Standard block array self-calibration system and method for multi-action haptic deficit retest

Abstract

The application provides a standard block array self-calibration system and method for multi-action tactile defect review, relates to the technical field of tactile detection, and comprises at least two types of known parameter standard blocks through a calibration base, a tactile end and a block array calibration coordinate system are established by means of a probe positioning reference, a detection execution unit is automatically controlled to contact a target block according to a preset path, corresponding calibration actions such as point pressing, sliding, scraping, rolling, clamping, tapping and thermal contact are performed on different blocks, multi-modal calibration signals are synchronously collected, a calibration parameter set is generated according to the deviation between the measured signals and the known parameters of the blocks, and the calibration parameter set is stored in a structured manner according to the end type, the action type and the calibration version, and then output to a multi-action tactile defect review process, which is used for signal correction, feature normalization, determination threshold updating and fusion determination, and can improve the repeatability, comparability and accuracy of the multi-action tactile defect review result.

Description

Technical Field

[0001] This invention relates to tactile detection technology, and more particularly to a standard sample array self-calibration system and method for re-inspection of tactile defects in multi-action movements. Background Technology

[0002] With the deepening development of industrial automation and intelligent manufacturing, tactile sensors, as core components for robots to accurately perceive the surface material, hardness, texture and defect characteristics of objects, are playing an increasingly crucial role in the quality inspection and defect re-inspection of high-end manufacturing.

[0003] Currently, existing tactile sensor calibration schemes generally only target a single physical channel. For example, standard weights are used to perform single-point or multi-point static calibration of the normal force channel, or standard displacement blocks are used to test the linearity of the displacement channel. This calibration method severs the connection between active detection action and physical attribute perception, and cannot establish specific calibration benchmarks for different detection actions such as pressing, scraping, rolling, and clamping. When the tactile sensor performs different actions to perceive different physical attributes, the same set of universal calibration coefficients cannot compensate for the specific deviations of the sensor response under different action modes, resulting in poor comparability and consistency of multi-action tactile detection results between different actions.

[0004] Therefore, how to achieve automated calibration of multi-motion tactile sensors and improve the accuracy of industrial defect re-inspection has become an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a standard sample array self-calibration system and method for re-inspection of multi-action tactile defects, which can realize the automated calibration of multi-action tactile sensors and improve the accuracy of industrial defect re-inspection.

[0006] A first aspect of the present invention provides a standard sample array self-calibration system for re-examination of multi-action tactile defects, comprising: Calibration base, used to support and position multiple standard sample blocks; A standard sample array, fixed on the calibration base, includes at least two types of standard sample blocks with known physical parameters or known defect characteristics; The probe positioning reference is set on the calibration base to establish a calibration coordinate system between the tactile end and the standard sample array; A calibration execution interface is used to cooperate with a detection execution unit to enable the detection execution unit to contact one or more target standard samples in the standard sample array according to a preset calibration path. An automatic calibration control module is used to control the detection execution unit to perform corresponding tactile calibration actions for different standard samples, and to simultaneously acquire multimodal calibration signals; The multimodal calibration signal acquisition module is used to simultaneously acquire one or more calibration signals from normal force, tangential force, displacement, vibration, temperature, or tactile images during the calibration process. The calibration parameter generation module is used to calculate and generate calibration parameters based on the deviation between the multimodal calibration signal and the known physical parameters or known defect characteristics of the corresponding standard sample. The calibration parameter storage and version management module is used to store calibration parameters in a structured manner and supports version management and expiration management of calibration parameters. The calibration parameter output interface is used to provide the calibration parameters to signal correction, feature normalization, judgment threshold update or fusion judgment in the re-examination of tactile defects; A second aspect of the present invention provides a self-calibration method for a standard sample array for re-examination of multi-action tactile defects, comprising: Obtain the current tactile end type of the detection execution unit and construct a calibration coordinate system between the tactile end and the standard sample array; Based on the type of the tactile end, a target standard sample is determined from the standard sample array. Based on the calibration coordinate system, the tactile end is controlled to perform corresponding calibration operations on each target standard sample and multimodal calibration signals are acquired simultaneously. The calibration operation includes one or more of the following: tapping, sliding, scraping, rolling, clamping, tapping, and thermal contact. A set of calibration parameters associated with each target standard sample and its corresponding calibration operation is generated based on the signal deviation between the multimodal calibration signal corresponding to each target standard sample and the target standard parameter. Based on the tactile end type corresponding to the defect re-inspection task and the action type corresponding to the calibration operation, the matching calibration parameter set is invoked to calibrate the tactile detection signal collected during the defect re-inspection process.

[0007] Optionally, in one possible implementation of the second aspect, obtaining the current tactile end-effector type of the detection execution unit and constructing a calibration coordinate system between the tactile end-effector and the standard sample array includes: Read the type identifier corresponding to the currently installed tactile end on the detection execution unit to determine the type of tactile end; The tactile end is controlled to move to a preset positioning reference and then approach the target along a direction perpendicular to the plane of the standard sample array. The contact force signal of the tactile end is collected in real time, and the approach operation is stopped when the contact force signal reaches a preset contact threshold. The location point corresponding to the tactile end when the approach operation stops is determined as the calibration origin; A calibration coordinate system is constructed based on the calibration origin and the preset positional relationship of the sample blocks.

[0008] Optionally, in one possible implementation of the second aspect, determining the target standard sample from the standard sample array based on the tactile end type includes: Based on the type of tactile end, a preset operation mapping relationship is retrieved to determine the calibration operation type corresponding to the tactile end type; Based on the preset correspondence between the calibration operation type and the standard sample block, the standard sample block corresponding to each calibration operation type is selected from the standard sample block array as the target standard sample block.

[0009] Optionally, in one possible implementation of the second aspect, the step of controlling the tactile end to perform corresponding calibration operations on each of the target standard samples based on the calibration coordinate system, and simultaneously acquiring multimodal calibration signals, includes: Based on the calibration coordinate system, determine the target contact position corresponding to each target standard sample block; Obtain preset operation parameters corresponding to the calibration operation type of the target standard sample block; According to the preset operating parameters, the tactile end is controlled to perform corresponding calibration operations on each target standard sample at each target contact position; During the calibration operation, multiple signal acquisition channels are triggered to start synchronously, and the physical quantity signals corresponding to each signal acquisition channel are acquired. The physical quantity signals that meet the preset quality conditions are determined as multimodal calibration signals.

[0010] Optionally, in one possible implementation of the second aspect, determining the target contact position corresponding to each of the target standard samples based on the calibration coordinate system includes: The relative position information of each target standard sample block with respect to the calibration origin is converted into the sample block position coordinates of each target standard sample block in the calibration coordinate system; Based on the calibration operation type of each target standard sample, determine the contact direction and contact posture of the tactile end contacting the target standard sample; Based on the sample block position coordinates, the contact direction, and the contact posture, the target contact position corresponding to each target standard sample block is determined.

[0011] Optionally, in one possible implementation of the second aspect, generating a calibration parameter set associated with each target standard sample and its corresponding calibration operation based on the signal deviation between the multimodal calibration signal corresponding to each target standard sample and the target standard parameter includes: The physical quantity signals corresponding to each signal acquisition channel in the multimodal calibration signal are preprocessed to obtain the processed signals corresponding to each signal acquisition channel. From the processed signal, identify signal feature points associated with the target standard parameters corresponding to the target standard sample block, and extract the signal feature values ​​corresponding to the signal feature points; The signal feature value is compared with the target standard parameter to determine the signal deviation between the signal feature value and the target standard parameter; A calibration parameter set is generated based on the signal deviation, and the calibration parameter set is stored in association with the end type, action type, sample type, calibration version, or validity period.

[0012] Optionally, in one possible implementation of the second aspect, generating the calibration parameter set based on the signal deviation includes: The signal deviation corresponding to the determination of the target standard parameter as the reference value is the zero-point offset compensation amount of the tactile end under the current calibration operation; The signal deviation corresponding to multiple target standard parameters with different values ​​is determined as candidate difference quantities. Based on the numerical correspondence between the target standard parameters and each candidate difference quantity, the gain correction coefficient of the tactile end under the current calibration operation is fitted and calculated. The intermediate value between the signal feature values ​​corresponding to two adjacent target standard parameters with different values ​​is determined as the defect judgment threshold of the tactile end under the current calibration operation. The zero-point offset compensation, the gain correction coefficient, and the defect judgment threshold are associated with the action type of the current calibration operation to generate a calibration parameter set associated with the current target standard sample and the corresponding calibration operation.

[0013] Optionally, in one possible implementation of the second aspect, the step of calling a matching calibration parameter set to calibrate the tactile detection signal collected during the defect re-inspection process based on the tactile end-effector type corresponding to the defect re-inspection task and the action type corresponding to the calibration operation includes: Obtain the tactile end type currently corresponding to the defect re-inspection task, and the action type currently performing the re-inspection action; Based on the obtained tactile end type and action type, retrieve and call the matching calibration parameter set, and extract the zero-point offset compensation amount, gain correction coefficient and defect judgment threshold from it; Based on the zero-point offset compensation and the gain correction coefficient, the tactile detection signal collected during the defect re-inspection process is subjected to offset processing and scale correction to obtain the calibrated target detection signal. The judgment boundary of the anomaly judgment module in the defect re-inspection task is dynamically updated according to the defect judgment threshold, and the defect re-inspection result is output based on the target detection signal and the updated judgment boundary.

[0014] Alternatively, in one possible implementation of the second aspect, it also includes: During the calibration operation, the contact force signal of the tactile end and the signal quality indicators of the multimodal calibration signal are monitored in real time. When the contact force signal is detected to exceed the preset overload threshold, the tactile end is controlled to exit the current calibration operation and re-execute the approach operation before retrying. When the signal quality index is detected to be unsatisfactory, the current target standard sample block is marked as calibration abnormal and a prompt message is generated. When the number of retries for the same target standard sample exceeds the preset threshold, the target standard sample is marked as not calibrated and the signal deviation corresponding to this calibration operation is prohibited from being used to generate the calibration parameter set. Continue performing the corresponding calibration operation on the next target standard sample block until all the target standard samples blocks have been calibrated.

[0015] The beneficial effects of this invention are as follows: 1. This invention establishes a mapping relationship between tactile end type, standard sample type, and calibration operation type, enabling precise calibration of different active tactile detection actions. By pre-establishing the operation mapping relationship, the range of calibration operation types applicable to the current tactile end is determined. Then, target standard samples are selected according to the preset correspondence between the calibration operation type and the standard sample, so that each tactile end performs a calibration action that matches its physical properties on the correct standard sample. This allows for the generation of corresponding calibration parameters for different action types such as pressing, sliding, scraping, clamping, and thermal contact, thereby improving the accuracy of detection signal calibration under different action types.

