Vehicle-mounted refrigerator gap detection method, device and equipment and storage medium

Abstract

The invention relates to a vehicle-mounted refrigerator gap detection method, device and equipment and a storage medium, and the method comprises the steps: building product coordinate information according to a surface mark point of a vehicle-mounted refrigerator, and determining a corresponding gap detection starting point in the product coordinate information; generating a product detection track corresponding to the clearance area of the vehicle-mounted refrigerator according to the product coordinate information, calling a detection sensor to move to a clearance detection starting point, and performing clearance detection on the vehicle-mounted refrigerator along the product detection track; acquiring gap detection data corresponding to the gap detection process of the vehicle-mounted refrigerator, and performing real-time filtering processing on the gap detection data to obtain filtered gap data; and inputting the filtering gap data into a defect identification model pre-trained to a convergence state so as to correspondingly obtain a defect identification result representing the gap error of the vehicle-mounted refrigerator according to the defect confidence output by the defect identification model. According to the invention, an efficient, accurate and stable vehicle-mounted refrigerator gap detection effect is realized, and the gap defect risk of the vehicle-mounted refrigerator is effectively controlled.

Description

Technical Field

[0001] The present application relates to vehicle refrigerator technology, and in particular to a vehicle refrigerator gap detection method, device, equipment and storage medium. Background Art

[0002] In the production process of vehicle refrigerators, gap detection is a critical step in ensuring product quality. With the advancement of automation technology, automating the gap detection process is crucial for improving production efficiency, ensuring detection accuracy, and reducing labor costs. Traditional gap detection methods often rely on manual measurement, resulting in low efficiency, unstable accuracy, and difficulty in data traceability. Furthermore, they lack multi-model compatibility, real-time data exchange, and dynamic adjustment of detection parameters, making them difficult to meet the complex inspection requirements of vehicle refrigerators.

[0003] Furthermore, when faced with the need to inspect multiple product models, current automated inspection equipment often requires complex mechanical adjustments or reprogramming to accommodate the sizes and shapes of different products. This not only increases the complexity and maintenance difficulty of the equipment, but also leads to wasted time and reduced efficiency during model switching. Furthermore, these automated inspection devices typically only operate under preset, fixed programs, making it difficult to acquire and process data from the inspection process in real time, and unable to promptly adjust inspection strategies based on the inspection results. Consequently, they are unable to respond quickly to slight changes in product gaps or sudden inspection anomalies, which can easily lead to missed detections or misjudgments. Furthermore, existing automated inspection equipment struggles to accurately identify gap change trends and potential defect patterns. For example, during the inspection process, factors such as ambient light interference, product surface reflections, or stains can cause noise and deviation in the inspection data, seriously affecting the reliability of the inspection results. Traditional gap detection methods also often struggle to accurately model and measure complex gap shapes or nonlinear gap changes, failing to meet the needs of high-precision inspection.

[0004] Therefore, in order to overcome the defects of the above-mentioned vehicle refrigerator gap detection scheme, there is an urgent need for a technical means that can realize efficient, accurate and stable detection of the vehicle refrigerator gap, so as to meet the requirements of the complex detection process of the vehicle refrigerator and improve product quality and production efficiency. Summary of the Invention

[0005] The purpose of this application is to solve the above problems and provide a vehicle refrigerator gap detection method and its corresponding device, equipment, non-volatile readable storage medium, and computer program product.

[0006] According to one aspect of the present application, a method for detecting a gap between a vehicle-mounted refrigerator and the like is provided, comprising:

[0007] Establish product coordinate information according to the surface marking points of the vehicle refrigerator, and determine a gap detection starting point corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information;

[0008] generating a product detection track corresponding to the gap area of ​​the vehicle refrigerator according to the product coordinate information, calling a preset detection sensor to move to the gap detection starting point, and performing gap detection on the vehicle refrigerator along the product detection track;

[0009] Collecting gap detection data corresponding to the gap detection process of the vehicle refrigerator, performing real-time filtering processing on the gap detection data, and obtaining filtered gap data;

[0010] The filtered gap data is input into a defect recognition model that has been pre-trained to a convergence state, so as to obtain a defect recognition result that characterizes the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model.

[0011] According to another aspect of the present application, a vehicle refrigerator gap detection device is provided, comprising:

[0012] a gap determination module configured to establish product coordinate information based on surface marking points of the vehicle refrigerator, and determine a gap detection starting point corresponding to a gap area of ​​the vehicle refrigerator based on the product coordinate information;

[0013] a gap detection module configured to generate a product detection trajectory corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information, call a preset detection sensor to move to the gap detection starting point, and then perform gap detection on the vehicle refrigerator along the product detection trajectory;

[0014] A data processing module is configured to collect gap detection data corresponding to a gap detection process of the vehicle-mounted refrigerator, perform real-time filtering on the gap detection data, and obtain filtered gap data;

[0015] The defect recognition module is configured to input the filtered gap data into a defect recognition model pre-trained to a convergence state, so as to obtain a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence output by the defect recognition model.

[0016] According to another aspect of the present application, a vehicle-mounted refrigerator gap detection device is provided, including a visual camera, a detection sensor, a manipulator, a data processing unit and a control unit, wherein the visual camera is used to obtain an image of the outer surface of the vehicle-mounted refrigerator and identify surface marking points; the detection sensor is used to perform gap detection on the vehicle-mounted refrigerator along a product detection trajectory and collect gap detection data; the manipulator is used to move the detection sensor to any surface area of ​​the vehicle-mounted refrigerator; the data processing unit is used to perform real-time filtering processing on the collected gap detection data to generate filtered gap data, and output defect confidence according to a defect recognition model deployed in the background; the control unit includes a central processing unit and a memory, and the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the method described in the present application.

[0017] According to another aspect of the present application, a non-volatile readable storage medium is provided, which stores a computer program implemented according to the vehicle refrigerator gap detection method in the form of computer-readable instructions. When the computer program is called and executed by a computer, the steps included in the method are executed.

[0018] According to another aspect of the present application, a computer program product is provided, comprising a computer program / instruction, which implements the steps of the method when executed by a processor.

[0019] Compared with the traditional gap detection technology of vehicle-mounted refrigerators, the present application achieves efficient, accurate and stable gap detection effects. By establishing product coordinate information through the surface marking points of the vehicle-mounted refrigerator, the starting point of the gap detection can be accurately determined, ensuring that the gap detection can start from the correct starting position, effectively improving the accuracy of the detection and reducing the repeatability of the detection; at the same time, a product detection trajectory corresponding to the gap area of ​​the vehicle-mounted refrigerator is generated according to the product coordinate information, and the preset detection sensor is called to detect along this trajectory, which can fully cover the gap area, avoid detection omissions, and improve the integrity of the detection; the gap detection data during the gap detection process of the vehicle-mounted refrigerator is collected and filtered in real time, which can effectively remove the noise and deviation generated in the detection process. Obtaining more accurate filtered gap data significantly improves the reliability and availability of detection data, providing a solid data foundation for subsequent analysis and judgment; finally, the filtered gap data is input into the defect recognition model that has been pre-trained to a convergent state, and the defect recognition result is obtained through the defect confidence output by the model, thereby realizing the automated evaluation of the gap error of the vehicle refrigerator, which not only improves the gap detection efficiency of the vehicle refrigerator and reduces manual intervention, but also provides more objective and accurate gap defect detection results by quantifying the defect confidence, and ultimately realizes efficient and accurate detection of the gap of the vehicle refrigerator, effectively improving the detection quality and production efficiency of the vehicle refrigerator, and providing a strong guarantee for enterprises to produce high-quality vehicle refrigerator products. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of an embodiment of a method for detecting a gap between a vehicle-mounted refrigerator according to the present application;

[0021] Figure 2 This is a principle block diagram of the vehicle refrigerator gap detection device of this application;

[0022] Figure 3 This is a structural diagram of another vehicle refrigerator gap detection device used in this application. DETAILED DESCRIPTION

[0023] The present application is to implement gap detection of a vehicle refrigerator through a vehicle refrigerator gap detection device. The vehicle refrigerator gap detection device mainly includes key components such as a visual camera, a detection sensor, a manipulator, and a control unit including a central processing unit and a memory. These components cooperate with each other through electrical connections and data transmission to jointly complete the vehicle refrigerator gap detection task.

[0024] The central processing unit (CPU), the brains of the control unit, determines the efficiency of the entire vehicle refrigerator gap detection system through its computing power. It possesses high-speed data processing capabilities and the ability to execute multiple tasks in parallel to handle the massive amounts of data and complex control logic generated during gap detection. The CPU is equipped with a dedicated image processing accelerator, such as a GPU, to accelerate the real-time processing of vehicle refrigerator surface images and gap detection data, improving detection efficiency. Furthermore, the CPU features expandable interfaces for subsequent functional upgrades and integration with other systems, such as an MES interface, to achieve seamless integration of production data.

[0025] The memory is responsible for storing the operating system, applications and test data. It can use large-capacity high-speed flash memory to quickly store and read images and data files generated during the test process, ensuring the continuity and stability of the test process. It also has data encryption function to ensure the security and integrity of the test data.

[0026] The visual camera is used to capture images of the vehicle refrigerator's exterior surface, ensuring the accuracy of surface marker recognition and subsequent product coordinate information. In this application, the visual camera can be a high-resolution, high-frame-rate industrial camera equipped with a wide-angle lens and autofocus to accommodate the inspection needs of different vehicle refrigerator models. The camera also features excellent low-light performance and dynamic range, ensuring stable operation in a variety of ambient lighting conditions, regardless of changes in workshop lighting conditions.

[0027] The detection sensor is responsible for detecting gaps in the vehicle refrigerator along the product detection track and collecting data. In this application, the detection sensor can use a non-contact laser displacement sensor, which can measure tiny gaps with high precision, reaching micron-level measurement accuracy, with low repeatability error, and has real-time temperature compensation to reduce the impact of workshop ambient temperature changes on measurement results. This application can also be equipped with multiple sets of sensors to achieve simultaneous detection of gaps in different locations, improving the efficiency of gap detection in the vehicle refrigerator.

