Industrial heterogeneous part appearance automatic detection device and use method thereof

Abstract

The invention discloses an industrial heterogeneous part appearance automatic detection device, and the device comprises a camera obscura frame which is used for forming a camera obscura environment for isolating external light; the collaborative robot is arranged in the camera obscura frame, and an image acquisition module is mounted on the end effector; the workpiece bearing and rotating mechanism is arranged in the camera obscura frame and is used for bearing and driving a workpiece to rotate; and the processor is connected with the collaborative robot, the workpiece bearing and rotating mechanism and the image acquisition module, and is configured to control the workpiece bearing and rotating mechanism and the collaborative robot to move cooperatively, so that the image acquisition module performs multi-angle image acquisition on the rotating workpiece, and performs defect identification and classification on the acquired images by adopting a deep learning algorithm. An intelligent analysis system is constructed in the closed camera obscura, so that imaging interference of complex curved surfaces and materials of heterogeneous parts is overcome, high-precision identification and classification of various tiny defects are realized, and the efficiency and reliability of industrial appearance detection are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material detection, and in particular to an industrial heterogeneous component appearance automatic detection device and a use method thereof. BACKGROUND

[0002] In the field of high-end equipment manufacturing and operation, such as gas turbine blades, aero-engine rotors, precision valves and other industrial key components, due to their complex curved surface morphology and heterogeneous characteristics, the surface integrity directly determines the performance, safety and service life of the whole machine. During manufacturing, processing, spraying or long-term operation, these components are prone to complex appearance defects such as micron-level fine scratches, local corrosion thinning (also known as "decreasing meat"), material protrusion or foreign matter attachment.

[0003] For a long time, the industry has mainly relied on traditional manual visual inspection or contact measurement methods for the detection of such defects. The manual visual inspection method is low in efficiency, and the detection results are easily affected by personnel experience, fatigue and subjective judgment, lacking objective and unified quantitative standards, resulting in high missed detection rate and difficult to trace the results. The contact measurement method, such as three-coordinate scanning, not only has low detection efficiency, but also has the risk of damaging the surface of the precision component due to contact, which cannot meet the non-contact, high-precision and high-efficiency requirements of quality control under the modern high-efficiency production rhythm.

[0004] To address the above challenges, automated detection schemes based on machine vision have gradually been applied. Among them, the most similar prior art to the present application is a monocular two-dimensional vision detection system based on fixed parameters. The system is usually composed of an industrial camera, a fixed-angle and brightness light source, and an industrial computer. Its working process is to collect two-dimensional images of the component surface under preset lighting conditions, and then identify defects through traditional image processing algorithms. However, this scheme has significant limitations. First, its optical system has poor adaptability, and the fixed light source configuration cannot adapt to the complex curved surface morphology, various materials and different optical properties of different components, resulting in unstable imaging quality in strong light or deep shadow areas, which is the main reason for false detection and missed detection. Second, its recognition algorithm has limited capacity, and the traditional image processing technology it relies on is insufficient in extracting small defect features with low background contrast and variable morphology, making it difficult to accurately distinguish between real defects and false defects, and even more difficult to achieve automatic classification of multiple defect types. Finally, the overall process has low automation and standardization, and still relies on a large amount of manual intervention from component positioning, image acquisition to result review, which not only is low in efficiency, but also is difficult to ensure the consistency and traceability of the detection results.

[0005] Therefore, the industry urgently needs an appearance defect detection solution that can balance high adaptability, high precision, high intelligence level and full-process automation. SUMMARY

[0006] Based on the above problems, the embodiment of the present application provides an industrial heterogeneous part appearance automatic detection device and a use method thereof, which can realize high-precision identification and classification of various micro defects and improve the efficiency and reliability of industrial appearance detection.

[0007] In a first aspect, the embodiment of the present application provides an industrial heterogeneous part appearance automatic detection device, which comprises: A dark box frame is used to form a dark box environment isolated from external light. A collaborative robot is arranged in the dark box frame, and an image acquisition module is installed on the end effector. A workpiece bearing and rotating mechanism is arranged in the dark box frame and is used to bear and drive the workpiece to rotate. A processor is connected with the collaborative robot, the workpiece bearing and rotating mechanism, and the image acquisition module, and is configured to control the workpiece bearing and rotating mechanism and the collaborative robot to move cooperatively, so that the image acquisition module collects multi-angle images of the rotating workpiece, and a deep learning algorithm is used to identify and classify defects of the collected images.

