Valve body inner cavity defect detection device based on machine vision

Abstract

The invention discloses a valve body inner cavity defect detection device based on machine vision, and relates to the technical field of valve body visual detection, the valve body inner cavity defect detection device comprises a box body, a three-axis sliding table module and a clamping platform, a sliding block of the three-axis sliding table module is connected with a rotating cylinder, the lower end of the three-axis sliding table module is fixedly connected with a detection rod, the detection rod is provided with a ventilation assembly, a positioning ring and a cleaning assembly, and a visual detection control system is further arranged. And the controller is electrically connected with each detection and execution component. The ventilation assembly is matched with the cleaning assembly and the air blowing and dust collecting box to clean impurities in an inner cavity, the cleaning assembly can protect a detection camera lens, and a proximity switch achieves anti-collision protection. The visual detection control system comprises a multifunctional module, realizes image acquisition and processing, defect judgment and motion control, and is also provided with a self-learning optimization module for optimizing an algorithm and executing parameters. According to the device, linkage control of cleaning, protection and detection is achieved, the problems that an existing device is not thorough in impurity cleaning, has no protection structure and is poor in detection adaptability are solved, and the detection stability and accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of valve body visual inspection equipment technology, and more particularly to a valve body internal cavity defect detection device based on machine vision. Background Technology

[0002] In the industrial manufacturing sector, the valve body, as a core component of fluid control systems, directly determines the safety and stability of equipment operation due to the quality of its internal cavity machining. Therefore, accurate detection of internal cavity defects has become a crucial aspect of valve body production quality control. With the development of intelligent manufacturing technology, machine vision, with its advantages of automation, high precision, and high efficiency, is gradually replacing manual visual inspection and becoming the mainstream technology for detecting defects in valve body cavities. The application of technologies such as three-axis motion positioning and image acquisition and analysis has further provided technical support for the automation upgrade of valve body inspection, meeting the needs of industrial production lines for batch and rapid quality inspection.

[0003] While existing machine vision inspection devices for valve body cavities have achieved basic automated inspection, they still have many technical shortcomings. The inspection process is easily interfered with by residual particles, dust, and other impurities in the cavity. These impurities not only obstruct the inspection lens, affecting image acquisition accuracy, but also easily scratch the lens and damage the equipment during cleaning. At the same time, the device lacks an anti-collision structure, and the inspection rod is prone to bumping into the inner wall of the valve body during movement. Furthermore, there is no coordinated process for blowing, cleaning, and dust extraction, resulting in incomplete removal of impurities. In addition, the detection algorithm has poor adaptability, leading to a high rate of missed and false detections, making it difficult to meet the requirements of high-precision quality inspection. Given these industry pain points, there is an urgent need to develop an integrated valve body cavity defect detection device that combines cleaning, protection, and intelligent detection. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a machine vision-based valve body cavity defect detection device.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a valve body cavity defect detection device based on machine vision, comprising a housing, a three-axis slide module fixedly installed on the upper end of the housing, a slider slidably installed on the three-axis slide module, and a clamping platform installed at the lower end of the housing, characterized in that: a rotary cylinder is fixedly installed on one side of the slider, the lower end of the rotary cylinder is coaxially fixedly connected to the detection rod, the detection rod rotates circumferentially with the rotary cylinder, a ventilation component is fixedly installed on one side of the detection rod, two positioning rings are fixedly sleeved on the lower side wall of the detection rod, the two positioning rings are spaced apart along the axial direction of the detection rod, and a cleaning component is assembled on the outer wall of the detection rod between the two positioning rings;

[0006] It is also equipped with a vision inspection and control system. The inspection rod is equipped with inspection components. The vision inspection and control system establishes electrical connections with the inspection components, the three-axis slide module, the rotary cylinder, the ventilation component and the cleaning component. The vision inspection and control system outputs control commands to each execution component to realize the linkage action control of each component. At the same time, it receives the inspection data collected by the inspection components to complete the machine vision recognition, analysis and judgment of defects in the valve body cavity.

[0007] Preferably, the ventilation assembly includes a ventilation pipe fixedly installed on one side of the detection rod, two branch pipes connected to the upper end of the ventilation pipe, and multiple through holes equidistantly opened on the outside of the positioning ring.

