Method and system for detecting surface defects of main pump motor in refueling vehicle and refueling vehicle

Abstract

The invention provides a method and system for detecting surface defects of a main pump motor in a refueling vehicle and the refueling vehicle. The method for detecting the surface defects of the main pump motor in the refueling vehicle comprises the steps of obtaining a surface image of the main pump motor; preprocessing the surface image to obtain a target detection area; the preprocessing comprises cutting; and determining a defect distribution condition in the target detection area through a single-stage real-time target detection model based on deep learning. After a surface image of a main pump motor is acquired to capture complete surface information, preprocessing including cutting operation is performed on the surface image to focus a target detection area, redundant interference is eliminated, and then defect distribution in the target detection area is accurately identified by means of a single-stage real-time target detection model based on deep learning. The problems of missing detection and false detection of defects are effectively avoided, surface defects of different types and different scales can be comprehensively captured, and the accuracy and comprehensiveness of surface defect detection of the main pump motor are improved.

Description

Technical Field

[0001] This invention relates to the field of refueling truck technology, and in particular to a method, system, and refueling truck for detecting surface defects of the main pump motor in a refueling truck. Background Technology

[0002] As the core drive component of a refueling truck, the integrity of the main pump motor's surface directly determines the operational stability, efficiency, and service life of the motor and, consequently, the truck itself, making it a crucial factor in ensuring continuous and reliable operation. Currently, the mainstream solutions for motor surface defect detection rely on manual inspection and traditional image processing algorithms. However, these methods have significant limitations in accuracy, failing to meet the stringent requirements of high-precision defect identification in refueling truck industrial inspections. Manual inspection relies entirely on operator experience, making it susceptible to visual fatigue and subjective perception differences. It suffers from weak identification capabilities for low-resolution defects such as minute cracks and slight wear, and carries a serious risk of missed or false detections, failing to consistently output results consistent with the actual defect state. While traditional image processing algorithms possess basic automated detection capabilities, they are weakly resistant to complex background interference in industrial environments and can only identify defect types with preset features. They lack adaptability to surface defects of various shapes and sizes, making accurate and comprehensive defect coverage detection difficult. Summary of the Invention

[0003] This application provides a method, system, and refueling truck for detecting surface defects in the main pump motor of a refueling truck, so as to improve the accuracy and comprehensiveness of the detection of surface defects in the main pump motor.

[0004] This application provides a method for detecting surface defects of the main pump motor in a refueling truck, comprising: obtaining a surface image of the main pump motor; preprocessing the surface image to obtain a target detection area; the preprocessing includes cropping; and determining the distribution of defects in the target detection area using a single-stage real-time target detection model based on deep learning.

[0005] Optionally, preprocessing the surface image to obtain the target detection region includes: denoising the surface image; enhancing the denoised surface image to improve its contrast and / or brightness; and cropping the enhanced surface image to obtain the target detection region.

[0006] Optionally, the distribution of defects in the target detection area is determined by using a single-stage real-time target detection model based on deep learning, including: identifying defects in the target detection area and obtaining the location information of the defects by using a single-stage real-time target detection model based on deep learning; determining defect areas based on the defects and their location information, wherein each defect area includes at least one complete defect.

[0007] Optionally, after determining the defect distribution in the target detection area using a single-stage real-time target detection model based on deep learning, the method further includes classifying each identified defect area to obtain the defect type corresponding to each defect area.

[0008] Optionally, the identified defect regions are classified to obtain the defect type corresponding to each defect region, including: extracting the feature vector of each defect region; and classifying each defect region according to the feature vector of each defect region using a pre-trained classification algorithm to obtain the defect type corresponding to each defect region.

[0009] Optionally, the distribution of defects in the target detection area can be identified using a single-stage real-time target detection model based on deep learning, including: identifying the distribution of defects in the target detection area using the YOLOv8 model.

[0010] Optionally, after determining the defect distribution in the target detection area using a single-stage real-time target detection model based on deep learning, the method further includes: determining the defect severity of the main pump motor based on the defect distribution; and issuing a defect alert based on the defect severity.

[0011] This application provides a system for detecting surface defects of the main pump motor in a refueling truck, including one or more processors for implementing the aforementioned method for detecting surface defects of the main pump motor in a refueling truck.