[0016] 2. By establishing a unified calibration coordinate system at the beginning of the calibration process, this invention can reduce the impact of tactile end-effector installation deviations on the consistency of multiple calibration results, ensuring that each calibration process is performed under the same spatial reference. This improves the positioning accuracy of the contact position during the calibration operation. At the same time, by synchronously triggering and acquiring multiple signal acquisition channels during the calibration operation, it ensures that the signals of each channel are aligned in the time dimension, thereby improving the synchronization accuracy between different physical quantity signals.

[0017] 3. This invention generates a multi-dimensional calibration parameter set based on the signal deviation, which can achieve a more comprehensive correction of the deviation of the tactile end system. It corrects the static zero-point drift of the sensor by the zero-point offset compensation amount, corrects the sensor sensitivity ratio deviation by the gain correction coefficient, and can further extend the temperature drift compensation coefficient to correct the influence of ambient temperature and the friction compensation coefficient to correct the influence of contact head wear. Thus, the calibrated detection signal can truly reflect the physical state of the measured object. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of a standard sample array self-calibration system for re-inspection of multi-action tactile defects provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating a standard sample array self-calibration method for re-inspection of multi-action tactile defects provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a typical active tactile action provided in an embodiment of the present invention; Figure 4 This is a block diagram of the functional modules of a standard sample array self-calibration system for re-inspection of multi-action tactile defects provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] See Figure 1 This is a schematic diagram of the structure of a standard sample array self-calibration system provided in an embodiment of the present invention. Figure 1 The overall structure of the standard sample array self-calibration system is demonstrated. It is divided into two main sections: the left side, consisting of the hardware entity unit calibration platform and standard sample array; and the right side, consisting of the software logic unit automatic calibration program and parameter output. The standard sample array self-calibration system provided by this invention comprises a calibration base 1, standard sample arrays 2–8, a probe positioning reference 9, and an automatic calibration control module 10. It works collaboratively with an external detection execution unit through a calibration execution interface. The calibration base 1 provides a stable installation reference for the entire system. The probe positioning reference 9 is located at a fixed position on the calibration base and is used to establish the origin of the calibration coordinate system and determine the coordinate axis direction. It serves as a reference point for the tactile end effector to perform positioning and alignment before calibration begins.

[0021] The calibration execution interface includes one or more of the following: mechanical positioning interface, motion control interface, communication interface, signal triggering interface, data transmission interface, or electrical connection interface. It is used to enable the standard sample array self-calibration system and the detection execution unit to collaboratively complete motion control, action triggering, calibration signal acquisition, and calibration parameter output. The automatic calibration program module 10, embedded in the system control software, is responsible for coordinating the motion control, calibration action triggering, calibration signal acquisition, standard value comparison, and calibration parameter generation of the detection execution unit. It is the core control logic unit of the entire calibration process.

[0022] like Figure 1 As shown, a standard sample array composed of multiple standard samples is mounted on a calibration base. This array includes the following types of standard samples, each with known physical parameters or known defect characteristics: standard hardness sample 2, standard roughness sample 3, standard burr sample 4, standard crack sample 5, standard elasticity sample 6, standard friction sample 7, temperature calibration block 8, probe positioning reference 9, and automatic calibration program module 10. Figure 1 As shown, the tactile probe, or tactile end effector, can complete the positioning reference alignment before calibration begins, determine the calibration coordinate system, and eliminate installation deviations. Then, it can be controlled to perform corresponding calibration operations on each standard sample block in the standard sample block array based on a preset calibration path. For example, P1 to P7 are sequentially contacted with each standard sample block in descending order to complete the corresponding calibration operation and obtain the corresponding multimodal calibration signal. The preset calibration path is pre-set and can be based on the layout of the standard sample block array. The tactile probe, or tactile end effector, performs calibration contact with each sample block and simultaneously transmits the multimodal physical quantity signals to the automatic calibration program module on the right through the calibration acquisition signal link. The module input can read the probe ID, tactile end effector type, and sample... The system automatically completes the following tasks: probe positioning and contact, synchronous acquisition of multi-channel signals, signal preprocessing, signal feature extraction, comparison of measured feature values ​​with target standard parameters to calculate signal deviation, and generation of various correction parameters. The final output includes a calibration parameter set containing zero-point offset compensation, gain correction coefficient, sensitivity coefficient, defect judgment boundary value, temperature drift compensation coefficient, and friction compensation coefficient. The calibration parameter application module can receive the calibration parameter set output by the automatic calibration program and complete signal preprocessing, feature normalization, dynamic updating of judgment boundaries, real-time temperature drift compensation, friction and wear compensation, and full-process traceability of calibration version in multi-action tactile defect re-inspection tasks.

[0023] See Figure 2 This is a flowchart illustrating a standard sample array self-calibration method for re-inspection of multi-action tactile defects provided in an embodiment of the present invention. Figure 2The method shown can be implemented by software and / or hardware devices. It includes steps S1 to S4, as detailed below: S1, obtain the current tactile end type of the detection execution unit, and construct the calibration coordinate system between the tactile end and the standard sample array.

[0024] The detection execution unit refers to an integrated device equipped with tactile end effectors that can drive them to perform multi-degree-of-freedom motion. The tactile end effector is a replaceable contact component installed at the front end of the detection execution unit, including different types such as soft fingertips, probe tips, scraper tips, and gripper tactile fingertips. The tactile end effector type refers to the category of the replaceable contact component installed at the front end of the detection execution unit. Common tactile end effector types include soft fingertips, probe tips, and scraper tips. It is a functional component within the detection execution unit responsible for directly contacting the object being measured and acquiring multimodal physical signals. Internally, it contains a normal force sensor and a tangential force sensor. The sensor system includes various types of sensing units such as sensors, displacement sensors, vibration sensors, and temperature sensors. A standard sample block refers to a standard reference object with known physical parameters. Each standard sample block is marked with calibrated known physical parameter values, such as hardness grade, roughness value, and temperature value. A standard sample block array refers to a set of calibration references composed of at least two types of standard samples with different known physical parameters arranged in a preset spatial layout. A calibration coordinate system refers to a three-dimensional spatial coordinate system that uses the standard sample block array as a reference to quantitatively describe the relative positional relationship between the tactile end and each standard sample block, as well as the posture of the tactile end itself.

[0025] Specifically, the type information corresponding to the tactile end currently installed in the detection execution unit can be obtained to determine the end category applicable to the calibration operation. Then, a spatial positioning operation can be performed between the tactile end and the standard sample array. By determining the spatial position and orientation of the tactile end relative to the standard sample array, the spatial relationship between the two can be established, and this spatial relationship can be determined as the calibration coordinate system that all calibration operations refer to.

[0026] In some embodiments, step S1 can be implemented as follows: S11, Read the type identifier corresponding to the currently installed tactile end on the detection execution unit to determine the type of tactile end.

[0027] Among them, the type identifier refers to the encoded information that can distinguish different types of tactile end devices. It can be stored on the tactile end device in the form of electronic chips, radio frequency identification tags, resistive coding, magnetic coding, optical coding or mechanical structure coding, and can be obtained through the corresponding reading interface.

[0028] Specifically, the end-effector identification module can be called through the data communication interface with the detection execution unit to read the type identifier corresponding to the tactile end currently installed at the front end of the detection execution unit, and the type of the tactile end can be determined based on the read identifier code.

[0029] S12, control the tactile end to move to the preset positioning reference, and perform an approach operation in a direction perpendicular to the plane where the standard sample array is located.

[0030] The preset positioning reference refers to a fixed spatial reference point set on the standard sample block array or calibration base. The approach operation refers to the movement process of the tactile end slowly approaching the reference from above the preset positioning reference in a specified direction until a reliable contact is established with the reference. During the approach process, the movement speed of the tactile end will be controlled to avoid impact damage to the positioning reference or the tactile end.

[0031] Specifically, after obtaining the type of tactile end, the detection execution unit is controlled to move the tactile end from its current position to above the preset positioning reference based on the spatial coordinates of the preset positioning reference. Then, the tactile end is controlled to perform a low-speed approaching action along a direction perpendicular to the plane where the standard sample array is located, so that the front end of the tactile end gradually approaches the positioning reference.

[0032] S13, collect the contact force signal of the tactile end in real time, and stop the approach operation when the contact force signal reaches the preset contact threshold.

[0033] Among them, the contact force signal refers to the normal force value collected by the force sensor integrated inside the tactile end when the tactile end comes into contact with an object. It can reflect the degree of contact between the tactile end and the contact surface. The preset contact threshold is a pre-set critical value of the contact force value, which is used to determine whether the tactile end has established a reliable and stable contact with the positioning reference.

[0034] Understandably, approach operations require a clear stopping condition. Relying solely on position control cannot accurately determine whether the tactile end has actually made contact with the positioning reference, due to factors such as installation errors and motion control precision. Using the contact force signal as the basis for determining the stopping condition is because the force signal directly reflects the physical contact state between the tactile end and the reference. When the force signal reaches a preset threshold, it indicates that the tactile end has established contact with the positioning reference to a certain degree. At this point, the spatial relationship between the two can be considered to be determined. Setting a preset contact threshold can prevent unreliable contact due to insufficient contact force, and can also prevent damage to the reference or tactile end due to excessive contact force.

[0035] Specifically, during the approach operation, the output value of the normal force sensor on the tactile end is read in real time and used as the contact force signal. The contact force signal is continuously compared with the preset contact threshold. When the contact force signal increases from zero or a minimum value to reach the preset contact threshold, it is determined that the tactile end has established a stable contact relationship with the positioning reference. At this time, a stop command can be sent to the motion control module immediately to terminate the approach operation and keep the tactile end in the current contact position.

[0036] S14, determine the position point corresponding to the tactile end when the approach operation stops as the calibration origin.

[0037] The calibration origin is the origin of the calibration coordinate system and serves as the starting reference point for calculating all spatial coordinates within the calibration coordinate system.

[0038] Specifically, after the approach operation stops, the position at which the approach operation stops is the physical position where the tactile end actually makes effective contact with the positioning reference surface. This position is not affected by equipment installation errors or motion errors, and is the most real and accurate reference position. Reading the current spatial coordinate value of the tactile end in the motion control system at this time and recording this coordinate value as the origin coordinate of the calibration coordinate system can reduce the impact of errors on the accuracy of the calibration position and complete the determination of the calibration origin.

[0039] S15, construct a calibration coordinate system based on the calibration origin and the preset positional relationship of the sample blocks.