[0008] Preferably, the inside of the detection rod has a cavity for connecting the vent pipe and the through hole, the upper ends of the two branch pipes are fixedly installed on the upper part of the box, and the side ends of the two branch pipes are fixedly installed with solenoid valves.

[0009] Preferably, the cleaning assembly includes a brush that is vertically slidably fitted onto the outer wall of a detection rod between two positioning rings and an electric push rod fixedly installed on the upper end of the brush, with an installation ring fixedly installed on the side wall of the detection rod at the lower end of the brush.

[0010] Preferably, detection cameras are fixedly installed on both sides of the fixed end of the electric actuator and the lower end of the detection rod, and proximity switches are installed on both sides of the multiple detection cameras.

[0011] Preferably, a blower box and a dust collection box are fixedly installed on both sides of the box, and multiple fans are installed inside the blower box.

[0012] Preferably, the vision inspection and control system includes an image acquisition module, a data transmission module, an image processing module, a motion control module, and a defect determination module. The image acquisition module is electrically connected to the inspection camera, the data transmission module is communicatively connected to both the image acquisition module and the image processing module, the motion control module is electrically connected to the three-axis slide module, the rotary cylinder, the ventilation assembly, and the cleaning assembly, and the defect determination module is communicatively connected to both the image processing module and the motion control module.

[0013] Preferably, the image acquisition module is used to acquire image information of the valve body cavity captured by the detection camera, and transmits the image information to the data transmission module after preliminary encoding processing; the image processing module is used to receive the image information from the data transmission module, perform preprocessing operations such as noise reduction, enhancement and segmentation on the image, and extract the feature information of the valve body cavity in the image.

[0014] Preferably, the defect judgment module has a built-in standard feature database of the valve body cavity. It compares and analyzes the feature information extracted by the image processing module with the data in the standard feature database, outputs the defect judgment result, and transmits the judgment result to the motion control module. The motion control module issues control commands based on the defect judgment result to realize the start and stop of the movement and parameter adjustment of the three-axis slide module, rotary cylinder, ventilation component and cleaning component.

[0015] Preferably, the vision inspection and control system also includes a human-computer interaction module and a data storage module. The human-computer interaction module is communicatively connected to the defect judgment module and the motion control module to realize the input of detection parameters, real-time display of the detection process, and visual output of defect judgment results. The data storage module is communicatively connected to the data transmission module and the defect judgment module to store the original image information acquired by the image acquisition module, the preprocessed data of the image processing module, and the final judgment result of the defect judgment module.

[0016] Compared with the prior art, the beneficial effects of the present invention are:

[0017] 1. This solution incorporates a ventilation component and a cleaning component, working in conjunction with a blower box and a dust collection box to achieve coordinated operation of air blowing and backflushing, air extraction and dust removal, and physical cleaning within the valve body cavity. This solves the problem of incomplete impurity removal in existing devices, eliminating residual particles and dust from the internal cavity and preventing impurities from obstructing the detection lens. Simultaneously, the ventilation component precisely controls the gas flow through a solenoid valve, enabling orderly switching between blowing and extraction to ensure a clear field of view for the detection lens, providing effective visual data for image acquisition and resolving the problem of detection being interfered with by impurities.

[0018] 2. This solution incorporates a proximity switch on the detection rod. When the detection rod extends into the valve body cavity, the proximity switch senses the distance to the valve body wall in real time. After the data is transmitted to the vision inspection and control system, the system dynamically adjusts the movement speed of the detection rod to prevent it from colliding with the valve body wall. The electric push rod of the cleaning component can push the brush into the mounting ring, causing the bristles to fold upwards and cover the inspection camera lens, preventing recoil particles from scratching the lens. This solves the problem of existing devices lacking anti-collision and lens protection structures, thus protecting the inspection components.

[0019] 3. This solution establishes a multi-module vision inspection and control system to achieve coordinated control of each inspection stage. The image acquisition module acquires and preprocesses the inspection images, the image processing module extracts feature information, the defect judgment module completes defect identification and grade classification, and the motion control module adjusts the parameters of each actuator based on the judgment results. The system also includes a self-learning optimization module, which can dynamically optimize the relevant parameters of image processing and defect judgment, and extract the optimal motion parameters for different inspection scenarios. This enables adaptive adjustment of the actuators, improving the adaptability of defect detection and reducing the rates of missed and false detections.

