Joint calibration method and system based on generalized regression neural network
By employing a joint calibration method based on a generalized regression neural network, a standard sphere is measured using visual sensing and contact sensing systems. The sensor position relationship is calculated, solving the calibration problem with large errors in existing technologies and achieving high-precision three-dimensional geometric feature measurement.
Patent Information
- Application Number
- CN202511346599.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-19
- Publication Date
- 2026-01-09
Smart Images
Figure CN121297660A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the field of precision instrument manufacturing and precision testing and metrology technology, and relates to a combined calibration method and system of visual sensing and contact sensing, in particular to a combined calibration method and system based on a generalized regression neural network. BACKGROUND
[0002] With the development of high-end equipment manufacturing industry towards high precision and miniaturization, the requirements for precision and efficiency of microstructure three-dimensional measurement are increasingly improved, and a single sensor system has been difficult to meet the demand. In order to quickly obtain high-precision coordinates and complete model information of the measured object, a composite measurement method emerges as the times require. The composite measurement method integrates optical sensors and contact sensors in the same system, combines the advantages of high-efficiency point cloud collection of optical sensors and high-precision measurement of contact sensors, and realizes the precision measurement of three-dimensional geometric features.
[0003] Due to different measurement principles and coordinate systems of different sensors, composite measurement needs to unify multi-source data to the same coordinate system through combined calibration. In this process, a physical reference part with specific geometric properties or combinations thereof is often used as a space calibration medium, for example, a three-coordinate measuring machine uses a spherical calibrator, a contact sensor fits the space center coordinates of the standard ball by measuring multiple points on the standard ball, and a visual sensor obtains the two-dimensional coordinates of the space ball by measuring the projection circle of the standard ball.
[0004] The current combined calibration method depends on a sufficient number of coincident points or common standards and features, and different measurements of the same calibrator are used to obtain the same feature points. However, due to different measurement principles and methods of different systems and different point cloud densities, it is usually very difficult for contact sensors and visual sensors to obtain the same reference points. If only a single reference point is relied on, system, human and accidental errors are easily introduced. This is the deficiency of the prior art.
[0005] Therefore, it is necessary to provide a combined calibration method and system based on a generalized regression neural network to solve the above-mentioned defects in the prior art. SUMMARY
[0006] The purpose of the present application is to provide a combined calibration method and system based on a generalized regression neural network to solve the above-mentioned technical problems in view of the defects of the prior art.
[0007] To achieve the above purpose, the present application provides the following technical scheme: A combined calibration method based on a generalized regression neural network comprises the following steps: Step S1, a step of initial visual calibration data acquisition, in which: four standard balls A1, A2, A3, A4 forming a square are measured by a visual sensing system to obtain the ball center coordinates Q1 in the visual coordinate system as the training set of the visual coordinate system; Step S2, a step of contact calibration data acquisition, in which: four standard balls A1, A2, A3, A4 forming a square are measured by a contact sensing system to obtain the ball center coordinates Q2 in the contact sensor coordinate system as the training set of the contact sensor coordinate system; Step S3, a step of center ball test coordinate acquisition, in which: a standard ball A5 located at the center of the five-ball calibrator is measured, and the measured ball center coordinates Q3 are taken as the test data set; Step S4, a step of calibration parameter solving, in which: The training set coordinate data is input into a generalized regression neural network model for pre-training, and then the test set sample is input into the trained model, so as to obtain the parameter Y representing the position relationship between the visual sensing and the contact sensing, and the joint calibration is completed.
[0008] As a preferred, the step S1 specifically comprises: The Z-direction linear translation axis, the Y-direction linear translation axis and the X-direction linear translation axis are moved, the visual sensing system with a visual imaging component is moved above the calibrator, four standard balls A1, A2, A3, A4 forming a square are measured, and the ball center coordinates Q1 in the visual coordinate system are obtained as the training set of the visual coordinate system.
[0009] As a preferred, the step S2 specifically comprises: The Z-direction linear translation axis, the Y-direction linear translation axis, the X-direction linear translation axis are moved, the sensing system with a contact probe is moved above the calibrator, four standard balls A1, A2, A3, A4 forming a square are measured, and the ball center coordinates Q2 in the contact sensor coordinate system are obtained as the training set of the contact sensor coordinate system.
[0010] As a preferred, the step S3 specifically comprises: The standard ball A5 located at the center of the five-ball calibrator is measured by the visual sensing system and the contact sensing system respectively, and the measured ball center coordinates Q3 are taken as the test data set.
