Surface topography sliding sensing method based on flexible tactile sensor

Abstract

The application discloses a surface morphology sliding sensing method based on a flexible tactile sensor, and comprises the following steps: acquiring flexible tactile sensor signals of surface morphologies of objects; performing signal interval judgment and data extraction operations on the flexible tactile sensor signals based on signal characteristics to obtain multi-channel characteristic signal data corresponding to the surface morphologies of the objects; performing multi-channel tactile sensing data processing on the flexible tactile sensor signal data according to the multi-channel characteristic signal data, and performing gray mapping to obtain corresponding tactile image classification training sample sets of the surfaces of the contacted objects; and learning and training a machine learning classification model of the surface morphologies of the objects according to the tactile image classification training sample sets to obtain the trained machine learning classification model of the surface morphologies of the objects. The machine learning classification model is obtained by extracting the sliding signal characteristics of the tactile sensor and performing training, so that the problems of low accuracy and few use scenarios of the existing tactile sensor are solved.

Description

Technical Field

[0001] This invention belongs to the field of pressure sensor technology, specifically relating to a surface topography sliding sensing method based on a flexible tactile sensor. Background Technology

[0002] With the widespread application of human-computer interaction, intelligent robots, and intelligent prosthetics in complex environments, there is an urgent need for tactile sensing. Among the diverse tactile information, the surface features of an object, including its contours, patterns, and textures, play a crucial role in determining the basic characteristics of an object's appearance.

[0003] However, the common method for surface topography recognition is to analyze the mapping distribution of signals from a tactile sensor array, i.e., to measure the distribution of static contact force. However, this method has a drawback: the sensing resolution is often affected by the excessively large size of the sensing units, making it difficult to achieve high resolution for surface topography. Furthermore, signals between individual sensing units in the array may interfere with each other, requiring complex acquisition circuits to suppress crosstalk. This also necessitates a large number of miniature thin-film transistors, which not only affects the reliability of the sensor but also typically results in high and complex manufacturing costs.

[0004] Currently, in the application of smart prosthetics, tactile sensors can identify surface textures by sensing surface vibrations. When a tactile sensor comes into contact with a surface, it can detect the surface's vibration patterns and determine the surface texture type by analyzing these patterns. However, complex surface textures can obscure the characteristic signals of vibration signals from existing tactile sensors, making it difficult to perceive surface pattern morphology.

[0005] Therefore, existing technologies still need improvement. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a surface morphology sliding sensing method based on a flexible tactile sensor to identify and classify various complex surface morphology patterns, thereby solving the problems of low accuracy and limited application scenarios of existing tactile sensors.

[0007] The technical solution adopted by this invention to solve the technical problem is as follows:

[0008] In a first aspect, the present invention provides a surface topography sliding sensing method based on a flexible tactile sensor, comprising:

[0009] Acquire flexible tactile sensor signals for the surface morphology of various objects;

[0010] The signal interval judgment and data extraction operations based on signal characteristics of the flexible tactile sensor signal are performed to obtain multi-channel feature signal data corresponding to the surface morphology of each object.

[0011] The flexible tactile sensor signal data is processed using multi-channel feature signal data, and grayscale mapping is performed to obtain the corresponding tactile image classification training sample set of the contact object surface.

[0012] Based on the tactile image classification training sample set, a machine learning classification model for object surface morphology is trained to obtain the trained machine learning classification model for object surface morphology.

[0013] In one implementation, acquiring the flexible tactile sensor signal of the surface morphology of each object includes:

[0014] The robotic arm is controlled to perform contact and sliding operations on the surfaces of various objects, and the corresponding flexible tactile sensor signals are obtained by using the flexible tactile sensor on the robotic arm; the flexible tactile sensor signals are signals obtained based on the sliding operation of the corresponding object surface pattern.

[0015] In one implementation, the step of performing signal interval judgment and data extraction operations based on signal characteristics on the flexible tactile sensor signal to obtain multi-channel feature signal data corresponding to the surface morphology of each object includes:

[0016] The signal from the flexible tactile sensor is low-pass filtered, and the maximum and minimum values ​​are extracted from the analog voltage signal curve after low-pass filtering.

