Automatic inspection equipment and method for assembly quality of missile compartment parts

Abstract

Disclosed is an automated inspection device and method for the assembly quality of missile compartment components. The hardware unit comprises a base platform, a motion device, a visual sensing system, and a control system; the software unit comprises a management terminal and an on-site execution terminal. The method employs multiple visual sensors and a target detection algorithm, combined with automated mechanisms such as a linear motion mechanism and a manipulator, to automatically inspect the assembly quality of missile compartment components. The method achieves a high degree of inspection automation, high inspection judgment accuracy, strong accessibility to inspection locations, and digitally traceable inspection results.

Description

Technical Field

[0001] The invention belongs to the technical field of missile assembly inspection, and in particular relates to an automated inspection method and equipment for the assembly quality of missile compartment parts. Background Art

[0002] Missile compartment component assembly is a key process in missile final assembly. This involves assembling standard parts, cable networks, and equipment within the compartment. This includes equipment installation, cable routing and bundling, and electrical connector docking. Assembly quality issues such as missing fasteners, incorrect installation orientation, improperly mated electrical connector plugs and sockets, damaged or overbent cables, incorrect cable routing, and incorrect cable bundling directly impact missile performance and functionality. Missile assembly processes are demanding, complex, and difficult to repair. These quality issues pose significant risks and challenges to missile production quality control. Therefore, implementing compartment component assembly inspection is an effective measure to control assembly quality defects and maintain control over missile manufacturing performance.

[0003] Missile compartment parts are complex in shape and diverse in variety, with dozens of types and about a hundred pieces. Currently, compartment assembly inspections are all completed by manual visual inspections. Inspectors need to complete the inspection of dozens of assembled products of the same model before they can leave the site to check the process and drawings and master the specific process inspection requirements. This approach places high demands on personnel's cognitive, comprehension, and memory abilities, relying heavily on experience and requiring a long training cycle. Secondly, the current missile manufacturing model utilizes parallel production along the same production line based on missile caliber series. The supporting products and assembly requirements for different models vary significantly, leading to confusion and mismatched inspection requirements between different products during the inspection process. This further increases the high level of competence required of inspectors to master the assembly inspection of multiple models, exacerbating the reliance on highly qualified and experienced personnel and increasing the workload of the limited number of inspectors. Thirdly, the small cross-sectional area and long longitudinal extension of the assembly space in missile compartments limit visual access, leading to blind inspections and quality risks during assembly quality inspections. Furthermore, physical penetration inspections create additional safety hazards. Furthermore, the current method for recording inspection results requires inspectors to visually inspect and then take photos with a handheld camera. The photos are then exported and manually labeled. This approach has significant drawbacks: consistent image quality relies on personnel skill, labeling large volumes of image files can lead to low-level human errors such as omissions and errors, significantly increases post-inspection workload, and tracking and tracing quality issues takes a long time, making data analysis challenging. Therefore, in order to meet the urgent needs of rapid response manufacturing and zero-defect product quality under high-capacity, multi-model, and multi-state co-line assembly of missiles, it is urgent to explore digital and automated inspection technology methods. Summary of the Invention

[0004] The technical problem solved by the present invention is to overcome the shortcomings of the existing technology and provide an automated inspection method and equipment for the assembly quality of missile compartment parts, so as to solve the disadvantages of manual visual inspection methods, such as confusion in inspection content, high labor intensity, quality and safety risks, strong dependence on human experience, and difficulty in manual photography and tracing.

[0005] The technical solution of the present invention is: an automated inspection device for the assembly quality of missile compartment parts, including a software unit and a hardware unit, wherein the hardware unit includes a base platform, a motion device, a visual sensor system and a control system;

[0006] The base platform serves as the foundation for the entire equipment, providing support for other parts, component storage, and overall mobility. The motion device is the equipment's operating actuator, executing all mechanical movements during operation. The visual sensing system collects image information from compartment inspection locations and operator gestures. The control system utilizes an industrial Ethernet bus for signal acquisition and control signal output from the motion device and visual sensing system.

[0007] The software unit includes a management end and a field execution end;

[0008] The management end adopts a B / S architecture to perform data maintenance, query and authority management for process and quality management personnel; the on-site execution end adopts a C / S architecture to perform on-site inspection business; the management end and the on-site execution end realize command interaction through the underlying database.

[0009] The base platform includes an assembly table, a platform cabinet, and universal brake industrial casters; four universal brake industrial casters are installed on the bottom of the platform cabinet through connectors; and the assembly table is installed on the upper surface of the platform cabinet through bolts.

[0010] The motion device includes a linear motion mechanism and a six-joint manipulator; the linear motion mechanism is fixedly installed on the assembly table; the six-joint manipulator is connected to the connecting slider on the linear motion mechanism through a flange, so that the six-joint manipulator can move as a whole following the linear motion mechanism.

[0011] The visual sensing system includes a detection visual sensing device and a USB computer camera; the detection visual sensing device is used to collect image information of the cabin inspection position, and includes a deep binocular industrial camera, a monocular industrial camera, a ring light source, and a bracket; the deep binocular industrial camera, the monocular industrial camera, and the ring light source are combined and fixedly installed on the bracket; the bracket is assembled and fixed to the end flange of the six-joint manipulator; the USB computer camera is used to collect the operator's interactive gesture images, and its own bracket base is bonded to the control system.