[0011] As a preferred, the generalized regression neural network model in the step S4 comprises: First, in the input layer of the generalized regression neural network topology model, the union of the training sets Q1 and Q2 is taken as the training sample, and the corresponding output expected value is taken as . The number of input layer neurons is the dimension of the input vector; Second, in the mode layer of the generalized regression neural network topology model, the number of neurons is equal to the number of samples n of the training set, and the transfer function of the mode layer is a Gaussian function, The Gaussian value of each input test data and training data is calculated, and the transfer function is used to calculate the Euclidean distance of the input test sample And the training sample The mathematical expression is as follows:
[0012] Third, in the summation layer of the generalized regression neural network topology model, the number of nodes of the neurons is more than the dimension of the input sample by 1, and the output of the summation layer includes two parts of arithmetic sum and weighted sum; in the arithmetic sum, the weight factor of the neuron and the mode layer is 1, and all the neurons in the mode layer need to be arithmetically summed, and the mathematical expression is as follows:
[0013] Fourth, in the weighted sum, the weight factors of the neurons are not equal, the connection weight factor of the i-th node in the mode layer and the j-th node in the summation layer is the j-th element value of the i-th output sample of the training set Then, weighted summation is performed, and the mathematical expression is as follows:
[0014] Fifth, in the output layer Y of the generalized regression neural network topology model, the parameter to be estimated is The number of output neurons is equal to the dimension of the output vector, and the output value of the node is the ratio of the arithmetic sum and the weighted sum of all nodes in the summation layer, and the mathematical expression is as follows:
[0015] Wherein, the parameter to be estimated is seven parameters to be solved in calibration, including the scale factor , the rotation angles around the x, y and z axes , and the translation matrix .
[0016] In addition, the present application also provides a joint calibration system based on the generalized regression neural network, comprising: An initial visual calibration data acquisition module, in which: four standard ball A1, A2, A3, A4 calibrators constituting a square are measured by a visual sensing system, and the ball center coordinates Q1 in the visual coordinate system are obtained as the training set of the visual coordinate system; The contact calibration data acquisition module, in which: the four standard balls A1, A2, A3, A4 forming a square are measured by the contact sensing system, and the ball center coordinates Q2 in the contact sensor coordinate system are obtained as the training set of the contact sensor coordinate system; The center ball test coordinate acquisition module, in which: the standard ball A5 located at the center of the five-ball calibrator is measured, and the measured ball center coordinates Q3 are taken as the test data set; The calibration parameter solving module, in which: The training set coordinate data is input into the generalized regression neural network model for pre-training, and then the test set sample is input into the trained model, so as to obtain the parameter Y representing the position relationship between the visual sensing and the contact sensing, and complete the joint calibration.
[0017] As preferred, the initial visual calibration data acquisition module specifically includes: The Z-direction linear translation axis, the Y-direction linear translation axis and the X-direction linear translation axis are moved to move the visual sensing system with the visual imaging component directly above the calibrator, the four standard balls A1, A2, A3, A4 forming a square are measured, and the ball center coordinates Q1 in the visual coordinate system are obtained as the training set of the visual coordinate system.
[0018] As preferred, the contact calibration data acquisition module specifically includes: The Z-direction linear translation axis, the Y-direction linear translation axis, the X-direction linear translation axis are moved to move the sensing system with the contact probe directly above the calibrator, the four standard balls A1, A2, A3, A4 forming a square are measured, and the ball center coordinates Q2 in the contact sensor coordinate system are obtained as the training set of the contact sensor coordinate system.
[0019] As preferred, the center ball test coordinate acquisition module specifically includes: The standard ball A5 located at the center of the five-ball calibrator is measured by the visual sensing system and the contact sensing system respectively, and the measured ball center coordinates Q3 are taken as the test data set.