[0020] In summary, this solution achieves efficient cleaning of the valve body cavity through structural design and system construction, completes collision protection and lens protection for the testing equipment, and builds an interconnected and intelligent visual inspection and control system. It solves the core problems of existing valve body cavity defect detection devices, such as incomplete impurity cleaning, lack of protective structure, and poor detection adaptability. It realizes protection of detection components, effective acquisition of detection data, accurate defect judgment, and adaptive adjustment of execution components, ensuring the stable operation of valve body cavity defect detection and improving the overall detection effect and adaptability. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0022] Figure 1 This is a schematic diagram of the overall three-dimensional structure proposed in this invention;

[0023] Figure 2 This is a schematic diagram of the three-dimensional structure of the detection rod proposed in this invention;

[0024] Figure 3 This is a three-dimensional structural diagram of the cleaning component proposed in this invention;

[0025] Figure 4 This is a schematic diagram of the three-dimensional structure of the detection camera proposed in this invention;

[0026] Figure 5 This is a connection diagram of the modules of the vision inspection and control system proposed in this invention.

[0027] The numbers in the diagram are: 1. Housing; 2. Three-axis slide module; 3. Vent pipe; 4. Detection rod; 5. Clamping platform; 6. Solenoid valve; 7. Rotary cylinder; 8. Positioning ring; 9. Brush; 10. Detection camera; 11. Electric actuator; 12. Mounting ring; 13. Proximity switch. Detailed Implementation

[0028] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0029] See Figures 1 to 5This invention discloses a machine vision-based valve body cavity defect detection device, comprising a housing 1, a three-axis slide module 2 fixedly mounted on the upper end of the housing 1, a slider slidably mounted on the three-axis slide module 2, and a clamping platform 5 installed inside the lower end of the housing 1. Multiple positioning cameras are installed on the inner wall of the housing 1 for identifying the position of the detection rod 4. The housing 1 also contains a pressure sensor and a dust concentration sensor, used to detect the blowing / squeezing air pressure value and the dust concentration inside the valve body cavity, respectively. All sensors are electrically connected to the vision detection control system. The mounting holes on the holding platform 5 allow the corresponding clamping device to be installed inside the housing 1. A rotary cylinder 7 is fixedly installed on one side of the slider, and a detection rod 4 is coaxially fixed to the lower end of the rotary cylinder 7. A ventilation component is installed on one side of the detection rod 4, and two positioning rings 8 are equidistantly fixed to the lower side wall of the detection rod 4. A cleaning component is installed between the two positioning rings 8. The housing 1 provides a closed detection environment. The three-axis slide module 2 drives the detection rod 4 to achieve three-dimensional movement. The clamping platform 5 fixes the valve body to ensure stable detection. The positioning camera accurately positions the detection rod 4, and the rotary cylinder 7 drives... The detection rod 4 rotates to achieve 360° all-around detection. The ventilation component completes the blowing and dust extraction of the inner cavity, the cleaning component removes impurities and protects the lens, and various sensors collect parameters of the detection environment and process to ensure a stable and accurate detection process. The ventilation component includes a ventilation pipe 3 fixedly installed on one side of the detection rod 4, two branch pipes connected to the upper end of the ventilation pipe 3, and multiple through holes equidistantly opened on the outside of the positioning ring 8. The ventilation pipe 3 is a telescopic flexible hose that can freely extend and retract without jamming as the detection rod 4 moves. The branch pipes divert gas to switch between blowing and extraction, and the through holes ensure that the gas is sprayed out evenly. In addition to inhalation, it enhances the cleaning effect of impurities in the internal cavity; the inside of the detection rod 4 has a cavity for connecting the air pipe 3 and the through hole. The upper ends of the two branch pipes are fixedly installed on the upper part of the box 1, and the side ends of the two branch pipes are fixedly installed with solenoid valves 6; after the detection rod 4 is inserted, the inflation device connected to the outside of one branch pipe is activated first, and the gas backflow in the through hole blows the particulate matter in the valve body out of the valve body. Then the cleaning component is activated, and at the same time the suction device connected to the outside of the other branch pipe is activated, and the dust is adsorbed through the through hole. The solenoid valve 6 precisely controls the gas flow, realizing the orderly switching of blowing and suction.