[0017] The voltage data for each sensor channel is scanned using a sliding window, and the signal maxima and minima for each channel are calculated. The threshold trigger line is then calculated using the following threshold calculation formula:

[0018] θ=minS i +(maxS i +minS i )×δ

[0019] Where Si is the voltage signal value of sensor unit i, and δ is the empirically selected threshold;

[0020] The flexible tactile sensor signal is used to determine the signal range based on signal characteristics according to the threshold trigger line, wherein the determination condition is:

[0021] Based on the threshold trigger line, the data of the entire feature signal area, the data of the first preset proportion after the rising area, and the sampling point data of the second preset proportion before the falling segment are taken to form the tactile signal data corresponding to the surface morphology of the object, and the corresponding feature signal area, rising signal area and falling signal area are obtained.

[0022] In one implementation, the step of performing multi-channel tactile sensing data processing on the flexible tactile sensor signal data based on multi-channel feature signal data, and performing grayscale mapping to obtain a corresponding tactile image classification training sample set for the contact object surface, includes:

[0023] Multi-channel characteristic signal data is obtained, and the voltage signal of each sensing unit is converted into a uniform pressure change data matrix through the pressure-voltage fitting function of each flexible tactile sensor.

[0024] The pressure change data for each multi-channel is standardized so that the range of pressure change data for each multi-channel is mapped to between 0 and 255.

[0025] In one implementation, the step of performing grayscale mapping to obtain the corresponding tactile image classification training sample set for the surface of the contact object includes:

[0026] Create a grayscale image, fill the corresponding columns with the standardized data row by row, and convert the data values ​​into corresponding grayscale values ​​according to the mapping rules between data values ​​and grayscale levels to obtain a grayscale tactile image of the surface of the object being touched.

[0027] The grayscale tactile image is processed using bilinear interpolation to obtain a corresponding set of tactile image classification training samples for the surface of the contact object.

[0028] In one implementation, the step of training a machine learning classification model of object surface morphology based on a set of tactile image classification training samples to obtain a trained machine learning classification model of object surface morphology includes:

[0029] The DNN supervised learning algorithm is used to train the tactile image classification training sample set to obtain a machine learning classification model of the object surface morphology corresponding to the contact-sliding operation.

[0030] In one implementation, the flexible tactile sensor signal is a continuous signal when the flexible tactile sensor and the surface pattern morphology make sliding contact.

[0031] In a second aspect, the present invention also provides a terminal, a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the surface topography sliding sensing method based on a flexible tactile sensor as described in the first aspect.

[0032] Thirdly, the present invention also provides a storage medium having a computer program stored thereon, characterized in that the program is executed by a processor to implement the surface topography sliding sensing method based on a flexible tactile sensor as described in the first aspect.

[0033] The present invention, by employing the above technical solution, has the following effects:

[0034] This invention mounts a flexible sliding tactile sensor on the fingertip of a single-finger robotic arm, combining the robotic arm's contact and sliding operations on an object's surface to obtain pressure tactile information. After analyzing and processing the tactile information, a tactile image of the object's surface pattern is reconstructed. This simulates the process by which humans convert time-scanning sequence signals into spatial perception during touch. Furthermore, a deep neural network model is used to train and learn the tactile image, enabling tactile perception of the width, depth, height, and sensor loading force of the measured pattern, as well as the reconstruction and classification of tactile images of various object surface patterns. This invention solves the problems of low accuracy and limited application scenarios of existing tactile sensors by extracting and training the sliding signal features of the tactile sensor to obtain a machine learning classification model. Attached Figure Description

[0035] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0036] Figure 1 This is a flowchart of a surface topography sliding sensing method based on a flexible tactile sensor in one implementation of the present invention.

[0037] Figure 2 This is a schematic diagram of a flexible tactile sensor device in one implementation of the present invention.

[0038] Figure 3 This is a schematic diagram of the signal acquisition circuit of the flexible tactile sensor in one implementation of the present invention.

[0039] Figure 4 This is a functional schematic diagram of the terminal in one implementation of the present invention.

[0040] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0042] Exemplary methods

[0043] In the process of surface morphology recognition, a common method is to analyze the mapping distribution results of the signals of the tactile sensor array. However, the perception resolution is often affected by the excessive size of the sensing unit, making it difficult to form a high resolution capability for surface morphology.