[0028] The robot arm is responsible for moving the detection sensor to any surface area of ​​the vehicle refrigerator. In this application, the robot arm is capable of high repeatability and a large load capacity to stably carry the detection sensor and accurately execute complex three-dimensional motion trajectories. The robot arm's control system is deeply integrated with the central processing unit, supporting real-time motion control and trajectory adjustment to meet the testing requirements of different product models. The robot arm's joints feature high-precision torque feedback control, which effectively suppresses vibration during movement and ensures the stability of detection data.

[0029] The data processing unit works in conjunction with the central processing unit to perform real-time filtering on the collected gap detection data to remove noise and deviations, ensuring the accuracy and reliability of the filtered gap data. The data processing module integrates multiple filtering algorithms, such as Kalman filtering and wavelet filtering, and can automatically select the optimal filtering strategy based on different detection scenarios and data characteristics. At the same time, the module uses a parallel processing architecture, which can simultaneously process multi-channel detection data, greatly improving data processing speed and providing a strong guarantee for real-time detection. The data processing unit also has a built-in dedicated graphics processing architecture that complements the image processing acceleration unit of the central processing unit. The data processing unit is capable of running a defect recognition model that has been pre-trained to a convergence state. This defect recognition model is trained to a convergence state based on a deep learning algorithm. It can extract and analyze features from the filtered gap data and output a defect confidence level, thereby achieving automatic identification and assessment of gap errors in vehicle refrigerators. At the same time, the data processing unit is equipped with a high-speed data interface for fast data transmission and communication with other equipment components to ensure smooth and real-time data flow. It has good scalability and compatibility and can be easily integrated into the system architecture of the entire vehicle refrigerator gap detection equipment to work with other modules to achieve an efficient gap detection process.

[0030] The robot's motion control system is deeply integrated with the detection sensor's data acquisition system, enabling synchronized control of data acquisition and robot motion, ensuring data accuracy and relevance. The robot's motion trajectory planning software automatically generates the optimal motion path based on the vehicle refrigerator's three-dimensional geometric model, improving detection efficiency and reducing mechanical wear.

[0031] The trigger system of the visual camera is closely connected with the control signal of the central processor, ensuring that the image can be captured in time when the robot carrying the detection sensor moves to the optimal shooting position. Its image acquisition software has automatic exposure and white balance adjustment functions, which can optimize image quality in real time and improve the recognition rate of surface marking points. At the same time, the mounting bracket of the visual camera adopts an adjustable design, which can quickly adapt to different models of car refrigerators to ensure the optimal angle and distance of image acquisition.

[0032] The mounting fixture of the detection sensor not only ensures the stability and measurement accuracy of the sensor during the movement of the robot, but can also be quickly disassembled and replaced to adapt to different types of detection tasks. Its internal wiring adopts an anti-interference design, which can effectively shield electromagnetic noise and ensure the stability of signal transmission. At the same time, the casing of the detection sensor adopts a dustproof and waterproof design to adapt to the changing environmental conditions of the workshop.

[0033] The control software run by the central processing unit has a human-machine interaction interface. Operators can use the touch screen to set parameters, start and stop programs, and perform other operations. Its built-in fault diagnosis system can monitor the equipment's operating status in real time, promptly detect and warn of potential faults, and reduce downtime. At the same time, the central processing unit is seamlessly connected to the factory's MES system, and can upload test data and results in real time, realizing full traceability and quality control of the production process.

[0034] The above-mentioned vehicle refrigerator gap detection equipment can not only efficiently execute each step of the vehicle refrigerator gap detection method, but also ensure the stability, reliability and flexibility of the detection process, and meet the strict requirements of vehicle refrigerator quality control on modern automated production lines. Through the close collaboration between various components, the vehicle refrigerator gap detection equipment can realize the full process automation of vehicle refrigerator gap detection, providing strong guarantees for gap defect control and high-quality production of vehicle refrigerators.

[0035] See also Figure 1 According to a vehicle refrigerator gap detection method provided by the present application, the method can be implemented as a computer program product installed in a vehicle refrigerator gap detection device and run. In some embodiments, the method includes the following steps:

[0036] Step S1001: establishing product coordinate information according to surface marking points of the vehicle refrigerator, and determining a gap detection starting point corresponding to a gap area of ​​the vehicle refrigerator based on the product coordinate information;

[0037] When the vehicle refrigerator enters the gap detection step, after calling the preset slide mechanism to fix the vehicle refrigerator to the gap detection station, the visual camera is called to obtain the outer surface image of the vehicle refrigerator. The visual camera can use a high-resolution, high-frame-rate industrial camera equipped with a wide-angle lens and autofocus function to meet the detection needs of different models of vehicle refrigerators. The visual camera can also have good low-light performance and dynamic range detection performance, so that it can work stably under various ambient light conditions and is not affected by changes in workshop lighting conditions. After the visual camera obtains the outer surface image of the vehicle refrigerator, the image can be pre-processed. The pre-processing process can include one or more of grayscale processing, binarization processing, and edge enhancement processing. After the pre-processing operation improves the contrast and clarity of the outer surface image, it can facilitate subsequent recognition and analysis of the outer surface image.

[0038] Furthermore, the preprocessed outer surface image is identified according to a preset image recognition algorithm, and the coordinate information of the marker points corresponding to one or more surface marker points is extracted. Surface marker points are high-contrast physical markers pre-set on the surface of the vehicle refrigerator for positioning and identification; the marker point coordinate information includes the number, characteristics, and coordinate distribution data of the marker points. The model information of the vehicle refrigerator can be determined through the surface marker points and the marker point coordinate information. Then, based on the surface marker points, the preset product coordinate system parameters corresponding to the vehicle refrigerator model information are called to construct the product coordinate system of the vehicle refrigerator model. The coordinate origin and coordinate axis direction of the product coordinate system are determined based on the marker point coordinate information, thereby generating complete product coordinate information from the coordinate origin and coordinate axis direction of the product coordinate system, which serves as the basis for gap positioning and detection navigation in the subsequent gap detection process.

[0039] After obtaining the product coordinate information, the gap area of ​​the vehicle refrigerator is determined based on the product coordinate information. The gap area refers to the specific part of the vehicle refrigerator that needs to be gap detected, such as the gap between the door and the main body, the gaps at various parts of the door, etc. The gap area has a clear position and range in the product coordinate system, so that the area of ​​the vehicle refrigerator that needs gap detection can be detected more accurately. When the specific position of the gap area of ​​the vehicle refrigerator in the product coordinate system corresponding to the product coordinate information is obtained, the gap detection starting point can be determined based on the position and range of the gap area. In this application, the gap detection starting point is the initial position where the corresponding detection sensor starts to detect the gap when the vehicle refrigerator needs to be gap detected. It is usually selected at the starting end of a certain surface or a certain edge corresponding to the gap area, and sometimes it can also be set at the center position. The selection of the gap detection starting point can ensure that the detection sensor can start from this position and fully cover the gap area along the predetermined detection trajectory.

[0040] Through the above steps, the product coordinate information is established based on the surface marking points of the vehicle refrigerator, and the starting point of the gap detection is determined, which provides a precise starting position and detection positioning basis for the subsequent gap detection process, thereby ensuring the accuracy and reliability of gap detection, and laying a solid foundation for achieving more efficient and accurate vehicle refrigerator gap detection.

[0041] Step S1002: generating a product detection track corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information, calling a preset detection sensor to move to the gap detection starting point, and performing a gap detection on the vehicle refrigerator along the product detection track;

[0042] After determining the gap area and gap detection starting point of the vehicle refrigerator, a product detection trajectory is generated for gap detection of the vehicle refrigerator. The product detection trajectory is the path followed by the detection sensor during gap detection. This detection path fully covers the gap area of ​​the vehicle refrigerator to ensure the completeness and accuracy of gap detection.

[0043] In this embodiment, the geometric shape characteristics and gap type of the gap area in the vehicle refrigerator that needs to be inspected are first determined based on the product coordinate information and the pre-stored three-dimensional geometric model of the vehicle refrigerator. Geometric shape characteristics include the length, width, and curvature of the gap to be inspected; the gap type refers to whether the gap to be inspected is a straight segment gap or a curved segment gap. For example, after the vehicle refrigerator is fully installed, there will be a gap between the door and the refrigerator body. For example, if one side of the vehicle refrigerator is a quadrilateral, there will be four corresponding straight sides and four corners. Each corner can be a right angle or a curve. In this case, the four straight sides of the quadrilateral correspond to straight segment gaps, and the four corners correspond to curved segment gaps. Furthermore, based on the geometric features of the gap detection starting point and the gap area, a product detection trajectory is generated. This product detection trajectory includes multiple equally spaced track points. The spacing between adjacent track points is adjusted according to the gap width corresponding to the vehicle refrigerator model. When a straight gap is detected, the track points are set using a first preset spacing value; when a curved gap is detected, the track points are set using a second preset spacing value that is smaller than the first preset spacing value. For example, for a straight gap, the preset spacing value can be set to 1 mm; for a curved gap, the preset spacing value can be set to 0.5 mm. This ensures that the detection sensor can collect data more densely at the curved gap, thereby improving detection accuracy.

[0044] The detection sensor, which is invoked after the product inspection trajectory is generated, can be a non-contact laser displacement sensor. Its measurement accuracy can reach the micron level, with low repeatability errors and real-time temperature compensation to reduce the impact of ambient temperature fluctuations in the workshop on measurement results. In actual settings, the detection sensor can also have high repeatability and a large load capacity to ensure stable detection capabilities even during the movement of the robot. The robot is responsible for moving the detection sensor to the gap detection starting point. The robot's control system is deeply integrated with the central processing unit, supporting real-time motion control and trajectory adjustment. Its joints feature high-precision torque feedback control, which effectively suppresses jitter during movement and ensures the stability of detection data. Once the detection sensor reaches the gap detection starting point, the robot moves the detection sensor along the product inspection trajectory. During this movement, the detection sensor collects gap detection data, including a series of gap width measurement values, in real time for subsequent analysis and processing.

