AI visual inspection equipment and system for shell structure defect detection
By designing AI vision detection equipment and systems, using drive devices such as hydraulic lifting rods, servo motors and rotary motors to clean, feed and all-round visual inspection of the shell, solving the problems of low shell detection efficiency and low accuracy, and realizing automated detection and classification.
Patent Information
- Application Number
- CN202411774523.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-03-04
AI Technical Summary
When visual inspection of the shell, it is impossible to detect all its appearance at once, resulting in subsequent position change of the shell, which increases trouble, cost and inefficiency.
Design an AI vision detection device and system, including a preprocessing box, a vision detection component and a deep learning detection unit, and realizes cleaning, feeding and all-round visual inspection of the shell through driving devices such as hydraulic lifting rods, servo motors and rotating motors.
It improves the accuracy and efficiency of shell detection, reduces the need for subsequent position transformation, reduces the cost, and realizes the automated detection and classification of shell structural defects through deep learning models.
Smart Images
Figure CN119555691A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of AI visual inspection, and in particular to an AI visual inspection device and system for shell structure defect detection. Background Art
[0002] Visual inspection is to use machines instead of human eyes to make measurements and judgments. Visual inspection refers to the process of converting captured targets into image signals through machine vision products, transmitting them to a dedicated image processing system, and converting them into digital signals based on pixel distribution, brightness, color and other information. The image system performs various operations on these signals to extract the characteristics of the target, and then controls the equipment action on site based on the results of the judgment. It is a valuable mechanism for production, assembly or packaging. It has immeasurable value in detecting defects and preventing defective products from being delivered to consumers. When visually inspecting the shell, its entire appearance cannot be inspected at one time, resulting in the need to subsequently change the position of the shell to complete the inspection. This method is more cumbersome and will increase costs, and efficiency cannot be guaranteed.
[0003] Therefore, it is necessary to propose an AI visual inspection device and system for shell structure defect detection to solve the above problems. Summary of the invention
[0004] The main purpose of the present invention is to provide an AI visual inspection device and system for shell structure defect detection, which can effectively solve the problems in the background technology.
[0005] To achieve the above object, the technical solution adopted by the present invention is: An AI visual inspection device for shell structure defect detection, comprising a pretreatment box, a cleaning component movably connected to the top of the inner cavity of the pretreatment box, a brush roller rotatably connected to the cleaning component and the bottom of the inner cavity of the pretreatment box, a driving component installed on the outer wall of the pretreatment box, a collecting component installed on the top of the driving component, a collecting box movably connected to the inner cavity of the collecting component, and a fan blade rotatably connected to one end of the collecting component away from the pretreatment box; A material guide plate is installed on the right side of the pretreatment box, a detection component is installed on the right side of the material guide plate, a frame is installed on the top of the detection component, and visual detection components are installed on the top of the frame cavity, both ends of the frame cavity, and the bottom of the material guide plate; The two ends of the top of the inner cavity of the detection component are symmetrically and movably connected with the alignment components, and the right side of the alignment component is symmetrically installed with a driving rod.
[0006] Preferably, a hydraulic lifting rod is installed in the center of the top of the pretreatment box, and the bottom of the hydraulic lifting rod is installed in the center of the top of the cleaning component. A belt is installed in the inner cavity of the driving component, and a pulley is installed in the inner cavity of the belt. The pulley is connected to the brush roller at the bottom, and a servo motor is installed on the outer wall of the driving component. The servo motor is installed on the outer wall of the pulley on the right side through a first rotating shaft.
[0007] Preferably, the outer wall of the driving assembly is rotatably connected with a gear, the gear is installed on the outer wall of the middle pulley, the top of the gear is meshed with a gear ring, and the brush roller is installed on the outer wall of the fan blade.
[0008] Preferably, the collecting assembly is a cavity structure with two ends communicating with the outside, the collecting assembly and the collecting box are both communicated with the inner cavity of the pretreatment box, and a filter screen is installed on the outer wall of the collecting box.
[0009] Preferably, both ends of the top of the guide plate are symmetrically rotatably connected with unloading wheels, a transparent plate is installed in the middle of the guide plate, and one of the visual detection components is installed at the bottom of the transparent plate.