[0012] This application provides a system for detecting surface defects of the main pump motor in a refueling truck, comprising: an image acquisition module for acquiring surface images of the main pump motor; an image preprocessing module for preprocessing the surface images to obtain a target detection area; the preprocessing includes cropping; and an image processing module for determining the defect distribution in the target detection area using a single-stage real-time target detection model based on deep learning.

[0013] This application provides a refueling truck, including: a main pump motor; and the aforementioned system for detecting surface defects of the main pump motor in the refueling truck.

[0014] The method, system, and refueling truck for detecting surface defects of the main pump motor in this application acquire an image of the main pump motor surface to capture complete surface information. Then, the surface image undergoes preprocessing including cropping to focus on the target detection area and remove redundant interference. Finally, a single-stage real-time target detection model based on deep learning is used to accurately identify the defect distribution within the target detection area. This effectively avoids the problems of missed or false detections of defects and can comprehensively capture surface defects of different types and scales, thus improving the accuracy and comprehensiveness of main pump motor surface defect detection. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a method for detecting surface defects of the main pump motor in a refueling truck, provided in one embodiment of this application. Figure 2 This is a flowchart illustrating a method for detecting surface defects of the main pump motor in a refueling truck, provided in another embodiment of this application. Figure 3 This is a flowchart illustrating a method for detecting surface defects of the main pump motor in a refueling truck, provided in another embodiment of this application. Figure 4 This is a flowchart illustrating a method for detecting surface defects of the main pump motor in a refueling truck, provided in another embodiment of this application. Figure 5 This is a flowchart illustrating a method for detecting surface defects of the main pump motor in a refueling truck, provided in another embodiment of this application. Figure 6 This is a schematic diagram of the main pump motor provided in one embodiment of this application; Figure 7 This is a schematic diagram of the architecture of a system for detecting surface defects of the main pump motor in a refueling truck, provided in one embodiment of this application; Figure 8 This is a schematic diagram illustrating an application scenario of a method and system for detecting surface defects of the main pump motor in a refueling truck, provided in one embodiment of this application. Figure 9 This is a flowchart illustrating a method for detecting surface defects of the main pump motor in a refueling truck, provided in another embodiment of this application.

[0016] Figure Labels 11: Bearing cover; 12: End cover; 13: Junction box; 14: Stator core; 15: Stator winding; 16: Shaft; 17: Frame; 18: Bearing; 19: Rotor; 110: Fan; 111: Cover; 10: Image acquisition module; 20: Image preprocessing module; 30: Image processing module; 301: Defect detection module; 302: Defect classification module; 40: Result display and alarm module; 50: Data storage and management module; 61: High-resolution camera; 62: Main pump motor; 63: YOLOv8 neural network model; 64: Host computer; 65: Database; 66: Display terminal; 67: Staff. Detailed Implementation

[0017] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings.

[0018] Combination Figure 1 As shown in the figure, this application provides a method for detecting surface defects of the main pump motor in a refueling truck, including steps S10 to S30.

[0019] Step S10: Obtain a surface image of the main pump motor.

[0020] A separate device, such as a camera, can be set up to acquire images of the main pump motor's surface in real time. This ensures the timeliness and continuity of the images. The device can be installed directly on the main pump motor or in a nearby area, as long as it can acquire images of the motor's surface in real time. Specifically, for example, a high-resolution camera can be positioned on the main pump motor to ensure comprehensive image capture of the motor's surface.

[0021] Step S20: Preprocess the surface image to obtain the target detection area; the preprocessing includes cropping.

[0022] The initial surface image of the main pump motor covers a large area. Therefore, the surface image undergoes preprocessing, including at least cropping, to achieve area focusing.

[0023] Step S30: Determine the distribution of defects in the target detection area using a single-stage real-time target detection model based on deep learning.

[0024] The method for detecting surface defects of the main pump motor in a refueling truck provided in this application involves acquiring an image of the main pump motor surface to capture complete surface information, followed by preprocessing of the surface image including cropping to focus on the target detection area and remove redundant interference. Then, a single-stage real-time target detection model based on deep learning is used to accurately identify the defect distribution within the target detection area, effectively avoiding the problems of missed or false detections of defects. This method can comprehensively capture surface defects of different types and scales, which is beneficial to improving the accuracy and comprehensiveness of the detection of surface defects of the main pump motor.