[0044] To address the aforementioned technical problems, this invention provides a surface topography sliding sensing method based on a flexible tactile sensor. By extracting sliding signal features from the flexible tactile sensor and training them to obtain a machine learning classification model, this method solves the problems of low accuracy and limited application scenarios of existing tactile sensors.

[0045] like Figure 1 As shown, this embodiment of the invention provides a surface topography sliding sensing method based on a flexible tactile sensor, comprising the following steps:

[0046] Step S100: Acquire flexible tactile sensor signals of the surface morphology of each object.

[0047] In this embodiment, the surface topography sliding sensing method based on a flexible tactile sensor is applied to the application scenario of surface topography recognition. By installing a tactile sensor that can adapt to sliding operations between robotic arms, the robotic arms are used to perform contact and sliding operations on various object surfaces, thereby obtaining tactile sensing data of sliding operations corresponding to the surface patterns of various objects.

[0048] like Figure 2 As shown, the flexible sensor in this embodiment has a multi-layer structure, consisting of, from top to bottom: a deformation sensing layer, a flexible adhesive layer, and an electrode layer.

[0049] The deformation sensing layer is made of a flexible polymer, such as thermoplastic polyurethane elastomer or polydimethylsiloxane, as the sensor substrate. Thermoplastic polyurethane elastomer and nano-conductive carbon black particles are preferred in a specific mass ratio (e.g., 1:4) to achieve excellent sensitivity, robustness, and high stability in pressure sensing. The flexible adhesive layer is formed by coating with adhesive materials, including silicone adhesives, polyurethane adhesives, vulcanizable silicone rubber adhesives, and polydimethylsiloxane. Vulcanizable silicone rubber adhesives are preferred because their curing can be controlled at room temperature, and they maintain excellent flexibility even after curing. The electrode layer is prepared by printing eight rectangular or spiral concentric interdigital electrodes on a polyimide film, with each flexible electrode having a connection point. The connection point provides analog electrical signal acquisition; the resistance signal is converted into a corresponding digital signal after passing through the acquisition circuit, allowing for further tactile data analysis.

[0050] In this embodiment, the flexible sensor data is acquired and received by the data acquisition card NI-USB-6215, such as... Figure 2 As shown, the data acquisition card receives a 1000Hz voltage signal generated by the sensor and transmits it to the host computer via a communication protocol for subsequent data processing and classification. The position and direction of movement of the robotic arm are directly controlled by the host computer.

[0051] Specifically, in one implementation of this embodiment, step S100 includes the following steps:

[0052] Step S101: Control the robotic arm to perform contact and sliding operations on the surfaces of various objects, and use the flexible tactile sensors on the robotic arm to obtain the corresponding flexible tactile sensor signals; the flexible tactile sensor signals are signals obtained based on the sliding operations of the corresponding object surface patterns.

[0053] Step S102, the flexible tactile sensor signal is a continuous signal when the flexible tactile sensor and the surface pattern morphology make sliding contact.

[0054] In this embodiment, it is necessary to obtain a flexible tactile sensor signal, which is a continuous signal when the flexible tactile sensor makes sliding contact with the surface pattern morphology.

[0055] In this embodiment, the signal change of the flexible tactile sensor includes: the sensor signal acquisition circuit is an 8-channel voltage divider circuit, the sensor acts as a variable resistor connected in series with a fixed resistor, and the potential difference across the fixed resistor is the voltage signal of the sensor. Figure 3 As shown, when the tactile sensor unit contacts and slides on a flat surface, it enters a compressed state, causing its resistance to decrease. This change in resistance leads to an increase in the analog voltage signal acquired from the voltage divider circuit. When the sensor slides over a grooved texture, it transitions between a relaxed and compressed state, resulting in a sharp drop followed by a rise in the acquired analog voltage signal. When sliding on a convex surface, the sensor undergoes further compression compared to its initial compressed state, leading to an increase in the voltage signal. When the sensor releases from the contact surface, the voltage signal decreases. The changes in the sensor voltage signal caused by the overall compression and relaxation states of the sensor can directly reflect the millimeter-scale surface pattern morphology of the object.