[0010] Preferably, a visual detection camera is installed at one end of the visual detection component, and movable grooves are symmetrically opened at the top and bottom of the outer wall of the visual detection component. A first bidirectional screw is rotatably connected in the inner cavity of the movable groove at the top, and a guide rod is installed in the inner cavity of the movable groove at the bottom. A rotating motor is installed on the side of the visual detection component, and the rotating motor is installed on the outer wall of the first bidirectional screw through a second rotating shaft. The visual detection component also includes an infrared thermal imager and a depth sensor.
[0011] Preferably, the top and bottom of the adjustment component near one end of the visual detection component are movably connected in the inner cavity of the movable groove, the top of the adjustment component is threadedly connected to the outer wall of the first bidirectional screw, the bottom of the adjustment component is sleeved on the outer wall of the guide rod, and a rotating cylinder is installed on the top of the adjustment component, and an adjusting light source is installed on the bottom of the rotating cylinder through a third rotating shaft, and the adjusting light source is rotatably connected in the inner cavity of the adjustment component.
[0012] Preferably, a conveyor belt is installed in the inner cavity of the detection component, an adjustment groove is opened at the bottom of the right side of the detection component, a second bidirectional screw rod is connected in the inner cavity of the adjustment groove, a drive motor is installed on the outer wall of the detection component, and the drive motor is installed on the outer wall of the cleaning component through a fourth rotating shaft.
[0013] Preferably, the bottom of the driving rod is movably connected in the inner cavity of the adjusting groove and is threadedly connected to the outer wall of the second bidirectional screw, and rollers are rotatably connected in the inner cavities on opposite sides of the squaring assemblies at both ends.
[0014] An AI visual inspection system for shell structure defect detection, applied to the above-mentioned AI visual inspection device for shell structure defect detection, includes an image acquisition unit, a data preprocessing unit, a deep learning detection unit and a result display and alarm module, wherein the image acquisition unit includes a detail information capture module, an internal structure capture module and a three-dimensional shape capture module; The detail information capture module captures the texture and color information of the shell surface based on the visual detection camera. The internal structure capture module detects the temperature difference caused by the internal structure defects and uneven surface material of the shell based on the infrared thermal imager to assist in discovering potential defects. The three-dimensional shape capture module obtains the three-dimensional shape information of the shell based on the depth sensor to identify the defect type and provide spatial coordinates for the location of the defect. The data preprocessing unit includes an image registration and fusion module, an image enhancement and standardization module; The image registration and fusion module is used to register the collected multi-modal images, align the shell structures in different modal images in space, and then perform a fusion operation; The image enhancement and standardization module is used to enhance the fused image, including contrast adjustment and noise removal, to improve the quality and clarity of the image, and to standardize the image to meet the input requirements of the deep learning model; The deep learning detection unit includes a defect detection neural network model and a model training and optimization module; The defect detection neural network model is used to construct a convolutional neural network model based on deep learning. The convolutional neural network model based on deep learning includes three branch structures, which are used to process different information features in the fused image respectively. The first branch is used to process the texture and color features in the visible light image, and extract the characteristic patterns of scratches and holes through multi-layer convolution and pooling operations; the second branch is used to analyze the temperature features in the infrared image, identify the temperature abnormality area caused by the defect, and learn its association with different defect types; the third branch uses the shape information of the depth image to detect the deformation of the shell; The convolutional neural network model based on deep learning finally merges the features extracted by the three branches and classifies them through the fully connected layer to determine whether there are defects and the specific type of defects, and outputs the location information of the defects in the image; The model training and optimization module collects a large number of multimodal image samples containing various shell structural defects, trains the constructed deep learning model, uses the cross entropy loss function as the objective function of the training, and adjusts the parameters of the model based on the optimization algorithm; During the training process, data augmentation technology is used to expand the training data set, and the early stopping method is used to monitor the performance of the model on the validation set. When the performance no longer improves, training is stopped to obtain the optimal model parameters; The result display and alarm module includes a visualization interface and an alarm and communication module, wherein the visualization interface includes a visualization user interface, which intuitively displays the detection results to the user, displays the original image of the shell structure, the fused image, and the detected defect location and type information on the interface, highlights the defect area with a marking box of different colors, and provides a detailed defect description report, including the size and severity of the defect; Alarm and communication module: used to automatically trigger the alarm device when serious defects are detected, and send the detection results to the remote monitoring center and the mobile devices of relevant managers based on network communication.