[0040] Among them, the sample block position relationship refers to the pre-stored spatial position relationship data of each standard sample block relative to the calibration origin, which describes the offset distance and arrangement order of the installation position of each standard sample block on the calibration base relative to the origin in the X-axis, Y-axis and Z-axis directions.

[0041] It is understandable that the arrangement of the standard sample array on the calibration base is fixed, and the relative positional relationship of each sample relative to the preset positioning reference is known and stable over a long period of time. After determining the calibration origin, it is not necessary to measure the position of each standard sample individually. By converting the preset relative positional relationship, the position of all standard samples in the calibration coordinate system can be quickly obtained.

[0042] Specifically, after determining the calibration origin, preset sample block position relationship data can be retrieved from the storage module. This data contains the spatial offset information of each standard sample block relative to the calibration origin. The coordinate values ​​of the calibration origin and the relative offset of each standard sample block are superimposed to calculate the position coordinates of each standard sample block in the calibration coordinate system, thus completing the construction of the calibration coordinate system. After the calibration coordinate system is completed, its origin is located at the positioning reference, and the direction of each coordinate axis is determined by the layout direction of the sample block array. The contact position of each standard sample block can be accurately described by coordinate values.

[0043] The above implementation method can reduce the impact of tactile end-effector installation deviation on the consistency of multiple calibration results, so that each calibration process is performed under the same spatial reference, which can improve the positioning accuracy of the contact position and the calibration efficiency during the calibration operation.

[0044] S2, determine the target standard sample from the standard sample array according to the type of the tactile end, control the tactile end to perform corresponding calibration operations on each of the target standard sample based on the calibration coordinate system, and simultaneously collect multimodal calibration signals.

[0045] Among them, the target standard sample block refers to the standard sample block selected from the standard sample block array according to the current tactile end type, which needs to be calibrated one by one. The range of target standard sample blocks corresponding to different types of ends is different. For example, the gripper tactile fingertip corresponds to the standard gripping sample block, the thermal tactile head corresponds to the temperature calibration block, and the soft fingertip may correspond to various samples such as hardness sample block, elasticity sample block, roughness sample block, etc. The calibration operation refers to the specific active tactile action performed by the tactile end on the target standard sample block, including various types such as pressing, sliding, scraping, rolling, gripping, tapping, and thermal contact. The multimodal calibration signal refers to the general term for various physical quantity signals that are synchronously activated and acquired by multiple signal acquisition channels during the calibration operation, including multiple signals such as normal force signal, tangential force signal, displacement signal, vibration signal, temperature signal, or tactile image signal. The signals of each channel are aligned in time.

[0046] It is understandable that tactile defect re-inspection involves various active tactile actions such as pressing, sliding, scratching, and clamping. The sensor response mode of the tactile end is different under different actions. If only a general calibration process is used to cover all actions, it may not be possible to capture the unique signal response characteristics and deviation patterns of each action mode. Different types of tactile ends have different functional application ranges and can perform a limited number of detection actions. Therefore, standard samples applicable to the end can be selected as calibration objects according to the end type. Simultaneous acquisition of multimodal calibration signals is because tactile detection involves multiple physical dimensions such as force, displacement, vibration, and temperature. The signal of a single channel can only reflect part of the response characteristics, and there is a coupling relationship between signals of different channels. By acquiring them simultaneously, the correlation characteristics between signals of each channel can be fully captured.

[0047] Specifically, the acquired tactile end type can be used as a query condition to select a set of target standard samples that match the end type from the standard sample array. Then, guided by the calibration coordinate system, the tactile end is controlled to move sequentially to each target standard sample. For each target standard sample, a calibration action matching its physical characteristics is performed. The execution order of each target standard sample can be performed sequentially according to a preset access path. While performing the calibration action, multiple signal acquisition channels can be triggered to acquire multimodal calibration signals containing multiple physical quantities such as force, displacement, vibration, and temperature.

[0048] See Figure 3 This is a schematic diagram of a typical active tactile action provided by an embodiment of the present invention. The diagram uses six sub-diagrams (AF) to illustrate six core active tactile actions: tap detection, slide detection, swipe detection, roll detection, grip detection, and tap detection, as well as the corresponding tactile end-effector movement mode, contact direction, and contact force direction for each action. Figure 3 This can be used to illustrate the correspondence between different standard samples and different active tactile calibration operations. For example, a standard hardness sample can correspond to the tapping action shown in sub-figure A, where the force F is collected by vertically pressing the tip of a soft fingertip. n The springback curve between the displacement z and the hardness is determined. The standard roughness sample can correspond to the sliding action shown in sub-figure B or the rolling action shown in sub-figure D. The frictional force F is collected by the friction sliding head and the end of the roller, respectively. t Roughness calibration is performed using vibration signals. The standard burr sample block corresponds to the scraping action shown in sub-figure C. Burr calibration is completed by collecting the peak value of tangential force change by sweeping the end of the scraper along the edge of the sample block. The standard crack sample block corresponds to the scraping action shown in sub-figure C or the tapping action shown in sub-figure F. Crack and hollow features are calibrated by probing displacement response, edge force change characteristics, and vibration attenuation differences, respectively. The standard elastic sample block corresponds to the pressing action shown in sub-figure A or the tapping action shown in sub-figure F. Elastic parameters are calibrated by hysteresis curve and dynamic vibration response. The standard friction sample block corresponds to the constant force sliding action shown in sub-figure B. The friction coefficient is calibrated by calculating the measured friction coefficient by the ratio of tangential force to normal force.

[0049] In some embodiments, step S2, "determining the target standard sample from the standard sample array according to the type of the tactile end," includes the following steps: S21, retrieve the preset operation mapping relationship according to the tactile end type, and determine the calibration operation type corresponding to the tactile end type.

[0050] Among them, the operation mapping relationship refers to the pre-established and stored correspondence table between the type of tactile end and the calibrated operation type that the end can perform. For example, the operation mapping relationship corresponding to the fingertip of the gripper is the gripping operation, the operation mapping relationship corresponding to the thermal tactile head is the thermal contact operation, and the operation mapping relationship corresponding to the soft fingertip may include various operations such as pressing, sliding, and tapping. The calibrated operation type refers to the classification of tactile actions performed by the tactile end on the standard sample block, including pressing, sliding, scraping, rolling, gripping, tapping, and thermal contact.

[0051] Specifically, after obtaining the current tactile end type, the end type can be used as a search condition to query the pre-stored operation mapping relationship data table, extract all calibration operation types that have a mapping relationship with the current end type from the data table, and determine these calibration operation types as the calibration operation range that the current end needs to perform.

[0052] S22, according to the preset correspondence between the calibration operation type and the standard sample block, select the standard sample block corresponding to each calibration operation type from the standard sample block array as the target standard sample block.

[0053] Among them, the preset correspondence refers to the data table that is established in advance to describe the correspondence between different calibration operation types and standard sample types.

[0054] Understandably, the standard sample array contains various types of standard samples, each with different types of physical parameters marked on it. If the calibration operation type is not properly matched with the standard sample type, the acquired signal will not be able to establish a meaningful correspondence with the physical parameters marked on the sample. For example, the hardness sample is marked with a hardness grade value. Only by acquiring the contact force-displacement curve through a point-press operation can the measured signal characteristics be compared with the hardness value. If a clamping operation is performed on the hardness sample, the clamping force signal is acquired, which lacks a direct physical correlation with the hardness value and cannot be used to generate effective calibration parameters. Therefore, after determining which calibration operation types need to be performed at the current end, it is necessary to determine on which type of standard sample each operation type should be performed to ensure that there is a clear correspondence between the acquired signal and the physical parameters marked on the sample.

[0055] Specifically, based on the calibration operation types determined in S21, the preset correspondence between the calibration operation types and the standard samples can be queried item by item to determine the standard sample type corresponding to each calibration operation type. Then, all matching standard samples can be found and selected from the standard sample array, and these standard samples can be determined as the target standard sample set.

[0056] In some embodiments, step S2, "controlling the tactile end to perform corresponding calibration operations on each of the target standard samples based on the calibration coordinate system, and simultaneously acquiring multimodal calibration signals," includes the following steps: S23, Based on the calibration coordinate system, determine the target contact position corresponding to each target standard sample block.

[0057] The target contact position refers to the precise position that the tactile end needs to reach when performing calibration operations on each target standard sample block, which includes spatial coordinates and contact direction information. This position determines the spatial conditions under which the tactile end contacts the standard sample block.

[0058] Understandably, after determining the target standard samples, it's necessary to know the specific location of each sample and the posture in which the tactile end effector should contact it. The calibration coordinate system, already established in previous steps, describes the spatial layout of the entire array of standard samples. Therefore, the contact position corresponding to each target standard sample can be calculated using this coordinate system. Specifically, based on the established calibration coordinate system and the layout information of each target standard sample in the array, the target contact position corresponding to each target standard sample can be determined. This position information includes the spatial coordinates that the tactile end effector should reach when contacting the sample, as well as the contact direction and posture to be adopted.

[0059] Based on the above embodiments, step S23 can be implemented in the following ways: S231, convert the relative position information of each target standard sample block with respect to the calibration origin into the sample block position coordinates of each target standard sample block in the calibration coordinate system.

[0060] Among them, relative position information refers to the spatial offset of each standard sample block with the calibration origin as the reference point. The sample block position coordinates refer to the absolute coordinate values ​​of each target standard sample block in the established calibration coordinate system after combining the relative position information with the coordinate values ​​of the calibration origin. These coordinate values ​​can be directly used to control the tactile end for precise motion positioning.

[0061] It is understandable that the arrangement of the standard sample array on the calibration base is fixed, and the relative position information of each sample relative to the calibration origin is known and stable over a long period of time. By obtaining the position coordinates of the sample through coordinate transformation, it is not necessary to measure the position of each sample individually during each calibration process, which can improve calibration efficiency and ensure the accuracy of the sample position.

[0062] Specifically, the relative position information of each target standard sample block can be superimposed with the coordinates of the calibration origin using the origin of the calibration coordinate system as a reference to obtain the three-dimensional coordinates of the center point of each target standard sample block in the calibration coordinate system, i.e., the sample block position coordinates.

[0063] S232, determine the contact direction and contact posture of the tactile end contacting the target standard sample according to the calibration operation type of each target standard sample.

[0064] Among them, the contact direction refers to the spatial movement direction along which the tactile end approaches and contacts the target standard sample, such as perpendicular to the sample surface downwards, along the horizontal direction of the sample surface, or along the oblique direction of the sample edge, etc. The contact posture refers to the spatial angle and orientation of the tactile end's own axis when it contacts the target standard sample, describing the tilt angle and orientation at which the tactile end contacts the sample surface.