[0025] In some embodiments, the method for detecting surface defects of the main pump motor in a refueling truck is applied to a comprehensive testing platform, where the main pump motor is the main pump motor of the refueling truck comprehensive testing platform.

[0026] In some embodiments, obtaining a surface image of the main pump motor includes obtaining a complete surface image of the main pump motor. This helps to avoid missing some defects, thereby ensuring the comprehensiveness of defect detection. During implementation, the integrity of the obtained surface image of the main pump motor is ensured by setting the number and / or location of the devices used to acquire the surface image.

[0027] In some embodiments, a single-stage real-time target detection model based on deep learning is used to identify the defect distribution in the target detection area, including: identifying the defect distribution in the target detection area using a YOLOv8 model. Here, the single-stage real-time target detection model based on deep learning is specifically defined as the YOLOv8 model. The YOLOv8 model's powerful feature extraction capabilities, inference speed, and detection accuracy, through its convolutional neural network structure, can more accurately capture the features of minute and complex-shaped defects on the surface of the main pump motor, reducing missed and false detections due to insufficient model detection capabilities. Simultaneously, it efficiently completes defect distribution identification, further improving detection efficiency while ensuring detection accuracy and comprehensiveness, making defect detection more adaptable to the real-time requirements of industrial scenarios.

[0028] In some embodiments, preprocessing the surface image to obtain the target detection region includes cropping the surface image to obtain the target detection region. Further, cropping the surface image to obtain the target detection region includes determining different target detection regions corresponding to different components in the main pump motor.

[0029] Combination Figure 2 As shown, in some embodiments, the aforementioned step S20, which preprocesses the surface image to obtain the target detection area, includes steps S201 to S203.

[0030] Step S201: Denoise the surface image.

[0031] Denoising can remove noise interference from surface images, preventing noise from being misjudged as defects or masking real defects.

[0032] Step S202 involves enhancing the denoised surface image to improve its contrast and / or brightness. This enhancement process strengthens the difference between defects and the undefected background in the surface image, making the defects more clearly visible.

[0033] Step S203: The enhanced surface image is cropped to obtain the target detection area.

[0034] Cropping allows for focusing on key identification areas. These key identification areas can be preset or determined based on the enhanced surface image. Areas with more obvious defects after enhancement are used as the target detection areas.

[0035] In this way, denoising and enhancement processing ensures that surface defects of the main pump motor are more easily identified. Furthermore, combining this with cropping operations helps remove redundant information and reduces the computational load of subsequent calculations. This preprocessing optimizes image quality, providing a solid data foundation for subsequent defect detection, effectively solving the detection bias problem caused by poor original image quality, and further improving the accuracy of main pump motor surface defect detection.

[0036] Combination Figure 3 As shown, the aforementioned step S30 determines the distribution of defects in the target detection area through a single-stage real-time target detection model based on deep learning, including steps S301 to S302.

[0037] Step S301: Using a single-stage real-time target detection model based on deep learning, identify defects in the target detection area and obtain the location information of the defects.

[0038] Step S302: Determine the defect area based on the defect and its location information. Each defect area includes at least one complete defect.

[0039] In this way, precise defect location is achieved through location information. Defect areas are divided based on the defect and its location information, ensuring that each area contains at least one complete defect, which helps avoid defect splitting or omission. This approach clarifies the specific location of the defect while guaranteeing the completeness of defect information, thus improving the accuracy of defect detection.

[0040] Understandably, in real-world applications, defects may have different sizes and shapes. Obtaining the defect's location information can involve identifying the point where the defect exists, and then determining the complete defect based on the defect and its location information, thus dividing the defect area. This avoids dividing a single defect into multiple parts, ensuring the completeness of defect identification.

[0041] Combination Figure 4 As shown, the aforementioned step S30 determines the distribution of defects in the target detection area through a single-stage real-time target detection model based on deep learning, including steps S301 to S303.