[0030] In this invention, the cleaning assembly includes a brush 9 vertically sliding on the outer wall of a detection rod 4 between two positioning rings 8, and an electric push rod 11 fixedly installed on the upper end of the brush 9. An installation ring 12 is fixedly installed on the side wall of the detection rod 4 at the lower end of the brush 9. The electric push rod 11 controls the vertical sliding of the brush 9. The upper end of the installation ring 12 has an installation groove corresponding to the brush 9. Pushing the brush 9 into the installation groove prevents the brush 9 from blocking the discharge path of particles. Simultaneously, the brush bristles fold upwards to cover and protect the detection camera 10, preventing backflowing particles from scratching the lens. The brush 9 can clean stubborn impurities on the inner wall of the valve body, improving the cleaning effect. Detection cameras 10 are fixedly installed on both sides of the fixed end of the electric push rod 11 and the lower end of the detection rod 4. Proximity switches 13 are installed on both sides of the multiple detection cameras 10. The detection cameras 10 capture images of the valve body cavity to complete defect detection. The proximity switches 13 sense the distance to the inner wall of the valve body, preventing the detection rod 4 from interfering with the valve body. The valve body's inner wall was damaged, protecting the equipment. A blower box and a dust collection box are fixedly installed on both sides of housing 1. Multiple fans are installed inside the blower box; these fans are variable frequency fans, their speed adjusted by the visual inspection and control system based on feedback data from the dust concentration sensor. The dust collection box is equipped with a filter-type dust collection structure and a negative pressure fan to achieve efficient collection of impurities. The airflow generated by the fans cleans the particles discharged via backflushing, and the dust collection box collects impurities, preventing their diffusion within housing 1 and maintaining a clean detection environment. The rotary cylinder 7 used in this device is an MSQB series rotary cylinder, the electric actuators 11 are all LA-T8 series electric actuators, the camera is a Hikvision CU series camera, the proximity switch 13 is an FH-IF02FE series proximity switch sensor, the air pressure sensor is a diffused silicon pressure sensor, and the dust concentration sensor is a laser scattering dust sensor.

[0031] This invention also includes a vision inspection and control system. The detection rod 4 is equipped with a detection component. The vision inspection and control system establishes electrical connections with the detection component, the three-axis slide module 2, the rotary cylinder 7, the ventilation component, and the cleaning component. The vision inspection and control system outputs control commands to each execution component to realize the linkage action control of each component. At the same time, it receives the detection data and environmental parameters collected by each detection component. Through a hierarchical algorithm, it completes the machine vision recognition, analysis, and judgment of defects in the valve body cavity. Furthermore, it realizes the dynamic parameter adjustment of the execution component based on the detection data and environmental parameters, forming a closed-loop system of detection-control-feedback.

[0032] More specifically, the visual inspection and control system includes an image acquisition module, a data transmission module, an image processing module, a motion control module, a defect determination module, a human-machine interaction module, a data storage module, and a self-learning optimization module. Each module is a functional module combining hardware and software. It is built on an industrial control motherboard and an embedded operating system, and the modules communicate at high speed through an internal bus. The image acquisition module is electrically connected to the inspection camera 10. The data transmission module is connected to the image acquisition module, the image processing module, and each inspection component. The motion control module is connected to the three-axis slide module 2, the rotary cylinder 7, the solenoid valve 6, the electric actuator 11, and various fans. The defect determination module is connected to the image processing module and the motion control module. The human-machine interaction module is connected to the defect determination module, the motion control module, and the data storage module. The data storage module establishes communication connections with each module. The self-learning optimization module establishes bidirectional communication connections with the image processing module, the defect determination module, and the motion control module.

[0033] The image acquisition module acquires image data from the detection camera 10 at a preset frame rate. The preset frame rate can be manually set through the human-computer interaction module, ranging from 10 to 30 fps. The module performs format conversion and frame buffering on the image data. The format conversion follows the color space conversion rules of YUV420 to RGB888, converting the original YUV format image acquired by the camera into the universal RGB format for easier subsequent processing.

[0034] The frame buffer employs a first-in-first-out (FIFO) circular buffer mechanism with a buffer depth of 5-10 times the preset detection frame rate to avoid image data loss due to image processing speed lag. Simultaneously, the module adds timestamps and camera numbers to the image data from each of the 10 detection cameras, enabling synchronization and traceability of multi-camera images. The data transmission module uses a custom communication protocol based on TCP / IP, divided into internal data transmission and external data interaction. Internal transmission achieves high-speed data exchange between modules via the industrial control motherboard's internal bus, while external interaction enables signal transmission with various detection and execution components. The module adds frame headers, frame trailers, and checksums to all transmitted data. The frame header contains data type, transmission address, and data length information; the frame trailer is a data end marker; and the checksum is calculated using the CRC32 algorithm. Where poly is the preset generator polynomial and init is the initial value, the module verifies the integrity of the data through a checksum. If the verification fails, a retransmission command is issued to ensure the integrity and accuracy of the data transmission.