[0008] In a possible implementation, the dark box frame is a one-piece sheet metal sealing structure, and a plurality of sheet metal doors that can be opened and closed are arranged.

[0009] In a possible implementation, the device further comprises: A planar array vision camera is arranged above the inside of the dark box frame and is used to position the workpiece on the workpiece bearing and rotating mechanism again, and provide a basis for the processor to adjust the motion parameters of the collaborative robot.

[0010] In a possible implementation, the workpiece bearing and rotating mechanism comprises: A hollow rotating platform is arranged with a workpiece tray, and the hollow rotating platform is used to directly bear and drive the workpiece tray to rotate; A clamping cylinder is installed on the hollow rotating platform or a rack adjacent to the hollow rotating platform, and is used to clamp and fix the workpiece tray during detection.

[0011] In a possible implementation, the image acquisition module is a line scan camera.

[0012] In a possible implementation, the device further comprises: In a possible implementation, the feeding mechanism uses a chain wheel roller conveying method to feed the workpiece into the inside of the dark box frame.

[0013] In a second aspect, the embodiment of the present application provides a use method of an industrial heterogeneous part appearance automatic detection device, which comprises: Feeding the workpiece into the dark box environment and positioning the workpiece on the workpiece bearing and rotating mechanism; The workpiece carrying and rotating mechanism drives the workpiece to rotate, and a collaborative robot is controlled to drive an image acquisition module to move synchronously to acquire multi-angle images of the surface of the workpiece. The collected images are processed by using a deep learning algorithm in the processor to realize automatic identification and classification of defects.

[0014] In a possible implementation, after the workpiece is sent into the dark box environment and positioned on the workpiece carrying and rotating mechanism, before the workpiece carrying and rotating mechanism drives the workpiece to rotate and a collaborative robot is controlled to drive an image acquisition module to move synchronously to acquire multi-angle images of the surface of the workpiece, the method further comprises: The industrial heterogeneous component appearance automatic detection device further comprises a planar array vision camera arranged above the inside of the dark box frame, used to position the workpiece on the workpiece carrying and rotating mechanism again, and provide a basis for adjusting the motion parameters of the collaborative robot by the processor; the workpiece positioned on the workpiece carrying and rotating mechanism is positioned again by the planar array vision camera. According to the result of the second positioning, the motion parameters of the collaborative robot are adjusted.

[0015] The technical scheme provided by some embodiments of the present application has at least the following beneficial effects: an integrated light-shielded dark box is designed to provide a stable optical environment for detection; a collaborative robot is introduced to carry a line-scan camera, and cooperates with a workpiece table that can rotate precisely to realize flexible image acquisition without dead angles on a complex curved surface, and solve the imaging adaptability problem; advanced deep learning algorithms are integrated to replace traditional image processing, and the system is endowed with the ability to learn defect features autonomously from massive data, thereby realizing high-precision identification and intelligent classification of various complex and tiny defects; finally, by integrating feeding, positioning, rotation, scanning, analysis and result output into a set of fully automatic and coherent processes, manual intervention is minimized, thereby improving the detection efficiency to the level of online full inspection, ensuring the standardization of the operation process and the traceability of the detection results, and providing high-reliability technical support for industrial quality control. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0017] Figure 1 A structural schematic diagram of an industrial heterogeneous component appearance automatic detection device provided by the embodiments of the present application is shown in the figure. Fig. 2(a) is a perspective view of an industrial heterogeneous parts appearance automatic detection device according to an embodiment of the present application; Fig. 2(b) is a front view of an industrial heterogeneous parts appearance automatic detection device according to an embodiment of the present application; Fig. 2(c) is a schematic diagram of the internal structure of an industrial heterogeneous parts appearance automatic detection device according to an embodiment of the present application; Fig. 3(a) is a front view of a 2D shell frame partial structure according to an embodiment of the present application; Fig. 3(b) is a side view of a 2D shell frame partial structure according to an embodiment of the present application; Figure 4 Fig. 4(a) is a schematic diagram of an ABB collaborative robot partial structure according to an embodiment of the present application; Figure 5 Fig. 5(a) is a schematic diagram of a feeding track partial structure according to an embodiment of the present application; Figure 6 Fig. 6(a) is a schematic diagram of a dark box internal structure according to an embodiment of the present application; Figure 7 Fig. 7(a) is a schematic diagram of an electrical partial structure according to an embodiment of the present application; Figure 8 Fig. 8 is a flowchart of a method for using an industrial heterogeneous parts appearance automatic detection device according to an embodiment of the present application; Figure 9 Fig. 9 is a timing flowchart of an industrial heterogeneous parts appearance automatic detection device according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the features and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0019] The following description refers to the accompanying drawings. Unless otherwise indicated, same or similar elements in different drawings are denoted by same or similar reference numerals. The embodiments described in the following exemplary embodiments do not represent all the embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0020] In the description of the present application, it is understood that the terms "first", "second" and the like are only for the purpose of description and cannot be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances. In addition, in the description of the present application, "a plurality of" means two or more, unless otherwise specified. The association relationship of the associated objects is described, which means that there can be three kinds of relationships, for example, A and / or B can represent three cases: A exists alone, A and B exist together, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects before and after it.