[0015] Compared with the prior art, the present invention provides an AI visual inspection device and system for shell structure defect detection, which has the following beneficial effects: The AI visual inspection equipment and system for shell structure defect detection can adjust the height of the top brush roller through the cleaning component by starting the set hydraulic lifting rod, and cooperate with the brush roller set at the bottom to clean the shell that needs to be inspected for defects, thereby improving the accuracy of subsequent inspections. By starting the set servo motor, the belt and pulley can be driven to drive the bottom brush roller to rotate, so that the brush roller can also feed the shell. The AI visual inspection equipment and system for shell structural defect detection can drive the gear to rotate through the rotating pulley, and the gear can engage with the gear ring to drive the fan blade to rotate. The fan blade can then have the effect of exhausting air, and can clean the waste generated by cleaning in the inner cavity of the pretreatment box so that it can be sucked into the inner cavity of the collection box for storage. The waste can be collected, thereby improving the cleaning efficiency. The AI visual inspection equipment and system for shell structure defect detection can detect defects on the shell through the set visual inspection components, and can adjust the brightness of the shell according to the detection requirements through the set adjustable light source, thereby ensuring the accuracy of the detection results. The AI visual inspection equipment and system for shell structure defect detection can drive the first bidirectional screw to rotate by starting the set rotating motor, so that the adjustment component can be threadedly connected with it. At this time, the distance between the two adjustment components can be conditioned, so that the position of the adjustment light source can be adjusted. By starting the set rotating cylinder, the adjustment light source can be driven to rotate, so that the angle of the adjustment light source can be adjusted. When the shell is inspected, the accuracy of the shell inspection result can be further guaranteed. The AI visual inspection equipment and system for shell structural defect detection uses a transparent plate to enable the visual inspection component at the bottom to perform visual inspection on the bottom of the shell, and the remaining visual inspection components are arranged at the top and both ends of the frame cavity. During inspection, the shell can be visually inspected in all directions. The AI visual inspection equipment and system for shell structure defect detection can drive the second bidirectional screw to rotate by starting the set driving motor, so that the bottom of the driving rod can be connected with its thread. At this time, the two alignment components can be driven to move to the opposite side. When the shell is inspected, it can be aligned, thereby improving the efficiency of inspection and the convenience of subsequent material unloading. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 It is a schematic diagram of the overall structure of the present invention; Figure 2 It is a schematic diagram of the structure of the collection component of the present invention; Figure 3 is a schematic diagram of the structure of the visual detection component of the present invention; Figure 4 The present invention Figure 1 The enlarged view of point A in the middle; Figure 5 It is a system block diagram of the present invention.
[0017] In the figure: 1. pretreatment box; 2. hydraulic lifting rod; 3. cleaning component; 4. brush roller; 5. driving component; 6. servo motor; 7. belt; 8. pulley; 9. gear; 10. gear ring; 11. fan blade; 12. collection component; 13. collection box; 14. filter; 15. guide plate; 16. unloading wheel; 17. transparent plate; 18. visual inspection component; 19. visual inspection camera; 20. rotating motor; 21. movable groove; 22. first bidirectional screw rod; 23. guide rod; 24. adjustment component; 25. rotating cylinder; 26. adjusting light source; 27. detection component; 28. frame; 29. conveyor belt; 30. driving motor; 31. straightening component; 32. roller; 33. driving rod; 34. adjustment groove; 35. second bidirectional screw rod. DETAILED DESCRIPTION
[0018] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods. Example
[0019] like Figure 1 , Figure 2As shown, an AI visual inspection device for shell structure defect detection includes a pretreatment box 1, a cleaning component 3 is movably connected to the top of the inner cavity of the pretreatment box 1, and a brush roller 4 is rotatably connected to the cleaning component 3 and the bottom of the inner cavity of the pretreatment box 1, a driving component 5 is installed on the outer wall of the pretreatment box 1, a collecting component 12 is installed on the top of the driving component 5, a collecting box 13 is movably connected in the inner cavity of the collecting component 12, and a fan blade 11 is rotatably connected to the end of the collecting component 12 away from the pretreatment box 1, a hydraulic lifting rod 2 is installed in the middle of the top of the pretreatment box 1, and the bottom of the hydraulic lifting rod 2 is installed in the middle of the top of the cleaning component 3, and the driving component A belt 7 is installed in the inner cavity of 5, a pulley 8 is installed in the inner cavity of the belt 7, the pulley 8 is connected to the brush roller 