[0045] Therefore, this application generates a product detection trajectory corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information, and calls the preset detection sensor to perform gap detection along the trajectory, ensuring that the detection sensor can start from a precise starting position and fully cover the gap area along the predetermined detection trajectory, thereby achieving efficient and accurate gap detection, and providing a reliable data basis for subsequent data processing and gap defect identification.

[0046] Step S1003: collecting gap detection data corresponding to the gap detection process of the vehicle refrigerator, performing real-time filtering processing on the gap detection data, and obtaining filtered gap data;

[0047] During gap detection on vehicle refrigerators, gap detection data is first collected as the detection sensor moves along the product detection trajectory. This data is then filtered in real time. This real-time filtering removes noise and outliers from the data, improving its accuracy and reliability. In practice, a Kalman filter can be used to process dynamically changing gap data. Through prediction and update steps, the state of the gap detection process is estimated in real time, and corrections are made based on the measured data to effectively remove noise from the data. For example, if the gap width of a vehicle refrigerator exhibits regular changes during detection, a Kalman filter can predict the current gap width based on previous data and update it based on the actual measured value, resulting in a more accurate estimate. For real-time filtering during gap detection on vehicle refrigerators, wavelet filtering can also be used to decompose the digital signal corresponding to the gap detection data and remove high-frequency noise while preserving key features. The wavelet basis and decomposition level of the Daubechies wavelet can be selected to achieve refined processing of the gap detection data. For example, for gap detection data on curved segments, wavelet filtering can remove high-frequency noise caused by sensor vibration while preserving detailed features of the gap variation.

[0048] In one embodiment, Kalman filtering and wavelet filtering can be combined to perform two-stage filtering. First, Kalman filtering is used to remove high-frequency noise, and then wavelet filtering is used to smooth the data to further improve data stability and reliability.

[0049] During the filtering process, abnormal trajectory points are identified and removed to ensure more accurate and reliable processed data. These are defined as points where the measured values ​​collected by the detection sensor significantly deviate from the normal range during the gap detection process. These points' values, when compared to adjacent points, exceed a preset threshold. These abnormal points may be caused by sensor noise, electromagnetic interference from the environment, surface stains on the refrigerator, or mechanical vibrations from the gap detection station. During the identification of abnormal trajectory points, based on the measured values ​​of adjacent trajectory points, if the difference between a trajectory point's measured value and those of its preceding and following adjacent points exceeds a preset threshold, for example, if the difference between the gap measured value of a trajectory point and its preceding and following adjacent points exceeds 10%, the point is identified as an abnormal point. For example, if the gap measured value of a trajectory point is 0.5 mm, while the gap values ​​of its preceding and following adjacent points are both 0.45 mm, and the threshold is set to 10%, the point is identified as abnormal due to exceeding the threshold and is removed or corrected accordingly. The corresponding filtered gap data is then obtained after filtering the gap detection data.

[0050] In one embodiment, after removing abnormal trajectory points, the filtered gap detection data can be further verified. First, the gap width measurement values ​​collected by the detection sensor at each trajectory point on the product detection trajectory are checked for integrity. Each trajectory point corresponds to a data point. By verifying whether the number of data points is consistent with the expected number of trajectory points, it is ensured that no trajectory point data is lost or interrupted. Then, the deviation between the measured value of the current trajectory point and the pre-stored standard data is compared. If the deviation exceeds a preset deviation threshold, it is determined whether the detection sensor has drift or calibration issues. Finally, the gap of the vehicle refrigerator can be repeatedly tested. By calculating whether the standard deviation of multiple measurements at the same trajectory point exceeds the standard deviation threshold, it is determined whether there is mechanical vibration or environmental interference in the gap detection station of the vehicle refrigerator.

[0051] By filtering the gap detection data, the continuity and consistency of the corresponding measurement value sequence can be further ensured. The filtered gap data obtained in this way is used for subsequent data analysis and processing, making the gap detection data more accurate and reliable.

[0052] Step S1004: input the filtered gap data into a defect recognition model that has been pre-trained to a convergence state, so as to obtain a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model.

[0053] Before inputting the filtered gap data into the defect recognition model, a corresponding gap feature vector must be generated. This gap feature vector also contains the sequence of gap width measurements and the corresponding product coordinate information. For example, a gap feature vector can be composed of the sequence of gap width measurements and the coordinates of the corresponding trajectory points, resulting in a single vector containing the measured gap widths of the vehicle refrigerator and the coordinates of the corresponding trajectory points. The defect recognition model is trained to convergence using a deep learning algorithm and includes a variety of labeled gap error feature samples. The training dataset consists of gap feature vectors and defect labels corresponding to a large amount of historical gap inspection data.

[0054] In this embodiment, the defect recognition model extracts gap error features through multi-level convolution kernels to generate the defect confidence of the vehicle refrigerator. For example, a convolutional neural network (CNN) is used as a defect recognition model. After the input layer of the convolutional neural network receives the gap feature vector, the hidden layer extracts features through the convolution kernel, and the final output layer generates the defect confidence. During the training of the convolutional neural network model, the model convergence is ensured by using the cross entropy loss function and the Adam optimizer. The training data set can contain 1000 samples, each sample containing the gap width measurement value and coordinate information of 100 detection trajectory points. During the model training process, the learning rate is set to 0.0005, the batch size is 32, and the number of training epochs is 50. The learning rate is adjusted accordingly according to the verification loss during the training process, and whether the number of training rounds needs to be increased is determined according to the convergence of the convolutional neural network model.

[0055] The gap feature vector generated corresponding to the filtered gap data is input into the trained defect recognition model, which then performs forward propagation calculations and outputs a defect confidence level. For example, the defect recognition model analyzes that the measured value of a certain trajectory point is 0.5 mm, which deviates from the standard value of 0.45 mm. Combined with the product coordinate information, the defect confidence level output for this point is 85%, indicating that there is a high probability of a gap error defect here. The final defect recognition result includes the gap defect type of the vehicle refrigerator, such as the gap between the door and the main body being too wide or the gaps in different parts of the door being uneven. If the defect confidence level output by the defect recognition model is 85%, the corresponding defect type is the gap between the door and the main body being too wide. The corresponding defect recognition result can provide clear guidance for subsequent vehicle refrigerator production quality control.

[0056] In one embodiment, during the data preprocessing stage, the filtered gap data is normalized to ensure that the data range is between [0, 1]. For example, the gap width measurement value is converted from millimeters to a normalized value. Assuming that the original measurement value range is 0-1 mm, the normalized range is [0, 1]. In addition, standardization can be used to convert the data into a distribution with a mean of 0 and a standard deviation of 1. When the gap feature vector is generated, the normalized gap width measurement value and product coordinate information are combined into a gap feature vector, so that the gap feature vector can be represented as a vector containing the gap width measurement value and trajectory point coordinate information. In addition, features such as the gap change rate and the timestamp of the trajectory point can be added, so that the gap feature vector can be expanded to a vector containing the gap width measurement value, product coordinate information, the gap change rate, and the timestamp of the trajectory point. During the model training stage, in addition to using a convolutional neural network (CNN), a recurrent neural network (RNN) or a long short-term memory (LSTM) network can also be used to process time series data. When the LSTM model is used to process the gap width measurement value sequence, it can better capture the temporal variation characteristics. During the model training process, data enhancement techniques can also be used, such as adding noise, randomly jittering trajectory point coordinates, etc., randomly adding Gaussian noise to the training data with a noise standard deviation of 0.05 mm, etc. to process the data to improve the robustness of the model.

[0057] The defect identification result of the vehicle refrigerator finally obtained in this way can realize the automatic evaluation of the gap error of the vehicle refrigerator very accurately. This application further processes the filtered data with a deep learning model. Regardless of whether the defect identification result is that the vehicle refrigerator does not have defects or that the vehicle refrigerator has defects of a certain type, the corresponding confidence results are very reliable and accurate. The defect identification result is used to determine whether there are gap defects in the vehicle refrigerator, which not only improves the efficiency and accuracy of gap detection, but also provides reliable detection support for the overall production process of the vehicle refrigerator, and provides very strong support for the efficient and high-quality production of vehicle refrigerators.

[0058] It is not difficult to understand from the above embodiments that the present application has achieved significant beneficial effects compared to the traditional vehicle refrigerator gap detection method, including but not limited to:

[0059] The vehicle refrigerator gap detection method proposed in this application achieves efficient, accurate and stable gap detection results compared to traditional manual visual inspection or visual camera recognition methods. By acquiring the outer surface image of the vehicle refrigerator through a visual camera and performing preprocessing operations, the contrast and clarity of the image can be effectively improved, laying the foundation for subsequent product mark point recognition and product coordinate information extraction. By constructing a product coordinate system that matches the vehicle refrigerator model and determining the gap detection starting point for gap detection based on the corresponding acquired product coordinate information, the adaptability and flexibility of the detection system are enhanced, and it can cope with vehicle refrigerators of different models and sizes. For example, the distribution and characteristics of the surface mark points of different models of vehicle refrigerators may be different. By quickly and accurately establishing the corresponding product coordinate system and determining the starting point of the gap detection. Furthermore, a product detection trajectory corresponding to the gap area of ​​the vehicle refrigerator is generated based on the product coordinate information, and the detection sensor is called to perform gap detection along the trajectory, ensuring that the detection sensor can start from a precise starting position and fully cover the gap area along the predetermined detection trajectory, thereby achieving efficient and accurate gap detection. By dynamically adjusting the spacing of the trajectory points, for example, using a larger spacing value for the gap in the straight segment and a smaller spacing value for the gap in the curved segment, the accuracy of the vehicle refrigerator gap detection is also improved.

