Machine tool accessory automatic adjustment control system based on visual identification

Abstract

The invention discloses a machine tool accessory automatic adjusting control system based on visual recognition, and relates to the technical field of automatic adjusting systems.The machine tool accessory automatic adjusting control system comprises a machine tool body, a multispectral visual module and three-axis driving mechanisms, and the three-axis driving mechanisms are fixedly installed on the periphery of the outer wall of the machine tool body; and a multispectral vision module is fixedly mounted on the outer wall of the three-axis driving mechanism. According to the invention, through the multispectral vision module, fitting multidimensional information fusion is realized, the problem of incomplete information acquisition caused by vision shielding is solved, the accuracy and reliability of machine tool fitting recognition are improved, and richer and more accurate data support is provided for subsequent adjustment and control.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic adjustment systems, and particularly to an automatic adjustment control system for machine tool accessories based on visual recognition. Background Art

[0002] Intelligent manufacturing emphasizes the coordinated development of upstream and downstream enterprises in the industrial chain. Building an intelligent factory requires automation and intelligence in all production links. With the continuous increase in labor costs, enterprises are urgently in need of automation technology to reduce costs, improve production efficiency and product consistency. Machine vision has developed from simple image processing in the early stage to a stage of deep integration with technologies such as artificial intelligence and deep learning;

[0003] However, due to the influence of the complex environment of machine tool processing, the vision system may misidentify objects or patterns in the background as the features of machine tool accessories, resulting in misidentification or reduced accuracy.

[0004] Therefore, it is very necessary for this application to propose an automatic adjustment control system for machine tool accessories based on visual recognition to solve the problems in the background.

[0005] Patent document CN118545613B discloses a hopper automatic initial positioning system based on visual recognition. The above patent realizes analyzing the image information obtained by the vision unit through the processing unit, and the driver can obtain accurate dimension information and rotation angle when the cold storage hopper and the cloth hopper are docked. It has the advantages of fast and accurate positioning. Through the spatial audio unit cooperating with the display module, the visual and auditory intuitiveness of the driver can be enhanced, making the operation of the tower crane more simple and intuitive.

[0006] To sum up, the above patent analyzes the image information obtained by the vision unit through the processing unit and is equipped with a display module. However, the feature recognition of machine tool accessories often leads to processing errors due to complex environment and normal wear and tear, resulting in the inability to obtain comprehensive reference information;

[0007] Therefore, this application proposes an automatic adjustment control system for machine tool accessories based on visual recognition that collects and calculates through multi-dimensional spectral vision fusion, combines temperature measurement to estimate the loss amount and three-dimensional penetration, and more comprehensively identifies and establishes the position information of machine tool accessories, so as to perform adjustment. Summary of the Invention

[0008] The purpose of the present invention is to provide an automatic adjustment control system for machine tool accessories based on visual recognition, so as to solve the technical problem that under the influence of the complex environment of machine tool processing, the vision system may misidentify objects or patterns in the background as the features of machine tool accessories, resulting in misidentification or reduced accuracy as mentioned in the above background art.

[0009] To achieve the above object, the present invention provides the following technical solution: An automatic adjustment control system for machine tool accessories based on visual recognition, including a machine tool body, a multispectral vision module, and a three-axis drive mechanism. The three-axis drive mechanism is fixedly installed around the outer wall of the machine tool body, and the multispectral vision module is fixedly installed on the outer wall of the three-axis drive mechanism;

[0010] The multispectral vision module includes an annular light supplement unit and a dual-camera array. The dual-camera array consists of a global camera and a microscopic camera. The global camera is fixedly installed on the gantry at the top of the machine tool body, and the microscopic camera is fixedly installed on the slide table on the side of the spindle. The annular light supplement unit is connected to the control center through a control line. The global camera is connected to the data processing unit and the control center through a data line. The microscopic camera transmits image data to the data processing unit through a data line.

[0011] Preferably, guide rails are fixedly installed around the machine tool body. A lifting column is fixedly installed on the side of the guide rail. A polarized industrial camera is fixedly installed at the top of the lifting column. The height of the industrial camera is aligned with the center line of the spindle of the machine tool body. The industrial camera and the dual-camera array are respectively connected to an optical fiber switch through a twisted pair. An annular light supplement light source is configured at the top of the industrial camera. The annular light supplement unit is provided with 12 groups of LED lights controlled by independent data lines. An adjustable filter is fixedly installed between the lens and the image sensor of the industrial camera. The adjustable filter switches visible light and near-infrared based on the Fabry-Perot interference principle;

[0012] Fiber Bragg gratings sensors are fixedly installed in the axial and circumferential directions of the spindle, on the length direction of the guide rail, and on the side of the outer wall of the machine tool body. The fiber Bragg gratings sensors are connected to a host computer through optical fibers.