[0020] As preferred, the generalized regression neural network model in the calibration parameter solving module includes: First, in the input layer of the generalized regression neural network topology model, the training set Q1 and Q2 are combined, and the training sample is denoted as , the expected value of the corresponding output is denoted as , and the number of input layer neurons is the dimension of the input vector; Second, in the mode layer of the generalized regression neural network topology model, the number of neurons is equal to the number of training samples n, and the transfer function of the mode layer is the Gaussian function The Gaussian value of each input test data and training data, and the transfer function is used to calculate the input test sample And the Euclidean distance of training sample The mathematical expression is as follows:
[0021] Third, in the summation layer of the generalized regression neural network topology model, the number of nodes of the neuron is more than the dimension of the input sample by 1, and the output of the summation layer includes two parts of arithmetic sum and weighted sum; in the arithmetic sum, the weight factor of the neuron and the mode layer is 1, and all the neurons in the mode layer need to perform arithmetic summation, and the mathematical expression is as follows:
[0022] Fourth, in the weighted sum, the weight factors of the neurons are not equal, and the connection weight factor of the i th node in the mode layer and the j th node in the summation layer is the j th element value of the i th output sample of the training set Then weighted summation is performed, and the mathematical expression is as follows:
[0023] Fifth, in the output layer Y of the generalized regression neural network topology model, the to-be-estimated parameter The number of output neurons is equal to the dimension of the output vector, and the output value of the node is the ratio of the arithmetic sum and the weighted sum of all nodes in the summation layer, and the mathematical expression is as follows:
[0024] Wherein, the to-be-estimated parameter is seven parameters to be solved in calibration, including the scale factor , the rotation angles around the x, y and z axes , and the translation matrix .
[0025] The beneficial effects of the present application are that by equally weighting the errors of measuring the same reference points, the error is reduced while the training model data set sample points are increased, the high-dimensional parameters can be obtained by using small sample data, and the joint calibration of the parameter model can be realized by measuring a standard sphere, so that the calibration precision is improved, and the present application has universal applicability.
[0026] In addition, the design principle of the present application is reliable, the structure is simple, and the present application has very wide application prospect.
[0027] Therefore, compared with the prior art, the present application has outstanding substantial characteristics and significant progress, and the beneficial effects of the implementation are also obvious. BRIEF DESCRIPTION OF DRAWINGS
[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0029] Figure 1 This is a schematic diagram of the joint calibrator structure provided by the present invention.
[0030] Figure 2 This is a diagram of the topology of the generalized regression neural network provided by this invention.
[0031] Figure 3 This is a flowchart of a joint calibration method based on a generalized regression neural network provided by the present invention.
[0032] Figure 4 This is a block diagram illustrating the principle of a joint calibration system based on a generalized regression neural network provided by the present invention.
[0033] Among them, 1-probe sensing system, 2-Z-linear translation axis, 3-Y-linear translation, 4-X-linear translation, 5-visual sensing system, 6-five-ball calibrator, 7-initial visual calibration data acquisition module, 8-contact calibration data acquisition module, 9-center ball test coordinate acquisition module, 10-calibration parameter calculation module. Detailed Implementation
[0034] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.
[0035] Example 1: like Figure 3 As shown in the figure, this embodiment provides a joint calibration method based on a generalized regression neural network, which includes the following steps: Step S1, the initial visual calibration data acquisition step, in which: Move the Z-axis linear translation 2, Y-axis linear translation 3 and X-axis linear translation 4 to move the visual sensing system 5 with visual imaging components directly above the calibrator 6. Measure the four standard spheres A1, A2, A3 and A4 that form a square to obtain the sphere center coordinates Q1 in the visual coordinate system, which serves as the training set for the visual coordinate system. Step S2, the step of acquiring contact calibration data, in which: Move the Z-axis linear translation 2, Y-axis linear translation 3, and X-axis linear translation 4 to move the sensing system 1 with the contact probe directly above the calibrator 6. Measure the four standard spheres A1, A2, A3, and A4 that form a square to obtain the sphere center coordinates Q2 in the contact sensor coordinate system, which serves as the training set for the contact sensor coordinate system. Step S3, the step of obtaining the test coordinates of the center sphere, in which: The standard ball A5, located at the center of the five-ball calibrator, was measured using both a visual sensing system and a contact sensing system. The measured center coordinates Q3 of the ball were used as the test dataset. Step S4, the calibration parameter calculation step, in which: The coordinate data of the training set is input into the generalized regression neural network model for pre-training, and then the test set samples are input into the trained model to obtain the parameter Y representing the positional relationship between visual sensing and contact sensing, thus completing the joint calibration. The generalized regression neural network model in step S4, such as Figure 2 As shown, it specifically includes: First, in the generalized regression neural network topology model, the input layer is the combination of the training sets Q1 and Q2, denoted as... The corresponding output is The number of neurons in the input layer is equal to the dimension of the input vector; Second, in the pattern layer, the number of neurons is equal to the number of samples n in the training set, and the transfer function of the pattern layer is a Gaussian function. For each input test data point, the Gaussian values are compared with the training data, and the transfer function is used to calculate the input test samples. and training samples The Euclidean distance is expressed mathematically as follows:
[0036] Third, in the summation layer, the number of neurons is one more than the dimension of the input samples. The output of the summation layer includes both an arithmetic sum and a weighted sum. In the arithmetic sum, the weight factor between the neurons and the pattern layer is 1. All neurons in the pattern layer need to be arithmetically summed, as shown in the following mathematical expression:
[0037] Fourth, in the weighted summation, the weight factors of each neuron are not equal. The connection weight factor between the i-th node in the pattern layer and the j-th node in the summation layer is the j-th element value of the i-th training set output sample. Then, a weighted sum is performed, as shown in the following mathematical expression:
[0038] Fifth, the output layer Y contains parameters to be estimated. The number of output neurons is equal to the dimension of the output vector. The output value of a node is the ratio of the arithmetic sum of all nodes in the summation layer to the weighted sum, as shown in the following mathematical expression:
[0039] Among them, the parameters to be estimated The seven parameters to be solved during calibration include the scaling factor. Rotation angles about the x, y, and z axes ( ), and translation matrix .