4 at the bottom, a servo motor 6 is installed on the outer wall of the drive assembly 5, the servo motor 6 is installed on the outer wall of the pulley 8 on the right side through the first rotating shaft, the outer wall of the drive assembly 5 is rotatably connected with a gear 9, the gear 9 is installed on the outer wall of the middle pulley 8, the top of the gear 9 is meshed with a gear ring 10, the brush roller 4 is installed on the outer wall of the fan blade 11, the collecting assembly 12 is a cavity structure with both ends communicating with the outside, the collecting assembly 12 and the collecting box 13 are both communicated with the inner cavity of the pretreatment box 1, and the outer wall of the collecting box 13 is installed with a filter screen 14; By starting the hydraulic lifting rod 2, the height of the top brush roller 4 can be adjusted through the cleaning component 3, and the shell that needs to be inspected for defects can be cleaned in cooperation with the brush roller 4 at the bottom, so as to improve the accuracy of subsequent inspections. By starting the servo motor 6, the belt 7 and the pulley 8 can be driven to drive the bottom brush roller 4 to rotate, so that the brush roller 4 can also feed the shell; By rotating the pulley 8, the gear 9 can be driven to rotate at the same time, and the gear 9 can be engaged with the gear ring 10, thereby driving the fan blades 11 to rotate. The fan blades 11 can then have the effect of exhausting air, and can clean the waste generated by cleaning in the inner cavity of the pretreatment box 1, so that it can be sucked into the inner cavity of the collection box 13 for storage, and the waste can be collected, thereby improving the cleaning efficiency. Example
[0020] like Figure 1 , Figure 3As shown, an AI visual inspection device for detecting structural defects of a shell body is shown, a guide plate 15 is installed on the right side of the pretreatment box 1, a detection component 27 is installed on the right side of the guide plate 15, a frame 28 is installed on the top of the detection component 27, the top of the inner cavity of the frame 28, the two ends of the inner cavity of the frame 28, and the bottom of the guide plate 15 are all installed with visual inspection components 18, the two ends of the top of the guide plate 15 are symmetrically connected to the unloading wheel 16, a transparent plate 17 is installed in the middle of the guide plate 15, one of the visual inspection components 18 is installed at the bottom of the transparent plate 17, a visual inspection camera 19 is installed at one end of the visual inspection component 18, and movable grooves 21 are symmetrically opened on the top and bottom of the outer wall of the visual inspection component 18, and a first movable groove 21 is rotatably connected in the inner cavity of the top movable groove 21. A bidirectional screw 22, a guide rod 23 is installed in the inner cavity of the bottom movable groove 21, a rotating motor 20 is installed on the side of the visual detection component 18, and the rotating motor 20 is installed on the outer wall of the first bidirectional screw 22 through the second rotating shaft. The visual detection component 18 also includes an infrared thermal imager and a depth sensor. The top and bottom of the adjustment component 24 near one end of the visual detection component 18 are both movably connected in the inner cavity of the movable groove 21, the top of the adjustment component 24 is threadedly connected to the outer wall of the first bidirectional screw 22, and the bottom of the adjustment component 24 is sleeved on the outer wall of the guide rod 23. A rotating cylinder 25 is installed on the top of the adjustment component 24, and an adjusting light source 26 is installed on the bottom of the rotating cylinder 25 through a third rotating shaft. The adjusting light source 26 is rotatably connected in the inner cavity of the adjustment component 24; By means of the visual inspection component 18, the shell can be inspected for defects, and by means of the adjustable light source 26, the brightness of the shell can be adjusted according to the inspection requirements, thereby ensuring the accuracy of the inspection results; By starting the rotating motor 20, the first bidirectional screw rod 22 can be driven to rotate, so that the adjusting component 24 can be threadedly connected thereto. At this time, the distance between the two adjusting components 24 can be adjusted, so that the position of the adjusting light source 26 can be adjusted. By starting the rotating cylinder 25, the adjusting light source 26 can be driven to rotate, so that the angle of the adjusting light source 26 can be adjusted. When the shell is inspected, the accuracy of the shell inspection result can be further guaranteed. By setting up the transparent plate 17, the visual inspection component 18 at the bottom thereof can perform visual inspection on the bottom of the shell, and the remaining visual inspection components 18 are arranged at the top and both ends of the inner cavity of the frame 28. During inspection, the shell can be visually inspected in all directions. Example