[0013] Preferably, a three-dimensional laser module is arranged in parallel below the cross beam of the machine tool body. The three-dimensional laser module moves synchronously with the X-axis. The three-dimensional laser module is connected to an edge computing node through an optical fiber. The three-dimensional laser module forms a 45° angle with the optical axis of the industrial camera;

[0014] Microbolometers are embedded on both sides of the spindle box installed on the side of the machine tool body. A transfer board is arranged in the optical path between the industrial camera and the three-dimensional laser module. A beam splitter prism is fixedly installed at the top of the transfer board. A synchronous signal source is arranged inside the electrical control cabinet. There are two output interfaces inside the synchronous signal source. The output interfaces respectively control the trigger pulses of the microbolometers and the three-dimensional laser module through a signal connection circuit. The microbolometers collect thermal imaging data and are connected to an electromagnetic actuator to generate thermal deformation compensation. The microbolometers are connected to the three-dimensional laser module, and the three-dimensional laser module is connected to the multispectral vision module. The microbolometers, the three-dimensional laser module, and the multispectral vision module together form an optical path acquisition component.

[0015] Preferably, the three-axis drive mechanism includes a closed-loop drive assembly and a Y / Z-axis drive assembly. The closed-loop drive assembly consists of an X-axis linear motor and a grating scale. The ball screw of the Y / Z-axis drive assembly is connected to a harmonic reducer. A piezoelectric ceramic micro-adjuster is fixedly installed at the end of the ball screw. The visual marking points adopt reflective ceramic balls. The driver arranged inside the piezoelectric ceramic micro-adjuster is connected in series with the ball screw.

[0016] Optical fiber gyroscopes are fixedly installed at the four corners of the base of the machine tool body. A magnetic levitation bearing is fixedly installed at the bottom end of the outer wall of the rotating platform. An annular encoder is arranged in the middle of the magnetic levitation bearing. 8 groups of visual marking points are equidistantly arranged on the edge of the tabletop of the rotating platform. A rotating shaft is fixedly installed at the bottom end of the outer wall of the rotating platform. An electromagnetic actuator is nested on the outer ring of the main shaft bearing of the rotating shaft. A shielding shell is arranged around the outer wall of the machine tool body. An alloy layer is arranged on the inner layer of the shielding shell, a conductive foam layer is arranged in the middle, and a crystal belt wound magnetic ring is arranged on the outer layer. The optical fiber gyroscope is embedded inside the crystal belt wound magnetic ring.

[0017] Preferably, the input ports of the edge computing node include Channel 1, Channel 2, and a bus port. Channel 1 accesses the image data of the dual-camera array and the industrial camera. Channel 2 accesses the laser point cloud of the three-dimensional laser module. The bus port collects the thermal imaging data and vibration signals. On the one hand, the output port of the edge computing node sends compensation instructions to the three-axis drive mechanism through the CAT bus. On the other hand, it connects to the PLC of the machine tool body through the UA protocol.

[0018] The edge computing node is connected to the multi-spectral vision module through an optical fiber. Based on the machine tool body to define the basic coordinates, Point 1 of the accessory 1 in the laser point cloud data is translated by rotation, and the square of the distance is calculated with the thermal imaging point 1 corresponding to Laser Point 1, and the two points are aligned in space. The multi-spectral vision module fuses visible light, near-infrared, and thermal features through an attention mechanism to construct a 5-layer feature level group.

[0019] Preferably, the global camera captures wide-angle images, extracts the workpiece contour features through a feature recognition algorithm, and matches them with a preset feature library to generate initial coordinates. The microscopic camera scans the surface of the workpiece along a spiral path, combines the point cloud data of the three-dimensional laser module to construct a sub-pixel level edge model, correlates the temperature gradient field detected by the microbolometer with the thermal expansion coefficient of the machine tool, divides the machine tool accessories into n small units, and each small unit calculates the deformation compensation amount according to the thermal expansion formula.

[0020] The filter is connected to the three-dimensional laser module and generates a compensation signal. The compensation signal is transmitted to the controller. The controller processes the compensation duration according to the Lyapunov function and transmits it to the digital mirror through an optical fiber. The digital mirror interacts with the machine tool PLC through the OC protocol. The machine tool PLC transmits the quantified accessory wear, cutting digital parameters, and machine tool process indicators to the digital mirror.

[0021] Preferably, the beam splitter prism reflects visible light for an industrial camera to identify the surface texture, and the transmitted laser path is received by a 3D laser module. The acquisition card built into the 3D laser module generates a laser point cloud through a point cloud generation algorithm, constructs a 3D point cloud model based on the laser point cloud. Thermal radiation irradiates the thermistor array inside the microbolometer through a diaphragm. The signal conditioning circuit inside the microbolometer converts the resistance change into a voltage signal. The synchronization signal source is based on the spindle encoder pulses. Every time the spindle rotates 5°, the spindle encoder outputs a pulse signal. The exposure time of the industrial camera is aligned with the rising edge of the laser pulse, and the thermal imaging acquisition frequency and the vibration signal sampling rate are maintained at 1:10 through a timer;

[0022] An interactive terminal is configured on the side of the outer wall of the electrical control cabinet. The video interface of the interactive terminal is connected to a computer to display the laser point cloud, thermal map, and visual recognition contour through a visualization interface, and an external device transmits signals in the reverse direction to operate the computer.