[0012] The control system includes a motion controller, an industrial computer, and a touch screen panel; the touch screen panel and the industrial computer communicate via a wired connection; the touch screen panel is used to display the software system interface and obtain touch interactive input instructions; the industrial computer and motion controller are installed and fixed in the platform cabinet; the touch screen panel is hung on the side of the base platform when not in use, and is placed on the work desk using its own bracket when in use.

[0013] The management end includes an inspection process model management module, an inspection result management module, a rights management module and a dictionary management module;

[0014] The inspection process model management is used to manage the established inspection process model and is the data source for the on-site execution end;

[0015] The inspection result management module is used to manage the inspection results, which include inspection conclusions and inspection images;

[0016] The authority management module is used to manage the authority of users logged in to the management terminal and the on-site execution terminal, and different authorities correspond to different module usage authorities;

[0017] The dictionary management module is used to define the standard names of various elements included in the product code and standardize the input of data.

[0018] The inspection process model consists of three parts: product code, calibration data, and inspection process, which are linked through a structured table; the product code refers to information that determines the status of the product to be inspected, including product model, drawing number, batch, and stage; the calibration data is the initial position data of the inspection equipment and the product to be inspected; the inspection process is the inspection process flow and inspection process standard, including process number, process name, work step name, component type, process text requirements, standard image template, and inspection position.

[0019] The on-site execution terminal includes a login module, a position calibration module, a standard image template formulation module, and a verification module;

[0020] The login module is used for login management of the on-site execution end. Login user information and product code information of the compartment to be inspected are input into the login module; the login user information is used to verify user permissions, and the product code information is used to retrieve the corresponding inspection process model data and inspection result data of the management end;

[0021] The position calibration module is used to obtain the position data of the inspection equipment and the product to be inspected, including calibration data and inspection position data; and push it to the inspection process model of the management end;

[0022] The standard image template formulation module uses an image annotation tool to annotate the inspection elements on a large number of imported qualified assembly quality images; after all images are annotated, an image training tool is used to train assembly quality features to obtain qualified recognition features and form a standard image template;

[0023] The inspection module obtains the inspection process model data of the management end based on the product code information input by the login module, drives the motion device to move to the obtained inspection position, controls the inspection visual sensor device to collect images, and calls the assembly quality image inspection tool to perform inspection and provide inspection results. The inspection results are also assigned to the inspection results of the management end; the assembly quality image inspection tool is encapsulated using the YOLO 4 algorithm.

[0024] The method for obtaining the position data is as follows: driving and debugging the motion device and the detection visual sensing device to the standard inspection preparation position, obtaining the position coordinates of the motion device and the product position coordinates, forming calibration data and pushing the calibration data of the inspection process model to the management end; driving and debugging the motion device and the detection visual sensing device to the inspection position that meets the process conditions, obtaining the coordinate position of the motion device, and pushing the inspection position data corresponding to the work step in the inspection process in the inspection process model to the management end.

[0025] The inspection module at the on-site execution end will expand the inspection task list each time it is started according to the inspection process update of the inspection process model, display the task completion status, inspection conclusion, inspection time in the list, and display the inspection image result of the selected list item on the interface.

[0026] The dynamic process data of the on-site execution end all comes from the inspection process model of the management end. When performing the current model task, you only need to create a new inspection process model on the management end and improve the data.

[0027] A method for automatically inspecting the assembly quality of missile compartment components using the above-mentioned automated inspection equipment comprises:

[0028] S1, establish the inspection process model: when it is used for the first time or the inspection process is changed, edit the inspection process model on the management side;

[0029] The inspection process model consists of three parts: product code, calibration data, and inspection process, which are linked through structured tables; the product code includes product model, drawing number, batch, and stage; the calibration data includes equipment calibration data and product calibration data; the inspection process includes process number, process name, step name, component type, process text requirements, standard image template, and inspection location; after completing the inspection process modeling of S1 on the management side, enter S2, S3, and S4 through the on-site execution side to obtain data to improve the model.

[0030] S2, conduct initial inspection position teaching: select a standard cabin section with qualified quality and place it on the assembly platform, start the automated inspection equipment, drive the linear motion mechanism and the six-joint manipulator, drive the inspection visual sensor device to move to the top of the standard cabin section and align it with the cabin section; visual recognition, calculation and display the actual three-dimensional spatial distance of the standard cabin section relative to the automated inspection equipment; adjust the posture of the standard cabin section until the relative spatial distance meets the inspection process conditions, and set the equipment position and the three-dimensional spatial distance of the cabin section relative to the equipment at this time as the initial position data of the inspection equipment and the initial position data of the product to be inspected, and push them to the management end for calibration data in the inspection process model respectively; enter S3 to further obtain the inspection position data in the inspection process model;