[0021] like Figure 1 , Figure 4As shown, an AI visual inspection device for detecting structural defects of a shell body, the two ends of the top of the inner cavity of the detection component 27 are symmetrically and movably connected with the alignment component 31, the right side of the alignment component 31 is symmetrically installed with a driving rod 33, a conveyor belt 29 is installed in the inner cavity of the detection component 27, an adjustment groove 34 is opened at the bottom of the right side of the detection component 27, a second bidirectional screw 35 is installed in the inner cavity of the adjustment groove 34, a driving motor 30 is installed on the outer wall of the detection component 27, and the driving motor 30 is installed on the outer wall of the cleaning component 3 through a fourth rotating shaft, the bottom of the driving rod 33 is movably connected in the inner cavity of the adjustment groove 34 and is threadedly connected to the outer wall of the second bidirectional screw 35, and rollers 32 are rotatably connected in the inner cavity on the opposite side of the alignment components 31 at both ends; By starting the set driving motor 30, the second bidirectional screw 35 can be driven to rotate, so that the bottom of the driving rod 33 can be threadedly connected thereto, and the two alignment components 31 can be driven to move to the opposite side. When the shell is inspected, it can be aligned, thereby improving the efficiency of the inspection and the convenience of subsequent material unloading. Example
[0022] like Figure 5 As shown, an AI visual inspection system for shell structure defect detection is applied to an AI visual inspection device for shell structure defect detection, including an image acquisition unit, a data preprocessing unit, a deep learning detection unit and a result display and alarm module, and the image acquisition unit includes a detail information capture module, an internal structure capture module, and a three-dimensional shape capture module; The detail information capture module captures the texture and color information of the shell surface based on the visual inspection camera 19. The internal structure capture module detects the temperature difference caused by the internal structure defects and uneven surface material of the shell based on the infrared thermal imager to assist in discovering potential defects. The three-dimensional shape capture module obtains the three-dimensional shape information of the shell based on the depth sensor to identify the defect type and provide spatial coordinates for the location of the defect. The data preprocessing unit includes an image registration and fusion module, an image enhancement and standardization module; The image registration and fusion module is used to register the acquired multimodal images, align the shell structures in different modal images in space, and then perform fusion operations; The image enhancement and standardization module is used to enhance the fused image, including contrast adjustment and noise removal, to improve the quality and clarity of the image, and to standardize the image to meet the input requirements of the deep learning model; The deep learning detection unit includes a defect detection neural network model and a model training and optimization module; The defect detection neural network model is used to build a convolutional neural network model based on deep learning. The convolutional neural network model based on deep learning includes three branch structures, which are used to process different information features in the fused image respectively. The first branch is used to process the texture and color features in the visible light image, and extract the characteristic patterns of scratches and holes through multi-layer convolution and pooling operations; the second branch is used to analyze the temperature features in the infrared image, identify the temperature abnormality area caused by defects, and learn its association with different defect types; the third branch uses the shape information of the depth image to detect the deformation of the shell; The convolutional neural network model based on deep learning finally merges the features extracted by the three branches and classifies them through the fully connected layer to determine whether there are defects and the specific type of defects, and outputs the location information of the defects in the image; The model training and optimization module collects a large number of multimodal image samples containing various shell structural defects, trains the constructed deep learning model, uses the cross entropy loss function as the training objective function, and adjusts the model parameters based on the optimization algorithm; During the training process, data augmentation technology is used to expand the training data set, and the early stopping method is used to monitor the performance of the model on the validation set. When the performance no longer improves, training is stopped to obtain the optimal model parameters; The result display and alarm module includes a visualization interface and an alarm and communication module, wherein the visualization interface includes a visualization user interface, which intuitively displays the detection results to the user, displays the original image of the shell structure, the fused image, and the detected defect location and type information on the interface, highlights the defect area with different color marking boxes, and provides a detailed defect description report, including the size and severity of the defect; Alarm and communication module: used to automatically trigger the alarm device when serious defects are detected, and send the detection results to the remote monitoring center and the mobile devices of relevant managers based on network communication.