[0023] Preferably, the edge computing node is set inside a heterogeneous processor, the heterogeneous processor is set inside an edge computing box, the edge computing box is fixedly installed on the side of the electrical control cabinet, and the edge computing box leads out a CAT bus to connect to an actuator;

[0024] The 3D laser module scans to generate point cloud data. A transformation matrix is established between the image coordinate system of the industrial camera and the machine tool coordinate system. The rotation matrix R is calculated through singular value decomposition, and the translation vector is fitted through the least squares method. The edge computing node extracts feature points from the global camera visible light image and the microscopic camera near-infrared image using the SF algorithm, and eliminates false matches using the RANSAC algorithm. Finally, sub-pixel level alignment is performed for registration. The temperature field data collected by the microbolometer is mapped to the 3D point cloud to establish a thermal deformation model. The thermal deformation model obtains a change value, and the change value drives a compensation instruction to be executed in two stages by a piezoelectric ceramic micro-adjuster. The first stage is coarse compensation, and the coarse compensation drives a ball screw. The second stage is fine compensation, and the fine compensation switches to a piezoelectric ceramic driver.

[0025] Preferably, the annular supplementary lighting unit detects reflection, switches the LED lamp to 30° low-angle illumination, the piezoelectric ceramic on the X-axis moves to change the cavity length of the filter, and a Peltier temperature control module is embedded inside the lens barrel of the industrial camera, and the Peltier temperature control module is associated with the microbolometer;

[0026] As the spindle temperature rises, the change in the bearing clearance causes vibration, and the synchronization signal source triggers thermal deformation compensation and an electromagnetic actuator.

[0027] Preferably, for the X-axis movement actuator, the actuator compares the readings of the grating scale with the data of the laser interferometer. An accelerometer is fixedly installed at the bottom of the machine tool body. The accelerometer is connected to the actuator, and the actuator uploads the model parameters to the central processor. The central processor uses a differential framework to add noise to the nodes of the edge model, thermal deformation model, and three-dimensional point cloud model and then performs aggregation to obtain safety parameters.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] 1. Through the multispectral vision module, the present invention realizes the multi-dimensional information fusion of the accessories, solves the problem of incomplete information acquisition due to visual occlusion, improves the accuracy and reliability of the identification of machine tool accessories, and provides richer and more accurate data support for subsequent adjustment and control;

[0030] 2. Through the microbolometer, the present invention realizes the monitoring of the thermal deformation of machine tool accessories caused by temperature changes, solves the problem of errors in the processing accuracy of accessories caused by thermal deformation, reduces the rejection rate caused by thermal deformation, and prolongs the service life of machine tool accessories;

[0031] 3. Through the closed-loop drive assembly, the present invention realizes the three-dimensional fine-tuning of the positioning accuracy of the accessories, solves the problem of insufficient motion control accuracy, and improves the processing accuracy and efficiency of the machine tool;

[0032] 4. Through the differential framework, the present invention realizes the aggregation of multiple models to form safety parameters, solves the problem of the lack of an effective evaluation mechanism for the operating safety of the machine tool, and reduces production accidents and downtime caused by equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is the front view structural schematic diagram of the present invention;

[0034] Figure 2 is the working process schematic diagram of the edge computing node of the present invention;

[0035] Figure 3 is the structural schematic diagram of the multispectral vision module of the present invention.

[0036] In the figure: 1, machine tool body; 2, multispectral vision module; 3, three-axis drive mechanism; 4, fiber optic switch; 5, lifting column; 6, three-dimensional laser module; 7, annular light supplement unit; 8, beam splitter prism; 9, host computer; 10, filter; 11, control center; 12, microscopic camera. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] Please refer to Figure 1 、 Figure 2 and Figure 3 For an embodiment provided by the present invention: an automatic adjustment control system for machine tool accessories based on visual recognition, which includes a machine tool body, a multi-spectral vision module, and a three-axis drive mechanism. The three-axis drive mechanism is fixedly installed around the outer wall of the machine tool body, and the multi-spectral vision module is fixedly installed on the outer wall of the three-axis drive mechanism; the multi-spectral vision module includes an annular supplementary light unit and a dual-camera array. The dual-camera array is composed of a global camera and a microscopic camera. The global camera is fixedly installed on the gantry of the top of the machine tool body, and the microscopic camera is fixedly installed on the side slide of the spindle. The annular supplementary light unit is connected to the control center through a control line. The global camera is connected to the data processing unit and the control center through a data line. The microscopic camera transmits image data to the data processing unit through a data line.