[0035] More specifically, the image processing module performs image preprocessing and feature extraction operations. Image preprocessing employs a weighted fusion multi-operator noise reduction and enhancement algorithm, first using Gaussian filtering and median filtering to reduce noise in the image. The Gaussian filtering formula is as follows: Median filtering uses a 3×3 filtering window to sort image pixels and take the median value. Then, the two noise reduction results are weighted and fused. The fusion formula is as follows: , where α+β=1, and α and β are weighting coefficients that are dynamically adjusted according to the type of image noise;

[0036] Image enhancement is achieved through adaptive histogram equalization, which divides the image into multiple non-overlapping sub-blocks. A histogram equalization transformation function is calculated for each sub-block separately, avoiding the loss of image details caused by overall equalization and improving the contrast between the cavity defect area and the background. Feature extraction employs a fusion feature extraction algorithm based on histogram of gradient orientation (HOG) and local binary pattern (LBP). First, HOG features are extracted from the image. The image is then divided into cells and blocks, and the gradient orientation histogram of each cell is calculated. L2-Hys normalization is applied to the features within each block. Finally, LBP features are extracted from the image. Calculate the LBP value of the pixel, where s() is the sign function. The grayscale value of the neighboring pixels. Let P be the gray value of the center pixel, P be the number of neighboring pixels, and R be the radius of the neighborhood. Finally, the HOG feature and LBP feature are concatenated dimensionally to form a fused feature vector, which is then normalized. The normalization adopts Min-Max standardization, which maps the feature values ​​to the [0,1] interval to improve the accuracy of feature matching.

[0037] More specifically, the defect determination module incorporates a standard feature database for the valve body cavity and an improved Euclidean distance matching algorithm. The standard feature database stores defect-free fusion feature vectors for valve bodies of different specifications, as well as fusion feature vectors for various typical defects such as cracks, dents, burrs, and pinholes. The database supports offline updates via a human-computer interaction module and online self-learning through a self-learning optimization module. The improved Euclidean distance matching algorithm performs weighted optimization on the traditional Euclidean distance formula, resulting in the following optimized formula: ,in The weight coefficient of the i-th feature is calculated by the variance contribution rate of the feature using the entropy weight method. The larger the variance contribution rate, the higher the weight coefficient, making the matching calculation more consistent with the feature distribution of defects in the valve body cavity.

[0038] The defect determination module matches the fused feature vector extracted by the image processing module with the defect-free feature vector in the standard feature database. When the matching distance is less than a preset threshold, it is determined to be defect-free. When the matching distance is greater than or equal to the preset threshold, it is matched a second time with the typical defect feature vector in the database to determine the defect type and defect level. The defect level is divided into three levels: light, medium and heavy, based on the matching similarity. At the same time, the module outputs the defect determination result, feature matching similarity and defect position coordinates. The defect position coordinates are jointly calibrated by the number of the detection camera 10, the rotation angle of the rotary cylinder 7 and the displacement data of the three-axis slide module 2.

[0039] The motion control module receives the judgment results from the defect judgment module and the data collected by each detection component. It issues control commands according to the preset control strategy. The control commands are output in pulse width modulation (PWM) mode, and the pulse frequency and duty cycle can be precisely adjusted. The module realizes dynamic adjustment of multiple parameters such as the moving speed of the three-axis slide module 2, the rotational angular velocity of the rotary cylinder 7, the flow orifice diameter of the solenoid valve 6, the moving speed and displacement of the electric push rod 11, and the speed of various fans.

[0040] When a defect is detected, the motion control module drives the three-axis slide module 2 and the rotary cylinder 7 to accurately locate the defect. At the same time, it controls the ventilation and cleaning components to stop moving to avoid impurities interfering with the secondary detection of the defect location. If the defect level is medium or severe, the module will trigger an audible and visual alarm to remind the staff to verify. When the dust concentration sensor detects that the dust concentration in the chamber 1 exceeds the preset value, the module automatically increases the speed of the blower and the negative pressure fan of the dust collection box to improve the efficiency of impurity cleaning. When the air pressure sensor detects abnormal blowing / exhausting air pressure, the module promptly adjusts the opening of the solenoid valve 6 and issues a fault prompt.