[0056] like Figure 1 As shown, in one implementation of the present invention, the surface topography sliding sensing method based on a flexible tactile sensor further includes the following steps:

[0057] Step S200: Perform signal interval judgment and data extraction operations based on signal characteristics on the flexible tactile sensor signal to obtain multi-channel feature signal data corresponding to the surface morphology of each object.

[0058] In this embodiment, the extraction of surface morphology features from the flexible tactile sensor signal includes a single-finger manipulator equipped with a sliding sensor making contact with the object and then sliding it to obtain a voltage data sequence of the pressure sensor changing over time.

[0059] In this embodiment, the specific process of performing signal range determination and data extraction based on signal characteristics of the flexible tactile sensor signal is as follows:

[0060] First, the voltage signal of the flexible tactile sensor is divided into an uncontacted signal region, a rising signal region, a falling signal region, and a characteristic signal region. The analog voltage signal obtained after the flexible tactile sensor slides on the object surface is then low-pass filtered, and the maxima and minima are extracted from the low-pass filtered analog voltage signal curve.

[0061] Then, the voltage data for each sensor channel is scanned using a sliding window, and the signal maxima and minima for each channel are calculated. The threshold trigger line is then calculated using the following threshold calculation formula:

[0062] θ=minS i +(maxS i +minS i )×δ

[0063] Where Si is the voltage signal value of sensor unit i, and δ is an empirically selected threshold (ranging from 0 to 0.2). The threshold determination method is as follows: Figure 4 As shown, the sliding window scans at intervals of 50 points. When the sensor signal meets the initial threshold condition θ, that is, from the first data point until the first signal value greater than θ is found, this segment is judged as an invalid segment that has not contacted the object, and this point is set as the starting point. Then the window starts to expand until the sensor signal value is less than θ, and this point is set as the ending point. Invalid segments after this point are removed. At this time, the feature data between the starting point and the ending point are the remaining valid segments.

[0064] Finally, the data from the entire feature signal region, along with a certain proportion of the data after the rising phase and a certain proportion of the sampling point data before the falling phase, are used to form the tactile signal data corresponding to the surface morphology of the object.

[0065] Specifically, in one implementation of this embodiment, step S200 includes the following steps:

[0066] Step S201: Perform low-pass filtering on the flexible tactile sensor signal, and extract the maximum and minimum values ​​from the low-pass filtered analog voltage signal curve;

[0067] Step S202: Perform low-pass filtering on the flexible tactile sensor signal, and extract the maximum and minimum values ​​from the analog voltage signal curve after low-pass filtering.

[0068] Step S203: Use a sliding window to scan the voltage data of each sensor channel, calculate the signal maximum and minimum values ​​for each channel, and calculate the threshold trigger line using the following threshold calculation formula:

[0069] θ=minS i +(maxS i +minS i )×δ

[0070] Where S i Let δ be the voltage signal value of sensor unit i, and δ be an empirically selected threshold.

[0071] Step S204: Based on the threshold trigger line, determine the signal range of the flexible tactile sensor signal according to signal characteristics, wherein the determination condition is:

[0072] Based on the threshold trigger line, the data of the entire feature signal area, the data of the first preset proportion after the rising area, and the sampling point data of the second preset proportion before the falling segment are taken to form the tactile signal data corresponding to the surface morphology of the object, and the corresponding feature signal area, rising signal area and falling signal area are obtained.

[0073] like Figure 1 As shown, in one implementation of the present invention, the surface topography sliding sensing method based on a flexible tactile sensor further includes the following steps:

[0074] Step S300: Perform multi-channel tactile sensing data processing on the flexible tactile sensor signal data based on the multi-channel feature signal data, and perform grayscale mapping to obtain the corresponding tactile image classification training sample set of the contact object surface.

[0075] In this embodiment, the specific process of performing multi-channel tactile sensing data processing on the flexible tactile sensor signal data based on multi-channel feature signal data is as follows:

[0076] First, the voltage change data of the eight sensor units in the extracted feature range are obtained. Each sensor has 3,000 data points, which can be represented as a matrix, where each column represents the voltage change of a channel. The voltage signal of each sensor unit is converted into a pressure change data matrix with uniformity through the pressure-voltage fitting function of each sensor.