[0021] As described above, in the field of high-end equipment manufacturing and operation, such as gas turbine blades, aircraft engine rotors, precision valves and other key components, they are collectively referred to as "heterogeneous components" because they usually have complex geometric curved surfaces, diverse material properties (such as metal, ceramic or composite materials) and different surface treatment processes (such as spraying and polishing). The surface integrity of these components, that is, whether there are scratches, pits, corrosion thinning (also known as "de-meat"), material protrusions or foreign matter adhesion and other microscopic defects, directly determines the aerodynamic performance, structural strength, safety and service life of the whole machine.

[0022] Traditional detection methods mainly rely on two paths: one is manual visual inspection, which not only has low efficiency and is difficult to meet the online full inspection rhythm of modern industry, but also its detection results are seriously dependent on the experience and mental state of the operator at the time, resulting in strong subjectivity of the results, high missed detection rate, and lack of unified, objective quantitative standards and traceable data records. The second is contact measurement, such as three-coordinate measuring machine, although the accuracy is high, but the scanning efficiency is very low, and there is a risk of scratching or contaminating the surface of the precision component due to the contact of the measuring head, which cannot be applied to large-scale production scenes.

[0023] To seek an automated solution, machine vision-based detection technology has emerged. Among them, the most representative existing technology is a fixed parameter-based monocular two-dimensional vision system. The system is usually composed of a fixedly installed industrial area array camera, a set of fixed-angle and fixed-brightness light source, and an industrial computer executing traditional image processing algorithms. However, this scheme has three fatal defects in practical application: first, the optical system has poor adaptability. The fixed light source configuration cannot adapt to the complex curvature of different parts, the light reflection characteristics and color changes, resulting in overexposure in strong light reflection areas and shadows in deep grooves, and the imaging quality is extremely unstable, which is the root cause of a large number of false positives and false negatives. Second, the defect recognition algorithm has low precision and insufficient intelligence. Its dependence on edge detection, threshold segmentation and other traditional algorithms has weak feature extraction capability for low-contrast, morphologically variable and micron-level defects (such as superficial scratches and gradual meat reduction), and it is difficult to accurately distinguish between real defects and oil stains, water stains and other pseudo-defects, and it is even more difficult to achieve automatic classification of defect types. Finally, the degree of automation and standardization is low. The entire detection process, including part feeding, positioning, multiple angle shooting and result review, still requires a large amount of manual intervention, and cannot form a closed-loop automated process, resulting in limited detection efficiency and inability to guarantee the consistency and traceability of the results.

[0024] In view of this, the present application provides an industrial heterogeneous part appearance automatic detection device and its use method. The present application provides a stable optical environment for detection by designing an integrated light-shielded dark box; by introducing a collaborative robot carrying a line scan camera and cooperating with a precisely rotatable workpiece table, flexible image acquisition without dead angles for complex curved surfaces is achieved, solving the imaging adaptability problem; by integrating advanced deep learning algorithms to replace traditional image processing, the system is given the ability to learn defect features from massive data autonomously, thereby achieving high-precision identification and intelligent classification of various complex and tiny defects; finally, by integrating feeding, positioning, rotation, scanning, analysis and result output into a set of fully automated and coherent process, human intervention is minimized, thereby improving detection efficiency to the level of online full inspection, ensuring the standardization of the operation process and the traceability of the detection results, and providing high-reliability technical support for industrial quality control.