[0040] Example 2: like Figure 4 As shown, this embodiment provides a joint calibration system based on a generalized regressive neural network, comprising: Initial visual calibration data acquisition module 7, in which: Move the Z-axis linear translation 2, Y-axis linear translation 3 and X-axis linear translation 4 to move the visual sensing system 5 with visual imaging components directly above the calibrator 6. Measure the four standard spheres A1, A2, A3 and A4 that form a square to obtain the sphere center coordinates Q1 in the visual coordinate system, which serves as the training set for the visual coordinate system. Contact calibration data acquisition module 8, in which: Move the Z-axis linear translation 2, Y-axis linear translation 3, and X-axis linear translation 4 to move the sensing system 1 with the contact probe directly above the calibrator 6. Measure the four standard spheres A1, A2, A3, and A4 that form a square to obtain the sphere center coordinates Q2 in the contact sensor coordinate system, which serves as the training set for the contact sensor coordinate system. Center sphere test coordinate acquisition module 9, in which: The standard ball A5, located at the center of the five-ball calibrator, was measured using both a visual sensing system and a contact sensing system. The measured center coordinates Q3 of the ball were used as the test dataset. Calibration parameter calculation module 10, in which: The coordinate data of the training set is input into the generalized regression neural network model for pre-training, and then the test set samples are input into the trained model to obtain the parameter Y representing the positional relationship between visual sensing and contact sensing, thus completing the joint calibration. The generalized regression neural network model in the calibration parameter calculation module 10, such as Figure 2 As shown, it specifically includes: First, in the generalized regression neural network topology model, the input layer is the combination of the training sets Q1 and Q2, denoted as... The corresponding output is The number of input layer neurons is the dimension of the input vector; Second, in the pattern layer, the number of neurons is equal to the number of samples n of the training set, and the transfer function of the pattern layer is a Gaussian function, The Gaussian value of each input test data and training data, and the transfer function is used to calculate the Euclidean distance of the input test sample and training sample The mathematical expression is as follows:
[0041] Third, in the summation layer, the number of neurons is one more than the dimension of the input sample, and the output of the summation layer includes two parts of arithmetic sum and weighted sum; in the arithmetic sum, the weight factor of the neuron and the pattern layer is 1, and all neurons in the pattern layer need to be arithmetically summed, and the mathematical expression is as follows:
[0042] Fourth, in the weighted sum, the weight factors of each neuron are not equal, and the connection weight factor between the i th node in the pattern layer and the j th node in the summation layer is the j th element value of the i th output sample of the training set Then weighted summation is performed, and the mathematical expression is as follows:
[0043] Fifth, the output layer Y is the parameter to be estimated The number of output neurons is equal to the dimension of the output vector, and the output value of the node is the ratio of the arithmetic sum and the weighted sum of all nodes in the summation layer, and the mathematical expression is as follows:
[0044] Wherein, the parameter to be estimated is the seven parameters to be solved in the calibration, including the scale factor , the rotation angles around the x, y and z axes , and the translation matrix .