[0077] Then, the pressure change data for each channel is standardized to map its range to between 0 and 255, creating an empty 8-column grayscale image, with each column representing a channel. For each channel, the standardized data is filled into the corresponding column row by row. According to the mapping rules between data values ​​and grayscale levels, the data values ​​can be converted into corresponding grayscale values, thus obtaining the grayscale tactile image of the surface of the object being touched.

[0078] Finally, to reduce the difference in pixel count between the row and column directions of the tactile image, bilinear interpolation is introduced for processing. The bilinear interpolation algorithm is used for resolution enhancement, interpolating an 8×3000 pixel grayscale tactile image into a 400×3000 pixel tactile image. The resolution interpolation process is as follows: First, the horizontal interpolation ratio is calculated, which is the target image width divided by the original image width. Next, each pixel position in the target image is traversed, and the column index of the target image is divided by the interpolation ratio. Then, the grayscale value of the current pixel is calculated using the bilinear interpolation algorithm. Finally, the calculated grayscale value is assigned to the corresponding position in the target image, interpolating the column direction from 8 pixels to 400 pixels, resulting in the interpolated 400×3000 pixel grayscale tactile image data.

[0079] Specifically, in one implementation of this embodiment, step S300 includes the following steps:

[0080] Step S301: Obtain multi-channel feature signal data, and convert the voltage signal of each sensing unit into a uniform pressure change data matrix through the pressure-voltage fitting function of each flexible tactile sensor.

[0081] Step S302: Standardize the pressure change data of each multi-channel so that the pressure change data range of each multi-channel is mapped to between 0 and 255.

[0082] Step S303: Create a grayscale image by filling the corresponding columns with the standardized data row by row. According to the mapping rules between data values ​​and grayscale levels, convert the data values ​​into corresponding grayscale values ​​to obtain a grayscale tactile image of the surface of the object being touched.

[0083] Step S304: Process the grayscale tactile image using bilinear interpolation to obtain a set of tactile image classification training samples for the corresponding contact object surface.

[0084] like Figure 1 As shown, in one implementation of the present invention, the surface topography sliding sensing method based on a flexible tactile sensor further includes the following steps:

[0085] Step S400: Based on the tactile image classification training sample set, the machine learning classification model of the object surface morphology is trained to obtain the trained machine learning classification model of the object surface morphology.

[0086] In this embodiment, the machine learning classification model for the object surface morphology is trained by using a DNN supervised learning algorithm to obtain a classification model of the object surface morphology features.

[0087] In one specific application scenario of this embodiment, the surface topography sliding sensing method based on a flexible tactile sensor includes the following steps:

[0088] (1) A single-finger manipulator equipped with a sliding sensor makes contact with an object and then slides it to obtain a voltage data sequence of the pressure sensor over time.

[0089] (2) The preprocessing of the change signal acquired by the flexible tactile sensor using the sliding operation is as follows:

[0090] 2.1 First, the analog voltage signal obtained by the flexible tactile sensor after sliding on the surface of the object is low-pass filtered, and the maximum and minimum values ​​are extracted from the low-pass filtered analog voltage signal curve.

[0091] Then, a sliding window is used to scan the voltage data of each sensor channel, and the signal maxima and minima of each channel are calculated. The threshold trigger line is then calculated using the following threshold calculation formula:

[0092] θ=minS i +(maxS i +minS i )×δ

[0093] Where Si is the voltage signal value of sensor unit i, δ is an empirically selected threshold (selected from 0 to 0.2), and the sliding window scans at intervals of 50 points. When the sensor signal meets the initial threshold condition θ, that is, from the first 5780 data point until the first signal value greater than θ is found, this segment is judged as an invalid segment that has not contacted the object, and this point is set as the start point. Then the window begins to expand until the sensor signal value is less than θ, and this point is set as the end point. Invalid segments after this point are removed; at this time, the feature data between the start point and the end point are the remaining valid segments.

