Tactile roughness measurement model training method, measurement method and equipment based on terahertz wireless perception

Abstract

The present invention provides a tactile roughness measurement model training method, measurement method, and device based on terahertz wireless sensing. The method comprises: obtaining terahertz reflection signals from multiple angles of multiple objects to be measured to construct a training sample set; sequentially performing data preprocessing, fast Fourier transform, and scattering feature extraction on the reflection signals to obtain a time domain signal and its time domain features, a frequency domain signal and its frequency domain features, and a scattering feature; obtaining a multi-material roughness inference network, which is a dual-input two-dimensional convolutional network that takes the time domain signal as its first input and the time domain features, frequency domain features, and scattering features as its second input; and training the multi-material roughness inference network using the training sample set to ultimately obtain a roughness measurement model. The present invention can extract roughness evaluation parameters of the surface of the object to be measured based on the roughness measurement model, achieving non-contact, high-precision surface roughness measurement of the object.

Description

Technical Field

[0001] The present invention relates to the field of wireless recognition technology, and in particular to a tactile roughness measurement model training method, a measurement method, and a device based on terahertz wireless perception. Background Art

[0002] Touch is the third sense of perception for humans to explore the physical world, in addition to vision and hearing. Human tactile perception mainly relies on tactile nerve cells throughout the body, which can perceive external environmental information such as temperature, humidity, and stimulation of physical contact of external objects on the skin surface, such as pain, pressure, vibration, etc. These external stimuli on the skin surface can help us recognize information such as roughness, shape, size, and weight of objects. This information is indispensable for human production and life activities.

[0003] Existing roughness measurement methods usually adopt contact solutions, which can be divided into the following three types. Solution 1 is a roughness measurement method based on a contact probe. This method requires the object to be measured and the measuring device to be fixed vertically, and can only be measured downward. In addition, different probe heads need to be configured to measure different types of objects, and the process is very cumbersome. Solution 2 is a roughness imaging method based on a deformable colloid based on vision. This method squeezes a deformable colloid onto the object to be measured, sets a camera behind the deformable colloid to take pictures, and uses a preset deep learning model to visualize the surface of the object. However, this method can only measure the surface qualitatively and cannot measure quantitatively. Solution 3 is a method based on white light interferometry. This method is mainly used in laboratory environments and is currently a relatively high-precision measurement method. The theoretical limit can measure the surface roughness of atomic grain size, but this technology is essentially optical 3D modeling. The required equipment is large and expensive, and difficult to miniaturize. When this method is implemented, the measurement time is long and the data processing is complex, so it is difficult to popularize it in actual measurements. Summary of the Invention

[0004] In view of this, the embodiments of the present invention provide a roughness measurement model training method, measurement method and equipment based on terahertz wireless sensing to eliminate or improve one or more defects existing in the prior art, and solve the problems of the existing roughness measurement technology such as long measurement time, cumbersome measurement process, complex data processing and the need for contact measurement.

[0005] In one aspect, the present invention provides a tactile roughness measurement model training method based on terahertz wireless perception, characterized in that the method comprises the following steps:

[0006] Acquire a training sample set, wherein the training sample set includes a plurality of samples, each sample including a reflected signal reflected back after the object to be tested receives a signal transmitted by a preset terahertz time-domain spectrometer at multiple angles; the object to be tested is fixed at a preset distance from a transceiver antenna of the terahertz time-domain spectrometer;

[0007] The reflected signal is subjected to data preprocessing to obtain a time domain signal and extract time domain features; the time domain signal is subjected to fast Fourier transform to generate a corresponding frequency domain signal and extract frequency domain features; the frequency domain signal is input into a preset scattering feature extractor to extract corresponding scattering features; the scattering feature extractor selects a Fresnel reflection model and is provided with a correction factor to evaluate the roughness of the surface of the object to be measured;

[0008] Obtaining a multi-material roughness inference network, wherein the multi-material roughness inference network is a two-dimensional convolutional network with two inputs; taking the time domain signal as a first input, and taking the time domain feature, the frequency domain feature, and the scattering feature as second inputs, and inputting them simultaneously into the multi-material roughness inference network to obtain a roughness evaluation parameter of the surface of the object to be measured;

[0009] The multi-material roughness inference network is trained using the training sample set until a preset performance index is met, and the trained multi-material roughness inference network is used as the final roughness measurement model.