[0045] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between each embodiment can be referred to each other. For the method disclosed in the embodiment, since it corresponds to the system disclosed in the embodiment, the description is relatively simple, and the relevant part can be referred to the method part.
[0046] Those skilled in the art will further appreciate that the functionality of the various examples described herein can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the
[0047] In several embodiments provided in the present application, it should be understood that the disclosed system, system and method can be implemented in other ways. For example, the above-described system embodiments are merely illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division, for example, multiple units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, systems or units, which can be electrical, mechanical or other forms.
[0048] The units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.
[0049] In addition, the functional modules in each embodiment of the present application can be integrated in one processing unit, or each module can be physically present separately, or two or more modules can be integrated in one unit.
[0050] Similarly, each processing unit in each embodiment of the present application can be integrated in one functional module, or each processing unit can be physically present separately, or two or more processing units can be integrated in one functional module.
[0051] The steps of the method or algorithm described in combination with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.
[0052] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0053] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.
Claims
1. A joint calibration method based on a generalized regression neural network, characterized in that, Includes the following steps: Step S1, the initial visual calibration data acquisition step, in which: the four standard spheres A1, A2, A3, A4 calibrators that make up the square are measured by the visual sensing system to obtain the sphere center coordinates Q1 in the visual coordinate system, which is used as the training set of the visual coordinate system; Step S2, the step of acquiring contact calibration data, in which: the four standard spheres A1, A2, A3, A4 calibrators that make up the square are measured by the contact sensing system to obtain the center coordinates Q2 of the spheres in the contact sensor coordinate system, which is used as the training set of the contact sensor coordinate system; Step S3, the step of obtaining the test coordinates of the center ball, in which: the standard ball A5 located at the center of the five-ball calibrator is measured, and the measured center coordinates Q3 of the ball are used as the test dataset; Step S4, the calibration parameter calculation step, in which: The coordinate data of the training set is input into the generalized regression neural network model for pre-training, and then the test set samples are input into the trained model to obtain the parameter Y representing the positional relationship between visual sensing and contact sensing, thus completing the joint calibration.
2. The joint calibration method based on a generalized regression neural network according to claim 1, characterized in that, Step S1 specifically includes: Move the Z-axis, Y-axis, and X-axis linear translation axes to position the visual sensing system with the visual imaging components directly above the calibrator. Measure the four standard spheres A1, A2, A3, and A4 that form a square to obtain the sphere center coordinates Q1 in the visual coordinate system, which serves as the training set for the visual coordinate system.
3. The joint calibration method based on a generalized regression neural network according to claim 1, characterized in that, Step S2 specifically includes: Move the Z-axis, Y-axis, and X-axis linear translation axes to move the sensing system with the contact probe directly above the calibrator. Measure the four standard spheres A1, A2, A3, and A4 that form a square to obtain the sphere center coordinates Q2 in the contact sensor coordinate system, which serves as the training set for the contact sensor coordinate system.
4. The joint calibration method based on a generalized regression neural network according to claim 1, characterized in that, Step S3 specifically includes: The standard ball A5, located at the center of the five-ball calibrator, was measured using both a visual sensing system and a contact sensing system. The measured center coordinates Q3 of the ball were used as the test dataset.
5. The joint calibration method based on a generalized regression neural network according to claim 1, characterized in that, The generalized regression neural network model in step S4 includes: First, in the input layer of the generalized regression neural network topology model, the training samples are the combined set of training sets Q1 and Q2, denoted as . The expected value of the corresponding output is denoted as The number of neurons in the input layer is equal to the dimension of the input vector; Second, in the pattern layer of the generalized regression neural network topology model, the number of neurons is equal to the number of samples n in the training set, and the transfer function of the pattern layer is a Gaussian function. For each input test data point, the Gaussian values are compared with the training data, and the transfer function is used to calculate the input test samples. and training samples The Euclidean distance is expressed mathematically as follows: Third, in the summation layer of the generalized regression neural network topology model, the number of neurons is one more than the dimension of the input samples. The output of the summation layer includes both an arithmetic sum and a weighted sum. In the arithmetic sum, the weight factor between the neurons and the pattern layer is 1. All neurons in the pattern layer need to be arithmetically summed, as shown in the following mathematical expression: Fourth, in the weighted summation, the weight factors of each neuron are not equal. The connection weight factor between the i-th node in the pattern layer and the j-th node in the summation layer is the j-th element value of the i-th training set output sample. Then, a weighted sum is performed, as shown in the following mathematical expression: Fifth, in the output layer Y of the generalized regression neural network topology model, the parameters to be estimated are... The number of output neurons is equal to the dimension of the output vector. The output value of a node is the ratio of the arithmetic sum of all nodes in the summation layer to the weighted sum, as shown in the following mathematical expression: Among them, the parameters to be estimated The seven parameters to be solved during calibration include the scaling factor. Rotation angles about the x, y, and z axes ( ), and translation matrix .