[0065] It is understandable that different types of calibration operations have different requirements for the spatial orientation of the tactile end. The pressing operation requires the tactile end to press down vertically in a direction perpendicular to the surface of the sample block so that the normal force sensor can accurately collect the contact force displacement curve. The scraping operation requires the tactile end to move along the edge of the sample block at a certain tilt angle so that the tangential force sensor can effectively capture the edge change signal. Therefore, the contact direction and contact posture that match the operation can be determined according to the calibration operation type corresponding to each target standard sample block.

[0066] Specifically, the calibration operation type corresponding to each of the determined target standard samples can be read one by one. Based on the standard contact method and spatial orientation parameters corresponding to each calibration operation type, the contact direction and contact posture that the tactile end should adopt when performing the calibration operation can be determined.

[0067] S233, based on the sample block position coordinates, the contact direction and the contact posture, determine the target contact position corresponding to each target standard sample block.

[0068] Understandably, the sample position coordinates typically represent the position of the center point of the target standard sample in the calibration coordinate system. However, when the tactile end effector actually contacts the target standard sample, its contact point is not always aligned with the center point. When the tactile end effector contacts the target standard sample vertically, the contact point is directly above the center point. But when the tactile end effector needs to perform a scraping operation at an angle or a sliding operation horizontally, the actual contact point of the tactile end effector will spatially offset relative to the center point of the target standard sample. If the sample position coordinates are directly used as the target contact position, it may cause the actual contact area between the tactile end effector and the target standard sample to deviate from the effective calibration surface, or even result in no contact with the target standard sample at all. The effective calibration surface refers to a specific surface portion on the target standard sample that is pre-designated as the calibration contact area. This surface is the actual physical bearing surface corresponding to the known physical parameters marked on the target standard sample. For example, the effective calibration surface of a standard hardness sample is the central area of ​​its upper surface. The tactile end collects the curve corresponding to the contact force and displacement by pressing on this area. The effective calibration surface of a standard burr sample is the surface where its edge protrusion structure is located. The tactile end collects the tangential force change signal by performing a scraping operation along this surface. By combining the contact direction and contact posture, the spatial offset correction of the sample position coordinates is performed so that the tactile end can accurately contact the effective calibration surface of the target standard sample, ensuring the accurate contact position between the tactile end and the target standard sample during the calibration operation.

[0069] Specifically, the coordinates of the sample block position can be used as a reference point. The spatial movement direction of the tactile end when it approaches the target standard sample block is determined according to the contact direction. The reference point is translated in the opposite direction of the spatial movement direction. The translation distance is determined according to the contact posture angle and the structural dimensions of the tactile end itself. This ensures that when the tactile end moves to the surface of the target standard sample block in the contact direction with the specified contact posture after translation, the contact area between the tactile end and the target standard sample block exactly covers the effective calibration surface of the target standard sample block. The three-dimensional coordinate point obtained after translation is the target contact position.

[0070] S24, obtain the preset operation parameters corresponding to the calibration operation type of the target standard sample block.

[0071] Among them, the preset operation parameters refer to a set of motion and control parameters that are pre-configured for each type of calibration operation to control the specific execution process of the calibration operation, including the magnitude of the target contact force, the speed of the action execution, the duration of the action, the displacement, the contact holding time, etc.

[0072] It is understandable that when the same type of calibration operation is performed on different target standard blocks, the requirements for parameters such as contact force, action speed and duration may differ. For example, when performing a point-press operation on a standard hardness block, it is necessary to set an appropriate target contact force to ensure that the contact force-displacement curve can be completely collected. When performing a point-press operation on a standard elastic block, it is necessary to set an appropriate contact holding time to collect the rebound curve and hysteresis area.

[0073] Specifically, based on the calibration operation type corresponding to each target standard sample block, preset operation parameters corresponding to that calibration operation type can be queried and extracted from the preset parameter configuration data. The extracted preset operation parameters, such as the magnitude of the target contact force, the speed of the action execution, the duration of the action, and the amount of displacement, can be used as the basis for subsequent operation instructions to control the tactile end to perform calibration operations.

[0074] S25, according to the preset operating parameters, control the tactile end to perform corresponding calibration operations on each of the target standard samples at each of the target contact positions.

[0075] It is understandable that step S23 has determined the target contact position corresponding to each target standard sample block and determined the spatial conditions under which the tactile end should contact each target standard sample block. Step S24 has obtained the preset operation parameters corresponding to each calibration operation type and clarified the force control conditions and movement rhythm under which the tactile end should perform the calibration operation. Therefore, by combining the target contact position and the preset operation parameters, the tactile end can be driven to move sequentially to the corresponding position of each target standard sample block and complete the calibration operation of each target standard sample block one by one according to the specified contact force, movement speed and duration.

[0076] For example, when the target standard sample is a hardness sample, the corresponding calibration operation can be a tapping operation; when the target standard sample is a roughness sample, the corresponding calibration operation can be a sliding operation; when the target standard sample is a standard burr sample, the corresponding calibration operation can be a scraping action; when the target standard sample is a standard crack / hollow sample, the corresponding calibration operation can be a probing, scraping, or tapping action; and when the target standard sample is a standard elastic sample, the corresponding calibration operation can be a tapping, holding, releasing, or tapping action.

[0077] S26, during the calibration operation, multiple signal acquisition channels are triggered to start synchronously, and physical quantity signals corresponding to each signal acquisition channel are acquired. The physical quantity signals that meet the preset quality conditions are determined as multimodal calibration signals.

[0078] Among them, the signal acquisition channel refers to the independent sensing unit in the tactile end that is responsible for acquiring signals of different physical quantities and its corresponding data acquisition link, such as the normal force acquisition channel, the tangential force acquisition channel, and the displacement acquisition channel. The physical quantity signal refers to the raw sensing data output in real time by each signal acquisition channel during the calibration operation. The preset quality condition refers to the pre-set index threshold used to judge whether the acquired physical quantity signal meets the usable standard, including the signal-to-noise ratio threshold, the signal amplitude range, and the signal integrity index. Signals that meet the conditions are considered to be of qualified quality and can be used for subsequent parameter generation. Signals that do not meet the conditions need to be rejected or triggered for re-acquisition.

[0079] It is understandable that tactile detection involves multiple physical dimensions such as force, displacement, vibration, and temperature. These different dimensions are correlated in time. If each channel is triggered independently, the signals will be out of sync in time, making it impossible to accurately analyze the correspondence between physical quantities at the same moment. Therefore, by sending trigger signals to all signal acquisition channels simultaneously, each channel can start acquisition at the same time, ensuring that the signals of each channel are aligned in the time dimension. Furthermore, during the acquisition process, the signals may be affected by accidental factors such as instantaneous noise and sensor jumps, causing some signals to fluctuate abnormally or have a low signal-to-noise ratio. If these poor-quality signals are used for subsequent parameter generation, it may affect the accuracy of the calibration parameters. By setting preset quality conditions to filter the acquired signals, unqualified signals can be filtered out, ensuring that the signals entering the subsequent processing stage have sufficient reliability.

[0080] Specifically, a synchronous trigger signal can be sent to all signal acquisition channels at the same time as the calibration operation begins, so that each channel starts data acquisition at the same moment, acquires the physical quantity signals output by each channel in real time, evaluates the quality of the acquired physical quantity signals of each channel, determines whether each signal meets the preset quality conditions such as signal-to-noise ratio, amplitude range and signal integrity, and retains the physical quantity signals that meet the conditions as multi-mode calibration signals for subsequent signal processing and parameter generation.

[0081] The above implementation method can improve the synchronization accuracy between different physical quantity signals and filter out unqualified signals, ensuring that the signals entering the subsequent processing stage have sufficient reliability.

[0082] S3, Generate a set of calibration parameters associated with each target standard sample and its corresponding calibration operation based on the signal deviation between the multimodal calibration signal and the target standard parameter corresponding to each target standard sample.

[0083] Among them, the target standard parameter refers to the known physical parameter values ​​pre-marked on each target standard sample block, which are used as the comparison benchmark during the calibration process. The signal deviation refers to the degree of deviation between the signal feature value extracted from the multimodal calibration signal and the corresponding target standard parameter. It can be reflected in various forms such as absolute difference, relative ratio, fitting residual, and system offset. The absolute difference is the difference obtained by directly subtracting the signal feature value from the target standard parameter. The relative ratio is the proportion obtained by dividing the signal feature value by the target standard parameter. The fitting residual is the difference between the signal feature value and the corresponding theoretical prediction value on the fitting curve in multi-point calibration. The system offset is the overall offset trend of the signal feature value relative to the target standard parameter under specific conditions. The calibration parameter refers to the correction data calculated and generated during the calibration process based on the deviation between the multimodal calibration signal collected by the tactile end and the known physical parameters of the standard sample block. It is used to correct the tactile detection signal in the defect re-inspection stage. The calibration parameter can include zero-point offset compensation, gain correction coefficient, defect judgment threshold, temperature drift compensation coefficient, friction compensation coefficient, etc. The calibration parameter set refers to a set composed of multiple calibration parameters.

[0084] It is understandable that the acquired multimodal calibration signal is the actual response of the tactile end under the current tactile end and the current calibration operation, while the target standard parameter is the objective true value marked on the standard sample. The difference between the two can reflect the degree of deviation of the measurement system under the current calibration operation conditions. These deviations may come from various factors such as sensor zero drift, ambient temperature drift, and contact head wear. The deviation patterns are different under different calibration operations. Therefore, the same set of deviation correction parameters cannot be applied to all operation types. The signal deviation can be used as the input basis for generating calibration parameters because the difference itself has already eliminated the standard true value and only retains the information of measurement deviation. The calibration parameters generated based on the difference can directly correct the deviation of the measurement system. Associating the generated set of calibration parameters with the target standard sample and the corresponding calibration operation is to give each set of parameters a clear applicable range label. When calling, the corresponding parameter set can be accurately located according to the end type and operation type, thereby avoiding parameter misuse.

[0085] Specifically, the acquired multimodal calibration signals can be preprocessed, such as by denoising and filtering. Signal feature values ​​that characterize the tactile end response characteristics can be extracted from the processed signals. Then, the extracted signal feature values ​​are compared with the target standard parameters marked on the corresponding standard sample block to calculate the signal deviation amount that reflects the degree of deviation between the two. A calibration parameter set containing correction information is generated based on the signal deviation amount, and the parameter set is associated with the corresponding target standard sample block and calibration operation type.

[0086] Based on the above embodiments, step S3 can be implemented in the following ways: S31, preprocess the physical quantity signals corresponding to each signal acquisition channel in the multimodal calibration signal to obtain the processed signals corresponding to each signal acquisition channel.