[0010] In some embodiments of the present invention, further comprising:

[0011] When performing data preprocessing on the reflected signal, obtaining the spectrum range of the terahertz time-domain spectrometer, and performing band-pass filtering on the reflected signal;

[0012] Before performing fast Fourier transform on the time domain signal, a 2-fold interpolation process is performed on the time domain signal to improve the resolution.

[0013] In some embodiments of the present invention, the time domain features include at least the first zero point position of the highest peak, the second zero point position, the position of the highest peak, the distance from the highest peak to the trough position for the first time, the height of the trough from the first time to the trough position, the distance from the highest peak to the trough, the height of the highest peak, the peak width at half of the highest peak and the integral; the frequency domain features include at least the position of the highest peak and the height of the highest peak.

[0014] In some embodiments of the present invention, the scattering feature extractor uses a Fresnel reflection model and is provided with a correction factor to evaluate the roughness of the surface of the object to be measured, and further includes:

[0015] If the correction factor is much smaller than 1, the surface roughness of the object to be measured is lower;

[0016] If the correction factor is close to 1, the surface roughness of the object to be measured is medium;

[0017] If the correction factor is much greater than 1, the surface roughness of the object to be measured is higher.

[0018] In some embodiments of the present invention, six convolutional layers are provided in the first input channel of the multi-material roughness inference network; four convolutional layers are provided in the second input channel; and a feature connection layer and a fully connected layer are provided in sequence after the first input channel and the second input channel.

[0019] On the other hand, the present invention also provides a tactile roughness measurement method based on terahertz wireless sensing, characterized in that the method comprises the following steps:

[0020] The transceiver antenna of the preset terahertz time-domain spectrometer is used to transmit signals to the objects to be measured placed at multiple angles, and receive reflected signals reflected back at corresponding angles; the objects to be measured are fixed at a preset distance from the transceiver antenna;

[0021] The reflected signal is subjected to data preprocessing to obtain a time domain signal and extract time domain features; the time domain signal is subjected to fast Fourier transform to generate a corresponding frequency domain signal and extract frequency domain features; the frequency domain signal is input into a preset scattering feature extractor to extract corresponding scattering features;

[0022] The time domain signal is used as the first input, the time domain feature, the frequency domain feature and the scattering feature are used as the second input, and the multi-material roughness measurement model in the tactile roughness measurement model training method based on terahertz wireless perception as described in any one of the above items is input at the same time to obtain the surface roughness evaluation parameter of the object to be measured.

[0023] On the other hand, the present invention further provides a tactile roughness measurement device based on terahertz wireless sensing, characterized in that the device is used to perform the steps of the tactile roughness measurement method based on terahertz wireless sensing as described in any one of the above items, and the device includes:

[0024] A terahertz source, used for providing a terahertz signal;

[0025] A transceiver antenna is used to transmit terahertz signals to the object to be measured and receive reflected signals;

[0026] An optical moving stage, used to carry the object to be measured and achieve horizontal movement and angle adjustment;

[0027] a data preprocessing module, configured to perform data preprocessing on the reflected signal to obtain a time domain signal and extract time domain features; perform fast Fourier transform on the time domain signal to generate a corresponding frequency domain signal and extract frequency domain features; and extract the scattering features of the frequency domain signal based on a preset scattering feature extractor;

[0028] The roughness reasoning module includes a trained roughness measurement model and is used to infer the roughness evaluation parameters of the surface of the object to be measured.

[0029] In some embodiments of the present invention, a pipe of a preset length is provided between the transceiver antenna and the object to be measured; the preset length of the pipe is determined by the focal length of a built-in lens of the transceiver antenna.

[0030] In some embodiments of the present invention, the data preprocessing module further includes a bandpass filter for performing bandpass filtering on the reflection signal.

[0031] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any one of the methods mentioned above when executed by a processor.