[0060] Real-time filtering of the collected gap detection data can effectively remove noise and outliers, improving data accuracy and reliability. Identifying and eliminating abnormal trajectory points ensures the accuracy of the filtered gap data. Inputting the filtered gap detection data into a defect recognition model that has been pre-trained to convergence automatically outputs defect confidence levels, enabling automated assessment of gap errors in vehicle refrigerators. The resulting defect recognition results include gap defect types, such as excessively wide gaps between the door and the main body or uneven gaps across the door. This provides clear guidance for production quality control of vehicle refrigerators, helping to promptly identify and resolve production problems and improve product quality.

[0061] Based on any embodiment of the method of the present application, establishing product coordinate information according to surface marking points of the vehicle refrigerator, and determining a gap detection starting point corresponding to a gap area of ​​the vehicle refrigerator based on the product coordinate information includes:

[0062] Step S2001: calling a preset visual camera to obtain an image of the outer surface of the vehicle refrigerator, and identifying the surface marking points on the outer surface image;

[0063] When using a vision camera to obtain images of the exterior surface of a vehicle refrigerator, a high-resolution, high-frame-rate industrial camera is typically used to ensure image clarity and detail capture. Equipped with a wide-angle lens, it can cover a larger field of view and accommodate different models of vehicle refrigerators. The camera's autofocus function also ensures clear images at varying distances, while its low-light performance and dynamic range ensure stable operation under various workshop lighting conditions.

[0064] After acquiring the exterior surface image, the corresponding surface markers are identified. These surface markers are high-contrast physical marks pre-placed on the refrigerator's surface for positioning and identification. They can be QR codes, barcodes, or other high-contrast patterns. Their model data and coordinate distribution are extracted to determine the refrigerator's corresponding model information and product coordinates. This ensures that during subsequent gap detection, the sensor can accurately detect gaps from the correct starting position, improving detection accuracy and reliability.

[0065] Step S2002: determining the model information of the vehicle refrigerator based on the surface marking points, constructing a product coordinate system for the vehicle refrigerator of the model, and obtaining the product coordinate information according to the product coordinate system;

[0066] After determining the surface marker points on the outer surface image of the vehicle refrigerator, a product coordinate system that matches the model of the vehicle refrigerator is then established based on the surface marker points. In practical applications, a template matching algorithm or a feature point matching algorithm can be used to identify and match surface marker points to compare the surface marker points in the outer surface image with the pre-stored marker point template to determine the model information and product coordinate information of the vehicle refrigerator. Once the model of the vehicle refrigerator is determined, a corresponding product coordinate system can be constructed based on the model. The coordinate origin and coordinate axis direction of the product coordinate system are determined based on the product coordinate information of the surface marker points. The coordinate origin can be set at a fixed position on the vehicle refrigerator, such as the center of the door or a corner. The direction of the coordinate axis can be determined based on the distribution of the surface marker points to ensure that the product coordinate system is aligned with the actual geometric structure of the vehicle refrigerator.

[0067] By constructing a product coordinate system, corresponding product coordinate information can be obtained. This product coordinate information is used to determine the specific areas of the vehicle refrigerator requiring gap detection. For example, the product coordinate system can be used to determine the specific locations of the gap between the door and the main body, as well as the gaps within the door itself. This product coordinate information further provides a precise positioning foundation for subsequent gap detection, ensuring that the detection sensor starts from the correct starting position and accurately performs gap detection, thereby improving detection accuracy and reliability.

[0068] Step S2003: Determine the gap area of ​​the vehicle refrigerator according to the product coordinate information, and determine the gap detection starting point in the gap area based on the product coordinate system.

[0069] After determining the product coordinate system and product coordinate information of the vehicle refrigerator, it is necessary to determine the gap area of ​​the vehicle refrigerator based on the product coordinate information, and determine the gap detection starting point for gap detection within the gap area. The gap area refers to the specific part of the vehicle refrigerator that requires gap detection, such as the gap between the door and the main body, the gaps at various parts of the door, etc. These areas have clear positions and ranges in the product coordinate system and can be accurately determined by the product coordinate information. For example, the specific position of the door edge can be defined by the coordinate value in the product coordinate system to determine the area where gap detection is required. After determining the gap area, it is necessary to select one or more gap detection starting points within the gap area. The gap detection starting point is the initial position where the detection sensor starts to detect the gap. It is usually selected at the starting end of a certain surface or edge of the gap area, and sometimes it can also be set at the center position. The purpose of selecting the detection starting point is to ensure that the detection sensor can start from this position and fully cover the gap area along the predetermined detection trajectory, thereby achieving efficient and accurate gap detection.

[0070] In this embodiment, the selection of the gap detection starting point can be determined based on factors such as the geometric shape of the gap area and the motion path planning of the detection sensor. For example, if the gap area is a rectangle, a corner of the rectangle can be selected as the detection starting point, and the detection sensor can be moved along the edge of the rectangle to cover the entire gap area. If the gap area is a complex curve, an endpoint of the curve can be selected as the detection starting point, and the detection path can be dynamically adjusted based on the shape characteristics of the curve. This ensures that during the subsequent gap detection process, the detection sensor can accurately perform gap detection starting from the correct starting position, thereby improving the accuracy and reliability of detection.

[0071] This embodiment uses a high-resolution and high-frame-rate industrial camera to obtain clear external surface images, and identifies surface marker points through a template matching algorithm or a feature point matching algorithm, thereby ensuring the accuracy of image recognition. The product coordinate system constructed based on the surface marker points can accurately determine the model information of the vehicle refrigerator, thereby providing an accurate positioning basis for subsequent gap detection. The gap area and its gap detection starting point are determined through the product coordinate system to ensure that the detection sensor can start from the correct starting position, fully cover the gap area, and achieve efficient and accurate gap detection. The above steps work together to significantly improve the accuracy and reliability of vehicle refrigerator gap detection, and provide strong support for the quality control of vehicle refrigerators.

[0072] Based on any embodiment of the method of the present application, identifying the surface marker points on the outer surface image, determining the model information of the vehicle refrigerator based on the surface marker points, constructing a product coordinate system for the vehicle refrigerator model, and obtaining the product coordinate information based on the product coordinate system includes:

[0073] Step S3001: performing a preprocessing operation on the outer surface image, wherein the preprocessing operation includes one or more of grayscale processing, binarization processing, and edge enhancement processing;

[0074] In this embodiment, after obtaining the outer surface image of the vehicle refrigerator, corresponding preprocessing operations can be performed to improve the image quality and the accuracy of feature extraction, wherein the preprocessing operations may include one or more of grayscale processing, binarization processing, and edge enhancement processing.

[0075] Specifically, if you want to grayscale an image of the exterior surface of a car refrigerator, you need to convert the corresponding color image into a grayscale image. Because color images contain information about the three color channels (red, green, and blue), while grayscale images only contain brightness information, grayscale processing can reduce image data redundancy and simplify subsequent image processing steps. For example, a weighted average method can be used to convert an RGB image of the exterior surface of a car refrigerator into a grayscale image, where the weights for red, green, and blue can be set to 0.299, 0.587, and 0.114, respectively.

[0076] Further binarization is the process of converting a color image or grayscale image into a black and white binary image. In binarization, the pixel values ​​in the image are set to one of two thresholds, usually 0 and 1, or black and white. By setting a grayscale threshold, the pixels in the image can be divided into two categories: pixels below the threshold are set to black, and pixels above the threshold are set to white, helping to highlight important features in the image and reduce noise interference.

[0077] Furthermore, edge enhancement processing can highlight edge information in the image, making the outlines of objects in the image clearer. An edge is the boundary between different areas in an image, usually corresponding to the outline or surface changes of an object. Edge enhancement can be achieved through a variety of methods, such as the Sobel operator and the Canny edge detection algorithm. The Sobel operator detects edges by calculating the gradient magnitude and direction of each pixel in the image, while the Canny edge detection algorithm combines multiple calculation methods such as Gaussian filtering, gradient calculation, non-maximum suppression, and dual threshold detection to more accurately detect edges in the image.

[0078] Through the above-mentioned preprocessing operation on the outer surface image, the contrast and clarity of the outer surface image of the vehicle refrigerator can be significantly improved to better meet the subsequent image recognition and analysis requirements.

[0079] Step S3002: identifying the preprocessed outer surface image according to a preset image recognition algorithm, extracting marker point coordinate information corresponding to one or more surface marker points to locate the surface marker points, and determining the model information of the vehicle refrigerator according to the marker point coordinate information, wherein the marker point marking information includes marker point quantity information, marker point feature information, and coordinate distribution data of the surface marker points;

[0080] The coordinate information of the surface markers in the pre-processed external surface image is identified and extracted using a preset image recognition algorithm. The image recognition algorithm can be selected and configured using a template matching algorithm or a feature point matching algorithm. The template matching algorithm compares a pre-stored marker template with the area in the image to identify the area that best matches the template, thereby determining the location of the marker. Alternatively, the feature point matching algorithm extracts feature points such as corners and edges from the image and compares them with pre-stored marker features to identify the location of the marker, completing the determination of the marker coordinate information for the surface marker. This allows the corresponding surface marker to be located in the product coordinate system.

[0081] After identifying the surface markers, the corresponding extracted marker coordinate information includes marker quantity information, marker feature information, and coordinate distribution data of the surface markers. The marker quantity information refers to the total number of surface markers identified in the external surface image, the marker feature information refers to features such as the shape, size, and orientation of the markers, and the coordinate distribution data refers to the specific position coordinates of the surface markers in the external surface image. The extracted marker coordinate information is compared with the pre-saved vehicle refrigerator model template. The model corresponding to the successfully matched vehicle refrigerator model template is determined as the model of the vehicle refrigerator. The matching method can be achieved by calculating the relative position relationship between the surface markers, the number and features of the surface markers, and other information to achieve matching. The vehicle refrigerator model information is thus determined based on the marker coordinate information corresponding to the surface markers.

[0082] Step S3003: Based on the surface marking points, call the preset product coordinate system parameters corresponding to the model information of the vehicle refrigerator to construct the product coordinate system of the vehicle refrigerator of this model, and determine the product coordinate information corresponding to the vehicle refrigerator of this model according to the coordinate origin and coordinate axis direction of the product coordinate system.