[0041] More specifically, the vision inspection and control system also includes a human-computer interaction module, a data storage module, and a self-learning optimization module. The human-computer interaction module is communicatively connected to the defect judgment module and the motion control module to realize the input of detection parameters, real-time display of the detection process, and visual output of defect judgment results. The real-time display includes the real-time image of the detection camera 10, the action parameters of the execution component, and the statistical data of defect detection. The visual output adopts a combination of text and graphics, including defect markings, defect location coordinates, and defect type and level.

[0042] The data storage module communicates with the data transmission module, defect judgment module, and self-learning optimization module. It adopts a distributed storage architecture to store the raw image information acquired by the image acquisition module, the preprocessed data of the image processing module, the final judgment result of the defect judgment module, and the optimization parameters of the self-learning optimization module. The stored image data is classified according to the naming rule of "equipment number-inspection time-valve body specification". The stored judgment results and optimization parameters support structured query and export.

[0043] The self-learning optimization module is a variant optimization structure of the vision inspection control system. It establishes bidirectional communication connections with both the image processing and defect detection modules. Through incremental learning, it dynamically optimizes the weight coefficients of image processing, the matching threshold for defect detection, and the feature weight coefficients. The self-learning optimization module incorporates a convolutional neural network model trained on a small sample basis, using manually reviewed defect detection data as training samples. It continuously adjusts the model parameters through forward and backward propagation. Backpropagation employs an improved algorithm of stochastic gradient descent, the Adam optimizer, with the following formula: , , ;in , For first-order and second-order moment estimation, , The attenuation coefficient is... For gradient, For learning rate, The minimum value is obtained; the optimized parameters obtained through model training are updated in real time to the image processing module and the defect judgment module, dynamically adjusting the Gaussian filter σ value, weighted fusion coefficients α and β, defect matching threshold, and feature weight coefficients. This enables the detection algorithm to self-optimize, improving the adaptability and accuracy of defect detection for valves of different working conditions and specifications, and reducing the rate of missed detections and false detections.

[0044] Meanwhile, the self-learning optimization module performs cluster analysis on the motion parameters of the actuators in the motion control module. Using the K-means clustering algorithm, it extracts the optimal combination of motion parameters for different detection scenarios (different valve body specifications, different impurity contents, different defect types) and stores it in the motion control module's parameter library. When detecting a new valve body, the module automatically calls upon the optimal combination of motion parameters based on the valve body specifications and previously detected impurity content data, achieving adaptive adjustment of the actuator's movements. This eliminates the need for manual, repeated adjustments by operators, improving detection efficiency.

[0045] The working principle of this invention is as follows: When using this invention, first turn on the power supply, and connect the inflation device and the vacuum device to the two branch pipe connection ports at the upper end of the housing 1 respectively. The inflation device is a high-pressure air pump, and the vacuum device is a negative pressure vacuum pump. The initial setting of the detection parameters is completed through the human-machine interaction module. At the same time, the vision detection control system automatically performs self-checks on each detection component and execution component, checks and ensures that each component is in good condition. If a fault is detected, an alarm prompt will be issued in time.

[0046] The valve body to be tested is then fixed on the clamping platform 5 by the clamping device installed through the preset mounting hole on the clamping platform 5. The calibration command is triggered by the human-machine interaction module. The vision inspection and control system controls the positioning camera on the inner wall of the housing 1, the detection camera 10 on the detection rod 4 and the proximity switch 13 to complete the joint calibration, determine the initial position of the detection rod 4 and the reference coordinates of the valve body. After the calibration is completed, the three-axis slide module 2 is started. The three-axis slide module 2 drives the slider that is slidably installed, the rotary cylinder 7 that is fixedly installed on one side of the slider and the detection rod 4 that is coaxially fixed to the lower end of the rotary cylinder 7 to move to directly above the valve body to be tested.