[0032] The beneficial effects of the present invention are at least:

[0033] The present invention provides a tactile roughness measurement model training method, measurement method, and device based on terahertz wireless sensing, comprising: obtaining terahertz reflection signals from multiple angles of multiple objects to be measured to construct a training sample set; sequentially performing data preprocessing, fast Fourier transform, and scattering feature extraction on the reflection signals to obtain a time domain signal and its time domain features, a frequency domain signal and its frequency domain features, and a scattering feature; obtaining a multi-material roughness inference network, which is a dual-input two-dimensional convolutional network that takes the time domain signal as its first input and the time domain features, frequency domain features, and scattering features as its second input; and training the multi-material roughness inference network using the training sample set to ultimately obtain a roughness measurement model. The present invention utilizes multi-angle measurement information and the relationship between roughness and scattering features to eliminate the influence of the material factors of the object to be measured itself. The roughness measurement model obtained through terahertz-based training evaluates the roughness evaluation parameters of the surface of the object to be measured from multiple aspects, ultimately achieving non-contact, high-precision surface roughness measurement of the object.

[0034] Additional advantages, objects, and features of the present invention will be set forth in part in the following description and will become apparent to those skilled in the art upon examination of the following or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained by the structures particularly pointed out in the description and drawings.

[0035] Those skilled in the art will understand that the purposes and advantages that can be achieved by the present invention are not limited to the above specific descriptions, and the above and other purposes that can be achieved by the present invention will be more clearly understood based on the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention. In the drawings:

[0037] Figure 1 Schematic diagram of the time domain signal waveform and frequency domain signal waveform of P80 sandpaper and a silver mirror in one embodiment of the present invention.

[0038] Figure 2 Schematic diagram of the steps of a tactile roughness measurement model training method based on terahertz wireless sensing in one embodiment of the present invention.

[0039] Figure 3 This is a flow chart of a tactile roughness measurement method based on terahertz wireless sensing in one embodiment of the present invention.

[0040] Figure 4 Schematic diagram of a local time domain signal waveform and its time domain characteristics, and a frequency domain signal waveform and its frequency domain characteristics of a certain object to be measured in one embodiment of the present invention.

[0041] Figure 5 FIG. 4 is a structural diagram of a roughness measurement model in one embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0043] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, the accompanying drawings only show structures and / or processing steps closely related to the solutions according to the present invention, while other details that are not closely related to the present invention are omitted.

[0044] It should be emphasized that the term "include / comprises" when used herein refers to the existence of features, elements, steps or components, but does not exclude the existence or addition of one or more other features, elements, steps or components.

[0045] It should also be noted that, unless otherwise specified, the term "connection" herein may refer not only to a direct connection but also to an indirect connection involving an intermediate.

[0046] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0047] It should be emphasized here that the step marks mentioned below do not limit the order of the steps, but it should be understood that the steps can be executed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be executed simultaneously.

[0048] Compared with existing sensing methods, such as Wi-Fi, ultra-wideband, and millimeter waves, terahertz sensing has the unique advantages of higher frequency, smaller spatial resolution, and wider bandwidth. Among them, the frequency band of terahertz is 0.1THz to 10THz, which is much larger than the existing wireless sensing bandwidth and can provide richer spectral fingerprint characteristics. At the same time, the surfaces of many indoor materials are considered optically rough in the terahertz band, and scattering has also become a part that cannot be ignored. By studying the influence of the degree of surface undulation of the object on the terahertz reflection signal, the ideas for the terahertz-based multi-material surface roughness perception of the present invention are provided.

[0049] like Figure 1 As shown, taking P80 sandpaper and silver mirror as examples, the terahertz time-domain spectrometer is used to measure the reflection signals of P80 sandpaper and silver mirror respectively, and the following are generated: Figure 1 The time domain signal waveform shown in (a) is obtained after fast Fourier transform as follows Figure 1 (b) shows the frequency domain signal waveform, combined with Figure 1 It can be clearly observed in (a) and (b) that when the surface roughness of the measured objects is similar, the reflection signal characteristics are significantly different due to the different surface roughness, and the greater the roughness, the more severe the signal scattering; due to the different materials, the characteristics of the reflection signal are also significantly different.