[0025] Please refer to Figure 1 , Figure 1 An industrial heterogeneous part appearance automatic detection device structure schematic diagram is provided for the embodiments of the present application. As shown in Figure 1As shown, the industrial heterogeneous part appearance automatic detection device 100 comprises: a dark box frame 110, a collaborative robot 120, a workpiece bearing and rotating mechanism 130, and a processor 140; wherein: the dark box frame 110 is used to form a dark box environment isolated from external light; the collaborative robot 120 is arranged in the dark box frame, and an image acquisition module 1201 is installed on the end effector; the workpiece bearing and rotating mechanism 130 is arranged in the dark box frame and is used to bear and drive the workpiece to rotate; and the processor 140 is connected with the collaborative robot, the workpiece bearing and rotating mechanism, and the image acquisition module, is configured to control the workpiece bearing and rotating mechanism to move cooperatively with the collaborative robot, so that the image acquisition module collects multi-angle images of the workpiece in rotation, and uses a deep learning algorithm to identify and classify defects of the collected images. In a possible implementation, the dark box frame 110 is a one-piece metal sealing structure and is provided with a plurality of openable and closable metal doors. Optionally, the dark box frame 110 can be further made of lightweight high-strength materials such as aluminum alloy to realize the lightweight of the device, facilitate transportation, installation, and deployment and application in specific locations (such as an upper workshop or a mobile detection platform) under the premise of ensuring the structural rigidity and light-shielding performance. The industrial heterogeneous part appearance automatic detection device 100 further comprises: a plane array vision camera arranged inside and above the dark box frame, used to reposition the workpiece on the workpiece bearing and rotating mechanism and provide a basis for the processor to adjust the motion parameters of the collaborative robot. The workpiece bearing and rotating mechanism 130 comprises: a hollow rotating platform provided with a workpiece tray, the hollow rotating platform being used to directly bear and drive the workpiece tray to rotate; and a clamping air cylinder installed on the hollow rotating platform or a rack adjacent to the hollow rotating platform, used to clamp and fix the workpiece tray during detection. The image acquisition module 1201 is a line scan camera. The industrial heterogeneous part appearance automatic detection device 100 further comprises: an inlet mechanism, which uses a chain wheel roller conveying mode to send the workpiece into the dark box frame.

[0026] Please refer to Figures 2(A) to 2(C), which are exemplary schematic diagrams from multiple perspectives of an automatic inspection device for the appearance of heterogeneous industrial parts provided in this application embodiment. As shown in Figures 2(A) to 2(C), the automatic inspection device for the appearance of heterogeneous industrial parts includes a 2D shell frame 1, an ABB collaborative robot 2, a feeding track 3, a dark box interior 4, and an electrical component 5. Specifically, the main structure of the 2D shell frame 1 is formed by an integrated sheet metal process using 2.5mm thick Q235 steel plate, which not only ensures the overall mechanical strength, but its fully enclosed design also provides the necessary dark box environment for visual inspection, effectively isolating external stray light interference. The four sides of the frame are equipped with openable sheet metal doors, which facilitates equipment maintenance, component debugging, and process monitoring. Figure 2(B) is an exemplary front view of the automatic inspection device for the appearance of heterogeneous industrial parts, while Figures 2(A) and 2(C) are exemplary schematic diagrams of the automatic inspection device for the appearance of heterogeneous industrial parts from different perspectives. Please refer to Figures 3(A) and 3(B), which are exemplary schematic diagrams of a 2D shell frame structure provided in the embodiments of this application. As shown in Figure 3(A), the 2D shell frame includes a tri-color light 1, a sheet metal door 2, a light-blocking curtain 3, a square tube frame 4, a cooling fan 5, casters 6, a hanging ring 7, a load-bearing foot cup 8, a main switch 9, and a door handle 10. As shown in Figure 3(B), the 2D shell frame 1 also includes a touch screen 11, operation buttons 12, a surface array vision device 13, a display 14, and a sheet metal keyboard 15. Because the inspection of heterogeneous components needs to be carried out in a dark box to ensure the isolation from other light pollution, 5050 square steel pipes are used as the frame, and sheet metal is welded to the outer surface to form a dark box. For easy debugging, multiple sheet metal doors are opened on all four sides. At the same time, for easy transportation, casters and load-bearing feet are installed at the bottom of the frame. To further block ambient light, a light-shielding curtain is installed at the inlet. For easy operation, one side is equipped with a touch screen, display, operation buttons and keyboard, which facilitates both control and debugging. The area array vision at the top of the frame is mainly used for secondary positioning of heterogeneous components inside the dark box to adjust the motion parameters of the collaborative robot.