[0047] Next, the three-axis slide module 2 is controlled to slowly extend the detection rod 4 into the valve body cavity. During this process, the proximity switches 13 installed on both sides of the detection camera 10 sense the distance between the detection rod 4 and the inner wall of the valve body in real time and transmit the data to the vision detection control system. The system dynamically adjusts the moving speed of the detection rod 4 according to the distance data to prevent the detection rod 4 from colliding with the inner wall of the valve body and to protect the equipment. After the detection rod 4 is extended to the designated position, the electric push rod 11 fixedly installed on the upper end of the brush 9 is activated. The electric push rod 11 pushes the brush 9 into the corresponding mounting groove opened on the upper end of the mounting ring 12 fixedly installed on the side wall of the detection rod 4. At this time, the bristles of the brush 9 are folded up to cover and protect the lens of the detection camera 10, so as to avoid scratching the lens by the subsequent recoil particles.

[0048] Subsequently, the visual inspection and control system opens the solenoid valve 6 on the inflation side, starts the external inflation device, and the gas is evenly sprayed out from multiple through holes on the outside of the positioning ring 8 through the air pipe 3 on one side of the detection rod 4 and the cavity inside the detection rod 4. The gas backflow blows the particulate matter out of the valve body. The air pressure sensor detects the blowing air pressure in real time, and the system dynamically adjusts the opening of the solenoid valve 6 according to the air pressure data to ensure the blowing effect. After the particulate matter is blown out, the visual inspection and control system starts the fans of the blower box and the dust collection box according to the feedback data of the dust concentration sensor to clean and collect the blown particulate matter. At the same time, the electric push rod 11 is started in reverse to remove the brush 9 from the mounting groove of the mounting ring 12, and the solenoid valve 6 on the extraction side is opened to start the external extraction device.

[0049] The vision inspection and control system activates the rotary cylinder 7 and the three-axis slide module 2, causing the detection rod 4 to slide up and down and rotate simultaneously within the valve body cavity. The brush 9 moves synchronously with the detection rod 4 to clean stubborn impurities from the inner wall of the valve body. At the same time, the through holes simultaneously adsorb dust from the valve body. The system dynamically adjusts the moving speed and rotational angular velocity of the detection rod 4 according to the valve body specifications to ensure cleaning effectiveness. During this process, the blower box and dust collection box continue to work to prevent impurities from spreading within the box 1 and to maintain a clean inspection environment.

[0050] After the impurities are cleaned, the dust concentration sensor detects that the dust concentration inside the chamber is lower than the preset threshold. The vision inspection and control system restarts the rotary cylinder 7 and the three-axis slide module 2, which drive the detection rod 4 to slide up and down along the preset trajectory in the valve body cavity and rotate at the same time. Multiple detection cameras 10 on both sides of the fixed end of the electric push rod 11 and the lower end of the detection rod 4 synchronously acquire images of the valve body cavity at the preset frame rate, completing the initial acquisition of image data.

[0051] The acquired image data is converted into a format and buffered by the image acquisition module, and then transmitted to the image processing module by the data transmission module. The module performs noise reduction, enhancement, and feature extraction on the image to generate a fused feature vector, which is then transmitted to the defect judgment module. The defect judgment module matches the fused feature vector with data in the standard feature database to identify, determine the type, and classify the level of defects in the valve body cavity. The judgment result and the defect location coordinates are then transmitted to the motion control module and the human-machine interaction module. The human-machine interaction module displays the judgment result in real time, marks the defect area, and triggers an audible and visual alarm if a medium or severe defect is detected.

[0052] After the inspection is completed, the vision inspection and control system controls the three-axis slide module 2 to move the inspection rod 4 out of the valve body cavity, and at the same time controls the ventilation component, cleaning component and various fans to stop moving. The staff removes the inspected valve body from the clamping platform 5 and sorts it according to the judgment results. All inspection data is stored locally and backed up in the cloud by the data storage module to achieve data traceability.

[0053] During long-term testing, the self-learning optimization module uses manually reviewed defect detection data as training samples to continuously optimize the detection algorithm. At the same time, it performs cluster analysis on the action parameters of the execution components to achieve adaptive adjustment of the action parameters, thereby continuously improving the accuracy and efficiency of the detection. After the entire testing operation is completed, the staff issues a stop command through the human-machine interaction module, shuts down each electrical device in sequence, cleans the residual impurities on the surface of the housing 1, dust collection box and detection rod 4, checks the working status of the brush 9, detection camera 10, proximity switch 13 and various sensors, and promptly maintains or replaces damaged or worn parts. Finally, the external power supply is cut off, completing the entire usage process of the device.