[0083] After identifying the surface markers and extracting their coordinates, the system uses the preset product coordinate system parameters corresponding to the vehicle refrigerator model to construct the product coordinate system for that model. This product coordinate system provides a precise positioning framework during gap detection, ensuring that the sensor starts from the correct starting position and accurately detects gaps within the vehicle refrigerator.

[0084] The construction of the product coordinate system relies on the coordinate information of the surface marker points. Its coordinate origin and coordinate axis direction are determined according to the distribution of the surface marker points. By constructing the product coordinate system, the product coordinate information can be obtained. The product coordinate information includes the position of the coordinate origin and the direction of the coordinate axis, so as to determine the specific area of ​​the vehicle refrigerator where gap detection is required based on the coordinate origin and the coordinate axis. In practical applications, the construction of the product coordinate system can be achieved in a variety of ways. For example, a coordinate transformation algorithm can be used to convert the coordinate information of the surface marker points into the coordinate information of the product coordinate system through operations such as translation, rotation, and scaling to ensure that the product coordinate system is aligned with the actual geometric structure of the vehicle refrigerator. Geometric modeling tools can also be used to automatically generate the product coordinate system based on the distribution characteristics of the surface marker points. The specific implementation can be flexibly performed by those skilled in the art. This ensures that in the subsequent gap detection process, the detection sensor can start from the correct starting position and accurately perform gap detection, thereby improving the accuracy and reliability of the gap detection of the vehicle refrigerator.

[0085] According to this embodiment, after obtaining a clear external surface image using a high-resolution and high-frame-rate industrial camera, the surface marker points are identified through a template matching algorithm or a feature point matching algorithm to ensure the accuracy of image recognition. The product coordinate system constructed based on the surface marker points can accurately determine the model information of the vehicle refrigerator, thereby providing an accurate positioning basis for subsequent gap detection. The gap area and its detection starting point are determined through the product coordinate system, ensuring that the detection sensor can detect the gap area that needs to be detected from the correct starting position, thereby achieving efficient and accurate gap detection. The above embodiment ensures the efficiency and accuracy of the vehicle refrigerator gap detection process, providing reliable technical support for the quality control of vehicle refrigerators.

[0086] Based on any embodiment of the method of the present application, generating a product detection trajectory corresponding to the gap area of ​​the vehicle refrigerator according to the product coordinate information, calling a preset detection sensor to move to the gap detection starting point, and performing gap detection on the vehicle refrigerator along the product detection trajectory, including:

[0087] Step S4001: Determine the geometric shape characteristics and gap type of the gap area based on the product coordinate information and the pre-stored three-dimensional geometric model of the vehicle refrigerator;

[0088] The geometric features and gap type of the gap area are determined based on the vehicle refrigerator's product coordinate information and a pre-stored three-dimensional geometric model. The three-dimensional geometric model is a digital representation of the vehicle refrigerator, containing its geometric shape and structural information. By comparing the product coordinate information with the three-dimensional geometric model, the specific area requiring gap detection and its geometric features can be determined. The geometric features of the gap area include parameters such as the gap's length, width, and curvature. For example, the gap can be a straight line segment or a curved line segment, and its length and width can be calculated using the coordinate data in the geometric model. The gap type refers to the geometric form of the gap, such as whether it is a straight line segment gap or a curved line segment gap. Straight line segment gaps typically occur on the straight edge of the door, while curved line segment gaps occur at the corners or curved portions of the door. By analyzing the three-dimensional geometric model of the vehicle refrigerator, the specific type of the gap area in the vehicle refrigerator can be determined.

[0089] Step S4002: Generate a product detection trajectory based on the gap detection starting point and the gap area, and control the detection sensor to move along the product detection trajectory, wherein the product detection trajectory includes a plurality of equally spaced trajectory points, and the spacing between adjacent trajectory points is adjusted according to the gap width corresponding to the model of the vehicle refrigerator. When the gap type is detected to be a straight segment gap, the trajectory point is set using a first preset spacing value; when a curved segment gap is detected, the trajectory point is set using a second preset spacing value that is smaller than the first preset spacing value.

[0090] After determining the geometric features and gap type of the gap area, a product detection trajectory is generated based on the gap detection starting point and the gap area. The product detection trajectory is the path followed by the detection sensor when performing gap detection on a vehicle refrigerator, ensuring that the detection sensor fully covers the gap area, thereby achieving efficient and accurate gap detection. Generating the product detection trajectory requires determining the distribution of trajectory points. Trajectory points are specific locations that the detection sensor needs to pass during the detection process. The distribution density of trajectory points can be adjusted based on the type and width of the gap area. Since the geometry of gap areas with straight-line gaps is relatively simple, the detection sensor can still accurately detect gap changes at larger gaps. Therefore, when the detection sensor detects that the gap type of the vehicle refrigerator gap area is a straight-line gap, a relatively large first preset spacing value can be used to set the trajectory points. However, the geometry of curved-line gaps is more complex. To enable the detection sensor to more accurately capture gap changes at smaller gaps, thereby improving detection accuracy, when the gap type of the detected gap area is a curved-line gap, a second preset spacing value, which is smaller than the first preset spacing value, is used to set the trajectory points. For example, for curve segment gap, the spacing of track points can be set to 0.5 mm, while for straight segment gap, the spacing of track points can be set to 1 mm.

[0091] After generating the product detection trajectory, the detection sensor can be controlled to move along the product detection trajectory. The movement of the detection sensor is controlled by the robot, and the robot's motion control system is deeply integrated with the central processing unit to support real-time motion control and trajectory adjustment. The joints of the robot have high-precision torque feedback control to effectively suppress jitter during movement and ensure the stability of product detection data.

[0092] It can be seen that this embodiment can significantly improve the accuracy and comprehensiveness of detection. By analyzing the product coordinate information and the three-dimensional geometric model, the geometric shape characteristics and gap type of the gap area are determined, thereby ensuring the accuracy of the detection process. The product detection trajectory generated based on the gap detection starting point and the gap area can adjust the spacing of the trajectory points in combination with different gap types to ensure that the detection sensor can fully cover and accurately detect the gap area, effectively improving the production efficiency and measurement accuracy of the gap detection of vehicle-mounted refrigerators.

[0093] Based on any embodiment of the method of the present application, collecting gap detection data corresponding to the gap detection process of the vehicle refrigerator, performing real-time filtering processing on the gap detection data, and obtaining filtered gap data, including:

[0094] Step S5001: collecting the gap detection data along the product detection track according to the detection sensor, wherein the gap detection data includes a measurement value sequence representing the gap width of the gap area;

[0095] After determining the product detection trajectory and controlling the detection sensor to move along the product detection trajectory, it is necessary to use the detection sensor to collect gap detection data. The measurement value sequence is a set of gap width values ​​recorded in sequence by the detection sensor during the gap detection process. Each gap width measurement value corresponds to the gap width detected by the detection sensor at a trajectory point on the product detection trajectory. In order to collect the gap width measurement value, the detection sensor will pause or move at a constant speed at each trajectory point during the movement process, and record the gap width at that position. The gap width measurement value can thus reflect the width change of the gap area at different positions. For example, if the product detection trajectory covers the gap between the door and the main body of a car refrigerator, the detection sensor will record the corresponding gap width values ​​at different positions of the gap, forming a series of gap width measurement values.

[0096] In this embodiment, the measurement value sequence not only provides the specific numerical value of the gap width, but also reflects the changing trend and distribution of the gap width of the vehicle-mounted refrigerator. By analyzing the measurement value sequence, the corresponding gap area characteristics can be identified, such as the gap width change of the gap area, and whether there is abnormal widening or narrowing, thereby providing a data basis for the gap width measurement value for subsequent data processing and defect identification.

[0097] Step S5002: performing a two-stage filtering process on the acquired gap detection data in real time, removing high-frequency noise from the gap monitoring data according to a preset data change threshold, and identifying and removing abnormal trajectory points with fluctuations exceeding a preset threshold based on the measured values ​​of adjacent trajectory points;

[0098] In this embodiment, after the gap detection data is collected, real-time filtering of the gap detection data is completed through a two-stage filtering operation that removes high-frequency noise and identifies and removes abnormal trajectory points. In the first stage, a filtering algorithm is used to remove high-frequency noise from the gap detection data. High-frequency noise is usually caused by electronic interference from the detection sensor itself or environmental factors, which will affect the accuracy of the collected gap detection data. Commonly used filtering algorithms include Kalman filtering and wavelet filtering. Kalman filtering estimates the state of the data in real time through prediction and update steps, and corrects it in combination with the measurement data, thereby effectively removing high-frequency noise. For example, during the detection process, if the gap width shows regular changes, the Kalman filter can predict the current gap width based on the previous data and update it in combination with the actual measurement value to obtain a more accurate estimate.

[0099] In the second stage, abnormal trajectory points need to be identified and removed. Abnormal trajectory points are points where the measured gap width value significantly deviates from the normal range. This may usually be caused by sensor noise, environmental electromagnetic interference, or surface stains on the car refrigerator. To identify these abnormal trajectory points, this embodiment sets a data change threshold. When the measured value of a trajectory point differs from that of adjacent trajectory points by more than this data change threshold, the trajectory point is identified as an abnormal trajectory point. This two-stage filtering process can significantly improve the accuracy and reliability of the collected gap detection data, ensuring more precise subsequent data analysis and defect identification.

[0100] Step S5003: When the continuous fluctuation amplitude of the gap detection data after filtering exceeds the tolerance threshold corresponding to the vehicle refrigerator model, calling the detection sensor to re-collect gap detection data corresponding to the gap area exceeding the tolerance threshold according to the local re-measurement trajectory;

[0101] After completing the filtering process of the gap detection data, it is checked whether the continuous fluctuation amplitude of the gap detection data exceeds the preset tolerance threshold. The tolerance threshold is preset according to the model and design requirements of the vehicle refrigerator, and is used to determine whether the gap width of the vehicle refrigerator is within an acceptable range. If the continuous fluctuation amplitude of the filtered gap detection data in a certain area exceeds the tolerance threshold, it means that the gap width of the gap area may have abnormal changes. At this time, by calling the corresponding detection sensor, according to the pre-set local re-measurement trajectory, the gap detection data corresponding to the gap area that exceeds the tolerance threshold is re-collected. The local re-measurement trajectory refers to the path that the detection sensor needs to move again and collect data after detecting the area of ​​abnormal fluctuation. This path usually covers the specific location of the abnormal fluctuation, and can also be set to include the adjacent area of ​​the location of the abnormal fluctuation area to ensure comprehensive detection of abnormal fluctuations.