[0042] Step S301: Using a single-stage real-time target detection model based on deep learning, identify defects in the target detection area and obtain the location information of the defects; Step S302: Determine the defect area based on the defect and its location information. Each defect area includes at least one complete defect.

[0043] Step S303: Classify the identified defect regions to obtain the defect type corresponding to each defect region.

[0044] In this way, based on the discovery of defects, they can be further classified to clarify the defect type. Compared with schemes that only detect the distribution of defects, this method enriches the dimensions of the detection results, overcomes the limitation of traditional detection methods that cannot distinguish defect types, and makes the detection results more practically valuable while ensuring accuracy and comprehensiveness. Defect types include, for example, cracks, wear, and scratches.

[0045] Specifically, in some embodiments, the identified defect regions are classified to obtain the defect type corresponding to each defect region. This includes: extracting feature vectors for each defect region; and classifying each defect region using a pre-trained classification algorithm based on its feature vectors to obtain the corresponding defect type. In this way, by extracting feature vectors from defect regions, the essential differences in characteristics between different types of defects can be accurately captured. Furthermore, using a pre-trained classification algorithm to classify each defect region ensures accurate determination of the defect type. Compared to a general classification method, combining feature extraction with algorithmic classification helps reduce misjudgments and confusion between different types of defects, further improving the accuracy of defect classification, making the detection results more precise and reliable, and providing stronger support for subsequent targeted maintenance.

[0046] Combination Figure 5 As shown in the figure, this application provides a method for detecting surface defects of the main pump motor in a refueling truck, including steps S10 to S30.

[0047] Step S10: Obtain a surface image of the main pump motor.

[0048] Step S20: Preprocess the surface image to obtain the target detection area; the preprocessing includes cropping.

[0049] Step S30: Determine the distribution of defects in the target detection area using a single-stage real-time target detection model based on deep learning.

[0050] Step S40: Determine the degree of defect in the main pump motor based on the defect distribution.

[0051] Defect severity levels can be categorized as minor, moderate, or severe. The number of defect severity levels can be set according to actual needs.

[0052] Step S50: Issue a defect warning based on the degree of defect.

[0053] Determining the severity of defects based on their distribution allows for a quantitative assessment. Issuing defect alerts based on these alerts ensures accuracy, enabling staff to promptly identify and address existing defects. This approach, while maintaining accuracy and comprehensiveness in testing, enhances the practicality of the results and the efficiency of emergency response, providing more comprehensive protection for the safe operation of the main pump motor.

[0054] In some embodiments, the defect distribution includes defect parameters. Determining the defect severity of the main pump motor based on the defect distribution includes determining the defect severity of the main pump motor based on the defect parameters. These defect parameters are quantifiable; for example, in some embodiments, the defect parameters include the number of defects and / or the defect size. Here, the number of defects can be the number of defective regions, and the defect size can be the size of the defective regions.

[0055] Specifically, in some embodiments, determining the defect severity of the main pump motor based on defect distribution includes: determining the defect severity of the main pump motor based on the number of defects. Multiple correspondences between defect quantities and various defect severity levels can be established. Based on the current defect quantity, the corresponding defect severity is determined according to this correspondence, which is the current defect severity. Here, the defect quantity and defect severity are positively correlated; that is, the more defects, the more severe the defect. Specifically, the defect quantity can be divided into intervals using defect quantity thresholds, with each interval corresponding to a different defect severity. For example, if the defect quantity is greater than a first quantity threshold, the current defect severity is determined to be a severe defect. If it is less than the first quantity threshold, the current defect is determined to be a moderate defect.

[0056] In some embodiments, determining the defect severity of the main pump motor based on defect distribution includes determining the defect severity of the main pump motor based on defect size. Multiple correspondences between defect sizes and multiple defect severity levels can be established. Based on the current defect size, the corresponding defect severity is determined according to this correspondence, which is the current defect severity. Here, defect size and defect severity are positively correlated; that is, the larger the defect size, the more severe the defect. More specifically, defect size can be, for example, the coverage area of ​​the defect, the maximum length of the defect, etc. Specifically, defect size can be divided into intervals using defect size thresholds, with each interval corresponding to a different defect severity.