[0050] Therefore, in order to solve the problems of long measurement time, cumbersome measurement process, complex data processing and contact measurement in the existing roughness measurement technology, the present invention provides a roughness measurement model training method based on terahertz wireless sensing, such as Figure 2 and Figure 3 As shown, the method includes the following steps S101 to S104:

[0051] Step S101: Acquire a training sample set, which includes multiple samples, each of which includes reflected signals from a predetermined terahertz time-domain spectrometer after receiving a signal transmitted by the object under test at multiple angles. The object under test is fixed at a predetermined distance from the transceiver antenna of the terahertz time-domain spectrometer.

[0052] Step S102: The reflected signal is preprocessed to obtain a time domain signal and extract time domain features. The time domain signal is then fast Fourier transformed to generate a corresponding frequency domain signal and extract frequency domain features. The frequency domain signal is then input into a pre-set scattering feature extractor to extract the corresponding scattering features. The scattering feature extractor uses a Fresnel reflection model and includes a correction factor to evaluate the surface roughness of the object being measured.

[0053] Step S103: Obtain a multi-material roughness inference network, which is a dual-input two-dimensional convolutional network; use the time domain signal as the first input, and the time domain characteristics, frequency domain characteristics and scattering characteristics as the second input, and input them into the multi-material roughness inference network at the same time to obtain the roughness evaluation parameters of the surface of the object to be measured.

[0054] Step S104: The multi-material roughness inference network is trained using the training sample set until the preset performance index is met, and the trained multi-material roughness inference network is used as the final roughness measurement model.

[0055] In step S101 , a training sample set for training a multi-material roughness inference network is constructed.

[0056] In some embodiments, four oak boards of varying roughness, four plastic boards of varying roughness, four aluminum boards of varying roughness, and six nickel alloy boards of varying roughness are obtained, and these samples of different materials and varying roughness are sequentially fixed at a preset distance from a transceiver antenna of a preset terahertz time-domain spectrometer. The transceiver antenna of the preset terahertz time-domain spectrometer is used to transmit a signal to the sample and receive the reflected signal. During the measurement, each sample is adjusted at multiple angles. For example, the reflected signal of each sample is measured at angles of 0 degrees, 2 degrees, and 3 degrees relative to the normal of the sample surface. The multi-angle reflected signals of each sample are used as samples to construct a training sample set. The sample is fixed at a preset distance from the transceiver antenna of the terahertz time-domain spectrometer, and the preset distance is the sensing distance of the transceiver antenna, which is determined by the focal length of the lens built into the transceiver antenna.

[0057] Considering that the reflection coefficients of different materials vary greatly, it is difficult to determine whether the difference in signal spectrum is caused by the reflection coefficient of the material being tested or by the surface roughness by only detecting the reflected signal once at vertical incidence. Therefore, in the above, multi-angle measurements are performed on each sample. The multi-angle measurement information and the relationship between roughness and scattering characteristics described above are used to eliminate the influence of the material factors of the object being tested itself.

[0058] In some embodiments, the sensing distance between the transceiver antenna and the sample is set to 10 centimeters.

[0059] In some embodiments, the spectrum range of the terahertz time-domain spectrometer is preset to be 0.1 THz to 4 THz, and its transceiver antenna transmits a pulse signal with a spot diameter of 4 mm to the sample.

[0060] In step S102, data preprocessing and fast Fourier transform are performed on the reflection signal obtained in step S101, and corresponding time domain features and frequency domain features are extracted, and then the corresponding scattering features are extracted by a scattering feature extractor.

[0061] In some embodiments, data preprocessing includes bandpass filtering. According to the actual spectrum range of the terahertz device used in the present invention, 0.1 THz to 4 THz, bandpass filtering is performed to filter noise.

[0062] In some embodiments, after data preprocessing of the reflected signal, a pure time domain signal is obtained, and the waveform of the time domain signal is as follows: Figure 4 (a) shows that according to Figure 4 (a) Extracting time-domain features. Exemplarily, the time-domain features include at least the first zero point position a of the highest peak, the second zero point position c, the position b of the highest peak, the first distance d from the highest peak to the valley position, the height h of the valley position, the distance e from the highest peak to the valley position, the height g of the highest peak, the peak width f at half the highest peak, and the integral i. The position b of the highest peak is the distance from the lens of the built-in probe of the transceiver antenna to the object to be measured.