[0039] The three-axis drive mechanism includes a closed-loop drive component and a Y / Z-axis drive component. The closed-loop drive component is composed of an X-axis linear motor and a grating scale. The ball screw of the Y / Z-axis drive component is connected to a harmonic reducer. A piezoelectric ceramic micro-adjuster is fixedly installed at the end of the ball screw. The visual marking point uses a reflective ceramic ball. The driver arranged inside the piezoelectric ceramic micro-adjuster is connected in series with the ball screw; Fiber optic gyroscopes are fixedly installed at the four corners of the base of the machine tool body. A magnetic levitation bearing is fixedly installed at the bottom end of the outer wall of the rotating platform. An annular encoder is arranged in the middle of the magnetic levitation bearing. 8 groups of visual marking points are equidistantly arranged on the edge of the tabletop of the rotating platform. A rotating shaft is fixedly installed at the bottom end of the outer wall of the rotating platform. An electromagnetic actuator is nested outside the outer ring of the main shaft bearing of the rotating shaft. A shielding shell is arranged around the outer wall of the machine tool body. An alloy layer is arranged on the inner layer of the shielding shell, a conductive foam layer is arranged in the middle, and a crystal band winding magnetic ring is arranged on the outer layer. A fiber optic gyroscope is embedded inside the crystal band winding magnetic ring.

[0040] Further, first, the workpiece is placed on the rotating platform. The global camera collects a wide-angle image, extracts the contour features of the workpiece through a feature recognition algorithm, and matches them with a preset feature library to generate the initial coordinates of the workpiece.

[0041] Then, the microscopic camera scans the surface of the workpiece along a spiral path. At the same time, the three-dimensional laser module moves synchronously with the X-axis, scans the workpiece, and generates laser point cloud data. The industrial camera captures the surface image of the workpiece. The image coordinate system and the machine tool coordinate system are transformed by establishing a transformation matrix. The rotation matrix R is calculated by singular value decomposition, and the translation vector is obtained by least squares fitting, so that the image captured by the industrial camera is mapped into the machine tool coordinate system;

[0042] Next, the microbolometer collects the thermal imaging data of the workpiece. The thermistor array inside it receives thermal radiation, and the signal conditioning circuit converts the resistance change into a voltage signal. The synchronous signal source is based on the spindle encoder pulse. When the spindle rotates 5° each time, the spindle encoder outputs a pulse signal to control the trigger pulses of the microbolometer and the three-dimensional laser module, so that the thermal imaging data is synchronized with the laser scanning data. At the same time, fiber Bragg grating sensors fixed on the axial and circumferential directions of the spindle, the length direction of the guide rail, and the side wall of the machine tool body collect stress and strain, and are connected to the upper computer through optical fibers for analysis and processing.

[0043] Please refer to Figure 1 and Figure 3 , an embodiment provided by the present invention: an automatic adjustment control system for machine tool accessories based on visual recognition. Guide rails are fixedly installed around the machine tool body. A lifting column is fixedly installed on the side of the guide rail. A polarized industrial camera is fixedly installed at the top of the lifting column. The height of the industrial camera is aligned with the center line of the spindle of the machine tool body. The industrial camera and the dual-camera array are respectively connected to the fiber optic switch through twisted pairs. An annular supplementary light source is configured at the top of the industrial camera. The annular supplementary light unit is provided with 12 groups of LED lights controlled by independent data lines. A tunable filter is fixedly installed between the lens and the image sensor of the industrial camera. The tunable filter switches visible light and near-infrared based on the Fabry-Perot interference principle; Fiber Bragg grating sensors are fixedly installed in the axial and circumferential directions of the spindle, the length direction of the guide rail, and the side wall of the machine tool body. The fiber Bragg grating sensors are connected to the upper computer through optical fibers;

[0044] The input ports of the edge computing node include Channel 1, Channel 2, and a bus port. Channel 1 accesses the image data of the dual camera array and the industrial camera. Channel 2 accesses the laser point cloud of the 3D laser module. The bus port collects thermal imaging data and vibration signals. One aspect of the output port of the edge computing node sends compensation instructions to the three-axis drive mechanism through the CAT bus, and on the other hand, connects to the PLC of the machine tool body through the UA protocol. The edge computing node is connected to the multispectral vision module through an optical fiber. With the machine tool body defining the basic coordinates, Point 1 of the accessory 1 in the laser point cloud data undergoes rotation and translation, and the square of the distance is calculated with the corresponding thermal imaging point 1 in space, aligning the two points in space. The multispectral vision module constructs a 5-layer feature level group by fusing visible light, near-infrared, and thermal features through an attention mechanism.

[0045] Furthermore, first, according to the established thermal deformation model, the temperature gradient field detected by the microbolometer, and the thermal expansion coefficient of the machine tool, the machine tool accessories are divided into n small units. Each small unit calculates the deformation compensation amount according to the thermal expansion formula. The filter is connected to the 3D laser module to generate a compensation signal and transmits it to the controller.

[0046] Then, the controller calculates the machining compensation duration according to the Lyapunov function and transmits it to the digital mirror through an optical fiber. The digital mirror interacts with the machine tool PLC through the OC protocol. The machine tool PLC transmits the quantified accessory wear, cutting digital parameters, and machine tool process indicators to the digital mirror.