[0027] The core of the ABB collaborative robot utilizes a high-precision six-axis ABB collaborative robot. Leveraging its flexible, multi-degree-of-freedom motion characteristics, and equipped with a high-resolution line-scan camera mounted on a custom bracket on the end effector, it can accurately track complex curved surface contours, achieving multi-angle, dead-end-free image data acquisition. Please refer to [link / reference]. Figure 4 , Figure 4 This is an exemplary schematic diagram of a partial structure of an ABB collaborative robot provided in an embodiment of this application, such as... Figure 4As shown, the ABB collaborative robot includes a line-scan camera bracket 1, a line-scan camera 2, a line-scan coaxial light source 3, an ABB collaborative robot 4, a height pad 5, and a collaborative robot base 6. In order to better use the motion range of the collaborative robot, the robot base is added to make the bottom of the robot as flat as possible with the workpiece tray, and the height of the robot bottom is adjusted by processing height pads of different heights to ensure that the robot motion process is not limited; since the main way to detect heterogeneous parts is 2D vision, the end of the robot needs to carry a camera bracket, a camera, and a light source; thanks to the flexibility of the collaborative robot, the acquisition path of the robot is further programmed through the contour mapping of the heterogeneous parts, precise acquisition is realized, and some scattered surfaces on the heterogeneous parts will also become a plane through the acquisition of the line-scan camera, thereby better detecting surface defects.

[0028] The feeding track mainly transmits the tray workpiece into the dark box interior part through the chain wheel roller way; please refer to Figure 5 , Figure 5 An example of a feeding track part provided by the embodiment of the application is shown in Figure 5 As shown, the feeding track includes a peripheral sheet metal 1, a positioning cylinder 2, a square through bracket 3, a reflection sensor 4, a stainless steel roller 5, a limiting stop edge 6, an adjusting foot cup 7, and a workpiece tray 8. In the track design, an intensive layout shared by feeding and feeding is adopted, which effectively saves equipment space and structural complexity. In view of the characteristics of the heavy weight of the heterogeneous parts, the track selects a stainless steel roller with excellent bearing capacity and high structural strength to ensure the stability of the conveying process and the durability of long-term operation. In addition, the peripheral sheet metal protective cover of the feeding track not only plays a safety protection role, but also integrates the simple industrial design style with the overall appearance of the equipment, significantly improving the coordination and professionalism of the equipment.

[0029] The dark box interior mainly receives the tray workpiece transmitted by the feeding track, and moves the tray workpiece to the specified position to cooperate with the collaborative robot to scan the workpiece through rotation. Please refer to Figure 6 , Figure 6 An example of a dark box interior provided by the embodiment of the application is shown in Figure 6As shown, the dark box interior includes linear guide pair 1, ball screw 2, slide bracket 3, clamping cylinder 4, hollow rotating platform 5, belt conveying line 6, workpiece tray 7, heterogeneous parts 8 and heavy load plate 9. After the heterogeneous parts and the workpiece tray enter the dark box interior from the feeding track, they will first be connected with the conveying line inside the dark box. After the whole body is in place, the slide motor starts to drive the slide to slide. After reaching the specified position, the clamping cylinder acts to position and clamp the workpiece tray. Then the area array vision camera at the top of the dark box interior (i.e. the area array vision 13 of the 2D housing frame part) will perform a secondary positioning on the heterogeneous parts. Then the hollow rotating platform will start to cooperate with the collaborative robot to shoot and collect data according to the instructions sent by the electrical part PLC. The collected data will be sent to the internal industrial computer for deep learning image processing, and finally the results will be fed back to the display.

[0030] The electrical part mainly has a centralized layout of electrical parts, and integrates a robot controller, a motion control card, a light source controller and a safety protection circuit, to provide stable and reliable power and control protection for the whole detection process. Please refer to Figure 7 , Figure 7 An exemplary schematic diagram of an electrical part provided by the embodiment of the present application is shown in Figure 7 As shown, the electrical part includes a galvanized base plate 1, a wire slot 2, an air switch 3, a Siemens PLC 4, a servo driver 5, a contactor 6, a terminal block 7, a switching power supply 8 and an intermediate relay 9. All electrical components are on the galvanized base plate, and all electronic components are connected through the wire slot for interaction.