[0063] In some embodiments, before performing fast Fourier transform on the time domain signal, a 2x interpolation process is performed on the time domain signal to improve the resolution.

[0064] The pre-processed time domain signal is subjected to fast Fourier transform to generate the corresponding frequency domain signal. The waveform of the frequency domain signal is as follows: Figure 4 (b) shows that according to Figure 4 (b) Extracting frequency domain features. Exemplarily, the frequency domain features include at least the position j of the highest peak and the height k of the highest peak.

[0065] The frequency domain signal is then input into a preset scattering feature extractor to extract the corresponding scattering features. This preset scattering feature extractor is based on the KA (Kirchhoff Approximation) scattering model, which can be understood as a complex version of the Fresnel reflection model. This model incorporates a correction factor. This correction factor, based on the concept of optical roughness proposed by British physicist Rayleigh, takes into account the effect of surface roughness on signal scattering and is applied based on the received smooth signal intensity.

[0066] In some embodiments, the scattering feature extractor is provided with a correction factor to evaluate the roughness of the surface of the object to be measured, specifically expressed as follows: if the correction factor is much smaller than 1, the surface roughness of the object to be measured is lower; if the correction factor is close to 1, the surface roughness of the object to be measured is medium; if the correction factor is much larger than 1, the surface roughness of the object to be measured is higher.

[0067] In step S103, a multi-material roughness inference model of the roughness measurement model of the present invention is obtained. The model adopts a dual-input two-dimensional convolutional network, with 6 convolutional layers in the first input channel and 4 convolutional layers in the second input channel. A feature connection layer and a fully connected layer are sequentially provided after the first input channel and the second input channel.

[0068] In some embodiments, the convolution kernel sizes of the convolution layers in the first input channel and the second input channel are both 3*3.

[0069] like Figure 5 As shown, the time domain signal is used as the first input, representing the reflected signal of each terahertz pulse; the time domain features, frequency domain features, and scattering features are used as the second input. For example, the size of the first input is 1*3*3751, and the size of the second input is 1*3*12. Feature extraction is performed through the convolutional layers built into the first and second input channels, respectively. The features extracted from the two channels are concatenated and fused through the feature connection layer. Finally, the fully connected layer outputs the roughness evaluation parameter of the surface of the object to be measured. For example, the roughness evaluation parameter is denoted as Ra.

[0070] In some embodiments, the scattering features in the second input are given a higher weight as the main influencing factor of the roughness measurement.

[0071] In step S104, the multi-material roughness inference network is trained using the training sample set until the preset performance index is met, and the trained multi-material roughness inference network is used as the final roughness measurement model.

[0072] The present invention also provides a tactile roughness measurement method based on terahertz wireless sensing, which includes the following steps S201 to S203:

[0073] Step S201: Using a preset transceiver antenna of a terahertz time-domain spectrometer, signals are transmitted to an object to be measured at multiple angles, and reflected signals are received from corresponding angles. The object to be measured is fixed at a preset distance from the transceiver antenna.

[0074] Step S202: perform data preprocessing on the reflected signal to obtain a time domain signal and extract time domain features; perform fast Fourier transform on the time domain signal to generate a corresponding frequency domain signal and extract frequency domain features; input the frequency domain signal into a preset scattering feature extractor to extract the corresponding scattering features.

[0075] Step S203: The time domain signal is used as the first input, and the time domain characteristics, frequency domain characteristics and scattering characteristics are used as the second input. At the same time, the roughness measurement model in the roughness measurement model training method based on terahertz wireless sensing as described above is input to obtain the surface roughness evaluation parameters of the object to be measured.

[0076] This tactile roughness measurement method, based on a roughness measurement model and implemented through wireless terahertz sensing, offers a convenient method for measuring surfaces difficult to measure using existing contact-based measurement methods. This method achieves high-precision, non-contact roughness measurement. Experimental data demonstrates that the method can distinguish surface variations as small as 10µm. This method offers a new way for future robots to perceive the physical world and is applicable to material surface quality monitoring in industrial production.