[0057] During implementation, defect severity can be categorized, and corresponding defect parameter thresholds can be set for each categorization. Based on the quantitative relationship between the actual detected defect parameters and the defect parameter thresholds, the defect severity categorization corresponding to the current defect parameter is determined.

[0058] It is understandable that the main pump motor comprises multiple components, each located in a different position within the main pump motor. For example, as... Figure 6 As shown, the main pump motor provided in this application embodiment includes a bearing cover 11, an end cover 12, a junction box 13, a stator core 14, a stator winding 15, a rotating shaft 16, a frame 17, a bearing 18, a rotor 19, a fan 110, and a casing 111.

[0059] In some embodiments, determining the defect severity of the main pump motor based on defect distribution includes: determining the defect severity of the main pump motor based on the component where the defect is located and the defect parameters. That is, the specific process for determining the defect severity based on defect parameters differs for different components. Differentiated defect severity judgment standards are preset based on the functional characteristics of each component. For example, crack defects on bearing 18, regardless of size, are judged as high-risk defects; wear defects on stator winding 15 with an area exceeding 5 mm² are high-risk defects, while scratch defects on frame 17 with a length not exceeding 10 mm are low-risk defects.

[0060] Specifically, in some embodiments, determining the defect severity of the main pump motor based on defect distribution includes: determining the defect severity of the main pump motor based on the component where the defect is located and the number of defects. There is a corresponding relationship between the number of defects and the defect severity. The number of defects can be divided into intervals using a defect number threshold, with each interval corresponding to a different defect severity. Different defect number thresholds can be set for different components. The defect number threshold is positively correlated with the size of the component, and / or negatively correlated with the importance of the component.

[0061] In some embodiments, determining the defect severity of the main pump motor based on defect distribution includes: determining the defect severity of the main pump motor based on the component where the defect is located and the defect size. There is a correspondence between defect size and defect severity. Defect sizes can be divided into intervals using defect size thresholds, with each interval corresponding to a different defect severity. Different defect size thresholds can be set for different components. Specifically, the defect size threshold is positively correlated with the component size, and / or negatively correlated with the component's importance.

[0062] During implementation, the importance of each component in the main pump motor can be categorized. For example, some components can be classified as important or unimportant. The defect parameter thresholds set for important components are more sensitive to ensure timely detection of critical defects. Important components have a relatively large impact on the main pump motor, such as bearing 18, stator winding 15, and rotor 19, while unimportant components have a relatively smaller impact, such as the frame 17 and housing 111.

[0063] In some embodiments, determining the defect severity of the main pump motor based on defect distribution includes: determining the defect severity of each component based on defect distribution, and then determining the defect severity of the main pump motor based on the defect severity of each component. Functional weight coefficients for each component of the main pump motor are preset, and the overall defect severity of the main pump motor is calculated based on these functional weight coefficients and the defect severity of each component. During implementation, the defect severity can be identified by numerical value for calculation and analysis.

[0064] In some embodiments, a defect alert is issued based on the degree of the defect, including issuing different defect alerts for different degrees of defect. This allows staff to directly determine the current degree of defect through the defect alert, enabling timely follow-up processing.

[0065] For example, minor defects are displayed on the terminal with text information about the affected parts, quantity, and dimensions, with a low priority. Moderate defects are displayed with detailed data in a combination of text and icons, triggering a low-intensity audio-visual alert, with a medium priority. Critical defects are displayed in full screen with defect details and safety risk warnings, activating a high-decibel audio-visual alarm, simultaneously pushing an emergency notification to mobile devices, recording alarm information and storing it in the database, with a high priority.

[0066] Furthermore, targeted inspection recommendations can be provided based on different defect severity levels. For example, minor defect alerts focus on issues that do not require immediate attention but necessitate regular monitoring. Moderate defect alerts emphasize the need for timely maintenance. Critical defect alerts highlight the need for immediate shutdown and handling to ensure rapid response from staff.

[0067] This application provides a system for detecting surface defects of the main pump motor in a refueling truck, including one or more processors for implementing the aforementioned method for detecting surface defects of the main pump motor in a refueling truck.