[0087] Among them, the processed signal refers to the signal obtained after preprocessing the original physical quantity signals acquired by each signal acquisition channel, which has improved quality. The preprocessing operations include removing random noise from the signal, filtering out high-frequency interference components, correcting baseline drift, and removing abnormal jump points. The processed signal waveform is smoother and can more accurately reflect the real process of physical interaction between the tactile end and the target standard sample.

[0088] It is understandable that the raw physical quantity signals acquired during the calibration process will inevitably be affected by factors such as the sensor's own electronic noise, environmental vibration interference, and electromagnetic interference, resulting in defects such as glitches, random fluctuations, or baseline offsets superimposed on the signal waveform. These defects will cause the true physical response characteristics in the signal waveform to be masked or distorted by noise, affecting the accuracy of the calibration parameters. Since the types of physical quantities acquired by each channel in the multimodal calibration signal are different, the sources of noise and interference characteristics are also different. For example, force signals are easily affected by mechanical vibration interference, and temperature signals are easily affected by environmental thermal radiation interference. Therefore, it is necessary to perform targeted preprocessing on each channel signal to make the effective characteristics in the signal waveform that are associated with the target standard parameters clearer.

[0089] Specifically, the determined multimodal calibration signal can be separated according to the signal acquisition channel to obtain an independent original physical quantity signal sequence for each channel. Preprocessing operations such as denoising, filtering and baseline correction are then performed on the original physical quantity signal of each channel in sequence to remove random noise and high-frequency interference in the signal, correct the baseline offset of the signal, and make the effective characteristics of the signal waveform more prominent, so as to obtain the processed signal corresponding to each signal acquisition channel.

[0090] S32, Identify signal feature points from the processed signal that are associated with the target standard parameters corresponding to the target standard sample block, and extract the signal feature values ​​corresponding to the signal feature points.

[0091] Among them, signal feature points refer to specific morphological positions on the waveform curve of the processed signal that correspond to the physical properties represented by the target standard parameters, such as the maximum displacement on the contact force-displacement curve, the frequency of the main peak on the vibration power spectrum, and the sudden peak on the tangential force time sequence signal. Signal feature values ​​refer to the specific values ​​read at the signal feature point positions. These values ​​directly reflect the magnitude of the measurement response of the tactile end to the specific physical properties on the target standard sample under the current calibration operation, such as the maximum contact force value, the main peak frequency value, and the sudden amplitude of the tangential force.

[0092] It is understandable that the normal force time series, tangential force time series, displacement time series, vibration time series, temperature time series, etc., contained in the multimodal calibration signal are essentially continuous data sequences formed by arranging physical quantities continuously collected by the sensor in chronological order during the calibration process. When plotted on the time axis or displacement axis, they appear as a continuously changing waveform curve. The physical properties represented by the target standard parameter are not reflected at any arbitrary position on the entire waveform curve, but rather at specific morphological positions on the curve. For example, in hardness calibration, the target standard parameter is the hardness grade value, and the physical response characteristic associated with this parameter is the contact force-displacement ratio. The force value corresponding to the maximum displacement in the displacement curve is the signal feature point. In roughness calibration, the target standard parameter is the roughness value. The physical response feature associated with this parameter is the frequency value corresponding to the main peak in the vibration power spectrum. The frequency position of the main peak in the power spectrum is the signal feature point. Therefore, the accurate position of the signal feature point associated with the target standard parameter can be located in the waveform curve of the processed signal first, and then the corresponding value can be extracted from that position as the signal feature value. This can ensure that the extracted feature value can correctly reflect the measurement response of the tactile end to the specific physical properties on the target standard sample.

[0093] S33, compare the signal feature value with the target standard parameter to determine the signal deviation between the signal feature value and the target standard parameter.

[0094] Understandably, signal characteristic values ​​are measured values ​​obtained by locating feature points and extracting values ​​from the processed signal. They reflect the magnitude of the physical quantity actually perceived by the tactile end in the current state. Target standard parameters, on the other hand, are known physical parameter values ​​pre-marked on a standard sample block, representing the objective true value of that physical quantity. Ideally, the signal characteristic value measured by a perfectly accurate measurement system should be equal to the target standard parameter, with zero difference between the two. However, in practical use, due to factors such as sensor zero drift, sensitivity offset, ambient temperature drift, and contact head wear, signal characteristic values ​​often deviate from the target standard parameter. The purpose of comparing signal characteristic values ​​with the target standard parameter is to separate the measurement deviation from the measured value, making the deviation an independent and quantifiable value. Depending on the different calibration parameter generation requirements, the deviation can be expressed as... Different representations are used. For example, when calculating zero-point offset compensation, the absolute difference can be used to intuitively reflect the magnitude of the offset between the measured value and the standard value. When calculating gain correction coefficient, the relative ratio can be used to reflect the proportional relationship between the measured value and the standard value. When evaluating calibration quality or performing multi-point fitting, the fitting residual can be used to reflect the degree of deviation of a single measurement point from the overall response trend. When analyzing the overall offset law of the sensor under specific conditions, the system offset can be used. By comparing the signal characteristic value with the target standard parameter and calculating the signal deviation, the measurement deviation can be separated from the measured value, making the deviation an independent and quantifiable value. This facilitates the subsequent generation of correction parameters such as zero-point offset compensation, gain correction coefficient, and defect judgment threshold.

[0095] S34, Generate a calibration parameter set based on the signal deviation, and store the calibration parameter set in association with the end type, action type, sample type, calibration version, or validity period.

[0096] Based on the above embodiments, step S34 can be implemented in the following ways: S341, when the target standard parameter is determined as the reference value, the corresponding signal deviation is the zero-point offset compensation amount of the tactile end under the current calibration operation.

[0097] Among them, the reference value is a known reference value used in the calibration process to detect whether the sensor of the tactile end has zero-point drift. The selection of the reference value depends on the type of physical quantity being calibrated. If the physical quantity has a clear zero state, such as the force signal should be zero when the tactile end is not in contact with any object, then the reference value is zero. If the physical quantity is classified by grade and does not have a natural zero state, such as hardness grade or roughness grade, then the reference value is the parameter value of the lowest grade marked on the standard sample. The current calibration operation refers to the specific calibration action that the tactile end is performing on the current target standard sample, such as a tapping operation on the standard hardness sample or a sliding operation on the standard roughness sample. The zero-point offset compensation amount refers to the compensation value measured under the reference state, which is used to correct the output deviation of the tactile end sensor at the reference starting position.

[0098] It is understandable that sensors may experience zero-point drift after long-term use. This means that when the tactile end is not in contact with any object or with a standard sample in a reference state, the sensor output value deviates from the theoretical value that it should output. To measure this drift, the tactile end can be measured in a known reference state. The signal deviation between the measured signal characteristic value and the reference value is used as the zero-point offset compensation amount. This is because, in the reference state, this difference is directly equal to the sensor's output deviation at the reference starting point, without the need for additional conversion. In subsequent defect re-inspection, this compensation amount is used to correct the detection signal, which can calibrate the sensor's output starting point back to the correct position and reduce the impact of zero-point drift on the measurement results.

[0099] S342, determine the signal deviation corresponding to multiple target standard parameters with different values ​​as candidate difference quantities, and based on the numerical correspondence between the target standard parameters and each candidate difference quantity, fit and calculate the gain correction coefficient of the tactile end under the current calibration operation.

[0100] Among them, the candidate difference quantity refers to the set of signal deviation quantities selected from all signal deviation quantities that correspond to multiple different numerical target standard parameters. The numerical correspondence relationship refers to the numerical correspondence distribution relationship formed in a two-dimensional coordinate system with the numerical value of the target standard parameter as the independent variable and the numerical value of the candidate difference quantity as the dependent variable. Each set of data points can be connected to form a curve reflecting the trend of the difference quantity with the standard parameter. The fitting calculation refers to calculating a curve that best represents the overall change law of the data points based on the distribution trend of the data points in the numerical correspondence relationship using a preset mathematical fitting method, and extracting the gain correction coefficient from the characteristic parameters of the curve. The gain correction coefficient is a multiplicative correction coefficient used to correct the measurement sensitivity deviation of the tactile end under different input levels. The gain correction coefficient can reflect the degree to which the sensitivity of the measurement system deviates from the ideal sensitivity.

[0101] Understandably, the signal deviation corresponding to a single target standard parameter can only reflect the magnitude of the deviation at that specific level, and cannot reflect the overall pattern of the deviation changing with the physical quantity. The essence of gain deviation is that the sensitivity of the measurement system deviates from the ideal value, which is manifested as a deviation in the proportional relationship between the signal characteristic value and the target standard parameter. This deviation will be regularly reflected in multiple standard parameters at different levels. The signal deviation corresponding to multiple target standard parameters with different values ​​is selected as candidate difference quantities because these difference quantities contain information about the change of sensitivity deviation with the input level. By constructing a numerical correspondence with the target standard parameter as the horizontal axis and the candidate difference quantities as the vertical axis, and performing fitting calculations on each data point, the gain correction coefficient reflecting the sensitivity deviation can be extracted from the overall data distribution, ensuring the accuracy of gain correction.

[0102] S343, the judgment boundary value between the signal feature values ​​corresponding to two adjacent target standard parameters with different values ​​is determined as the defect judgment threshold of the tactile end under the current calibration operation.

[0103] Among them, the judgment boundary value refers to a dividing boundary determined according to the numerical distribution relationship between the signal characteristic values ​​corresponding to the target standard parameters of two adjacent different levels. This boundary is used to effectively separate the measurement response intervals of two adjacent levels. The defect judgment threshold refers to the judgment boundary value used to distinguish different defect levels of the tested object. When the value of the detection signal collected in the defect re-inspection falls on one side of the threshold, it is judged as a certain level, and when it falls on the other side, it is judged as another level.

[0104] Understandably, target standard samples typically have multiple levels of known physical parameters. For example, a standard crack sample might be labeled with four levels: no crack, micro-crack, medium crack, and large crack. Each level corresponds to a signal characteristic value acquired through calibration. To distinguish these four levels, a boundary line needs to be set between adjacent levels, i.e., three defect judgment thresholds. The determination of the judgment boundary value can be done in several ways. For example, the average value of the signal characteristic values ​​of two adjacent levels can be taken as the boundary value, so that the boundary value is exactly in the middle of the measurement response of the two levels, maximizing the balance between the two levels. Alternatively, it can be based on the statistical distribution characteristics of the signal characteristic values ​​of the two levels. The value that minimizes the error rate in distinguishing between the two levels is taken as the cutoff value. Alternatively, a cutoff value can be set near a certain level based on actual testing needs to reduce the risk of missed or false judgments at that level. The cutoff value is determined as the defect judgment threshold because it is directly based on the signal feature values ​​actually collected during the calibration operation and matches the current response characteristics of the tactile end. This more accurately reflects the device's current ability to distinguish between different levels of physical attributes than using a preset default threshold. The more known physical parameter levels of the target standard sample, the more defect judgment thresholds are generated. These thresholds together constitute a set of judgment boundaries used to distinguish different levels of the physical attribute under the current calibration operation.