[0031] The embodiment of the present application provides an automatic appearance detection device for industrial heterogeneous parts. A one-piece light-shielded dark box is designed to provide a stable optical environment for detection. A collaborative robot is introduced to carry a line scanning camera, and cooperates with a workpiece table that can be precisely rotated to realize flexible image collection without dead angles on complex curved surfaces, and solve the imaging adaptability problem. Advanced deep learning algorithms are integrated to replace traditional image processing, and the system is given the ability to learn defect features from massive data autonomously, thereby realizing high-precision recognition and intelligent classification of various complex and tiny defects. Finally, by integrating feeding, positioning, rotation, scanning, analysis and result output into a set of fully automatic and coherent processes, manual intervention is minimized, thereby improving detection efficiency to the level of online full inspection, ensuring the standardization of the operation process and the traceability of the detection results, and providing high-reliability technical support for industrial quality control.

[0032] Please refer to Figure 8 , Figure 8 A flowchart of a use method of an automatic appearance detection device for industrial heterogeneous parts provided by the embodiment of the present application is shown in

[0033] AsFigure 8 As shown, the method for using the industrial heterogeneous component appearance automatic detection device can at least include: S801, the workpiece is sent into the dark box environment and positioned on the workpiece bearing and rotating mechanism.

[0034] Specifically, the workpiece is first placed on a special workpiece tray, and then carried by the chain wheel cylinder conveying line of the feeding mechanism, and starts to be conveyed into the dark box. When the tray travels to the dark box entrance, it will pass through a light curtain, which effectively isolates the interference of external environmental light, ensuring the purity and stability of the internal imaging environment. After the tray enters the dark box, it is continued to be carried by the internal belt conveying line until it reaches the preset docking station. At this time, the through sensor detects the tray positioning signal, and the slide table motor is started to drive the precision slide table composed of ball screws and linear guide pairs to move forward, carrying the tray and the workpiece on it to the center area of the hollow rotating platform as the core of the workpiece bearing and rotating mechanism. After reaching, the clamping cylinder immediately acts to firmly clamp and fix the workpiece tray from the side or bottom, completing the preliminary mechanical positioning of the workpiece. Then, in order to compensate for the possible small positional deviation of the workpiece on the tray and lay the foundation for the subsequent accurate scanning of the robot, the area vision camera installed at the top of the dark box is activated to perform secondary visual positioning on the clamped workpiece; the deviation between the current actual position and the theoretical position of the workpiece is accurately calculated through image recognition technology, and the deviation data is sent to the processor, which immediately calculates the compensation amount according to the data and pre-adjusts the scanning motion trajectory parameters of the collaborative robot with high precision.

[0035] Please refer to Figure 9 , Figure 9 The timing flow chart of the industrial heterogeneous component appearance automatic detection device provided by the embodiment of the application is as shown in Figure 9 The workpiece is first placed on a special tray, the core controller sends the slide rail reset or receiving instruction to the precision slide table, the chain wheel cylinder conveying line of the feeding mechanism is conveyed to the dark box, and after passing through the dark box entrance light curtain to isolate external stray light, the internal belt conveying line of the dark box is received to the preset docking station; At this time, the through sensor detects the tray positioning signal and feeds back to the core controller, which immediately drives the precision slide table to move the tray and the workpiece to the center area of the hollow rotating platform, and the clamping cylinder is started to realize the preliminary mechanical fixation of the tray after reaching the initial positioning.

[0036] In one possible implementation, after the workpiece is transported by the feeding mechanism and preliminarily fixed on the hollow rotating platform by the clamping cylinder, the core controller triggers the area array image acquisition instruction, the fixed area array vision camera installed at the top end inside the dark box frame is triggered to perform rapid image acquisition on the workpiece in the stable clamping state, and the position deviation data is returned to the core controller. The core purpose of this shooting is to accurately compensate the small position and angle deviations of the workpiece on the tray, which are caused by the placement error of the previous process or the slight movement in the conveying process. The system compares and matches the collected workpiece image with the pre-stored standard template through the real-time running image recognition algorithm, so as to calculate the offset amount between the current actual coordinate system of the workpiece and the theoretical coordinate system used for robot motion programming in the X, Y and rotation angle Rz directions. The offset amount is sent to the processor of the system as a key compensation parameter; the core controller (such as PLC or industrial computer) in the processor calculates and updates all scanning path point coordinates of the collaborative robot in real time according to the data through the preset kinematics model and coordinate transformation matrix. After the adjustment, the original motion program of the robot is given an adaptive compensation value, which ensures that the linear scanning camera carried at the end of the robot can ignore the initial placement deviation and always accurately move along the equidistant trajectory of the real contour of the workpiece in the subsequent collaborative scanning, thereby improving the accuracy.