[0068] Combination Figure 7 As shown, this application provides a system for detecting surface defects of the main pump motor in a refueling truck, including an image acquisition module 10, an image preprocessing module 20, and an image processing module 30. In some embodiments, the aforementioned method for detecting surface defects of the main pump motor in a refueling truck is applied to this system.

[0069] Image acquisition module 10 is used to acquire surface images of the main pump motor. Image preprocessing module 20 is used to preprocess the surface images to obtain the target detection region; preprocessing includes cropping. Image processing module 30 is used to determine the defect distribution in the target detection region using a single-stage real-time target detection model based on deep learning.

[0070] The system for detecting surface defects of the main pump motor in a refueling truck, as provided in this application, acquires an image of the main pump motor surface to capture complete surface information. Then, it performs preprocessing on the surface image, including cropping, to focus on the target detection area and remove redundant interference. Finally, it uses a single-stage real-time target detection model based on deep learning to accurately identify the defect distribution within the target detection area. This effectively avoids the problems of missed or false detections of defects and can comprehensively capture surface defects of different types and scales, thus improving the accuracy and comprehensiveness of main pump motor surface defect detection.

[0071] The image acquisition module 10 can directly acquire surface images of the main pump motor, or it can be connected to an external acquisition device such as a camera. The camera can be, for example, a high-resolution camera.

[0072] In some embodiments, a data transmission module is connected between the image acquisition device and the image preprocessing device to transmit the surface image of the main pump motor to the image preprocessing module 20 wirelessly or via wired means.

[0073] In some embodiments, the image preprocessing module 20 is used to denoise the surface image, enhance the denoised surface image to improve the contrast and / or brightness of the surface image, and crop the enhanced surface image to obtain the target detection region.

[0074] In some embodiments, the image processing module 30 includes a defect detection module 301. The defect detection module 301 determines the distribution of defects in the target detection region using a single-stage real-time target detection model based on deep learning. Specifically, it employs a YOLOv8 model for defect detection, using the model's convolutional neural network structure to quickly and accurately detect defective regions in the image. More specifically, the preprocessed image is input into the trained YOLOv8 model for defect detection. The feature extraction capabilities of the YOLOv8 model are utilized to identify defective regions in the image and generate detection results.

[0075] In some embodiments, the image processing module 30 further includes a defect classification module 302. The defect classification module 302 is used to classify the identified defect regions to obtain the defect type corresponding to each defect region.

[0076] In some embodiments, the system further includes a result display module for displaying detection information in real time based on the detection results. In some embodiments, the system further includes an alarm module for issuing prompt information based on the detection results. Alternatively, in some embodiments, the system includes a result display and alarm module 40. The result display and alarm module 40 is used to display detection information in real time based on the detection results.

[0077] In some embodiments, the system further includes a data storage and management module 50. The data storage and management module 50 is used to store and manage detection data and images, supporting data querying and analysis. Specifically, it is divided into data storage functions and data management functions. The data storage function can be implemented through a database, storing detection data and images in the database to support subsequent data analysis and querying. The data management function can be implemented through a data management system. Using the data management system, data is organized, archived, and retrieved, ensuring data security and integrity.

[0078] As an example, this paper provides a method for detecting surface defects of the main pump motor 62 in a refueling truck, and the application scenario of the system. Combined with... Figure 8 As shown, the scene includes a high-resolution camera 61, a main pump motor 62, a YOLOv8 neural network model 63, a host computer 64, a database 65, a display terminal 66, and staff 67. Staff 67 could be, for example, a refueling truck engineer.

[0079] Corresponding to this scenario, combined with Figure 9 As shown in the figure, this application provides a method for detecting surface defects of the main pump motor in a refueling truck, including steps S101 to S106.

[0080] Step S101: High-resolution images of the surface of the main pump motor 62 are acquired in real time using a high-resolution camera 61.

[0081] A high-resolution camera 61 is installed at a critical location on the main pump motor 62 to ensure comprehensive capture of various details on the motor surface. The acquired images are transmitted to the host computer 64 via a data transmission module.

[0082] In step S102, the host computer 64 performs preprocessing on the acquired images, such as denoising, enhancement, and cropping, to ensure the purity and quality of the images.

[0083] The specific operations include removing noise from the image, improving the image's contrast and brightness to make defects more obvious, focusing on the detection area, and generating high-quality image data suitable for input into the YOLOv8 neural network model 63.