[0031] S3, conduct motion trajectory data teaching: First, for each inspection point required by the process inspection conditions, drive the linear motion mechanism and the six-joint manipulator to the inspection position that meets the process conditions, turn on the inspection visual sensor device, and the posture of the six-joint manipulator end ensures that the inspection visual sensor device can collect all inspection elements. The coordinate values ​​of the linear motion mechanism and the six-joint manipulator at this time are recorded to form the inspection position corresponding to the work step, and the inspection position of the inspection point of the work step in the process model of the management end is pushed; repeat step S3 to complete the motion trajectory teaching of all inspection points; enter S4 to further obtain the data of the standard image template in the inspection process model;

[0032] S4, create a standard inspection template: The standard image template refers to the standard state template of a qualified product. It is used to compare and analyze the assembly state of the product to be inspected for consistency. It is a standard image template for quality inspection. A qualified product identical to the product to be inspected is used to demonstrate the standard quality of assembly at a certain step required by the process model. Images of multiple products and multi-angles with qualified assembly quality are collected. Image annotation tools and image training tools are used to annotate and train the inspection objects on the images. The training results are pushed to the management end to create the standard image template for the inspection process model.

[0033] S5, inspection preparation: Start the inspection equipment, input the code information of the product to be inspected, the system automatically retrieves the matching data in the inspection process model, and completes the data preparation; push the missile compartment to be inspected to the inspection station, call the calibration data in the inspection process model and drive the linear motion mechanism and the six-joint manipulator to the initial inspection position, and display the image of the compartment to be inspected captured by the monocular industrial camera in real time on the touch screen panel. The binocular depth industrial camera recognizes and calculates the coordinate deviation between the current position and the product calibration data in the inspection process model, provides a standard position virtual frame and adjustment parameters, and adjusts the compartment to be inspected into the virtual frame. After completing the physical inspection preparation, enter S6;

[0034] S6, Inspection Execution: Select the work step to be inspected, and after receiving the human-machine interaction command to start the inspection, read the inspection position data in the inspection process model, drive the motion device to the inspection position, automatically turn on the ring light source and monocular industrial camera of the inspection visual sensor device to collect images, call the assembly quality image inspection tool to compare the feature data of the image with the standard image template for consistency, and enter S7 after obtaining the consistency comparison result;

[0035] S7, test results: the consistency comparison results, including the test conclusion and the test image; the test results are pushed to the management end; the test images include two types: the acquisition image and the judgment image; the acquisition image is the original image captured by the monocular industrial camera; the judgment image marks the objects that pass the test with a green frame, and the objects that fail the test with a red frame; repeat S6 until all test steps are completed, and enter S8;

[0036] S8, inspection is completed; push the compartment to the next assembly position.

[0037] The visual recognition, calculation and display of the actual three-dimensional space distance of the standard cabin section relative to the automated inspection equipment in S2 include: setting a corner point of the assembly table of the automated inspection equipment as the origin O of the world coordinate system, using the deep binocular industrial camera of the detection visual sensor device to identify the center of the cabin section reference hole, deducing and transforming based on stereo vision positioning and robot kinematics, calculating the three-dimensional space distance of the front end face of the cabin section relative to the automated inspection equipment, and displaying the standard position of the reference circular hole on the screen with a frame for graphical guidance.

[0038] The image annotation of S4 adds a “tiny” attribute annotation dimension for parts in the cabin that are smaller than 1 cm, and further refines the convolution division for the detection images with “tiny” attributes.

[0039] The assembly quality image inspection tool in S6 is encapsulated using the YOLO 4 algorithm. For cable path detection, the YOLO 4 algorithm is first used to identify the missile barrel cross-section and cables. Cable color features and a threshold algorithm are then used to filter out other background feature information within the frame to obtain the cable's appearance outline. This outline is then refined to remove redundant features. A skeleton extraction algorithm is used to find the cable's skeleton line, which is then pushed to a feature recognition operator to detect and determine the length and position of the cables in the area and the image.

[0040] The order of S2, S3, and S4 can be changed, but they must all be completed before entering S5. S1 is completed at the management end, and S2-S8 are completed at the on-site execution end.

[0041] The present invention has the following beneficial effects:

[0042] 1. The present invention fills the gap in automated inspection of the assembly quality of missile compartment components, effectively improves the automation level of the missile assembly manufacturing process, and lays a technical foundation for unmanned inspection.

[0043] 2. The present invention utilizes automated inspection equipment inspection software to establish a structured process model and structurally binds the working data of the inspection equipment to the process model, thereby realizing the operation of automated equipment driven by the process model and improving the ability to control the effectiveness of the automated equipment process.

[0044] 3. The automated inspection equipment inspection software of the present invention adopts a method of separating the management end from the on-site execution end, which effectively improves the efficiency of issuing process data and collecting inspection results of multiple automated equipment on site, and realizes one-to-many one-time processing; at the same time, users can use it more targetedly and the operation is simpler and more convenient.

[0045] 3. The invention eliminates manual process participation by combining automated movement of actuators, clear observation of visual sensors, and accurate judgment of artificial intelligence detection algorithms. It not only ensures high accuracy of inspection and judgment, but also improves the accessibility of inspection and observation, effectively eliminating the quality and safety risks caused by the original manual vision.