[0105] Specifically, all target standard parameters are sorted by numerical value. The signal feature values ​​corresponding to each of the two adjacent target standard parameters after sorting are extracted. A judgment boundary value is determined between these two signal feature values. This boundary value is used to distinguish between the two adjacent levels. The method of determining the judgment boundary value can be selected according to actual needs. For example, the average value of the two signal feature values ​​can be calculated as the boundary value, or the value that achieves the best distinction effect can be selected based on the distribution characteristics of the signal feature values ​​of the two levels. The determined judgment boundary value is used as the defect judgment threshold between the adjacent levels. This operation is performed sequentially for all adjacent levels to obtain a set of defect judgment thresholds under the current calibration operation.

[0106] S344, associate the zero-point offset compensation amount, the gain correction coefficient, and the defect judgment threshold with the action type of the current calibration operation to generate a calibration parameter set associated with the current target standard sample and the corresponding calibration operation.

[0107] Understandably, different calibration operations correspond to different action types, and the response characteristics and deviation patterns of the tactile end effector vary under different action types. Therefore, the generated zero-point offset compensation, gain correction coefficient, and defect judgment threshold are also different. Associating these calibration parameters with the action type of the current calibration operation is to give each set of calibration parameters a clear label of its scope of use. In the subsequent defect re-inspection stage, when a detection operation of a certain action type is performed, the calibration parameter set associated with it can be directly retrieved and called according to the action type, ensuring that the calibration uses parameters specifically generated for that action type, rather than general default parameters or erroneous parameters associated with other action types.

[0108] In some embodiments, the calibration parameters also include temperature drift compensation coefficient, friction compensation coefficient, etc.

[0109] Temperature drift compensation coefficient refers to a compensation parameter used to correct the impact of changes in ambient temperature on the output of non-temperature sensors in tactile extremities. Changes in ambient temperature can alter the physical characteristics of internal sensor components, causing force sensors, displacement sensors, and other sensors to output different signal values ​​when measuring the same physical quantity. The temperature drift compensation coefficient describes the offset of the sensor's output signal for every unit change in temperature.

[0110] Understandably, the internal components of force sensors and displacement sensors in tactile excision devices are highly sensitive to temperature changes. When the ambient temperature changes, even if the tactile excision device is in contact with the exact same object, the signal values ​​output by the sensors will drift. For example, when the ambient temperature rises, the zero point of the force sensor may drift in the positive direction, resulting in a systematically larger measured force value. This temperature-introduced error is different in nature from the measurement bias of the sensor itself and requires separate correction. By performing a thermal contact operation on a temperature calibration block, the signal output of the tactile excision device under different temperature conditions can be obtained. By comparing these signals with the known temperature values ​​of the temperature calibration block, the pattern of signal change with temperature can be calculated, thereby obtaining the temperature drift compensation coefficient.

[0111] The friction compensation coefficient is a parameter used to correct deviations in friction measurement values ​​caused by wear on the contact head surface or changes in material properties when the tactile end performs actions involving friction, such as sliding or scraping.

[0112] Understandably, during long-term use, the surface of the contact head of a tactile sensor will wear down due to repeated friction. The surface roughness and frictional characteristics of the worn contact head differ from its factory condition, leading to a deviation between the measured tangential frictional force and the actual value under the same normal force. For example, after wear, the surface becomes rougher, and the frictional force measured under the same normal force will be greater than that of a new contact head. If the initial calibration parameters are still used for correction, it will lead to misjudgment of the frictional characteristics of the tested object's surface. By performing a constant-force sliding operation on a standard friction sample and comparing the currently measured frictional force signal with the theoretical frictional force value corresponding to the known friction coefficient marked on the standard friction sample, the friction compensation coefficient under the current contact head condition can be calculated.

[0113] Specifically, during the calibration process, the tactile end is controlled to perform a constant force sliding operation on a standard friction sample block, and normal force signals and tangential force signals are collected. The ratio of the tangential force signal to the normal force signal is used as the measured friction coefficient. The measured friction coefficient is compared with the known friction coefficient marked on the standard friction sample block, and the friction compensation coefficient is calculated based on the difference between the two.

[0114] In addition, calibration parameters can be stored in a structured manner according to end type, action type, and calibration version, recording calibration timestamps and expiration dates, and supporting historical version traceability. That is, the generated calibration parameter set can also be associated with calibration timestamps, ambient temperature records, and parameter expiration dates to form a unique calibration version identifier. Furthermore, the most recent N historical calibration versions can be retained for subsequent traceability and reproduction of test data. When the calibration parameters exceed their expiration date, the recalibration process will be automatically triggered.

[0115] Through the above implementation methods, a more comprehensive correction of the deviation of the tactile end system can be achieved, so that the calibrated detection signal can truly reflect the physical state of the object being measured.

[0116] S4. Based on the tactile end type corresponding to the defect re-inspection task and the action type corresponding to the calibration operation, call the matching calibration parameter set to calibrate the tactile detection signal collected during the defect re-inspection process.

[0117] Among them, the defect repair task refers to the inspection task in actual industrial production, which is to conduct contact physical confirmation of suspected defect areas marked by the initial inspection on the surface of the product. During the re-inspection process, corresponding tactile detection actions need to be performed according to different defect types. The action type refers to the category of calibration operation performed by the tactile end on each target standard sample during the calibration process, which is used to distinguish different active tactile detection methods. Action types include pressing, sliding, scraping, rolling, clamping, tapping and thermal contact, etc. The tactile detection signal refers to the original physical quantity signal collected in real time by each signal acquisition channel when the tactile end contacts the surface of the tested product during the defect re-inspection process.

[0118] Understandably, the generated calibration parameter set has already established a correlation with the type of tactile end and the action type of the calibration operation. The defect re-inspection task also has these two attributes, namely, each re-inspection task uses a specific tactile end and performs a specific detection action. Therefore, these two attributes can be used as retrieval conditions to accurately locate the parameter set that matches the current task from the stored calibration parameter set. The tactile detection signal collected during the defect re-inspection process is the raw signal, which contains systematic errors introduced by various factors such as sensor zero drift, environmental temperature drift, and contact head wear. If the raw signal is used directly for defect judgment, the systematic errors will be superimposed on the real physical response, leading to misjudgment or missed judgment. The raw detection signal can be calibrated by matching calibration parameters to reduce systematic errors before defect judgment, making the detection results closer to the real physical state of the tested object. Furthermore, the calibrated detection results obtained under different times, different ends, and different operating conditions are comparable.

[0119] Specifically, the type of tactile end effector currently used in the defect re-inspection task and the action type of the re-inspection action currently being performed can be obtained first. Based on the obtained tactile end effector type and action type, a set of calibration parameters that simultaneously meets the two matching conditions can be retrieved from the stored calibration parameter set. Using the correction parameters contained in the called calibration parameter set, the tactile detection signal collected in real time during the defect re-inspection process can be corrected to obtain the calibrated detection signal. Furthermore, the defect can be judged based on the calibrated detection signal, and the defect re-inspection result can be output.

[0120] Based on the above embodiments, step S4 can be implemented in the following ways: S41, obtain the tactile end type corresponding to the defect re-inspection task, and the action type of the current re-inspection action.

[0121] Understandably, when a defect re-inspection task is executed, it will use a specific type of tactile end effector and perform a specific type of detection action. The type of tactile end effector is obtained because different types of tactile end effectors have different contact characteristics and deviation patterns. The calibration stage has already generated a dedicated set of calibration parameters for each end effector type. The action type is obtained because even for the same type of tactile end effector, the sensor response characteristics are different when performing different detection actions. The calibration stage has already generated corresponding correction parameters for each action type. Only by obtaining both of these pieces of information can the parameter set that perfectly matches the current re-inspection task be accurately retrieved from the stored set of calibration parameters, thus avoiding parameter misuse.

[0122] S42, based on the obtained tactile end type and action type, retrieve and call the matching calibration parameter set, and extract the zero-point offset compensation amount, gain correction coefficient and defect judgment threshold from it.

[0123] Understandably, the calibration parameter set was already associated with the tactile end type and motion type during generation. Each calibration parameter record is labeled with its applicable end type and motion type. Using the tactile end type and motion type as joint search conditions, the set of parameters that perfectly matches the current re-inspection task can be found in all the stored calibration parameter sets. The extracted parameters include three items: zero-point offset compensation, gain correction coefficient, and defect judgment threshold. This is because these three parameters have different correction functions: the zero-point offset compensation is used to correct the zero-position deviation of the sensor, the gain correction coefficient is used to correct the sensitivity deviation of the sensor, and the defect judgment threshold is used to update the boundary standard for defect judgment.

[0124] S43, based on the zero-point offset compensation amount and the gain correction coefficient, the tactile detection signal collected during the defect re-inspection process is subjected to offset processing and scale correction to obtain the calibrated target detection signal.

[0125] Among them, bias processing refers to the operation of subtracting the zero-point bias compensation amount from the value of the tactile detection signal, which is used to reduce the fixed offset error introduced by the zero-point drift of the sensor and bring the reference zero point of the signal back to the correct position. Scale correction refers to the operation of multiplying the signal value after bias processing by the gain correction coefficient, which is used to reduce the proportional error introduced by the sensor sensitivity deviation and restore the signal change amplitude to be consistent with the change amplitude of the real physical quantity.

[0126] It is understandable that the tactile detection signal collected during the defect re-inspection process is the raw signal, which contains systematic errors such as sensor zero drift and sensitivity deviation. Zero drift causes the signal to deviate by a fixed amount. Even if no defect is detected, the signal value may deviate from zero. Sensitivity deviation causes the signal change ratio to be distorted. For example, if the real physical quantity changes by 10 units, the sensor output signal may only change by 8 units. These two types of errors are different in nature and are corrected in different ways. Specifically, the tactile detection signal collected in real time by each signal acquisition channel during the defect re-inspection process is first subtracted from the extracted zero-point offset compensation amount to eliminate the zero-point offset of the signal and obtain the offset-processed signal. Then, the offset-processed signal is multiplied by the gain correction coefficient to correct the signal response ratio and obtain the calibrated target detection signal. The target detection signal obtained after offset processing and scale correction has a correct zero-point reference and accurate change ratio, which can truly reflect the physical state of the measured object.