[0023] It should be noted that the present invention is an AI visual inspection device and system for detecting structural defects of a shell. When in use, the shell is placed in the inner cavity of a pretreatment box 1, the hydraulic lifting rod 2 is started, and the brush roller 4 on the top is driven downward by the cleaning component 3 to make it fit on the top of the shell. The servo motor 6 is started to drive the belt 7 and the pulley 8, so that the pulley 8 can drive the brush roller 4 at the bottom and the gear 9 outside the driving component 5 to rotate. The brush roller 4 can then feed the shell, and the gear 9 will mesh with the gear ring 10, so that the fan blade 11 can rotate. At this time, the waste scraps cleaned by the brush roller 4 in the inner cavity of the pretreatment box 1 will be sucked into the inner cavity of the collection box 13; When the cleaned moving body moves to the right side, it will slide to the conveyor belt 29 through the transparent plate 17, and after passing the visual inspection component 18 at the bottom of the transparent plate 17, the bottom of the shell can be visually inspected; When the shell moves to the conveyor belt 29, the drive motor 30 is started to drive the second bidirectional screw 35 to rotate, so that the drive rod 33 can drive the alignment component 31 to move, so that the alignment component 31 can be attached to the two ends of the shell through the roller 32, and the conveyor belt 29 is started to drive the shell to move. At this time, the visual inspection component 18 at the top of the inner cavity of the frame 28 and at its two ends can inspect the top and both ends of the shell, thereby completing the inspection of the shell as a whole.
[0024] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. An AI visual inspection device for shell structure defect detection, comprising a pretreatment box (1), characterized in that: The top of the inner cavity of the pretreatment box (1) is movably connected to a cleaning component (3), and the bottom of the cleaning component (3) and the inner cavity of the pretreatment box (1) are both rotatably connected to brush rollers (4). The outer wall of the pretreatment box (1) is mounted with a driving component (5), and the top of the driving component (5) is mounted with a collecting component (12). The inner cavity of the collecting component (12) is movably connected to a collecting box (13), and the end of the collecting component (12) away from the pretreatment box (1) is rotatably connected to a fan blade (11); A material guide plate (15) is installed on the right side of the pretreatment box (1), a detection component (27) is installed on the right side of the material guide plate (15), a frame (28) is installed on the top of the detection component (27), and visual detection components (18) are installed on the top of the inner cavity of the frame (28), both ends of the inner cavity of the frame (28), and the bottom of the material guide plate (15); The two ends of the top of the inner cavity of the detection component (27) are symmetrically and movably connected with the alignment components (31), and the right side of the alignment component (31) is symmetrically installed with a driving rod (33).
2. The AI visual inspection device for shell structure defect detection according to claim 1 is characterized in that: A hydraulic lifting rod (2) is installed in the middle of the top of the pretreatment box (1), and the bottom of the hydraulic lifting rod (2) is installed in the middle of the top of the cleaning component (3). A belt (7) is installed in the inner cavity of the driving component (5), and a pulley (8) is installed in the inner cavity of the belt (7). The pulley (8) is connected to the brush roller (4) at the bottom. A servo motor (6) is installed on the outer wall of the driving component (5), and the servo motor (6) is installed on the outer wall of the pulley (8) on the right side through a first rotating shaft.
3. The AI visual inspection device for shell structure defect detection according to claim 1 is characterized in that: The outer wall of the driving assembly (5) is rotatably connected to a gear (9), the gear (9) is mounted on the outer wall of the middle belt pulley (8), the top of the gear (9) is meshed with a gear ring (10), and the brush roller (4) is mounted on the outer wall of the fan blade (11).
4. The AI visual inspection device for shell structure defect detection according to claim 1, characterized in that: The collecting component (12) is a hollow structure with two ends communicating with the outside, the collecting component (12) and the collecting box (13) are both in communication with the inner cavity of the pretreatment box (1), and a filter screen (14) is installed on the outer wall of the collecting box (13).
5. The AI visual inspection device for shell structure defect detection according to claim 1, characterized in that: Both ends of the top of the guide plate (15) are symmetrically rotatably connected to unloading wheels (16), a transparent plate (17) is installed in the middle of the guide plate (15), and one of the visual detection components (18) is installed at the bottom of the transparent plate (17).
6. The AI visual inspection device for shell structure defect detection according to claim 1, characterized in that: A visual detection camera (19) is installed at one end of the visual detection component (18); movable grooves (21) are symmetrically provided at the top and bottom of the outer wall of the visual detection component (18); a first bidirectional screw rod (22) is rotatably connected in the inner cavity of the top movable groove (21); a guide rod (23) is installed in the inner cavity of the bottom movable groove (21); a rotating motor (20) is installed on the side of the visual detection component (18); the rotating motor (20) is installed on the outer wall of the first bidirectional screw rod (22) via a second rotating shaft; and the visual detection component (18) further comprises an infrared thermal imager and a depth sensor.