[0046] 4. The present invention targets the inspection result image. When the detection algorithm determines that the inspection conclusion is formed, the product batch information, the current inspection point and the assembly quality inspection conclusion information are automatically marked in the image file. At the same time, the image is saved and bound to the inspection database, completing the dual synchronization of inspection execution and result recording, realizing the digital traceability of the inspection results, liberating the manual marking work, and effectively eliminating the many disadvantages caused by the original manual handheld camera shooting and marking.

[0047] 5. The present invention can be further expanded to complete the inspection feature recognition and determination of other missile structures, and realize the automated inspection of the assembly quality of the entire missile. It can also adaptably expand the automated inspection equipment, such as upgrading the base platform to a full-drive AGV platform, or configuring multiple manipulators, to promote the establishment of missile automated detection units and automated inspection production lines, and realize the intelligent workshop of missile assembly and manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a schematic diagram of the overall structure of the automated inspection equipment for assembly quality of missile compartment parts of the present invention;

[0049] Figure 2 It is a partial enlarged structural schematic diagram of the visual sensing device of the present invention;

[0050] Figure 3 This is a schematic diagram of the main components of the inspection software system for the automated inspection equipment for assembly quality of missile compartment parts of the present invention;

[0051] Figure 4 Schematic diagram of the method for automated inspection of assembly quality of missile compartment parts when performing the S2 initial inspection position teaching and S5 inspection preparation steps of the present invention;

[0052] Figure 5 Schematic diagram of the method for automated inspection of assembly quality of missile compartment parts when the present invention performs S3 motion trajectory data teaching and S6 inspection execution steps;

[0053] Figure 6 This is a flowchart of the automated inspection method for assembly quality of missile compartment components of the present invention. DETAILED DESCRIPTION

[0054] The invention discloses an automatic inspection device for the assembly quality of missile compartment parts, comprising a software unit and a hardware unit.

[0055] The hardware unit includes a base platform, a motion device, a visual sensing system, and a control system;

[0056] The base platform serves as the foundation for the entire equipment, providing support for other parts, component storage, and overall mobility. The motion device is the equipment's operating actuator, executing all mechanical movements during operation. The visual sensing system collects image information from compartment inspection locations and operator gestures. The control system utilizes an industrial Ethernet bus for signal acquisition and control signal output from the motion device and visual sensing system.

[0057] like Figure 1 The hardware unit base platform shown here serves as the foundation for the entire device, providing support for the rest of the device, component storage, and overall mobility. From top to bottom, the base platform comprises an assembly surface 1, a platform cabinet 2, and universal brake industrial casters 3. Four universal brake industrial casters 3 are bolted to the bottom of the platform cabinet 2; the assembly surface 1 is also bolted to the top of the platform cabinet 2.

[0058] like Figure 2 The motion device shown includes a linear motion mechanism 4 and a six-joint manipulator 5; the linear motion mechanism 4 is fixedly mounted on the assembly table 1; the six-joint manipulator 5 is connected to the connecting joint slider on the linear motion mechanism 4 through a flange, so that the six-joint manipulator 5 can move as a whole following the linear motion mechanism 4.

[0059] The visual sensing system includes a detection visual sensing device 6 and a USB computer camera 13. The detection visual sensing device 6 is used to collect image information of the compartment inspection position, such as Figure 2The figure includes a depth binocular industrial camera 6.1, a monocular industrial camera 6.2, a ring light source 6.3, and a bracket 6.4; the depth binocular industrial camera 6.1, the monocular industrial camera 6.2, and the ring light source 6.2 are assembled and fixedly mounted on the bracket 6.4; the USB computer camera 13 is used to capture the operator's interactive gesture images, and its own bracket base is bonded to the top of the touch screen panel 9.

[0060] The control system includes a motion controller 7, an industrial control computer 8, and a touch screen panel 9; the touch screen panel 9 communicates with the industrial control computer 8 via a wired connection; the touch screen panel 9 is used to display the software system interface and obtain touch interactive input tasks; the industrial control computer 8 and the motion controller 7 are installed and fixed in the platform cabinet 2; the touch screen panel 9 is hung on the side of the base platform when not in use, and is placed on the work desk using its own bracket when in use.

[0061] The equipment inspection software system consists of two parts: the management end and the on-site execution end.

[0062] The management terminal utilizes a browser-based B / S architecture and is accessible via a browser. It provides data maintenance, querying, and permission management for process and quality management personnel. The on-site execution terminal utilizes a client-based C / S architecture and is installed in an industrial control computer. It integrates equipment hardware control and is used to perform on-site inspections. The management and on-site execution terminals exchange commands via an underlying database.

[0063] like Figure 3 The management end shown includes an inspection process model management module, an inspection result management module, a permission management module, and a dictionary management module.

[0064] The inspection process model management is used to manage the established inspection process model and is the data source for the on-site execution end. The inspection process model consists of three parts: product code, calibration data, and inspection process, which are associated through a structured table.

[0065] The product code refers to information that determines the status of the product to be inspected, including product model, drawing number, batch, and stage; the calibration data refers to the initial position data of the inspection equipment and the product to be inspected; the inspection process refers to the inspection process flow and inspection process standards, including process number, process name, step name, component type, process text requirements, standard image template, and inspection location;

[0066] The inspection result management module is used to manage the inspection results, which include inspection conclusions and inspection images;

[0067] The authority management is the authority management of users logged in at the management end and the on-site execution end, and different authorities correspond to different module usage authorities.