[0102] To remeasure the local gap area, the sensor needs to be repositioned to the starting point of the abnormal fluctuation area and move along the local remeasurement trajectory. During this movement, the sensor records the gap width measurement values ​​again, thus forming a new measurement value sequence. The newly acquired measurement value sequence is then compared and analyzed with the previously acquired measurement value sequence to determine whether the abnormal fluctuation gap area has a gap width anomaly.

[0103] The above-mentioned local retest mechanism ensures the accuracy and reliability of the detection data, and promptly detects and corrects abnormal data caused by sensor noise, environmental interference, or other factors. This improves the robustness of the gap detection process and ensures that the final detection results can truly reflect the gap status of the vehicle refrigerator.

[0104] Step S5004: verify the measurement value sequence in the gap detection data, and generate the filtered gap data according to the gap detection data corresponding to the verified measurement value sequence.

[0105] In this embodiment, after filtering the measurement sequence and checking the tolerance threshold in the gap detection data, the measurement sequence is further verified. This verification process includes verifying the continuity and consistency of the measurement sequence, as well as detecting missing data or outliers. The verification process first checks whether there are any missing trajectory points in the measurement sequence. Missing trajectory points may be caused by sensor failure or data transmission errors. Therefore, it is necessary to verify the existence of the measurement value of each trajectory point and ensure the integrity of the measurement sequence. Secondly, the verification process checks the continuity of the measurement sequence, that is, it is necessary to verify whether the variation between the measurement values ​​of adjacent trajectory points is within a reasonable range. If the variation between adjacent measurement values ​​exceeds a preset reasonable threshold, it may indicate an anomaly, such as sensor noise or environmental interference, and the anomalous measurement value will be marked accordingly. Furthermore, during the verification process, each measurement value is also checked for consistency to verify whether the measurement value conforms to the expected variation trend. For example, if the detection result of the gap width of a car refrigerator within a certain gap area should be gradually increasing, but the measurement value shows an abnormal decreasing trend, the corresponding inconsistent detection result needs to be identified and handled accordingly.

[0106] Finally, after completing the verification of the gap detection data, the corresponding measurement value sequence is generated as filtered gap data. This ensures the accuracy and reliability of the filtered gap data and provides a reliable basis for evaluating the gap status of the vehicle refrigerator.

[0107] This embodiment collects gap detection data in real time and generates a measurement value sequence, which can record the changes in gap width in detail. By adopting a two-stage filtering process, high-frequency noise and outliers are effectively removed, ensuring the cleanliness and stability of the data. When data fluctuations exceeding the tolerance threshold are detected, local retesting is automatically triggered to further verify the accuracy of the data. Finally, high-quality filtered gap data is generated by verifying the measurement value sequence, ensuring the efficiency and accuracy of the gap detection process. The highly reliable filtered gap data obtained in this way can also provide reliable technical support for the quality control of vehicle refrigerators.

[0108] Based on any embodiment of the method of the present application, the filtered gap data is input into a defect recognition model pre-trained to a convergence state, so as to obtain a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model, including:

[0109] Step S6001: generating a gap feature vector based on the measured value sequence of the filtered gap data, and inputting the gap feature vector into the defect recognition model, wherein the gap feature vector includes the gap width of the vehicle refrigerator and the corresponding product coordinate information, and the defect recognition model includes multiple labeled gap error feature samples;

[0110] After completing the filtering processing of the gap detection data, a gap feature vector is generated based on the filtered gap data. The generation of the gap feature vector involves combining the filtered gap width measurement value with its position information in the product coordinate system. It is a data structure used to describe the characteristics of the gap area, which contains a sequence of gap width measurement values ​​and its corresponding product coordinate information, wherein the product coordinate information provides the specific position of the gap area on the vehicle refrigerator. For example, the gap feature vector can contain the gap width value of a specific trajectory point and its coordinate information in the product coordinate system.

[0111] The generated gap feature vector is then input into a defect recognition model that has been pre-trained to convergence. This defect recognition model, trained using a deep learning algorithm, contains a variety of labeled gap error feature samples generated from a large amount of gap error data, covering a wide range of possible gap error types and characteristics. By learning from these gap error feature samples, the defect recognition model can identify different types of gap defects. Therefore, by inputting the gap feature vector into the defect recognition model, the defect recognition model can output the type of gap defect and its confidence probability, providing a final, reliable basis for subsequent defect assessment and classification, ensuring the accuracy and reliability of the detection results.

[0112] Step S6002: extract the gap error features corresponding to the vehicle refrigerator according to the multi-level convolution kernel of the defect recognition model, generate the defect confidence of the gap area accordingly, and determine the defect recognition result according to the defect confidence.

[0113] In this embodiment, after the gap feature vector is input into the defect recognition model, the defect recognition model extracts features from the input data through its multi-level convolution kernel. Multi-level convolution kernels are tools used in deep learning models to automatically extract data features. Each layer of convolution kernels can capture features of different scales and complexities. When processing the gap feature vector, the convolution kernel can identify gap width changes, edge features, and other patterns related to gap errors. The convolution kernel performs convolution operations on the input data in a sliding window manner to calculate the local feature response at each position. These response values ​​are then passed to subsequent network layers for further feature abstraction and combination. Through multi-level convolution operations, the defect recognition model can extract high-level feature representations from the original gap width measurement values ​​to identify gap errors.

[0114] After gap error feature extraction is complete, the defect recognition model generates a defect confidence score for the gap region based on the extracted features. This confidence score reflects the model's confidence that a defect exists in the gap region, typically expressed as a percentage. For example, if the defect recognition model outputs a defect confidence score of 85% for a trajectory point, it means that the defect recognition model is 85% confident that a gap error exists at that trajectory point.

[0115] To generate defect confidence, defect recognition models typically use a softmax function as the activation function of the output layer, converting the network's output into a probability distribution. This converts the model's raw output into probabilities for each category, ensuring that the sum of all category probabilities is 1. Finally, the final defect recognition result is determined based on the generated defect confidence. If the confidence probability of a trajectory point exceeds a preset threshold, the trajectory point is deemed defective, and its defect type and location are recorded. This provides a basis for subsequent defect assessment and classification, ensuring the accuracy and reliability of the test results while obtaining the gap detection results for the vehicle refrigerator.

[0116] This embodiment generates a gap feature vector containing gap width and product coordinate information and inputs it into a pre-trained defect recognition model. It can automatically extract the features of the gap area and calculate the defect confidence. The application of multi-level convolution kernels ensures that the defect recognition model can effectively identify various gap error features to generate high-confidence defect recognition results. Therefore, through systematic synergy, the accuracy and execution efficiency of the gap detection process are effectively guaranteed.

[0117] Based on any embodiment of the method of the present application, after the step of obtaining a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model, the method includes:

[0118] Step S7001: Compare the defect confidence level with a preset gap defect threshold. When the defect confidence level is equal to or higher than a preset first probability threshold, determine that a gap error defect exists in the current gap area, and generate a defect determination result including product coordinate information of the corresponding defect area and a defect type, wherein the defect type includes the gap between the door and the main body of the vehicle refrigerator being higher or lower than a preset gap range, and the gap width at each location of the door of the vehicle refrigerator being higher than a preset width range.

[0119] After the defect recognition model is output, the defect confidence level generated by the model is compared with a preset gap defect threshold. This threshold is set based on the design requirements and quality control standards of the vehicle refrigerator and is used to determine whether a gap defect exists. If the defect confidence level of a trajectory point is equal to or higher than the first probability threshold, a gap error defect is determined to exist in that area.

[0120] After determining that a defect exists, a defect determination result including the product coordinate information of the corresponding defective area and the defect type is generated, so that the product coordinate information provides the specific location of the defective area on the vehicle refrigerator, and the defect type provides the specific nature of the gap defect. For example, the defect type may include the gap between the door and the main body being higher or lower than a preset gap range, or the gap width at each location of the door being higher than a preset width range, wherein the gap range and width range are determined according to the design specifications and quality standards of the vehicle refrigerator. For example, the standard gap range between the door and the main body is 2 mm to 3 mm. If the detected gap width exceeds this range, the vehicle refrigerator is correspondingly determined to be defective. Similarly, the gap width at each location of the door also has a corresponding standard range, and a gap width exceeding this range will also be determined to be defective. This can also be flexibly set by those skilled in the art, thereby automatically identifying and determining defects in the gap area and providing detailed defect determination results.

[0121] Step S7002: When the defect confidence level is between the first probability threshold and a preset second probability threshold, a defect review event is triggered, and a defect review report is generated based on the product coordinate information and the gap detection data corresponding to the vehicle refrigerator. The vehicle refrigerator is then transferred to an abnormality pending area for defect review based on the defect review report.

[0122] When the defect confidence level falls between a first probability threshold and a second probability threshold, a defect review event is automatically triggered. The second probability threshold is set lower than the first probability threshold. This identifies gap areas with critical defect confidence levels, requiring further re-inspection or manual confirmation of gap defects. A detailed defect review report is generated based on the vehicle refrigerator's product coordinate information and the original gap detection data. This defect review report can include the following information: First, the location of the area to be reviewed, annotated based on the product coordinate information. Accurately annotated product coordinates are used to ensure accurate marking of the area to be reviewed, using millimeter-level accuracy. Second, a sequence of measurement values ​​for the gap area in the original gap detection data, including the gap width value for each trajectory point, can be included. This includes the confidence probability value output by the defect recognition model and the corresponding defect type prediction. For example, if the gap width measurement values ​​for a door edge fluctuate between 2.1mm and 2.5mm, and the defect recognition model outputs a defect confidence level of 72%, meaning the defect confidence level falls within the 60%-80% review threshold range, a defect review report will be generated.