[0094] 2.2 Extract the remaining valid data segments. Take the starting point as reference point 1. The X data points forward from the reference point constitute the ascending segment. Here, X is dynamically determined, based on the training results corresponding to different X values ​​during the training of the classification model. Similarly, backtrack from the ending point as reference point 2. The X data points forward from this point constitute the descending segment, where the X value is the same as that of the ascending segment. The remaining intermediate segments are the plateau segments.

[0095] 2.3 Extract approximately 3000 data points (sampling frequency 1000Hz) from the platform section for voltage-pressure calibration. Convert the sensor voltage signal into the magnitude of the loading force on the sensor to standardize and calibrate each sensor unit, thereby improving the uniformity of each channel.

[0096] 2.4 For each channel, the standardized tactile pressure data is filled into the corresponding column row by row and then standardized. The standardization method is as follows: subtract the minimum value from the data of each channel, then divide by the difference between the maximum and minimum values ​​to map the data to the range of 0 to 1, then multiply by 255 to map it to the grayscale value range of 0 to 255. According to the mapping rules between data values ​​and grayscale levels, the data values ​​are converted into corresponding grayscale values ​​and displayed as a grayscale tactile image.

[0097] 2.5 A bilinear interpolation algorithm is used for resolution enhancement, interpolating an 8×3000 pixel grayscale tactile image into a 400×3000 pixel tactile image.

[0098] 2.6 A deep neural network model is used for classification model training. The parameters are obtained using a grid search method to obtain the model parameters with the highest cross-validation accuracy. The input to the algorithm is the tactile image generated by the tactile sensor during contact and sliding on the object's surface. The output multi-class labels are the width, depth, and height of the pattern on the tested object's surface, the sensor's applied force, and the corresponding classification label of the pattern.

[0099] Specifically, in one implementation of this embodiment, step S400 includes the following steps:

[0100] Step S401: The DNN supervised learning algorithm is used to train the tactile image classification training sample set to obtain a machine learning classification model of the object surface morphology corresponding to the contact-sliding operation.

[0101] This embodiment achieves the following technical effects through the above technical solution:

[0102] This embodiment installs a flexible sliding tactile sensor on the fingertip of a single-finger robotic arm, combining the robotic arm's contact and sliding on the object's surface to obtain pressure tactile information. After analyzing and processing the tactile information, a tactile image of the object's surface pattern is reconstructed, simulating the process of converting a time-scanning sequence signal into spatial perception during human touch. A deep neural network model is used to train and learn the tactile image, enabling tactile perception of the width, depth, height, and sensor loading force of the measured pattern shape, as well as the reconstruction and classification of tactile images of various object surface patterns. The data processing method of this invention can reconstruct tactile images of object surface patterns at the millimeter and sub-millimeter level. Moreover, through tactile images and deep neural network training, the width, depth, height, sensor loading force, and corresponding classification label of the measured object surface pattern shape can be quantitatively predicted. The model trained using sliding tactile information is used for offline and online prediction, achieving a classification accuracy of 97% for various complex object surface patterns.

[0103] Exemplary device

[0104] Based on the above embodiments, the present invention also provides a terminal, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the surface topography sliding sensing method based on a flexible tactile sensor as described above.

[0105] In one embodiment, a storage medium is provided having a computer program stored thereon, characterized in that the program is executed by a processor to implement the surface topography sliding sensing method based on a flexible tactile sensor as described above.

[0106] When the computer program is executed by the processor, it is used to implement the operation of a surface topography sliding sensing method based on a flexible tactile sensor.

[0107] It will be understood by those skilled in the art that Figure 4 The principle block diagram shown is only related to the solution of this invention.

[0108] In summary, this invention provides a surface topography sliding sensing method based on a flexible tactile sensor. The method includes: acquiring flexible tactile sensor signals of various object surface topographys; performing signal interval judgment and data extraction operations on the flexible tactile sensor signals based on signal features to obtain multi-channel feature signal data corresponding to each object surface topography; performing multi-channel tactile sensing data processing on the flexible tactile sensor signal data based on the multi-channel feature signal data, and performing grayscale mapping to obtain a corresponding tactile image classification training sample set for the contact object surface; and training a machine learning classification model of the object surface topography based on the tactile image classification training sample set to obtain a trained machine learning classification model of the object surface topography. This invention solves the problems of low accuracy and limited application scenarios of existing tactile sensors by extracting sliding signal features from tactile sensors and training them to obtain a machine learning classification model.