[0068] The dictionary management module is used to define the standard names of various elements included in the product code and standardize the input of data.

[0069] The on-site execution terminal includes a login module, a position calibration module, a standard image template formulation module, and a verification module;

[0070] The login module is used for login management of the on-site execution end. The login user information and the product code information of the compartment to be inspected are entered in the login module; the login information is used to verify user permissions, and the product code information is used to retrieve the corresponding inspection process model data and inspection result data from the management end.

[0071] The position calibration module is used to obtain the position data of the inspection equipment and the product to be inspected, and push it to the inspection process model on the management side. Specifically, it includes calibration data and inspection position data. It drives and debugs the motion device and the inspection visual sensor device 6 to the standard inspection preparation position, obtains the position coordinates of the motion device and the product position coordinates, forms calibration data, and pushes the calibration data to the inspection process model on the management side. It drives and debugs the motion device and the inspection visual sensor device 6 to the inspection position that meets the process conditions, obtains the coordinate position of the motion device, and pushes it to the inspection position corresponding to the work step in the inspection process in the inspection process model on the management side.

[0072] The standard image template formulation module uses an image annotation tool to annotate the detection elements on a large number of imported qualified assembly quality images; after all images are annotated, an image training tool is used to train assembly quality features to obtain qualified recognition features and form a standard image template.

[0073] The inspection module retrieves the inspection process model data from the management end based on the product code information input by the login module, drives the motion device to the acquired inspection position, controls the inspection visual sensor device 6 to capture images, and invokes the assembly quality image inspection tool to perform inspection and generate inspection results. The inspection results are also assigned to the management end. The assembly quality image inspection tool is encapsulated using the YOLO 4 algorithm.

[0074] The dynamic process data of the on-site execution end all comes from the inspection process model of the management end. When performing the current model task, you only need to create a new inspection process model on the management end and improve the data.

[0075] The inspection module of the on-site execution end updates and expands the inspection task list each time it is started according to the inspection process of the inspection process model, displays the completion status of the task, inspection conclusion, inspection time in the list, and displays the inspection image result of the selected list item on the interface.

[0076] A method for automated inspection of the assembly quality of missile compartment components, such as Figure 6The following implementation steps are shown:

[0077] S1, establish the inspection process model: when it is used for the first time or the inspection process is changed, edit the inspection process model on the management side;

[0078] like Figure 3 The inspection process model shown consists of three parts: product code, calibration data, and inspection process, which are linked through structured tables. The product code includes product model, drawing number, batch, and stage. The calibration data includes equipment calibration data and product calibration data. The inspection process includes process number, process name, step name, component type, process text requirements, standard image template, and inspection location. After completing the inspection process modeling of S1 on the management side, enter S2, S3, and S4 through the on-site execution side to obtain data to improve the model.

[0079] S2, conduct initial inspection position teaching: select a standard cabin section 11 with qualified quality and place it on the assembly platform 10, start the automated inspection equipment, such as Figure 4 As shown, the linear motion mechanism 4 and the six-joint manipulator 5 are driven to drive the detection visual sensor device 6 to move to the top of the standard compartment 11 and align it with the compartment; visual recognition, calculation and display of the actual three-dimensional spatial distance of the standard compartment 11 relative to the automated inspection equipment; the posture of the standard compartment 11 is adjusted until the relative spatial distance meets the inspection process conditions, and the equipment position and the three-dimensional spatial distance of the compartment relative to the equipment at this time are set as the initial position data of the inspection equipment and the initial position data of the product to be inspected, and are pushed to the "equipment calibration data" and "product calibration data" of the calibration data in the inspection process model on the management side respectively; enter S3 to further obtain the inspection position data in the inspection process model.

[0080] S3, conduct motion trajectory data teaching: First, for each inspection point required by the process inspection conditions, such as Figure 5 As shown, the linear motion mechanism 4 and the six-joint manipulator 5 are driven to the inspection position that meets the process conditions, and the detection visual sensor device 6 is turned on. The posture of the end of the six-joint manipulator 5 ensures that the detection visual sensor device 6 can collect all the inspection elements, and the coordinate values ​​of the linear motion mechanism 4 and the six-joint manipulator at this time are recorded to form the inspection position corresponding to the work step, and push it to the management end. The inspection position of the inspection point of the work step in the process model; repeat step S3 to complete the motion trajectory teaching of all inspection points; enter S4 to further obtain the data of the standard image template in the inspection process model.

[0081] S4, prepare standard inspection templates: standard image templates refer to standard status templates of qualified products, which are used for consistency comparison and analysis with the assembly status of the product to be inspected. They are standard image templates for quality inspection. Qualified products identical to the product to be inspected are used to demonstrate the standard quality of assembly for a certain step required in the process model. Qualified images of multiple products and multi-angle assembly quality are collected. Image annotation tools and image training tools are used to annotate and train the inspection objects on the images. The training results are pushed to the standard image template of the inspection process model on the management side. The order of S2, S3, and S4 can be swapped, but they must all be completed before entering S5.