[0123] In this embodiment, after the defect review report is generated, the defect review report can be sent to the production management system through the production management system interface, and the automatic transfer instruction is triggered. The slide transfer mechanism corresponding to the gap detection station will transfer the car refrigerator to the abnormal pending area, and the car refrigerator transfer status information will be updated in real time during the transfer process. A dedicated station is set up in the abnormal pending area and equipped with professional detection equipment for quality inspectors to review and confirm. In this way, by setting up a reasonable confidence interval classification processing mechanism, while ensuring the detection efficiency, the risk of missed detection can also be effectively avoided. In actual applications, the review station in the abnormal pending area can be equipped with a high-precision laser measuring instrument and an industrial camera. After a secondary inspection of the gap area that needs to be re-inspected, its inspection results are compared and analyzed with the initial inspection data. Finally, the quality inspector confirms the defect status and updates the inspection record of the car refrigerator.

[0124] Step S7003: When the defect confidence is lower than the second probability threshold, a corresponding gap qualified result is generated.

[0125] When the defect confidence falls below the second probability threshold, a gap acceptance result is automatically generated. The second probability threshold is a pre-set quality criterion used to distinguish between acceptable and pending gap areas, ensuring that only gap areas with sufficiently low defect confidence levels are directly deemed acceptable. When generating a gap acceptance result, the inspection status of the gap area is also recorded based on the complete inspection data. For example, if the gap width measurement of a door edge area is stable between 2.3mm and 2.4mm, and the model outputs a defect confidence of 45%, which is below the pre-set 50% review threshold, the gap area is marked as acceptable. The corresponding gap acceptance result includes the vehicle refrigerator's product coordinate information, the measured gap width values ​​at each trajectory point, and the final acceptance signal, and is updated in real time to the production management system. In the production management system, qualified vehicle refrigerators are marked as "passed." The logistics system automatically plans subsequent process routes based on the defect identification results for the vehicle refrigerator. The inspection data for the vehicle refrigerator's gap area is also archived, including the complete series of measurements, calculated process parameters, and the final gap acceptance result, creating a traceable quality record.

[0126] The unique and significant technical advantage of this embodiment is that the gap area is divided into three states: defective, pending review and qualified through the first probability threshold and the second probability threshold. The gap area with a defect confidence level higher than the first threshold is directly judged as defective and a defect judgment result containing product coordinate information and defect type is generated. The gap area with a defect confidence level between the first probability threshold and the second probability threshold triggers the review process and generates a review report to transfer the abnormal area. The corresponding vehicle refrigerator with a gap area lower than the second probability threshold automatically generates a qualified result and updates the production management system. This hierarchical judgment mechanism not only ensures the immediate interception of major defects, but also reduces the risk of misjudgment through the review mechanism, while ensuring the rapid circulation of qualified products, significantly improving the detection efficiency and accuracy, and providing a scientific and reliable decision-making basis for the quality control of vehicle refrigerators.

[0127] See also Figure 2 According to one aspect of the present application, a vehicle refrigerator gap detection device is provided, including a gap determination module 8100, a gap detection module 8200, a data processing module 8300, and a defect recognition module 8400. The gap determination module 8100 is configured to establish product coordinate information based on surface marking points of the vehicle refrigerator, and determine a gap detection starting point corresponding to a gap area of ​​the vehicle refrigerator based on the product coordinate information; the gap detection module 8200 is configured to generate a product detection trajectory corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information, call a preset detection sensor to move to the gap detection starting point, and then perform gap detection on the vehicle refrigerator along the product detection trajectory; the data processing module 8300 is configured to collect gap detection data corresponding to the vehicle refrigerator gap detection process, perform real-time filtering on the gap detection data, and obtain filtered gap data; the defect recognition module 8400 is configured to input the filtered gap data into a defect recognition model pre-trained to a convergence state, so as to obtain a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model.

[0128] Based on any embodiment of the device of the present application, the gap determination module 8100 includes: an image acquisition module, configured to call a preset visual camera to obtain the outer surface image of the vehicle refrigerator and identify the surface marking points on the outer surface image; a model determination module, configured to determine the model information of the vehicle refrigerator based on the surface marking points, correspondingly construct a product coordinate system of the vehicle refrigerator of this model, and obtain the product coordinate information according to the product coordinate system; a starting point determination module, configured to determine the gap area of ​​the vehicle refrigerator according to the product coordinate information, and determine the gap detection starting point in the gap area based on the product coordinate system.

[0129] Based on any embodiment of the device of the present application, the gap determination module 8100 further includes: an image preprocessing module, configured to perform a preprocessing operation on the outer surface image, wherein the preprocessing operation includes one or more of grayscale processing, binarization processing and edge enhancement processing; a marker extraction module, configured to identify the preprocessed outer surface image according to a preset image recognition algorithm, extract the marker point coordinate information corresponding to one or more of the surface marker points to locate the surface marker points, and then determine the model information of the vehicle refrigerator according to the marker point coordinate information, wherein the marker point marking information includes the marker point quantity information, marker point feature information and coordinate distribution data of the surface marker points; a coordinate construction module, configured to call the preset product coordinate system parameters corresponding to the model information of the vehicle refrigerator based on the surface marker points to construct the product coordinate system of the vehicle refrigerator of this model, and determine the product coordinate information corresponding to the vehicle refrigerator of this model according to the coordinate origin and coordinate axis direction of the product coordinate system.

[0130] Based on any embodiment of the device of the present application, the gap detection module 8200 includes: a geometric determination module, configured to determine the geometric shape characteristics and gap type of the gap area based on the product coordinate information and the pre-stored three-dimensional geometric model of the vehicle refrigerator; a gap detection module, configured to generate a product detection trajectory based on the gap detection starting point and the gap area, and control the detection sensor to move along the product detection trajectory, wherein the product detection trajectory includes a plurality of equally spaced trajectory points, and the spacing between adjacent trajectory points is adjusted according to the gap width corresponding to the model of the vehicle refrigerator. When the gap type is detected to be a straight segment gap, the first preset spacing value is used to set the trajectory point, and when a curved segment gap is detected, the second preset spacing value smaller than the first preset spacing value is used to set the trajectory point.

[0131] Based on any embodiment of the device of the present application, the data processing module 8300 includes: a data acquisition module, configured to collect the gap detection data along the product detection trajectory according to the detection sensor, wherein the gap detection data includes a measurement value sequence representing the gap width of the gap area; a filtering processing module, configured to perform two-stage filtering processing on the acquired gap detection data in real time, remove high-frequency noise from the gap monitoring data according to a preset data change threshold, and identify and remove abnormal trajectory points exceeding a preset fluctuation change threshold based on the measurement values ​​of adjacent trajectory points; a tolerance judgment module, configured to call the detection sensor to re-collect gap detection data corresponding to the gap area exceeding the tolerance threshold according to a local re-measurement trajectory when the continuous fluctuation amplitude of the gap detection data after filtering exceeds the tolerance threshold corresponding to the model of the vehicle refrigerator; and a measurement verification module, configured to verify the measurement value sequence in the gap detection data and generate the filtered gap data based on the gap detection data corresponding to the verified measurement value sequence.

[0132] Based on any embodiment of the device of the present application, the defect recognition module 8400 includes: a vector generation module, configured to generate a gap feature vector based on the measurement value sequence of the filtered gap data, and input the gap feature vector into the defect recognition model, wherein the gap feature vector includes the gap width of the vehicle refrigerator and the corresponding product coordinate information, and the defect recognition model includes a variety of labeled gap error feature samples; a defect determination module, configured to extract the gap error features corresponding to the vehicle refrigerator according to the multi-level convolution kernel of the defect recognition model, generate the defect confidence of the gap area accordingly, and determine the defect recognition result according to the defect confidence.

[0133] On the basis of any embodiment of the device of the present application, it also includes: a first comparison module, which is configured to compare the defect confidence with a preset gap defect threshold. When the defect confidence is equal to or higher than the preset first probability threshold, it is determined that a gap error defect exists in the current gap area, and a defect determination result including product coordinate information and defect type of the corresponding defect area is generated, wherein the defect type includes the gap between the door and the main body of the vehicle refrigerator being higher or lower than a preset gap range, and the gap width at each location of the door of the vehicle refrigerator being higher than a preset width range; a second comparison module, which is configured to trigger a defect review event when the defect confidence is between the first probability threshold and the preset second probability threshold, and generate a defect review report according to the product coordinate information and the gap detection data corresponding to the vehicle refrigerator, so as to transfer the vehicle refrigerator to the abnormal waiting area for defect review according to the defect review report; a third comparison module, which is configured to generate a gap qualified result when the defect confidence is lower than the second probability threshold.

[0134] Another embodiment of the present application also provides a vehicle refrigerator gap detection device. Figure 3 Figure 2 shows a schematic diagram of the internal structure of a vehicle refrigerator gap detection device. The vehicle refrigerator gap detection device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable, non-volatile storage medium of the vehicle refrigerator gap detection device stores an operating system, a database, and computer-readable instructions. The database may store information sequences. When executed by the processor, the computer-readable instructions enable the processor to implement a vehicle refrigerator gap detection method.

[0135] The processor of the vehicle refrigerator gap detection device provides computing and control capabilities, supporting the operation of the entire vehicle refrigerator gap detection device. The memory of the vehicle refrigerator gap detection device may store computer-readable instructions that, when executed by the processor, cause the processor to execute the vehicle refrigerator gap detection method of the present application. The vehicle refrigerator gap detection device also has a network interface for connecting and communicating with a terminal.