7. The AI visual inspection device for shell structure defect detection according to claim 1, characterized in that: The top and bottom of the adjustment component (24) at one end close to the visual detection component (18) are movably connected in the inner cavity of the movable groove (21); the top of the adjustment component (24) is threadedly connected to the outer wall of the first bidirectional screw (22); the bottom of the adjustment component (24) is sleeved on the outer wall of the guide rod (23); a rotating cylinder (25) is installed on the top of the adjustment component (24); an adjustment light source (26) is installed on the bottom of the rotating cylinder (25) via a third rotating shaft; and the adjustment light source (26) is rotatably connected in the inner cavity of the adjustment component (24).
8. The AI visual inspection device for shell structure defect detection according to claim 1, characterized in that: A conveyor belt (29) is installed in the inner cavity of the detection component (27); an adjustment groove (34) is provided at the bottom of the right side of the detection component (27); a second bidirectional screw rod (35) is installed in the inner cavity of the adjustment groove (34); a drive motor (30) is installed on the outer wall of the detection component (27); and the drive motor (30) is installed on the outer wall of the cleaning component (3) via a fourth rotating shaft.
9. The AI visual inspection device for shell structure defect detection according to claim 1, characterized in that: The bottom of the driving rod (33) is movably connected in the inner cavity of the adjusting groove (34) and is threadedly connected to the outer wall of the second bidirectional screw (35). Rollers (32) are rotatably connected in the inner cavities on the opposite sides of the alignment components (31) at both ends.
10. An AI visual inspection system for shell structure defect detection, applied to the above-mentioned AI visual inspection device for shell structure defect detection, comprising an image acquisition unit, a data preprocessing unit, a deep learning detection unit and a result display and alarm module, characterized in that: The image acquisition unit includes a detail information capture module, an internal structure capture module, and a three-dimensional shape capture module; The detail information capture module captures the texture and color information of the shell surface based on a visual detection camera (19); the internal structure capture module detects the temperature difference of the shell caused by internal structural defects and uneven surface material based on an infrared thermal imager to assist in discovering potential defects; and the three-dimensional shape capture module obtains the three-dimensional shape information of the shell based on a depth sensor to identify the defect type and provide spatial coordinates for the location of the defect; The data preprocessing unit includes an image registration and fusion module, an image enhancement and standardization module; The image registration and fusion module is used to register the acquired multi-modal images, align the shell structures in different modal images in space, and then perform a fusion operation; The image enhancement and standardization module is used to enhance the fused image, including contrast adjustment and noise removal, to improve the quality and clarity of the image, and to standardize the image to meet the input requirements of the deep learning model; The deep learning detection unit includes a defect detection neural network model and a model training and optimization module; The defect detection neural network model is used to construct a convolutional neural network model based on deep learning. The convolutional neural network model based on deep learning includes three branch structures, which are used to process different information features in the fused image respectively. The first branch is used to process the texture and color features in the visible light image, and extract the characteristic patterns of scratches and holes through multi-layer convolution and pooling operations; the second branch is used to analyze the temperature features in the infrared image, identify the temperature abnormality area caused by the defect, and learn its association with different defect types; the third branch uses the shape information of the depth image to detect the deformation of the shell; The convolutional neural network model based on deep learning finally merges the features extracted by the three branches and classifies them through the fully connected layer to determine whether there are defects and the specific type of defects, and outputs the location information of the defects in the image; The model training and optimization module collects a large number of multimodal image samples containing various shell structural defects, trains the constructed deep learning model, uses the cross entropy loss function as the objective function of the training, and adjusts the parameters of the model based on the optimization algorithm; During the training process, data augmentation technology is used to expand the training data set, and the early stopping method is used to monitor the performance of the model on the validation set. When the performance no longer improves, training is stopped to obtain the optimal model parameters; The result display and alarm module includes a visualization interface and an alarm and communication module, wherein the visualization interface includes a visualization user interface, which intuitively displays the detection results to the user, displays the original image of the shell structure, the fused image, and the detected defect location and type information on the interface, highlights the defect area with a marking box of different colors, and provides a detailed defect description report, including the size and severity of the defect; Alarm and communication module: used to automatically trigger the alarm device when serious defects are detected, and send the detection results to the remote monitoring center and the mobile devices of relevant managers based on network communication.