[0054] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A machine vision-based valve body cavity defect detection device, comprising a housing (1), a three-axis slide module (2) fixedly mounted on the upper end of the housing (1), a slider slidably mounted on the three-axis slide module (2), and a clamping platform (5) mounted on the lower end of the housing (1), characterized in that: A rotary cylinder (7) is fixedly installed on one side of the slider. The lower end of the rotary cylinder (7) is coaxially fixed to the detection rod (4). The detection rod (4) rotates circumferentially with the rotary cylinder (7). A ventilation assembly is fixedly installed on one side of the detection rod (4). Two positioning rings (8) are fixedly sleeved on the lower side wall of the detection rod (4). The two positioning rings (8) are spaced apart along the axial direction of the detection rod (4). A cleaning assembly is installed on the outer wall of the detection rod (4) between the two positioning rings (8). It is also equipped with a vision inspection and control system. The inspection rod (4) is equipped with inspection components. The vision inspection and control system establishes electrical connections with the inspection components, the three-axis slide module (2), the rotary cylinder (7), the ventilation component and the cleaning component respectively. The vision inspection and control system outputs control commands to each execution component to realize the linkage action control of each component. At the same time, it receives the inspection data collected by the inspection component and completes the machine vision recognition, analysis and judgment of defects in the valve body cavity.

2. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The ventilation assembly includes a ventilation pipe (3) fixedly installed on one side of the detection rod (4), two branch pipes connected to the upper end of the ventilation pipe (3), and multiple through holes equidistantly opened on the outside of the positioning ring (8).

3. The machine vision-based valve body cavity defect detection device according to claim 2, characterized in that: The inside of the detection rod (4) is provided with a cavity for connecting the vent pipe (3) and the through hole. The upper ends of the two branch pipes are fixedly installed on the upper part of the box (1), and the side ends of the two branch pipes are fixedly installed with solenoid valves (6).

4. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The cleaning assembly includes a brush (9) that slides vertically between two positioning rings (8) on the outer wall of a detection rod (4) and an electric push rod (11) fixedly installed on the upper end of the brush (9). An installation ring (12) is fixedly installed on the side wall of the detection rod (4) at the lower end of the brush (9).

5. The machine vision-based valve body cavity defect detection device according to claim 4, characterized in that: Both sides of the fixed end of the electric push rod (11) and the lower end of the detection rod (4) are fixedly installed with detection cameras (10), and both sides of the multiple detection cameras (10) are equipped with proximity switches (13).

6. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The two sides of the box (1) are respectively fixedly installed with a blower box and a dust collection box, and multiple fans are installed inside the blower box.

7. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The visual inspection and control system includes an image acquisition module, a data transmission module, an image processing module, a motion control module, and a defect determination module. The image acquisition module is electrically connected to the inspection camera (10). The data transmission module is communicatively connected to the image acquisition module and the image processing module. The motion control module is electrically connected to the three-axis slide module (2), the rotary cylinder (7), the ventilation component, and the cleaning component. The defect determination module is communicatively connected to the image processing module and the motion control module.

8. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The image acquisition module is used to acquire the image information of the valve body cavity captured by the detection camera (10), and transmit the image information to the data transmission module after preliminary encoding processing; the image processing module is used to receive the image information from the data transmission module, perform preprocessing operations such as noise reduction, enhancement and segmentation on the image, and extract the feature information of the valve body cavity in the image.

9. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The defect judgment module has a built-in standard feature database of the valve body cavity. It compares and analyzes the feature information extracted by the image processing module with the data in the standard feature database, outputs the defect judgment result, and transmits the judgment result to the motion control module. The motion control module issues control commands according to the defect judgment result to realize the start and stop of the three-axis slide module (2), the rotary cylinder (7), the ventilation component and the cleaning component and the parameter adjustment.

10. The machine vision-based valve body cavity defect detection device according to claim 1, characterized in that: The vision inspection and control system also includes a human-computer interaction module and a data storage module. The human-computer interaction module is connected to the defect judgment module and the motion control module to realize the input of detection parameters, real-time display of the detection process, and visual output of defect judgment results. The data storage module is connected to the data transmission module and the defect judgment module to store the raw image information acquired by the image acquisition module, the preprocessed data of the image processing module, and the final judgment result of the defect judgment module.

Images