[0084] Step S103: Input the preprocessed image data into the YOLOv8 neural network model 63 for defect detection.

[0085] The YOLOv8 neural network model 63 leverages its powerful feature extraction capabilities to quickly and accurately detect defect regions in images and generate detection results. During the training phase, the model has been optimized using a large amount of labeled data to ensure its high recognition ability for various defect types.

[0086] In step S104, the detection results and defect classification information are stored in database 65.

[0087] Database 65 is used to manage and store detected defect data and images, supporting subsequent data queries and analysis. Through the data management system, data can be organized, archived, and retrieved, ensuring data security and integrity.

[0088] In step S105, the detection results are transmitted to the display terminal 66 via the host computer 64 to display the defect detection and classification results in real time.

[0089] Display terminal 66 provides an intuitive visual interface, showing the location and type of defects. When a serious defect is detected, the system will issue an alarm signal. This corresponds to the aforementioned results display module.

[0090] In step S106, staff member 67 makes fault diagnosis and maintenance decisions through the visual interface of display terminal 66.

[0091] The detection results and early warning information provided by the system can help engineers identify and address potential problems in advance, ensuring the safe operation and timely maintenance of the motor.

[0092] This application provides a refueling truck, including a main pump motor; and a system for detecting surface defects of the main pump motor in the refueling truck.

[0093] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are quite specific and detailed. However, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept of this application, and these modifications and improvements all fall within the protection scope of this application.

[0094] In the description of this application, it should be understood that 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. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for detecting surface defects in the main pump motor of a refueling truck, characterized in that, include: Obtain a surface image of the main pump motor; The surface image is preprocessed to obtain the target detection area; The preprocessing includes trimming; The distribution of defects in the target detection area is determined by a single-stage real-time target detection model based on deep learning.

2. The method according to claim 1, characterized in that, The preprocessing of the surface image to obtain the target detection region includes: The surface image is then denoised. Enhancement processing is applied to the denoised surface image to improve its contrast and / or brightness. The enhanced surface image is cropped to obtain the target detection area.

3. The method according to claim 1, characterized in that, The step of determining the defect distribution in the target detection area using a single-stage real-time target detection model based on deep learning includes: A single-stage real-time target detection model based on deep learning is used to identify defects in the target detection area and obtain the location information of the defects. Defect regions are determined based on the defects and their location information, and each defect region includes at least one complete defect.

4. The method according to claim 3, characterized in that, After determining the defect distribution in the target detection area using a single-stage real-time target detection model based on deep learning, the method further includes: The identified defect regions are classified to obtain the defect type corresponding to each defect region.

5. The method according to claim 4, characterized in that, The process of classifying the identified defect regions to obtain the defect type corresponding to each defect region includes: Extract the feature vectors of each defect region; Based on the feature vectors of each defect region, a pre-trained classification algorithm is used to classify each defect region, thereby obtaining the defect type corresponding to each defect region.

6. The method according to claim 1, characterized in that, The step of identifying the defect distribution in the target detection region using a single-stage real-time target detection model based on deep learning includes: The YOLOv8 model is used to identify the distribution of defects in the target detection area.

7. The method according to claim 1, characterized in that, After determining the defect distribution in the target detection area using a single-stage real-time target detection model based on deep learning, the method further includes: The degree of defect in the main pump motor is determined based on the defect distribution. A defect warning is issued based on the degree of the defect.

8. A system for detecting surface defects of the main pump motor in a refueling truck, characterized in that, It includes one or more processors for implementing the method for detecting surface defects of the main pump motor in a refueling truck as described in any one of claims 1 to 7.

9. A system for detecting surface defects of the main pump motor in a refueling truck, characterized in that, include: The image acquisition module is used to acquire surface images of the main pump motor; An image preprocessing module is used to preprocess the surface image to obtain the target detection region; the preprocessing includes cropping. The image processing module is used to determine the distribution of defects in the target detection area using a single-stage real-time target detection model based on deep learning.

10. A refueling truck, characterized in that, include: Main pump motor; and The system for detecting surface defects of the main pump motor in a refueling truck as described in claim 8 or 9.

Images