6. A joint calibration system based on a generalized regressive neural network, characterized in that, include: The initial visual calibration data acquisition module: In this module, the four standard spheres A1, A2, A3, and A4 that make up the square are measured by the visual sensing system to obtain the center coordinates Q1 of the spheres in the visual coordinate system, which is used as the training set of the visual coordinate system; The contact calibration data acquisition module measures the four standard spheres A1, A2, A3, and A4 that make up a square using a contact sensing system, and obtains the center coordinates Q2 of the spheres in the contact sensor coordinate system as the training set of the contact sensor coordinate system. The module for obtaining the center ball test coordinates includes: measuring the standard ball A5 located at the center of the five-ball calibrator, and using the measured center coordinates Q3 as the test dataset; The calibration parameter calculation module contains: The coordinate data of the training set is input into the generalized regression neural network model for pre-training, and then the test set samples are input into the trained model to obtain the parameter Y representing the positional relationship between visual sensing and contact sensing, thus completing the joint calibration.
7. A joint calibration system based on a generalized regressive neural network according to claim 6, characterized in that, The initial visual calibration data acquisition module specifically includes: Move the Z-axis, Y-axis, and X-axis linear translation axes to position the visual sensing system with the visual imaging components directly above the calibrator. Measure the four standard spheres A1, A2, A3, and A4 that form a square to obtain the sphere center coordinates Q1 in the visual coordinate system, which serves as the training set for the visual coordinate system.
8. A joint calibration system based on a generalized regressive neural network according to claim 7, characterized in that, The contact calibration data acquisition module specifically includes: Move the Z-axis, Y-axis, and X-axis linear translation axes to move the sensing system with the contact probe directly above the calibrator. Measure the four standard spheres A1, A2, A3, and A4 that form a square to obtain the sphere center coordinates Q2 in the contact sensor coordinate system, which serves as the training set for the contact sensor coordinate system.
9. A joint calibration system based on a generalized regressive neural network according to claim 8, characterized in that, The aforementioned center sphere test coordinate acquisition module specifically includes: The standard ball A5, located at the center of the five-ball calibrator, was measured using both a visual sensing system and a contact sensing system. The measured center coordinates Q3 of the ball were used as the test dataset.
10. A joint calibration system based on a generalized regressive neural network according to claim 9, characterized in that, The generalized regression neural network model in the calibration parameter calculation module includes: First, in the input layer of the generalized regression neural network topology model, the training samples are the combined set of training sets Q1 and Q2, denoted as . The expected value of the corresponding output is denoted as The number of neurons in the input layer is equal to the dimension of the input vector; Second, in the pattern layer of the generalized regression neural network topology model, the number of neurons is equal to the number of samples n in the training set, and the transfer function of the pattern layer is a Gaussian function. For each input test data point, the Gaussian values are compared with the training data, and the transfer function is used to calculate the input test samples. and training samples The Euclidean distance is expressed mathematically as follows: Third, in the summation layer of the generalized regression neural network topology model, the number of neurons is one more than the dimension of the input samples. The output of the summation layer includes both an arithmetic sum and a weighted sum. In the arithmetic sum, the weight factor between the neurons and the pattern layer is 1. All neurons in the pattern layer need to be arithmetically summed, as shown in the following mathematical expression: Fourth, in the weighted summation, the weight factors of each neuron are not equal. The connection weight factor between the i-th node in the pattern layer and the j-th node in the summation layer is the j-th element value of the i-th training set output sample. Then, a weighted sum is performed, as shown in the following mathematical expression: Fifth, in the output layer Y of the generalized regression neural network topology model, the parameters to be estimated are... The number of output neurons is equal to the dimension of the output vector. The output value of a node is the ratio of the arithmetic sum of all nodes in the summation layer to the weighted sum, as shown in the following mathematical expression: Among them, the parameters to be estimated The seven parameters to be solved during calibration include the scaling factor. Rotation angles about the x, y, and z axes ( ), and translation matrix .