the light source device 104, etc. can be powered through the power supply unit.

[0166] It should be noted that in other embodiments of the present application, the intelligent piston cleanliness evaluation device may not integrate a power supply unit, but obtain electric energy from the outside through a power interface. Alternatively, the intelligent piston cleanliness evaluation device is integrated with a power supply unit, and at the same time, it can also obtain electric energy from the outside through a power interface.

[0167] Unless otherwise defined, the technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field of the present application. The terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the present application. Terms such as "arranged" that appear herein can mean either that one component is directly attached to another component or that one component is attached to another component through an intermediate member. Features described in one embodiment herein can be applied to another embodiment alone or in combination with other features, unless the feature is not applicable or otherwise stated in that other embodiment.

[0168] The present application has been described through the above embodiments, but it should be understood that the above embodiments are only for the purpose of illustration and example, and are not intended to limit the present application to the scope of the described embodiments. Those skilled in the art can understand that according to the teachings of the present application, more variations and modifications can be made, and these variations and modifications all fall within the scope claimed by the present application.

Claims

1. An intelligent evaluation method for piston cleanliness, characterized in that The intelligent evaluation method for piston cleanliness includes: Obtaining image data of the piston; Preprocessing the image data of the piston to obtain the sediment image data of the piston after preprocessing; Inputting the sediment image data after preprocessing into a trained intelligent evaluation model based on a deep learning network. The intelligent evaluation model identifies the sediment feature information in the sediment image data and calculates and outputs an identification result according to the sediment feature information.

2. The intelligent evaluation method for piston detergency according to claim 1, wherein The step of obtaining image data of the piston includes: Dividing the area of the piston to obtain multiple sub-regions; Obtaining image data within each sub-region; The multiple sub-regions respectively correspond to the piston ring grooves, piston skirt, piston ring platform and piston inner top of the piston.

3. The intelligent evaluation method for piston detergency according to claim 2, characterized in that, The step of the intelligent evaluation model identifying the sediment feature information in the sediment image data and calculating and outputting an identification result according to the sediment feature information includes: The intelligent evaluation model identifying the sediment feature information within each sub-region; Calculating the identification result according to the sediment feature information within all sub-regions.

4. The intelligent evaluation method for piston cleanliness according to claim 1, characterized in that, The step of preprocessing the image data of the piston includes: Performing at least one of grayscale processing, image enhancement processing, filtering processing and binarization processing on the image data of the piston.

5. The intelligent evaluation method for piston detergency according to claim 1, characterized in that, The sediment feature information includes a sediment severity coefficient, a sediment property probability and a sediment area coefficient.

6. The intelligent evaluation method for piston cleanliness according to claim 1, wherein The identification result includes the cleanliness score or sediment score of the piston.

7. The intelligent evaluation method for piston cleanliness according to claim 1, characterized in that When the obtained image data of the piston includes video data, the step of preprocessing the image data of the piston further includes: Decomposing the video data to obtain picture data for inputting into the intelligent evaluation model.

8. The intelligent evaluation method for piston cleanliness according to any one of claims 1 to 7, characterized in that, The training of the intelligent evaluation model includes the following training steps: Obtaining a data set, the data set including a training set; Inputting the sample images in the training set into the intelligent evaluation model to be trained. The intelligent evaluation model to be trained performs the following operations: Performing feature extraction on the sample images to obtain sediment image features; Converting the sediment image features into sediment feature information; Identifying the sediment feature information, comparing the obtained result with the standard result of the sample image, and outputting model weight parameters based on the comparison result; Iterating the above training steps until the intelligent evaluation model to be trained converges to obtain the trained intelligent evaluation model based on a deep learning network.

9. The intelligent evaluation method for piston detergency according to claim 8, wherein Before inputting the sample images in the training set into the intelligent evaluation model to be trained, it further includes: Performing image processing operations on the sample images in the training set through image processing operations so as to input the sample images after image processing into the intelligent evaluation model to be trained; Wherein, the image processing operations include at least one of grayscale processing, image enhancement processing, filtering processing and binarization processing.

10. The intelligent evaluation method for piston cleanliness according to claim 8, characterized in that The data set further includes a validation set. The training steps of the intelligent evaluation model further include: The verification set is used to verify the intelligent evaluation model after multiple trainings, so as to adjust the parameters of the intelligent evaluation model.

11. The intelligent evaluation method for piston cleanliness according to claim 8, wherein The data set further includes a test set, and the test set is used to test the trained intelligent evaluation model based on a deep learning network to verify the recognition accuracy of the intelligent evaluation model.

12. An intelligent evaluation device for piston cleanliness, characterized in that, The piston cleanliness intelligent evaluation device includes: an image acquisition module for acquiring image data of a piston; a preprocessing module for preprocessing the image data of the piston to obtain the sediment image data of the piston after preprocessing; and an image recognition module for inputting the sediment image data after preprocessing into a trained intelligent evaluation model based on a deep learning network, and the intelligent evaluation model recognizes the sediment feature information in the sediment image data, calculates and outputs a recognition result according to the sediment feature information.

13. An intelligent evaluation device for piston cleanliness, comprising a memory and a processor, wherein a computer program run by the processor is stored on the memory, and is characterized in that When the computer program runs on the processor, the processor is caused to execute the piston cleanliness intelligent evaluation method according to any one of claims 1 to 11.

14. An intelligent evaluation system for piston detergency, characterized in that, The system includes: a collection device for collecting image data of a piston; and the piston cleanliness intelligent evaluation device according to claim 12 or 13, wherein the piston cleanliness intelligent evaluation device is electrically connected to the collection device to obtain the image data of the piston collected by the collection device.

15. The intelligent evaluation system for piston detergency according to claim 14, wherein The collection device includes a detection table and a positioning boss, the positioning boss is detachably connected to the detection table and is used for installing a piston to be evaluated, and the piston is configured to be rotatably connected to the positioning boss around a rotation axis.

16. The intelligent piston cleanliness evaluation system according to claim 15, wherein The collection device further includes: a support frame, the support frame is connected to the detection table, the support frame includes a cross beam and a longitudinal beam, the longitudinal beam extends in the vertical direction, and the cross beam is used to be arranged above the piston vertically; and a camera, the camera is arranged on the support frame, and the camera is configured to be able to move along the extension direction of the longitudinal beam, and the camera is further configured to be able to move along the extension direction of the cross beam.

17. The piston cleanliness intelligent evaluation system according to claim 15, characterized in that, The collection device further includes a light source device, and the light source device is arranged around the detection table to illuminate the piston on the detection table.

18. The piston cleanliness intelligent evaluation system according to claim 14, wherein A power supply unit is integrated on the piston cleanliness intelligent evaluation device; and / or The piston cleanliness intelligent evaluation device obtains electric energy from the outside.

19. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program runs on the processor, the processor is caused to execute the piston cleanliness intelligent evaluation method according to any one of claims 1 to 11.

Images