[0077] The present invention also provides a tactile roughness measurement device based on terahertz wireless sensing, which is used to perform the steps of the tactile roughness measurement method based on terahertz wireless sensing as described above, and the device includes:

[0078] A terahertz source is used to provide a terahertz signal.

[0079] The transceiver antenna is used to transmit terahertz signals to the object to be measured and receive the reflected signals.

[0080] The optical moving stage is used to carry the object to be measured and realize horizontal movement and angle adjustment. For example, the optical moving stage can adjust the angle of the object to be measured in the horizontal axis direction within an angle range of ±10 degrees.

[0081] The data preprocessing module is used to perform data preprocessing on the reflected signal to obtain the time domain signal and extract the time domain features; perform fast Fourier transform on the time domain signal to generate the corresponding frequency domain signal and extract the frequency domain features; and extract the scattering features of the frequency domain signal based on a preset scattering feature extractor.

[0082] The roughness inference module includes a trained roughness measurement model and is used to infer the roughness evaluation parameters of the surface of the object to be measured.

[0083] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a tactile roughness measurement model training method based on terahertz wireless perception and a tactile roughness measurement method based on terahertz wireless perception.

[0084] Corresponding to the above method, the present invention also provides a device, which includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, the processor is used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the device implements the steps of the method described above.

[0085] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aforementioned edge computing server deployment method. The computer-readable storage medium may be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the art.

[0086] In summary, the present invention provides a tactile roughness measurement model training method, measurement method, and device based on terahertz wireless sensing, including: obtaining terahertz reflection signals from multiple angles of multiple objects to be measured to construct a training sample set; performing data preprocessing, fast Fourier transform, and scattering feature extraction on the reflection signals in sequence to obtain time domain signals and time domain features, frequency domain signals and frequency domain features, and scattering features, respectively; obtaining a multi-material roughness inference network, which is a dual-input two-dimensional convolutional network that takes the time domain signal as its first input and the time domain features, frequency domain features, and scattering features as its second input; training the multi-material roughness inference network using the training sample set to ultimately obtain a roughness measurement model. The present invention utilizes multi-angle measurement information and the relationship between roughness and scattering features to eliminate the influence of the material factors of the object to be measured itself; and based on terahertz, the roughness measurement model obtained through training evaluates the roughness evaluation parameters of the surface of the object to be measured from multiple aspects, ultimately achieving non-contact, high-precision surface roughness measurement of the object.

[0087] It should be understood by those skilled in the art that the various exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is specifically performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0088] It should be understood that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, a detailed description of known methods is omitted. In the above embodiments, several specific steps are described and illustrated as examples. However, the method of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art may make various changes, modifications, and additions, or change the order of the steps after understanding the spirit of the present invention.

[0089] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or replace features of other embodiments.

[0090] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations to the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A tactile roughness measurement model training method based on terahertz wireless perception, characterized in that: The method comprises the following steps: Acquire a training sample set, wherein the training sample set includes a plurality of samples, each sample including a reflected signal reflected back after the object to be tested receives a signal transmitted by a preset terahertz time-domain spectrometer at multiple angles; the object to be tested is fixed at a preset distance from a transceiver antenna of the terahertz time-domain spectrometer; The reflected signal is subjected to data preprocessing to obtain a time domain signal and extract time domain features; the time domain signal is subjected to fast Fourier transform to generate a corresponding frequency domain signal and extract frequency domain features; the frequency domain signal is input into a preset scattering feature extractor to extract corresponding scattering features; the scattering feature extractor selects a Fresnel reflection model and is provided with a correction factor to evaluate the roughness of the surface of the object to be measured; Obtaining a multi-material roughness inference network, wherein the multi-material roughness inference network is a two-dimensional convolutional network with two inputs; taking the time domain signal as a first input, and taking the time domain feature, the frequency domain feature, and the scattering feature as second inputs, and inputting them simultaneously into the multi-material roughness inference network to obtain a roughness evaluation parameter of the surface of the object to be measured; The multi-material roughness inference network is trained using the training sample set until a preset performance index is met, and the trained multi-material roughness inference network is used as the final roughness measurement model.