[0136] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the vehicle-mounted refrigerator gap detection device to which the solution of the present application is applied. The specific vehicle-mounted refrigerator gap detection device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0137] In this embodiment, the processor is used to execute Figure 2 The memory stores the program code and various data required to execute the modules and submodules described above. The network interface facilitates data transmission between user terminals or servers. In this embodiment, the non-volatile, readable storage medium stores the program code and data required to execute all modules in the vehicle refrigerator gap detection device of this application. The server can call upon the server's program code and data to execute the functions of all modules.

[0138] The present application also provides a non-volatile readable storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the vehicle refrigerator gap detection method of any embodiment of the present application.

[0139] The present application also provides a computer program product, comprising a computer program / instruction, which implements the steps of the method described in any embodiment of the present application when executed by one or more processors.

[0140] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments of the method. The aforementioned storage medium can be a computer-readable storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0141] In summary, the present application realizes an efficient and accurate technical solution for detecting the gap between vehicle-mounted refrigerators through a complete intelligent detection process. It can automatically identify error defects between the door and the main body and in various gap areas, accurately locate the defect position by constructing a product coordinate system, and generate detailed judgment results including a gap width measurement value sequence, coordinate information and defect type, which significantly improves the accuracy of defect positioning. It also adopts a dynamic threshold grading judgment mechanism to intelligently divide the detection results into three categories: defects, pending review and qualified. It automatically triggers the review process for areas where the confidence level is at a critical value and generates a review report containing the original data, which effectively reduces the misjudgment rate while ensuring the rigor of quality control, and realizes the fully automated operation of the vehicle-mounted refrigerator gap detection process. From image acquisition, data processing to result judgment, the entire process can be done without human intervention, which greatly improves the detection efficiency. Moreover, the detection data is uploaded to the production management system in real time and automatically generates traceable quality records, which also provides complete data support for the quality improvement of vehicle-mounted refrigerators. The vehicle-mounted refrigerator gap detection technology solution of this application ensures the reliability of the detection results through a multi-level verification mechanism. The defect identification accuracy rate reaches more than 99.5%, effectively controlling the risk of defective product outflow. At the same time, the qualified product rapid circulation mechanism avoids the loss of production efficiency caused by excessive detection, and comprehensively improves the overall quality control level of the vehicle-mounted refrigerator production line.

Claims

1. A method for detecting gaps in a vehicle refrigerator, characterized in that: include: Establish product coordinate information according to the surface marking points of the vehicle refrigerator, and determine a gap detection starting point corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information; generating a product detection track corresponding to the gap area of ​​the vehicle refrigerator according to the product coordinate information, calling a preset detection sensor to move to the gap detection starting point, and performing gap detection on the vehicle refrigerator along the product detection track; Collecting gap detection data corresponding to the gap detection process of the vehicle refrigerator, performing real-time filtering processing on the gap detection data, and obtaining filtered gap data; The filtered gap data is input into a defect recognition model that has been pre-trained to a convergence state, so as to obtain a defect recognition result that characterizes the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model.

2. The vehicle refrigerator gap detection method according to claim 1, characterized in that: The step of establishing product coordinate information according to the surface marking points of the vehicle refrigerator, and determining a gap detection starting point corresponding to a gap area of ​​the vehicle refrigerator based on the product coordinate information includes: Calling a preset visual camera to obtain an outer surface image of the vehicle-mounted refrigerator, and identifying the surface marking points on the outer surface image; Determining the model information of the vehicle refrigerator based on the surface marking points, constructing a product coordinate system for the vehicle refrigerator of the model, and obtaining the product coordinate information according to the product coordinate system; The gap area of ​​the vehicle refrigerator is determined according to the product coordinate information, and the gap detection starting point in the gap area is determined based on the product coordinate system.

3. The vehicle refrigerator gap detection method according to claim 2, characterized in that: The identifying the surface marking points on the outer surface image, determining the model information of the vehicle refrigerator based on the surface marking points, constructing a product coordinate system for the vehicle refrigerator of the model, and obtaining the product coordinate information according to the product coordinate system includes: Performing a preprocessing operation on the outer surface image, wherein the preprocessing operation includes one or more of grayscale processing, binarization processing, and edge enhancement processing; Identifying the preprocessed outer surface image according to a preset image recognition algorithm, extracting marker point coordinate information corresponding to one or more surface marker points to locate the surface marker points, and determining the model information of the vehicle refrigerator according to the marker point coordinate information, wherein the marker point marking information includes marker point quantity information, marker point feature information, and coordinate distribution data of the surface marker points; Based on the surface marking points, the preset product coordinate system parameters corresponding to the model information of the vehicle refrigerator are called to construct the product coordinate system of the vehicle refrigerator of this model, and the product coordinate information corresponding to the vehicle refrigerator of this model is determined according to the coordinate origin and coordinate axis direction of the product coordinate system.

4. The vehicle refrigerator gap detection method according to claim 3, characterized in that: The method of generating a product detection track corresponding to the gap area of ​​the vehicle refrigerator according to the product coordinate information, calling a preset detection sensor to move to the gap detection starting point, and performing gap detection on the vehicle refrigerator along the product detection track includes: Determining the geometric shape characteristics and gap type of the gap area based on the product coordinate information and the pre-stored three-dimensional geometric model of the vehicle refrigerator; A product detection trajectory is generated based on the gap detection starting point and the gap area, and the detection sensor is controlled to move along the product detection trajectory, wherein the product detection trajectory includes a plurality of equally spaced trajectory points, and the spacing between adjacent trajectory points is adjusted according to the gap width corresponding to the model of the vehicle refrigerator. When the gap type is detected to be a straight segment gap, the first preset spacing value is used to set the trajectory point; when a curved segment gap is detected, the second preset spacing value smaller than the first preset spacing value is used to set the trajectory point.

5. The vehicle refrigerator gap detection method according to claim 3, characterized in that: The collecting of gap detection data corresponding to the gap detection process of the vehicle refrigerator, performing real-time filtering on the gap detection data, and obtaining filtered gap data includes: collecting the gap detection data along the product detection track according to the detection sensor, wherein the gap detection data includes a sequence of measurement values ​​representing a gap width of the gap area; Performing a two-stage filtering process on the acquired gap detection data in real time to remove high-frequency noise from the gap monitoring data according to a preset data change threshold, and identifying and removing abnormal trajectory points that exceed a preset fluctuation change threshold based on the measured values ​​of adjacent trajectory points; When the continuous fluctuation amplitude of the gap detection data after filtering exceeds the tolerance threshold corresponding to the vehicle refrigerator model, calling the detection sensor to re-collect the gap detection data corresponding to the gap area exceeding the tolerance threshold according to the local re-measurement trajectory; The measurement value sequence in the gap detection data is verified, and the filtered gap data is generated according to the gap detection data corresponding to the verified measurement value sequence.

6. The vehicle refrigerator gap detection method according to claim 5, characterized in that: Inputting the filtered gap data into a defect recognition model pre-trained to a converged state to obtain a defect recognition result representing the gap error of the vehicle refrigerator according to a defect confidence level output by the defect recognition model, includes: Generating a gap feature vector based on a measurement value sequence of the filtered gap data, and inputting the gap feature vector into the defect recognition model, wherein the gap feature vector includes the gap width of the vehicle refrigerator and corresponding product coordinate information, and the defect recognition model includes a plurality of labeled gap error feature samples; The gap error features corresponding to the vehicle refrigerator are extracted according to the multi-level convolution kernel of the defect recognition model, the defect confidence of the gap area is generated accordingly, and the defect recognition result is determined according to the defect confidence.

7. The method for detecting gaps in a vehicle refrigerator according to any one of claims 1 to 6, wherein: After the step of obtaining a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence level output by the defect recognition model, the method includes: Comparing the defect confidence with a preset gap defect threshold, when the defect confidence is equal to or higher than a preset first probability threshold, determining that a gap error defect exists in the current gap area, and generating a defect determination result including product coordinate information of the corresponding defect area and a defect type, wherein the defect type includes the gap between the door and the main body of the vehicle refrigerator being higher or lower than a preset gap range, and the gap width at each location of the door of the vehicle refrigerator being higher than a preset width range; When the defect confidence level is between the first probability threshold and a preset second probability threshold, a defect review event is triggered, and a defect review report is generated based on the product coordinate information and the gap detection data corresponding to the vehicle refrigerator, so that the vehicle refrigerator is moved to an abnormality pending area for defect review based on the defect review report; When the defect confidence is lower than the second probability threshold, a gap qualified result is generated accordingly.

8. A vehicle refrigerator gap detection device, characterized in that: include: a gap determination module configured to establish product coordinate information based on surface marking points of the vehicle refrigerator, and determine a gap detection starting point corresponding to a gap area of ​​the vehicle refrigerator based on the product coordinate information; a gap detection module configured to generate a product detection trajectory corresponding to the gap area of ​​the vehicle refrigerator based on the product coordinate information, call a preset detection sensor to move to the gap detection starting point, and then perform gap detection on the vehicle refrigerator along the product detection trajectory; A data processing module is configured to collect gap detection data corresponding to a gap detection process of the vehicle-mounted refrigerator, perform real-time filtering on the gap detection data, and obtain filtered gap data; The defect recognition module is configured to input the filtered gap data into a defect recognition model pre-trained to a convergence state, so as to obtain a defect recognition result representing the gap error of the vehicle refrigerator according to the defect confidence output by the defect recognition model.

9. A vehicle refrigerator gap detection device, characterized in that: The invention comprises a visual camera, a detection sensor, a manipulator, a data processing unit and a control unit, wherein the visual camera is used to obtain an image of the outer surface of a vehicle refrigerator and identify surface marking points; the detection sensor is used to perform gap detection on the vehicle refrigerator along a product detection trajectory and collect gap detection data; the manipulator is used to move the detection sensor to any surface area of ​​the vehicle refrigerator; the data processing unit is used to perform real-time filtering on the collected gap detection data to generate filtered gap data, and output defect confidence according to a defect recognition model deployed in the background; the control unit comprises a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to control the vehicle refrigerator gap detection device to perform the steps of the method according to any one of claims 1 to 7 to realize gap detection of the vehicle refrigerator.

10. A non-volatile readable storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images