[0127] S44, dynamically update the judgment boundary of the anomaly judgment module in the defect re-inspection task according to the defect judgment threshold, and output the defect re-inspection result based on the corrected target detection signal and the updated judgment boundary.

[0128] The anomaly determination module is a decision-making unit in the defect re-inspection process responsible for determining whether there are defects on the product surface and the level of defects based on the detection signals. It has multiple levels of determination boundaries to classify different defect types or levels. The determination boundary is the signal threshold dividing line used in the anomaly determination module to distinguish different defect levels or categories. Different determination conclusions are corresponding to different detection signals falling into different intervals. The defect re-inspection result is the conclusion of the anomaly determination module on the surface defect status of the tested product based on the target detection signal and the determination boundary, including whether there are defects, the type of defects, and the severity level of defects.

[0129] It is understandable that the anomaly detection module usually has a set of default detection boundaries. However, these default boundaries are determined during the initial setup. As the sensor state and environmental conditions change, the default boundaries may no longer match the actual response characteristics of the current tactile end. The defect detection threshold generated during the calibration stage is calculated based on the actual response characteristics of the current tactile end under specific operations, reflecting the device's current true signal discrimination capability. Updating the detection boundaries of the anomaly detection module with the defect detection threshold, instead of directly using the default boundaries, is to ensure that the detection criteria match the current actual state of the device and avoid misjudgment or missed judgment due to inaccurate criteria caused by device state drift.

[0130] It should be noted that the tactile detection signal collected during the defect re-inspection process is a continuous waveform that changes over time. Each moment on the waveform has a value, but most of these values ​​are only intermediate and are not directly used for judgment. What is actually used for judgment are one or more signal feature values ​​extracted from the calibrated target detection signal, such as the maximum contact force value in the tapping operation, the average friction force in the sliding operation, and the sudden peak value of the tangential force in the scraping operation. The judgment boundary is the threshold set for these signal feature values. A set of judgment boundaries divides the value of the signal feature into multiple intervals, and each interval corresponds to a judgment conclusion. For example, if the signal feature value falls in a lower interval, it is judged as normal, and if it falls in a higher interval, it is judged as a defect. Different intervals at different heights correspond to different levels of defects. Therefore, the judgment boundary is a set of fixed thresholds corresponding to a specific signal feature type. A defect re-inspection task usually corresponds to a set of judgment boundaries, which are used to compare with the signal feature values ​​corresponding to that task.

[0131] Specifically, the extracted defect judgment threshold can be written into the anomaly judgment module in the defect re-inspection task, replacing the original default judgment boundary of the module, and completing the update of the judgment boundary. Then, from the obtained calibrated target detection signal, the signal feature value corresponding to the current re-inspection action is extracted. The extracted signal feature value is compared with the updated judgment boundary one by one. The defect category and defect level of the surface of the tested product are determined according to the interval in which the signal feature value falls, and the defect re-inspection result is output.

[0132] Furthermore, calibration parameters can also be used for tactile detection feature normalization, contact force threshold setting, temperature drift compensation, friction compensation, confidence correction, and fusion judgment weight adjustment. For example, after feature extraction is completed, the system calls the sensitivity coefficient and friction compensation coefficient in the currently valid calibration parameters to normalize the extracted tactile detection features such as contact force-displacement slope, friction coefficient, and vibration power spectrum features, eliminating the absolute value differences of feature quantities between different probes and different ends, making the feature quantities comparable across probes. The contact force safety threshold in the calibration parameter set can be used to update the protection threshold of the compliant floating mechanism to avoid problems during detection. For workpieces or probes damaged by pressure, the temperature drift compensation coefficient and friction compensation coefficient can be used to dynamically correct the influence of ambient temperature and contact head wear on the detection signal. They can also be used to correct the confidence level of defect re-inspection results based on the fitting residuals and remaining validity period of the calibration parameters. The fitting residuals refer to the difference between the measured signal feature values ​​collected by the tactile end for different grade standard samples during multi-point calibration and the theoretical predicted values ​​of the corresponding points on the sensor's theoretical response curve obtained by fitting these measured values. Furthermore, the weights of each sensing channel can be dynamically adjusted in multimodal fusion judgment to further improve the accuracy of the detection results.

[0133] The above implementation method ensures that the judgment criteria always match the current actual response characteristics of the tactile end, thereby improving the accuracy and reliability of defect re-inspection of industrial products.

[0134] Based on the above steps, this solution also includes the following embodiments: A1, during the calibration operation, the contact force signal of the tactile end and the signal quality index of the multimodal calibration signal are monitored in real time.

[0135] Among them, signal quality indicators refer to several quantitative parameters used to evaluate whether the acquired multimodal calibration signals meet the usable standards, including signal-to-noise ratio, signal amplitude range, signal integrity, etc. Signal-to-noise ratio is the ratio of effective signal power to noise power, reflecting the proportional relationship between effective information and interference noise in the signal. Signal amplitude range is the difference between the maximum and minimum values ​​in the signal waveform, reflecting whether the dynamic response amplitude of the signal is within a reasonable range. Signal integrity refers to whether the signal waveform is continuous and complete, and whether there are data loss, breakpoints, or abnormal jumps.

[0136] Understandably, the calibration process is not always smooth and some common anomalies may occur. For example, abnormal contact force may occur. When the tactile end is close to the standard sample, the contact force may suddenly increase beyond the safe range due to positioning deviation, foreign objects on the sample surface, or mechanical failure. If not detected and handled in time, it may damage the tactile end or the standard sample. Abnormal signal quality may also occur. During the acquisition process, the sensor may be affected by factors such as transient electromagnetic interference, vibration and shock, or its own failure, resulting in excessive noise, abnormal amplitude, or incomplete waveform in the output signal. If these poor-quality signals enter the subsequent parameter generation stage, it will lead to inaccurate calibration parameters. Therefore, monitoring the contact force signal and signal quality indicators simultaneously during the calibration process can detect anomalies in time and provide triggering conditions for subsequent anomaly handling.

[0137] Specifically, during the entire calibration process, the contact force value of the normal force sensor on the tactile end is read in real time at a fixed sampling frequency. At the same time, the physical quantity signals output by each signal acquisition channel are evaluated in real time, and quality indicators such as signal-to-noise ratio, amplitude range, and integrity of each signal are calculated.

[0138] A2, when the contact force signal is detected to exceed the preset overload threshold, the tactile end is controlled to exit the current calibration operation and re-execute the approach operation before retrying.

[0139] Among them, the preset overload threshold refers to the pre-set safe upper limit of contact force. Exceeding this value means that the contact force between the tactile end and the target standard sample has exceeded the force control range required for normal calibration, which may cause damage to the tactile end or the standard sample.

[0140] Understandably, a contact force exceeding the preset overload threshold usually indicates an abnormal contact situation, such as a positioning deviation causing the tactile end to hit the edge or sidewall of the sample block instead of contacting the valid calibration surface of the sample block as expected. In this case, if the calibration operation continues, not only will the acquired signal fail to reflect the correct physical response, but the tactile end or the standard sample block may also be damaged due to excessive contact force. Therefore, it is necessary to immediately stop the current operation, allow the tactile end to return to a safe position, and then re-execute the approach operation to try to approach the target standard sample block again with the correct path. The retry mechanism exists because not all overloads are caused by hardware failures; many times it is just an occasional positioning deviation, which can be restored to normal by approaching again.

[0141] Specifically, when the value of the contact force signal detected in real time exceeds the preset overload threshold, an emergency stop command is immediately sent to the motion control module to control the tactile end to exit the current contact state in the opposite direction of the approach direction and return to a safe position above the target standard sample block. Then, the approach action is re-executed according to the standard approach operation procedure to complete the retry.

[0142] A3. When the signal quality index is detected to be unsatisfactory, the current target standard sample block is marked as calibration abnormal and a prompt message is generated.

[0143] Among them, calibration anomaly refers to the current calibration operation of the target standard sample being judged as unqualified due to signal quality problems. This indicates that the multimodal calibration signal acquired this time is unreliable. The prompt message is a notification issued to the operator or the upper control system, which includes the anomaly type, the target standard sample number where the anomaly occurred, and the suggested handling measures.

[0144] Understandably, preset quality conditions are a series of pre-defined criteria used to determine whether the acquired multimodal calibration signal meets usable standards. These criteria may include a signal-to-noise ratio not lower than a certain threshold, signal amplitude within a preset range, a complete signal waveform without breaks or missing segments, and the absence of abnormal jumps or abrupt peaks in the signal. If any of these signal quality indicators fails to meet the corresponding criteria, the signal quality is deemed substandard. Even with normal contact force, the acquired signal itself may still be substandard due to sensor transient noise, electromagnetic interference, or poor contact. For example, the sensor may experience excessive signal noise due to transient electromagnetic interference. Poor contact between the tactile end and the target standard sample can lead to excessively low signal amplitude or incomplete waveform. Transient failures in the data acquisition link can also cause signal interruptions or data loss. Substandard signal quality means that the effective information in the signal is masked by noise or the signal waveform is incomplete. The signal feature values ​​extracted from such signals will deviate from the true values. If they are still used for parameter generation, the calibration parameters will be inaccurate. Therefore, calibration operations with substandard signal quality need to be marked as abnormal and prompts should be generated to inform the operators so that they can check whether there is contamination on the surface of the target standard sample, whether the tactile end needs to be cleaned, or whether there are other hardware problems.

[0145] Specifically, when any signal quality index obtained from real-time evaluation exceeds the corresponding preset quality threshold, the current calibration operation of the target standard sample is marked as a calibration anomaly state, and a prompt message containing an anomaly type description, target standard sample number, and anomaly occurrence time is generated. This prompt message is sent to the operator or upper control system through the operation interface or communication interface.

[0146] A4. When the number of retries for the same target standard sample exceeds the preset threshold, the target standard sample is marked as not calibrated and the signal deviation corresponding to this calibration operation is prohibited from being used to generate the calibration parameter set.

[0147] Among them, the number of retries refers to the cumulative number of times the same calibration operation for the same target standard sample block is retried due to overload or abnormal signal quality. The preset number threshold refers to the maximum number of retries that are set in advance. After exceeding this number, no further retries will be made. The marked as incomplete means that the calibration operation of the current target standard sample block has failed after multiple retries and is judged as unable to complete the calibration.

[0148] Understandably, while retry mechanisms can resolve occasional anomalies, if the same target standard sample fails to retries multiple times consecutively, it indicates that the problem may not be sporadic but rather a persistent issue such as hardware failure, sample damage, or severe wear of the tactile end. Therefore, an upper limit on the number of retries can be set. Once the upper limit is exceeded, retries should be abandoned, and the sample should be marked as incomplete. At the same time, the signal deviation values ​​collected during this calibration operation should not be used to generate the calibration parameter set, because these signals were collected under abnormal conditions, and their signal deviation values ​​cannot truly reflect the normal deviation characteristics of the equipment.