2. The tactile roughness measurement model training method based on terahertz wireless perception according to claim 1 is characterized in that: Also includes: When performing data preprocessing on the reflected signal, obtaining the spectrum range of the terahertz time-domain spectrometer, and performing band-pass filtering on the reflected signal; Before performing fast Fourier transform on the time domain signal, a 2-fold interpolation process is performed on the time domain signal to improve the resolution.

3. The tactile roughness measurement model training method based on terahertz wireless perception according to claim 1 is characterized in that: The time domain features include at least the first zero point position of the highest peak, the second zero point position, the position of the highest peak, the distance from the highest peak to the trough position for the first time, the height of the trough from the first time to the trough position, the distance from the highest peak to the trough, the height of the highest peak, the peak width at half of the highest peak and the integral; the frequency domain features include at least the position of the highest peak and the height of the highest peak.

4. The tactile roughness measurement model training method based on terahertz wireless perception according to claim 1 is characterized in that: The scattering feature extractor uses a Fresnel reflection model and is provided with a correction factor to evaluate the roughness of the surface of the object to be measured, and further includes: If the correction factor is much smaller than 1, the surface roughness of the object to be measured is lower; If the correction factor is close to 1, the surface roughness of the object to be measured is medium; If the correction factor is much greater than 1, the surface roughness of the object to be measured is higher.

5. The tactile roughness measurement model training method based on terahertz wireless perception according to claim 1 is characterized in that: The first input channel of the multi-material roughness inference network is provided with 6 convolutional layers; the second input channel is provided with 4 convolutional layers; and a feature connection layer and a fully connected layer are sequentially provided after the first input channel and the second input channel.

6. A tactile roughness measurement method based on terahertz wireless sensing, characterized in that: The method comprises the following steps: The transceiver antenna of the preset terahertz time-domain spectrometer is used to transmit signals to the objects to be measured placed at multiple angles, and receive reflected signals reflected back at corresponding angles; the objects to be measured are fixed at a preset distance from the transceiver antenna; The reflected signal is subjected to data preprocessing to obtain a time domain signal and extract time domain features; the time domain signal is subjected to fast Fourier transform to generate a corresponding frequency domain signal and extract frequency domain features; the frequency domain signal is input into a preset scattering feature extractor to extract corresponding scattering features; The time domain signal is used as the first input, the time domain feature, the frequency domain feature and the scattering feature are used as the second input, and the roughness measurement model in the tactile roughness measurement model training method based on terahertz wireless perception as described in any one of claims 1 to 5 is input at the same time to obtain the surface roughness evaluation parameter of the object to be measured.

7. A tactile roughness measurement device based on terahertz wireless sensing, characterized in that: The device is used to perform the steps of the tactile roughness measurement method based on terahertz wireless sensing as described in claim 6, and the device includes: A terahertz source, used for providing a terahertz signal; A transceiver antenna is used to transmit terahertz signals to the object to be measured and receive reflected signals; An optical moving stage, used to carry the object to be measured and achieve horizontal movement and angle adjustment; a data preprocessing module, configured to perform data preprocessing on the reflected signal to obtain a time domain signal and extract time domain features; perform fast Fourier transform on the time domain signal to generate a corresponding frequency domain signal and extract frequency domain features; and extract the scattering features of the frequency domain signal based on a preset scattering feature extractor; The roughness reasoning module includes a trained roughness measurement model and is used to infer the roughness evaluation parameters of the surface of the object to be measured.

8. The tactile roughness measurement device based on terahertz wireless sensing according to claim 7, characterized in that: A pipe of preset length is provided between the transceiver antenna and the object to be measured; the preset length of the pipe is determined by the focal length of the built-in lens of the transceiver antenna.

9. The tactile roughness measurement device based on terahertz wireless sensing according to claim 7, characterized in that: The data preprocessing module further includes a bandpass filter for performing bandpass filtering on the reflected signal.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images