[0047] Next, the edge computing node sends compensation instructions to the three-axis drive mechanism through the CAT bus according to the calculated compensation amount. The piezoelectric ceramic micro-adjuster performs compensation actions in two stages according to the compensation instructions. In the first stage, rough compensation is carried out to drive the ball screw for a large stroke adjustment. In the second stage, fine compensation is carried out, and the piezoelectric ceramic driver is switched to perform precise adjustment of a small stroke to ensure that the position and posture of the machine tool accessories meet the machining requirements. At the same time, when the spindle temperature rises and the bearing clearance change causes vibration, the synchronous signal source triggers the thermal deformation compensation and the electromagnetic actuator, and the electromagnetic actuator generates thermal deformation compensation to ensure the machining accuracy of the machine tool.

[0048] Please refer to Figure 1 and Figure 2, an embodiment provided by the present invention: an automatic adjustment control system for machine tool accessories based on visual recognition. A three-dimensional laser module is arranged in parallel below the crossbeam of the machine tool body. The three-dimensional laser module moves synchronously with the X-axis. The three-dimensional laser module is connected to an edge computing node through an optical fiber. The three-dimensional laser module forms a 45° angle with the optical axis of the industrial camera. Micrombolometers are embedded on both sides of the spindle box installed on the side of the machine tool body. A transfer board is arranged in the middle of the optical paths of the industrial camera and the three-dimensional laser module. A beam splitter prism is fixedly installed at the top of the transfer board. A synchronous signal source is arranged inside the electrical control cabinet. Two output interfaces are arranged inside the synchronous signal source. The output interfaces respectively control the trigger pulses of the micrombolometer and the three-dimensional laser module through signal connection circuits. The micrombolometer collects thermal imaging data and connects to an electromagnetic actuator to generate thermal deformation compensation. The micrombolometer is connected to the three-dimensional laser module. The three-dimensional laser module is connected to a multi-spectral vision module. The micrombolometer, the three-dimensional laser module, and the multi-spectral vision module together form an optical path acquisition component;

[0049] The global camera collects wide-angle images, extracts workpiece contour features through a feature recognition algorithm, and matches them with a preset feature library to generate initial coordinates. The microscopic camera scans the surface of the workpiece along a spiral path, combines the point cloud data of the three-dimensional laser module to construct a sub-pixel edge model, correlates the temperature gradient field detected by the micrombolometer with the thermal expansion coefficient of the machine tool, divides the machine tool accessories into n small units, and calculates the deformation compensation amount for each small unit according to the thermal expansion formula; The filter is connected to the three-dimensional laser module and generates a compensation signal. The compensation signal is transmitted to the controller. The controller processes the compensation duration according to the Lyapunov function and transmits it to the digital mirror through an optical fiber. The digital mirror interacts with the machine tool PLC through the OC protocol. The machine tool PLC transmits the quantified accessory wear, cutting digital parameters, and machine tool process indicators to the digital mirror;

[0050] Further, first, the interaction terminal configured on the outer wall side of the electrical control cabinet is connected to a computer through a video interface, and displays the laser point cloud, thermal map, and visual recognition contour through a visualization interface. The operator can understand the processing status of the workpiece and the operation of the machine tool.

[0051] Then, an external device can operate the computer through a reverse transmission signal to achieve remote control and parameter adjustment of the machine tool. At the same time, an accelerometer fixedly installed at the bottom of the machine tool body is connected to an actuator. The actuator uploads model parameters to a central processor. The central processor uses a differential framework to add noise to the nodes of the edge model, thermal deformation model, and three-dimensional point cloud model and then perform aggregation to obtain safety parameters.

[0052] Finally, the system runs data collection, processing, adjustment, and monitoring.

[0053] Please refer to Figure 2 and Figure 3, an embodiment provided by the present invention: an automatic adjustment control system for machine tool accessories based on visual recognition. The annular supplementary lighting unit detects reflection and switches the LED lamp to 30° low-angle illumination. The piezoelectric ceramic in the X-axis moves to change the cavity length of the filter. A Peltier temperature control module is embedded inside the industrial camera lens barrel, and the Peltier temperature control module is associated with the microbolometer; when the spindle temperature rises, the bearing clearance changes, causing vibration, and the synchronous signal source triggers thermal deformation compensation and the electromagnetic actuator;

[0054] The beam splitter prism reflects visible light to the industrial camera to recognize the surface texture, and the transmitted laser path is received by the 3D laser module. The acquisition card built into the 3D laser module generates a laser point cloud through the point cloud generation algorithm, constructs a 3D point cloud model based on the laser point cloud. Thermal radiation irradiates the thermistor array inside the microbolometer through the aperture. The signal conditioning circuit inside the microbolometer converts the resistance change into a voltage signal. The synchronous signal source takes the spindle encoder pulse as a reference. Every time the spindle rotates 5°, the spindle encoder outputs a pulse signal. The exposure time of the industrial camera is aligned with the rising edge of the laser pulse, and the thermal imaging acquisition frequency and the vibration signal sampling rate are kept at 1:10 through the timer; An interactive terminal is configured on the side wall of the electrical control cabinet. The video interface of the interactive terminal is connected to the computer to display the laser point cloud, thermal map, and visual recognition contour through the visualization interface. The external device transmits signals in the reverse direction to operate the computer;