[0082] S5, inspection preparation: start the inspection equipment, enter the code information of the product to be inspected, the system automatically retrieves the matching data in the inspection process model, and completes the data preparation; push the missile compartment 12 to be inspected to the inspection station, call the calibration data in the inspection process model, such as Figure 4 As shown, the linear motion mechanism 4 and the six-joint manipulator 5 are driven to the initial inspection position, and the image of the compartment to be inspected captured by the monocular industrial camera 6.2 is displayed in real time on the touch screen panel 9. The binocular depth industrial camera 6.1 recognizes and calculates the coordinate deviation between the current position and the product calibration data in the inspection process model, gives a standard position virtual frame and adjustment parameters, adjusts the compartment to be inspected into the virtual frame, completes the physical inspection and prepares to enter S6.

[0083] S6, Inspection execution: Inspection execution: Select the work step to be inspected, and after receiving the human-computer interaction command to start the inspection, read the inspection position data in the inspection process model, such as Figure 5 As shown, the driving motion device reaches the inspection position, automatically turns on the ring light source 6.3 and the monocular industrial camera 6.2 of the inspection visual sensor device 6 to collect images, and calls the assembly quality image inspection tool to compare and analyze the image with the feature data of the standard image template for consistency, obtains the analysis results, and enters S7;

[0084] S7, Inspection Result: The inspection module generates the inspection result, including the inspection conclusion and inspection image. The inspection result is pushed to the management terminal. The inspection image includes two types: acquisition image and determination image. The acquisition image is the original image captured by the monocular industrial camera 6.2. The determination image marks the objects that pass the inspection with a green frame, and those that fail with a red frame. All inspection steps are completed until S6.

[0085] S8, inspection is completed; push the cabin section to the next assembly station.

[0086] The S1 is completed at the management end, and S2-S8 are completed at the on-site execution end.

[0087] The visual identification, calculation and display of the actual three-dimensional space distance of the standard cabin 11 relative to the automated inspection equipment in step S2 include: setting a corner point of the assembly table 1 of the automated inspection equipment as the origin O of the world coordinate system, using the depth binocular industrial camera 6.1 of the inspection visual sensor device 6 to identify the center of the cabin reference hole, deducing and transforming based on stereo vision positioning and robot kinematics, calculating the three-dimensional space distance of the front end face of the cabin relative to the automated inspection equipment, and displaying the standard position of the reference circular hole on the screen with a frame for graphical guidance.

[0088] The image annotation of step S4 adds a “tiny” attribute annotation dimension for small parts smaller than 1 cm in the cabin, and further refines the convolution division for the detection points of the “tiny” attribute.

[0089] Step S6, the assembly quality image inspection tool, is encapsulated using the YOLO 4 algorithm. For cable path detection, the YOLO 4 algorithm is first used to identify the missile barrel cross-section and cables. It then uses algorithms such as cable color features and thresholds to filter out other background features within the image to obtain the cable's outline. This outline is then refined to remove redundant features. A skeleton extraction algorithm is then used to find the cable's skeleton line, which is then fed into a feature recognition operator to detect and determine the cable's length and position in the area, as well as the image itself.

[0090] Products with the same product code as in the process model can be inspected directly using S5-S8. If the inspection process changes, the process model needs to be modified according to S1-S4, and the on-site execution terminal can automatically synchronize the data.

[0091] All modifications and improvements of the present invention in this field should be included in the protection scope of the claims of the present invention.

Claims

1. An automated inspection device for the assembly quality of missile compartment components, comprising a software unit and a hardware unit, characterized in that: The software unit controls the hardware unit through instruction interaction to complete the automated inspection process; The hardware unit includes a base platform, a motion device, a visual sensing system, and a control system; The base platform serves as the foundation for the entire equipment, providing support for other parts, component storage, and overall mobility. The motion device is the equipment's operating actuator, executing all mechanical movements during operation. The visual sensing system collects image information from compartment inspection locations and operator gestures. The control system utilizes an industrial Ethernet bus for signal acquisition and control signal output from the motion device and visual sensing system. The software unit includes a management end and a field execution end; The management end adopts a B / S architecture to perform data maintenance, query and authority management for process and quality management personnel; the on-site execution end adopts a C / S architecture to perform on-site inspection business; the management end and the on-site execution end realize command interaction through the underlying database; The on-site execution end includes a login module, a position calibration module, a standard image template formulation module, and an inspection module; the inspection module obtains the inspection process model data of the management end according to the product code information input by the login module, drives the motion device to move to the obtained inspection position, controls the inspection visual sensor device (6) to collect images, and calls the assembly quality image inspection tool to perform inspection and give the inspection result, and the inspection result is simultaneously assigned to the inspection result of the management end; the assembly quality image inspection tool is encapsulated using the YOLO 4 algorithm.

2. The automated inspection equipment for assembly quality of missile compartment parts according to claim 1, characterized in that: The base platform comprises an assembly table (1), a platform cabinet (2), and universal brake industrial casters (3); four universal brake industrial casters (3) are installed on the bottom of the platform cabinet (2) via connectors; and the assembly table (1) is installed on the upper surface of the platform cabinet (2) via bolts.