[0037] S802, control the workpiece carrying and rotating mechanism to drive the workpiece to rotate, and synchronously control the collaborative robot to drive the image acquisition module to move, so as to perform multi-angle image acquisition on the surface of the workpiece.

[0038] Specifically, after the workpiece completes accurate positioning, the core detection process of the system is started. Please continue to refer to Figure 9 For example, Figure 9As shown, the core controller (with PLC as the core) sends synchronous motion instructions to the workpiece bearing and rotating mechanism and the collaborative robot: the hollow rotating platform starts to rotate at a preset angular velocity matching the image acquisition frame rate under the precise control of the servo driver, thereby driving the workpiece on it to continuously and smoothly display its circumferential surface. At almost the same time, the six-axis collaborative robot starts to move according to the scan trajectory that has been subjected to pose compensation. In this process, the motion of the robot and the action of the rotating platform are strictly synchronized to form a coordinated motion system. It ensures that the linear scanning camera mounted at the end keeps a constant optimal imaging distance and angle with the workpiece measured points, while ensuring that the continuous triggering of the linear scanning and the exposure of the workpiece surface are perfectly matched in time sequence. Specifically, when the workpiece rotates, its complex curved surface (such as the twisted curved surface of the blade) will enter the field of view of the camera one by one. Through the cooperation of "rotary display" and "robot trajectory tracking", the linear scanning camera can collect the complex three-dimensional curved surface profile of the workpiece in the form of continuous and non-overlapping image strips. These image strips are real-time spliced into a complete and "unfolded" two-dimensional plane image in software, so that the defects (such as scratches across the ridge line) that may be blocked or distorted by the curved surface itself in the two-dimensional view are clearly presented on a plane. This coordinated motion and imaging method overcomes the imaging blind area and distortion problem caused by the single view angle and the curved surface blocking of the fixed camera system, and provides the most critical data basis for realizing high-consistency appearance defect detection.

[0039] Optionally, a seventh axis can be added to the collaborative robot, and by introducing external axis structures such as linear guides or ground rails, the working range of the robot can be further expanded, enabling it to cover larger-sized parts or implement multi-station detection, thereby improving the overall flexibility and applicability of the system. On the other hand, the image acquisition module can be changed to a 3D camera, which can not only identify two-dimensional appearance defects, but also detect the depth information of defects such as pits and protrusions, and realize more comprehensive three-dimensional topography analysis and quantitative evaluation.

[0040] S803, processing the collected images using a deep learning algorithm in the processor to realize automatic identification and classification of defects.

[0041] Specifically, after completing the full-range image acquisition of the workpiece surface through the coordinated motion system, the system enters the intelligent analysis and decision-making phase. Please continue to refer to Figure 9 For example, Figure 9As shown, all the continuous image sequences captured by the line-scan camera are transmitted in real time to a core controller integrated within the processor for processing. The core controller first performs a series of standardized pre-processing operations on the images, including noise filtering, illumination uniformity correction, and image enhancement, to eliminate interference that can be introduced during the imaging process and ensure the quality of the input data. Subsequently, the pre-processed images are fed into a pre-trained deep learning convolutional neural network model for inference. The network model has the ability to learn and extract subtle defect features autonomously, rather than relying on fixed rules set by humans, through training on a vast sample library containing various defect types. The network performs pixel-by-pixel analysis on the input images, not only detecting micron-level scratches, pits, or bumps that are difficult for the human eye to discover with high sensitivity, but also accurately distinguishing the categories of these defects according to their characteristic patterns and intelligently excluding false defect interference caused by stains, water marks, or slight reflections. Finally, the system outputs the identified real defects along with information such as their types, sizes, and position coordinates in a structured manner, displays them on the operation interface (e.g., a display) in real time, and generates a detection report, thereby completely realizing the full-process automation of intelligent diagnosis from image acquisition to defect recognition and classification, completely replacing the traditional reliance on human experience in the judgment process, and ensuring the objectivity, accuracy, and traceability of the detection results.

[0042] Optionally, the detection system of the present method supports parallel detection of multiple devices, and through a unified scheduling system, multiple detection stations are connected in series or in parallel to form an efficient detection assembly line, which can simultaneously process multiple workpieces or different parts of a workpiece in parallel, thereby greatly improving the overall detection efficiency and productivity.