[0055] Furthermore, the annular supplementary lighting unit illuminates the machining area. When reflection is detected, the control center controls the annular supplementary lighting unit to switch the LED lamp to the 30° low-angle illumination mode. The industrial camera collects the visible light image of the workpiece surface. At the same time, the piezoelectric ceramic telescopic device in the X-axis moves to change the cavity length of the tunable filter and collect images of different bands. The 3D laser module scans the workpiece, and the acquisition card built into it generates a laser point cloud through the point cloud generation algorithm and transmits the laser point cloud data to the edge computing node through the optical fiber;

[0056] The microbolometer collects the thermal imaging data of the workpiece and the machine tool spindle box, converts the resistance change into a voltage signal and transmits it to the edge computing node. At the same time, the vibration sensor collects the vibration signal of the machine tool and transmits it to the edge computing node. The edge computing node processes the image data of the industrial camera, extracts the texture features and contour information of the workpiece surface through the feature recognition algorithm, correlates and analyzes the thermal imaging data collected by the microbolometer with the vibration signal, and analyzes the rise of the machine tool spindle temperature and the bearing clearance change by establishing a thermal deformation model and a vibration model, and calculates the adjustment compensation amount of the machine tool accessories.

[0057] Please refer to Figure 1 、 Figure 2 and Figure 3, an embodiment provided by the present invention: an automatic adjustment control system for machine tool accessories based on visual recognition. The edge computing node is arranged inside a heterogeneous processor, and the heterogeneous processor is arranged inside an edge computing box. The edge computing box is fixedly installed on the side of the electrical control cabinet, and the edge computing box leads out a CAT bus to connect to the actuator; the 3D laser module scans to generate point cloud data. The image coordinate system of the industrial camera and the machine tool coordinate system establish a transformation matrix. The rotation matrix R is calculated by singular value decomposition, and the translation vector is fitted by the least squares method. The edge computing node extracts feature points from the visible light image of the global camera and the near-infrared image of the microscopic camera using the SF algorithm, and eliminates the mismatches using the RANSAC algorithm. Finally, sub-pixel-level alignment is performed for registration. The temperature field data collected by the microbolometer is mapped to the 3D point cloud to establish a thermal deformation model. The thermal deformation model obtains a change value, and the change value drives a compensation instruction to be executed in two stages by the piezoelectric ceramic micro-adjuster. The first stage is rough compensation, and the rough compensation drives the ball screw. The second stage is fine compensation, and the fine compensation switches to the piezoelectric ceramic actuator;

[0058] For the X-axis moving actuator, the actuator compares the readings of the grating scale with the data of the laser interferometer. An accelerometer is fixedly installed at the bottom of the machine tool body, and the accelerometer is connected to the actuator. The actuator uploads the model parameters to the central processor, and the central processor uses a differential framework to add noise points to the nodes of the edge model, thermal deformation model, and 3D point cloud model and then performs aggregation to obtain safety parameters;

[0059] Furthermore, the 3D laser module scans the machine tool accessories, generates point cloud data at a frequency of 100 Hz, and transmits the data to the edge computing node through an optical fiber. The industrial camera collects visible light and near-infrared images of the machine tool accessories and transmits the image data to the edge computing node through Gigabit Ethernet at a frame rate of 90 fps. The microbolometer collects the temperature field data of the machine tool accessories in real time and transmits the data to the edge computing node through the SPI bus;

[0060] After the edge computing node receives the point cloud data of the 3D laser module, the image data of the industrial camera, and the temperature field data of the microbolometer, it first extracts feature points from the visible light image and near-infrared image of the industrial camera using the SF algorithm, then uses the RANSAC algorithm to eliminate the mismatched points, and finally realizes sub-pixel-level alignment for registration. The temperature field data collected by the microbolometer is mapped to the 3D point cloud, and a thermal deformation model is established using the finite element analysis method to calculate the change value caused by thermal deformation. The edge computing node generates a compensation instruction according to the change value obtained from the thermal deformation model.

[0061] Working principle: First, the annular supplementary lighting unit detects reflection and switches the LED lamp to 30° low-angle illumination. The microbolometers on both sides of the spindle box collect thermal imaging data, the fiber Bragg grating sensor collects data such as strain, and the accelerometer collects the vibration value of the machine tool body;

[0062] Then, the three-dimensional laser module under the crossbeam of the machine tool body moves synchronously with the X-axis, forming a 45° angle with the optical axis of the industrial camera. It emits laser light, which is transmitted through the beam splitter prism and then received by the three-dimensional laser module. The point cloud generation algorithm of the acquisition card is used to generate the laser point cloud. Channel 1 of the edge computing node accesses the image data of the dual camera array and the industrial camera, and channel 2 accesses the laser point cloud of the three-dimensional laser module. The bus port collects the thermal imaging data and vibration signals, spatially aligns the corresponding points of the fittings in the laser point cloud and thermal imaging. The industrial camera processes the global camera and microscopic camera images by establishing a transformation matrix with the machine tool coordinate system, extracts feature points through the algorithm, and performs registration after eliminating false matches;