3. The automated inspection equipment for assembly quality of missile compartment parts according to claim 2, characterized in that: The motion device comprises a linear motion mechanism (4) and a six-joint manipulator (5); the linear motion mechanism (4) is fixedly mounted on an assembly table (1); the six-joint manipulator (5) is connected to a connecting slider on the linear motion mechanism (4) via a flange, so that the six-joint manipulator (5) can follow the overall movement of the linear motion mechanism (4).

4. The automated inspection equipment for assembly quality of missile compartment parts according to claim 3, characterized in that: The visual sensing system comprises a detection visual sensing device (6) and a USB computer camera (13); wherein the detection visual sensing device (6) is used to collect image information of the compartment inspection position, and comprises a deep binocular industrial camera (6.1), a monocular industrial camera (6.2), a ring light source (6.3), and a bracket (6.4); the deep binocular industrial camera (6.1), the monocular industrial camera (6.2), and the ring light source (6.3) are assembled and fixedly mounted on the bracket (6.4); the bracket (6.4) is assembled and fixed to the end flange of the six-joint manipulator (5); the USB computer camera (13) is used to collect interactive gesture images of the operator, and its own bracket base is bonded to the control system.

5. The automated inspection equipment for assembly quality of missile compartment parts according to claim 3, characterized in that: The control system comprises a motion controller (7), an industrial control computer (8), and a touch screen panel (9); the touch screen panel (9) and the industrial control computer (8) are connected and communicated via a wired connection; the touch screen panel (9) is used to display a software system interface and obtain touch interaction input instructions; the industrial control computer (8) and the motion controller (7) are installed and fixed in the platform cabinet (2); the touch screen panel (9) is hung on the side of the base platform when not in use, and is placed on the work desk using its own bracket when in use.

6. The automated inspection equipment for assembly quality of missile compartment parts according to claim 1, characterized in that: The management end includes an inspection process model management module, an inspection result management module, a rights management module and a dictionary management module; The inspection process model management is used to manage the established inspection process model and is the data source for the on-site execution end; The inspection result management module is used to manage the inspection results, which include inspection conclusions and inspection images; The authority management module is used to manage the authority of users logged in to the management terminal and the on-site execution terminal, and different authorities correspond to different module usage authorities; The dictionary management module is used to define the standard names of various elements included in the product code and standardize the input of data.

7. The automated inspection equipment for assembly quality of missile compartment parts according to claim 6, characterized in that: The inspection process model consists of three parts: product code, calibration data, and inspection process, which are associated through a structured table; The product code refers to information that determines the status of the product to be inspected, including product model, drawing number, batch, and stage; the calibration data refers to the initial position data of the inspection equipment and the product to be inspected; The inspection process is the inspection process flow and inspection process standards, including process number, process name, component type, process text requirements, standard image template, and inspection location.

8. The automated inspection equipment for assembly quality of missile compartment parts according to claim 6, characterized in that: The login module is used for login management of the on-site execution end. Login user information and product code information of the compartment to be inspected are input into the login module; the login user information is used to verify user permissions, and the product code information is used to retrieve the corresponding inspection process model and inspection results of the management end; The position calibration module is used to obtain the position data of the inspection equipment and the product to be inspected, including calibration data and inspection position data; and push it to the inspection process model of the management end; The standard image template formulation module uses an image annotation tool to annotate the detection elements on a large number of imported qualified assembly quality images; after all images are annotated, an image training tool is used to train assembly quality features to obtain qualified recognition features and form a standard image template.

9. The automated inspection equipment for assembly quality of missile compartment parts according to claim 8, characterized in that: The method for obtaining the position data is as follows: driving and debugging the motion device and the detection visual sensor device (6) to a standard inspection preparation position, obtaining the position coordinates of the motion device and the product position coordinates, forming calibration data and pushing the calibration data of the inspection process model to the management end; driving and debugging the motion device and the detection visual sensor device (6) to an inspection position that meets the process conditions, obtaining the coordinate position of the motion device, and pushing the inspection position data corresponding to the process in the inspection process in the inspection process model to the management end.

10. The automated inspection equipment for assembly quality of missile compartment parts according to claim 8, characterized in that: The inspection module at the on-site execution end will expand the inspection task list each time it is started according to the inspection process update of the inspection process model, display the task completion status, inspection conclusion, inspection time in the list, and display the inspection image result of the selected list item on the interface.

11. The automated inspection equipment for assembly quality of missile compartment parts according to claim 8, characterized in that: The dynamic process data of the on-site execution end all comes from the inspection process model of the management end. When executing model tasks, you only need to create a new inspection process model on the management end and improve the data.