[0149] Specifically, the number of retries for the same target standard sample is accumulated and counted. The count is incremented by one for each retrieval. When the count exceeds the preset threshold, the retrieval process for the target standard sample is stopped, the target standard sample is marked as incomplete, and the signal deviation collected in this calibration operation is discarded and not used as input data for generating the calibration parameter set in the future.

[0150] A5, continue to perform the corresponding calibration operation on the next target standard sample block, until all the target standard samples have been calibrated.

[0151] Understandably, a standard sample array typically contains multiple target standard samples. The calibration operation of each sample is relatively independent. If the calibration of one sample is abnormal or incomplete, it should not affect the calibration process of other normal samples. If the entire calibration process is stopped because the retries of a certain sample are exhausted, it will cause the calibration of other samples to fail, reducing calibration efficiency. Therefore, regardless of whether the calibration of the current target standard sample is successfully completed, marked as abnormal, or marked as incomplete, the process should continue to the next target standard sample until all target standard samples have been calibrated at least once. This way, as many calibration tasks as possible can be completed without affecting the overall process if the abnormality of a single target standard sample is not affected.

[0152] Specifically, after processing the calibration results of the current target standard sample block, the process pointer is moved to the next target standard sample block according to the preset target standard sample block access order. The calibration operation and anomaly monitoring process are repeated for the next target standard sample block until all target standard samples blocks have undergone at least one calibration operation process.

[0153] This calibration method can be automatically triggered by the system or manually triggered by the user. Automatic triggering conditions include: no valid calibration parameters for the current end after the system is powered on or restarted; a change in type is detected after the tactile end is replaced; the preset periodic calibration time is reached; the difference between the ambient temperature and the last calibration time exceeds a preset threshold; the validity period of the current calibration parameters expires; and the fitting residual exceeds a quality threshold. When any of the triggering conditions is met, the system will automatically start this calibration process.

[0154] See Figure 4 This is a functional module block diagram of a standard sample array self-calibration system for re-inspection of multi-motion tactile defects provided in an embodiment of the present invention. The data processing system based on the standard sample array self-calibration system for re-inspection of multi-motion tactile defects includes: An automatic calibration control module is used to control the detection execution unit to perform corresponding tactile calibration actions for different standard samples, and to simultaneously acquire multimodal calibration signals; The multimodal calibration signal acquisition module is used to simultaneously acquire one or more calibration signals from normal force, tangential force, displacement, vibration, temperature, or tactile images during the calibration process. The calibration parameter generation module is used to calculate and generate calibration parameters based on the deviation between the multimodal calibration signal and the known physical parameters or known defect characteristics of the corresponding standard sample. The calibration parameter storage and version management module is used to store calibration parameters in a structured manner and supports version management and expiration management of calibration parameters.

[0155] Figure 4 The apparatus of the illustrated embodiment can be used to perform corresponding actions. Figure 2 The steps in the method embodiments shown are implemented in a similar manner and have similar technical effects, and will not be repeated here.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A self-calibration method for a standard sample array for re-examination of multi-action tactile defects, characterized in that, include: Obtain the current tactile end type of the detection execution unit and construct a calibration coordinate system between the tactile end and the standard sample array; Based on the type of the tactile end, target standard samples are determined from the array of standard samples. The tactile end is then controlled to perform corresponding calibration operations on each target standard sample based on the calibration coordinate system, and multimodal calibration signals are acquired simultaneously, including: Based on the type of tactile end, a preset operation mapping relationship is retrieved to determine the calibration operation type corresponding to the tactile end type; According to the preset correspondence between the calibration operation type and the standard sample block, the standard sample block corresponding to each calibration operation type is selected from the standard sample block array as the target standard sample block; Based on the calibration coordinate system, determine the target contact position corresponding to each target standard sample block; Obtain preset operation parameters corresponding to the calibration operation type of the target standard sample block; According to the preset operating parameters, the tactile end is controlled to perform corresponding calibration operations on each target standard sample at each target contact position; During the calibration operation, multiple signal acquisition channels are triggered to start synchronously, and the physical quantity signals corresponding to each signal acquisition channel are acquired. The physical quantity signals that meet the preset quality conditions are determined as multimodal calibration signals. The calibration operation includes multiple actions such as tapping, sliding, scraping, rolling, clamping, tapping, and thermal contact. A set of calibration parameters associated with each target standard sample and its corresponding calibration operation is generated based on the signal deviation between the multimodal calibration signal corresponding to each target standard sample and the target standard parameter. Based on the tactile end type corresponding to the defect re-inspection task and the action type corresponding to the calibration operation, the matching calibration parameter set is invoked to calibrate the tactile detection signal collected during the defect re-inspection process.

2. The method according to claim 1, characterized in that, The process of obtaining the current tactile end-effector type of the detection execution unit and constructing a calibration coordinate system between the tactile end-effector and the standard sample array includes: Read the type identifier corresponding to the currently installed tactile end on the detection execution unit to determine the type of tactile end; The tactile end is controlled to move to a preset positioning reference and then approach the target along a direction perpendicular to the plane of the standard sample array. The contact force signal of the tactile end is collected in real time, and the approach operation is stopped when the contact force signal reaches a preset contact threshold. The location point corresponding to the tactile end when the approach operation stops is determined as the calibration origin; A calibration coordinate system is constructed based on the calibration origin and the preset positional relationship of the sample blocks.

3. The method according to claim 1, characterized in that, The step of determining the target contact position corresponding to each target standard sample block based on the calibration coordinate system includes: The relative position information of each target standard sample block with respect to the calibration origin is converted into the sample block position coordinates of each target standard sample block in the calibration coordinate system; Based on the calibration operation type of each target standard sample, determine the contact direction and contact posture of the tactile end contacting the target standard sample; Based on the sample block position coordinates, the contact direction, and the contact posture, the target contact position corresponding to each target standard sample block is determined.

4. The method according to claim 1, characterized in that, The step of generating a calibration parameter set associated with each target standard sample and its corresponding calibration operation based on the signal deviation between the multimodal calibration signal corresponding to each target standard sample and the target standard parameters includes: The physical quantity signals corresponding to each signal acquisition channel in the multimodal calibration signal are preprocessed to obtain the processed signals corresponding to each signal acquisition channel. From the processed signal, identify signal feature points associated with the target standard parameters corresponding to the target standard sample block, and extract the signal feature values ​​corresponding to the signal feature points; The signal feature value is compared with the target standard parameter to determine the signal deviation between the signal feature value and the target standard parameter; A calibration parameter set is generated based on the signal deviation, and the calibration parameter set is stored in association with the end type, action type, sample type, calibration version, or validity period.

5. The method according to claim 4, characterized in that, The step of generating a calibration parameter set based on the signal deviation includes: The signal deviation corresponding to the determination of the target standard parameter as the reference value is the zero-point offset compensation amount of the tactile end under the current calibration operation; The signal deviation corresponding to multiple target standard parameters with different values ​​is determined as candidate difference quantities. Based on the numerical correspondence between the target standard parameters and each candidate difference quantity, the gain correction coefficient of the tactile end under the current calibration operation is fitted and calculated. The judgment boundary value between the signal feature values ​​corresponding to two adjacent target standard parameters with different values ​​is determined as the defect judgment threshold of the tactile end under the current calibration operation. The zero-point offset compensation, the gain correction coefficient, and the defect judgment threshold are associated with the action type of the current calibration operation to generate a calibration parameter set associated with the current target standard sample and the corresponding calibration operation.

6. The method according to claim 1, characterized in that, The step of calibrating the tactile detection signals collected during the defect re-inspection process by calling a matching set of calibration parameters based on the tactile end-effector type corresponding to the defect re-inspection task and the action type corresponding to the calibration operation includes: Obtain the tactile end type currently corresponding to the defect re-inspection task, and the action type currently performing the re-inspection action; Based on the obtained tactile end type and action type, retrieve and call the matching calibration parameter set, and extract the zero-point offset compensation amount, gain correction coefficient and defect judgment threshold from it; Based on the zero-point offset compensation and the gain correction coefficient, the tactile detection signal collected during the defect re-inspection process is subjected to offset processing and scale correction to obtain the calibrated target detection signal. The judgment boundary of the anomaly judgment module in the defect re-inspection task is dynamically updated according to the defect judgment threshold, and the defect re-inspection result is output based on the target detection signal and the updated judgment boundary.

7. The method according to claim 1, characterized in that, Also includes: During the calibration operation, the contact force signal of the tactile end and the signal quality indicators of the multimodal calibration signal are monitored in real time. When the contact force signal is detected to exceed the preset overload threshold, the tactile end is controlled to exit the current calibration operation and re-execute the approach operation before retrying. When the signal quality index is detected to be unsatisfactory, the current target standard sample block is marked as calibration abnormal and a prompt message is generated. When the number of retries for the same target standard sample exceeds the preset threshold, the target standard sample is marked as not calibrated and the signal deviation corresponding to this calibration operation is prohibited from being used to generate the calibration parameter set. Continue performing the corresponding calibration operation on the next target standard sample block until all the target standard samples blocks have been calibrated.

8. A standard sample array self-calibration system for re-inspection of multi-motion tactile defects, used to perform the method according to any one of claims 1-7, characterized in that, include: Calibration base, used to support and position multiple standard sample blocks; A standard sample array, fixed on the calibration base, includes at least two types of standard sample blocks with known physical parameters or known defect characteristics; The probe positioning reference is set on the calibration base to establish a calibration coordinate system between the tactile end and the standard sample array; A calibration execution interface is used to cooperate with a detection execution unit, enabling the detection execution unit to contact multiple target standard blocks in the standard sample block array according to a preset calibration path. An automatic calibration control module is used to control the detection execution unit to perform corresponding tactile calibration actions for different standard samples and to simultaneously acquire multimodal calibration signals; The multimodal calibration signal acquisition module is used to simultaneously acquire multiple calibration signals from normal force, tangential force, displacement, vibration, temperature, or tactile images during the calibration process. The calibration parameter generation module is used to calculate and generate calibration parameters based on the deviation between the multimodal calibration signal and the known physical parameters or known defect characteristics of the corresponding standard sample. The calibration parameter storage and version management module is used to store calibration parameters in a structured manner and supports version management and expiration management of calibration parameters. The calibration parameter output interface is used to provide the calibration parameters to signal correction, feature normalization, judgment threshold update or fusion judgment in the re-examination of tactile defects.

Images