[0063] Finally, the edge computing node sends compensation instructions to the three-axis drive mechanism through the CAT bus. The piezoelectric ceramic micro-adjuster is driven to execute in two stages: rough compensation and fine compensation. The rough compensation drives the ball screw, and the fine compensation switches to the piezoelectric ceramic driver. The X-axis movement actuator compares the readings of the grating scale with the data of the laser interferometer, and the actuator uploads the model parameters to the central processor. The central processor uses the differential framework to add noise points to the relevant model nodes and then aggregates them to obtain the safety parameters.

[0064] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

Claims

1. An automatic adjustment and control system for machine tool accessories based on visual recognition, characterized in that: It includes a machine tool body, a multi-spectral vision module and a three-axis driving mechanism, wherein the three-axis driving mechanism is fixedly installed around the outer wall of the machine tool body, and the multi-spectral vision module is fixedly installed on the outer wall of the three-axis driving mechanism; The multi-spectral vision module includes an annular fill light unit and a dual camera array, the dual camera array consists of a global camera and a microscopic camera, the global camera is fixedly installed on the gantry at the top of the machine tool body, the microscopic camera is fixedly installed on the side slide of the spindle, the annular fill light unit is connected to the control center through a control line, the global camera is connected to the data processing unit and the control center through a data line, and the microscopic camera transmits image data to the data processing unit through the data line.

2. According to claim 1, the automatic adjustment and control system for machine tool accessories based on visual recognition is characterized in that: The machine tool body is fixedly installed with guide rails around it, and lifting columns are fixedly installed on the sides of the guide rails. A polarized industrial camera is fixedly installed on the top of the lifting column. The height of the industrial camera is aligned with the center line of the main axis of the machine tool body. The industrial camera and the dual-camera array are respectively connected to the optical fiber switch through twisted pair cables. The top of the industrial camera is equipped with an annular fill light source, and the annular fill light unit is provided with 12 groups of LED lights controlled by independent data lines. A tunable filter is fixedly installed between the lens and the image sensor of the industrial camera. The tunable filter switches between visible light and near infrared based on the Fabry-Perot interference principle. The fiber grating sensor is fixedly installed in the axial and circumferential directions of the main shaft, the fiber grating sensor is installed in the length direction of the guide rail and the fiber grating sensor is fixedly installed on the side of the outer wall of the machine tool body, and the fiber grating sensor is connected to the host computer through optical fiber.

3. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 1 is characterized in that: A three-dimensional laser module is arranged in parallel below the crossbeam of the machine tool body. The three-dimensional laser module moves synchronously with the X-axis. The three-dimensional laser module is connected to the edge computing node through an optical fiber. The three-dimensional laser module forms a 45° angle with the optical axis of the industrial camera. Microbolometers are embedded on both sides of the spindle box installed on the side of the machine tool body, an adapter plate is arranged in the middle of the optical path between the industrial camera and the three-dimensional laser module, a dichroic prism is fixedly installed on the top of the adapter plate, a synchronization signal source is arranged inside the electrical control cabinet, and two output interfaces are arranged inside the synchronization signal source. The output interfaces control the trigger pulses of the microbolometer and the three-dimensional laser module through the signal connection circuit respectively. The microbolometer collects thermal imaging data and connects to the electromagnetic actuator to generate thermal deformation compensation. The microbolometer is connected to the three-dimensional laser module, and the three-dimensional laser module is connected to the multi-spectral vision module. The microbolometer, the three-dimensional laser module and the multi-spectral vision module together form an optical path acquisition component.

4. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 1 is characterized in that: The three-axis drive mechanism includes a closed-loop drive component and a Y / Z-axis drive component. The closed-loop drive component is composed of an X-axis linear motor and a grating ruler. The ball screw of the Y / Z-axis drive component is connected to a harmonic reducer. A piezoelectric ceramic fine-tuner is fixedly installed at the end of the ball screw. The visual marking point adopts a reflective ceramic ball. The driver set inside the piezoelectric ceramic fine-tuner is connected in series with the ball screw. Fiber optic gyroscopes are fixedly installed at the four corners of the machine tool base, a magnetic bearing is fixedly installed at the bottom of the outer wall of the rotating platform, a ring encoder is arranged in the middle of the magnetic bearing, 8 groups of visual marking points are equidistantly arranged on the edge of the table of the rotating platform, a rotating shaft is fixedly installed at the bottom of the outer wall of the rotating platform, an electromagnetic actuator is embedded in the outer ring of the spindle bearing of the rotating shaft, a shielding shell is arranged around the outer wall of the machine tool body, an alloy layer is arranged in the inner layer of the shielding shell, a conductive foam layer is arranged in the middle, and a crystal ribbon is arranged on the outer layer. The fiber optic gyroscope is embedded in the crystal ribbon.

5. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 3 is characterized in that: The input port of the edge computing node includes channel 1, channel 2 and a bus port. Channel 1 is connected to the image data of the dual camera array and the industrial camera, channel 2 is connected to the laser point cloud of the three-dimensional laser module, and the bus port collects thermal imaging data and vibration signals. The output port of the edge computing node sends compensation instructions to the three-axis drive mechanism through the CAT bus on the one hand, and connects to the PLC of the machine tool body through the UA protocol on the other hand; The edge computing node is connected to the multispectral vision module through optical fiber, and the basic coordinates are defined by the machine tool body. The point one of the accessory in the laser point cloud data is rotated and translated, and the distance to the thermal imaging point one corresponding to the laser point one is squared, and the two points are aligned in space. The multispectral vision module fuses visible light, near-infrared and thermal features through the attention mechanism to construct a 5-layer feature level group.

6. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 1 is characterized in that: The global camera collects wide-angle images, extracts workpiece contour features through feature recognition algorithms, and matches them with a preset feature library to generate initial coordinates. The microscopic camera scans the workpiece surface along a spiral path, and constructs a sub-pixel edge model in combination with point cloud data of a three-dimensional laser module. The temperature gradient field detected by a microbolometer is associated with the thermal expansion coefficient of the machine tool, and the machine tool accessories are divided into n small units. The deformation compensation amount of each small unit is calculated according to the thermal expansion formula; The filter is connected to the three-dimensional laser module and generates a compensation signal, which is transmitted to the controller. The controller transmits the compensation time to the digital mirror through optical fiber according to the Lyapunov function processing. The digital mirror interacts with the machine tool PLC through the OC protocol. The machine tool PLC transmits the quantification of accessory wear, cutting digital parameters and machine tool process indicators to the digital mirror.

7. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 3 is characterized in that: The beam splitter prism reflects visible light to the industrial camera to identify the surface texture, the transmitted laser route is received by the three-dimensional laser module, the acquisition card built into the three-dimensional laser module generates a laser point cloud through a point cloud generation algorithm, and a three-dimensional point cloud model is constructed based on the laser point cloud. The thermal radiation is irradiated to the thermistor array inside the microbolometer through the aperture, and the signal conditioning circuit inside the microbolometer converts the resistance change into a voltage signal. The synchronization signal source is based on the spindle encoder pulse. The spindle encoder outputs a pulse signal every time the spindle rotates 5°. The exposure time of the industrial camera is aligned with the rising edge of the laser pulse. The thermal imaging acquisition frequency and the vibration signal sampling rate are maintained at 1:10 by a timer. An interactive terminal is configured on the side of the outer wall of the electrical control cabinet. The video interface of the interactive terminal is connected to the computer to display the laser point cloud, thermal map and visual recognition contour through a visual interface. The external device transmits signals in reverse to operate the computer.

8. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 3 is characterized in that: The edge computing node is arranged inside the heterogeneous processor, the heterogeneous processor is arranged inside the edge computing box, the edge computing box is fixedly mounted on the side of the electrical control cabinet, and the edge computing box leads out the CAT bus to connect the actuator; The three-dimensional laser module scans and generates point cloud data. The image coordinate system of the industrial camera and the machine tool coordinate system establish a transformation matrix. The rotation matrix R is calculated through singular value decomposition, and the translation vector is fitted through the least squares method. The edge computing node uses the SF algorithm to extract feature points from the visible light image of the global camera and the near-infrared image of the microscope camera. The RANSAC algorithm eliminates mismatches and finally performs sub-pixel alignment for registration. The temperature field data collected by the micro-bolometer is mapped to the three-dimensional point cloud, and a thermal deformation model is established. The thermal deformation model obtains the change value, and the change value drives the compensation instruction to be executed through the piezoelectric ceramic fine-tuner in two stages. The first stage is coarse compensation, and the coarse compensation drives the ball screw. The second stage is fine compensation, and the fine compensation is switched to the piezoelectric ceramic driver.

9. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 1, characterized in that: The annular fill light unit detects reflections, switches the LED light to 30° low-angle lighting, moves the piezoelectric ceramic on the X-axis to change the filter cavity length, embeds a Peltier temperature control module inside the industrial camera lens barrel, and the Peltier temperature control module is associated with a microbolometer; The spindle temperature rises, the bearing clearance changes and causes vibration, and the synchronous signal source triggers the thermal deformation compensation and electromagnetic actuator.

10. The automatic adjustment and control system for machine tool accessories based on visual recognition according to claim 3, characterized in that: The X-axis moves the actuator, and the actuator compares the grating ruler reading with the laser interferometer data. An accelerometer is fixedly installed at the bottom of the machine tool body, and the accelerometer is connected to the actuator. The actuator uploads the model parameters to the central processor. The central processor uses a differential framework to add noise points to the nodes of the edge model, thermal deformation model and three-dimensional point cloud model, and then performs aggregation to obtain safety parameters.

Images