12. A method for automated inspection of the assembly quality of missile compartment parts using the automated inspection equipment according to claim 5, characterized in that: include: S1, establish the inspection process model: when it is used for the first time or the inspection process is changed, edit the inspection process model on the management side; The inspection process model consists of three parts: product code, calibration data, and inspection process, which are linked through a structured table; The product code includes product model, drawing number, batch and stage; Calibration data includes equipment calibration data and product calibration data; inspection process includes process number, process name, component type, process text requirements, standard image template, and inspection location; after completing the S1 inspection process modeling on the management side, access S2, S3, and S4 through the on-site execution side to obtain data and improve the model; S2, conduct initial inspection position teaching: select a standard cabin section (11) of qualified quality and place it on the assembly platform (10), start the automated inspection equipment, drive the linear motion mechanism (4) and the six-joint manipulator (5), drive the inspection visual sensor device (6) to move to the top of the standard cabin section (11) and align it with the cabin section; visually identify and calculate the actual three-dimensional space distance of the standard cabin section (11) relative to the automated inspection equipment; adjust the posture of the standard cabin section (11) until the relative space distance meets the inspection process condition requirements, and set the equipment position and the three-dimensional space distance of the cabin section relative to the equipment at this time as the initial position data of the inspection equipment and the initial position data of the product to be inspected, and push them to the calibration data in the inspection process model of the management end respectively; enter S3, and further obtain the inspection position data in the inspection process model; S3, conduct motion trajectory data teaching: first, for each inspection point required by the process inspection conditions, drive the linear motion mechanism (4) and the six-joint manipulator (5) to the inspection position that meets the process conditions, turn on the inspection visual sensor device (6), the posture of the end of the six-joint manipulator (5) ensures that the inspection visual sensor device (6) can collect all inspection elements, record the coordinate values ​​of the linear motion mechanism (4) and the six-joint manipulator at this time, form the inspection position corresponding to the process, and push the inspection position of the inspection point of the process to the management end process model; repeat S3 operation to complete the motion trajectory teaching of all inspection points; Enter S4 to further obtain data of the standard image template in the inspection process model; S4, create a standard inspection template: The standard image template refers to the standard state template of a qualified product. It is used to compare and analyze the assembly state of the product to be inspected for consistency. It is a standard image template for quality inspection. A qualified product identical to the product to be inspected is used to demonstrate the standard quality of assembly of a certain process required by the process model. Images of multiple products and multi-angles with qualified assembly quality are collected. Image annotation tools and image training tools are used to annotate and train the inspection objects on the images. The training results are pushed to the management end to inspect the standard image template of the process model. S5, Inspection Preparation: Start the inspection equipment, enter the code information of the product to be inspected, and the system automatically retrieves the matching data in the inspection process model to complete the data preparation; Push the missile compartment (12) to be inspected to the inspection station, call the calibration data in the inspection process model and drive the linear motion mechanism (4) and the six-joint manipulator (5) to the initial inspection position, display the image of the compartment to be inspected collected by the monocular industrial camera (6.2) in real time on the touch screen panel (9), and the deep binocular industrial camera (6.1) recognizes and calculates the coordinate deviation between the current position and the product calibration data in the inspection process model, gives a standard position virtual frame and adjustment parameters, adjusts the compartment to be inspected into the virtual frame, and enters S6 after completing the physical inspection preparation; S6, inspection execution: select the process to be inspected, obtain the human-machine interaction command to start the inspection, read the inspection position data in the inspection process model, drive the motion device to the inspection position, automatically turn on the ring light source (6.3) and the monocular industrial camera (6.2) of the inspection visual sensor device (6) to collect images, call the assembly quality image inspection tool to compare the image with the feature data of the standard image template for consistency, and enter S7 after obtaining the consistency comparison result; S7, test results: the consistency comparison results, including the test conclusion and the test image; The inspection results are pushed to the management end; the inspection images include two types: acquisition images and judgment images; the acquisition images are original images captured by the monocular industrial camera (6.2); objects that pass the inspection are marked with a green frame on the judgment image, and objects that fail are marked with a red frame; repeat S6 until all inspection processes are completed, and then enter S8; S8, inspection is completed; push the compartment to the next assembly position.

13. The automated inspection method according to claim 12, wherein: The visual recognition, calculation and display of the actual three-dimensional space distance of the standard cabin section (11) relative to the automated inspection equipment in S2 include: setting a corner point of the assembly table (1) of the automated inspection equipment as the origin O of the world coordinate system, using a deep binocular industrial camera (6.1) of the inspection visual sensor device (6) to identify the center of the cabin section reference hole, deducing and converting based on stereo vision positioning and robot kinematics, calculating the three-dimensional space distance of the front end face of the cabin section relative to the automated inspection equipment, and displaying the standard position of the reference circular hole on the screen with a frame for graphical guidance.

14. The automated inspection method according to claim 12, wherein: The S4 image annotation adds a "tiny" attribute annotation dimension for parts in the cabin that are smaller than 1 cm, and further refines the convolution division for the detection images with "tiny" attributes.

15. The automated inspection method according to claim 12, wherein: The assembly quality image inspection tool in S6 is encapsulated using the YOLO 4 algorithm. For cable path detection, the YOLO 4 algorithm is first used to identify the missile barrel cross-section and cables. Cable color features and a threshold algorithm are then used to filter out other background feature information within the frame to obtain the cable's appearance outline. This outline is then refined to remove redundant features. A skeleton extraction algorithm is used to find the cable's skeleton line, which is then pushed to a feature recognition operator to detect and determine the length and position of the cables in the area and the image.

16. The automated inspection method according to any one of claims 12 to 15, characterized in that: The order of S2, S3, and S4 can be changed, but they must all be completed before entering S5; S1 is completed at the management end, and S2-S8 are completed at the on-site execution end.

Images