[0043] The embodiment of the present application provides a use method of an industrial heterogeneous part appearance automatic detection device. The method first completely isolates external stray light interference through a closed dark box environment, providing stable and controllable imaging basic conditions for visual detection. Then, through the precise cooperative motion of the collaborative robot and the rotary workbench, multi-angle and dead-angle-free image acquisition of complex curved surface parts is realized, effectively solving the imaging quality problem caused by curved surface reflection and shielding of traditional fixed visual systems. More importantly, the method introduces an intelligent recognition algorithm based on deep learning, which can automatically and accurately identify and classify various defect types such as scratches, pits, and bumps, greatly improving the accuracy and intelligence level of detection. At the same time, from automatic workpiece feeding, accurate positioning to complete process automation of cooperative scanning and intelligent analysis, not only the detection efficiency is improved to the level that can meet the requirements of online full inspection, but also the consistency and traceability of the detection results are ensured through standardized operation processes, providing high-reliability technical support for industrial quality control.

[0044] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium. When the program is executed, the program can include the processes of the above-mentioned embodiment methods. The aforementioned storage medium includes read only memory (ROM), random access memory (RAM), magnetic disk or optical disk and various storage media that can store program codes. In the case of no conflict, the technical features in the embodiments and the implementation forms can be combined arbitrarily.

[0045] The above description of disclosed embodiments enables those skilled in the art to implement or use the present application. It will be obvious to those skilled in the art that various modifications can be made to the embodiments, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An automatic appearance inspection device for industrial heterogeneous parts, characterized in that, The device includes: A darkroom frame, which is used to form a darkroom environment that isolates external light; A collaborative robot is set up inside the dark box frame, and an image acquisition module is installed on the end effector; The workpiece bearing and rotation mechanism is set inside the dark box frame and is used to bear and drive the workpiece to rotate; The processor, connected to the collaborative robot, the workpiece carrying and rotating mechanism, and the image acquisition module, is configured to control the workpiece carrying and rotating mechanism to move in coordination with the collaborative robot, enabling the image acquisition module to acquire multi-angle images of the rotating workpiece, and to use a deep learning algorithm to identify and classify defects in the acquired images.

2. The apparatus as claimed in claim 1, characterized in that, The dark box frame is an integrated sheet metal sealed structure and is equipped with multiple openable sheet metal doors.

3. The apparatus as described in claim 1, characterized in that, The device further includes: An area array vision camera is installed inside the upper part of the dark box frame to perform secondary positioning of the workpiece on the workpiece bearing and rotation mechanism, and to provide a basis for the processor to adjust the motion parameters of the collaborative robot.

4. The apparatus as claimed in claim 1, characterized in that, The workpiece bearing and rotation mechanism includes: A hollow rotating platform, wherein a workpiece tray is provided on the hollow rotating platform, and the hollow rotating platform is used to directly support and drive the workpiece tray to rotate; A clamping cylinder is installed on the hollow rotating platform or on a frame adjacent to the hollow rotating platform, and is used to clamp and fix the workpiece tray during inspection.

5. The apparatus as claimed in claim 1, characterized in that, The image acquisition module is a line scan camera.

6. The apparatus as claimed in claim 1, characterized in that, The device further includes: The feeding mechanism uses a sprocket and roller conveyor to feed the workpiece into the dark box frame.

7. A method of using an automatic appearance inspection device for industrial heterogeneous parts, characterized in that, The method is applicable to the apparatus according to any one of claims 1 to 6, and the method comprises: The workpiece is fed into the dark chamber environment and positioned on the workpiece bearing and rotating mechanism; The workpiece bearing and rotation mechanism is controlled to drive the workpiece to rotate, and the collaborative robot is simultaneously controlled to drive the image acquisition module to move, so as to acquire multi-angle images of the workpiece surface. The acquired images are processed using deep learning algorithms in the processor to achieve automatic defect identification and classification.

8. The method as described in claim 7, characterized in that, The method, which includes, after sending the workpiece into the dark chamber environment and positioning it on the workpiece carrying and rotating mechanism, controlling the workpiece carrying and rotating mechanism to drive the workpiece to rotate, and simultaneously controlling the collaborative robot to move the image acquisition module to acquire multi-angle images of the workpiece surface, further includes: The automatic inspection device for the appearance of heterogeneous industrial parts also includes an area array vision camera, which is located inside and above the dark box frame. It is used to perform secondary positioning of the workpiece on the workpiece bearing and rotation mechanism, and to provide a basis for the processor to adjust the motion parameters of the collaborative robot. The area array vision camera performs secondary positioning of the workpiece already positioned on the workpiece bearing and rotation mechanism. Based on the results of the secondary positioning, the motion parameters of the collaborative robot are adjusted.

Images