[0109] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A surface topography sliding sensing method based on a flexible tactile sensor, characterized in that, include: Acquire flexible tactile sensor signals for the surface morphology of various objects; The signal interval judgment and data extraction operations based on signal characteristics of the flexible tactile sensor signal are performed to obtain multi-channel feature signal data corresponding to the surface morphology of each object. The flexible tactile sensor signal data is processed using multi-channel feature signal data, and grayscale mapping is performed to obtain the corresponding tactile image classification training sample set of the contact object surface. Based on the set of training samples for tactile image classification, a machine learning classification model for object surface morphology is trained to obtain the trained machine learning classification model for object surface morphology. The process of dividing the flexible tactile sensor signal into signal regions and extracting data based on signal features yields multi-channel feature signal data corresponding to the surface morphology of each object, including: The signal from the flexible tactile sensor is low-pass filtered, and the maximum and minimum values ​​are extracted from the analog voltage signal curve after low-pass filtering. The voltage data for each sensor channel is scanned using a sliding window, and the signal maxima and minima for each channel are calculated. The threshold trigger line is then calculated using the following threshold calculation formula: ； in, For sensor unit The voltage signal value, Select a threshold based on experience; The flexible tactile sensor signal is used to determine the signal range based on signal characteristics according to the threshold trigger line, wherein the determination condition is: Based on the threshold trigger line, the data of the entire feature signal area, the data of the first preset proportion after the rising area, and the sampling point data of the second preset proportion before the falling segment are taken to form the tactile signal data corresponding to the surface morphology of the object, and the corresponding feature signal area, rising signal area and falling signal area are obtained. The step of performing multi-channel tactile sensing data processing on the flexible tactile sensor signal data based on multi-channel feature signal data, and performing grayscale mapping to obtain a corresponding tactile image classification training sample set for the contact object surface, includes: Multi-channel characteristic signal data is obtained, and the voltage signal of each sensing unit is converted into a uniform pressure change data matrix through the pressure-voltage fitting function of each flexible tactile sensor. The pressure change data for each multi-channel is standardized so that the range of pressure change data for each multi-channel is mapped to between 0 and 255.

2. The surface topography sliding sensing method based on a flexible tactile sensor according to claim 1, characterized in that, The acquisition of flexible tactile sensor signals for the surface morphology of each object includes: The robotic arm is controlled to perform contact and sliding operations on the surfaces of various objects, and the corresponding flexible tactile sensor signals are obtained by using the flexible tactile sensor on the robotic arm; the flexible tactile sensor signals are signals obtained based on the sliding operation of the corresponding object surface pattern.

3. The surface topography sliding sensing method based on a flexible tactile sensor according to claim 1, characterized in that, The grayscale mapping process to obtain the corresponding tactile image classification training sample set for the contact object surface includes: Create a grayscale image, fill each row of the standardized multi-channel pressure change data into the corresponding column, and convert the data values ​​into corresponding grayscale values ​​according to the mapping rules between data values ​​and grayscale levels to obtain a grayscale tactile image of the surface of the object being touched. The obtained grayscale tactile images are processed using bilinear interpolation to obtain a corresponding set of tactile image classification training samples for the contact object surface.

4. The surface topography sliding sensing method based on a flexible tactile sensor according to claim 1, characterized in that, The step of training a machine learning classification model of object surface morphology based on a set of training samples from tactile images to obtain a trained machine learning classification model of object surface morphology includes: The DNN supervised learning algorithm is used to train the tactile image classification training sample set to obtain a machine learning classification model of the object surface morphology corresponding to the contact-sliding operation.

5. The surface topography sliding sensing method based on a flexible tactile sensor according to claim 1, characterized in that, The flexible tactile sensor signal is a continuous signal when the flexible tactile sensor and the surface pattern morphology make sliding contact.

6. A terminal, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement the surface topography sliding sensing method based on a flexible tactile sensor as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the surface topography sliding sensing method based on a flexible tactile sensor as described in any one of claims 1-5.

Images