Polarization-based tactile visual perception device and method thereof

Abstract

The application provides a polarization-based visual-haptic perception device, comprising a polarized light illumination unit for generating incident polarized light, a beam splitter for splitting the incident polarized light into an imaging beam and a sensing beam, a contact sensing unit for contacting a sample, and a detection imaging unit; the contact sensing unit comprises an elastic layer and a reflecting layer, the sensing beam passes through the elastic layer, is reflected by the reflecting layer, and returns to the original path, and forms a sensing exit light parallel to the optical axis of the imaging beam after passing through the beam splitter; the imaging beam and the sensing exit light enter the detection imaging unit in the same direction, and the detection imaging unit can take pictures and obtain a plurality of polarized images with different polarization phase angles. In addition, the application also provides a method using the polarization-based visual-haptic perception device. Compared with the prior art, the polarization-based visual-haptic perception device and method can obtain deformation without fixed shape and fine deformation, and are not affected by obstruction.

Description

Technical Field

[0001] This invention relates to the field of robot sensing technology, and in particular to a polarization-based visual-tactile sensing device and method. Background Technology

[0002] In many modern scenarios, robots play an increasingly important role, requiring precise operations in environments where they work closely with humans. Tactile information is particularly crucial in these operations. Therefore, it is extremely urgent to enable robots to acquire high-resolution tactile information and express it efficiently, ultimately achieving object recognition and attribute determination. Vision-tactile sensing is a key perception technology for tasks such as precise robot manipulation, dexterous grasping, and human-robot interaction. Acquiring contact force and its spatial distribution is one of the fundamental capabilities of vision-tactile sensing. Existing vision-tactile sensors typically observe the geometric deformation of an elastic body surface, the displacement of marker points, or changes in surface appearance using a camera, and then obtain contact force information through mechanical inversion or data-driven mapping.

[0003] Currently, the main force sensing technologies in the field of visual-tactile sensing are based on deformation fields, such as force sensing based on marker deformation and force sensing based on surface deformation. Although deformation field-based sensing methods have relatively mature contact sensing capabilities, and are quite adept at sensing contact forces such as normal forces and shear forces of observable deformation, some contact interaction scenarios, such as contact with liquids, colloids, or plastics without a fixed shape, do not produce observable and obvious deformations. It is difficult to reliably detect and sense the contact process of these behaviors using deformation fields. Secondly, methods based on marker deformation are also prone to occlusion problems, leading to the loss of local information and susceptibility to interference from occlusion. Summary of the Invention

[0004] To overcome the above shortcomings, this application provides a polarization-based visual-tactile sensing device, the technical solution of which is as follows:

[0005] A polarization-based visual-tactile sensing device includes a polarization illumination unit, a beam splitter, a contact sensing unit, and a detection imaging unit. The polarization illumination unit is used to generate incident polarized light; A beam splitter is used to separate the incident polarized light into an imaging beam and a sensing beam. The contact sensing unit is used to contact the sample, so that the sample loads the contact sensing unit. It includes an elastic layer and a reflective layer. After the sensing beam passes through the elastic layer, it is reflected by the reflective layer and then passes through the beam splitter to form a sensing output beam parallel to the optical axis of the imaging beam. The imaging beam and the sensing output beam enter the detection imaging unit in the same direction, and the detection imaging unit can capture and acquire several frames of polarization images with different polarization phase angles.

[0006] Compared with existing technologies, the polarization-based visual-tactile sensing device of this application improves the distinguishability, robustness and spatial resolution of contact states through reflective polarization imaging. Even in the presence of obstructions, it can still obtain the complete stress field distribution and can obtain the deformation of materials without fixed shapes.

[0007] Furthermore, the detection imaging unit includes a phase adjuster, a second polarizer, and an imaging camera arranged sequentially along the optical axis of the imaging beam toward the direction away from the beam splitter; the phase adjuster is capable of rotating about an axis parallel to the optical axis of the imaging beam to change the polarization phase angle; the imaging camera captures several frames of polarized images with different polarization phase angles as the phase adjuster rotates.

[0008] Furthermore, the optical axis angle between the imaging beam and the sensing beam is 180°.

[0009] Furthermore, the polarization illumination unit includes an incoherent monochromatic light source, a first polarizer, and a quarter-wave plate arranged sequentially along the incident polarized light path.

[0010] Furthermore, the polarization illumination unit includes a multispectral light source, an achromatic wideband waveplate, and a wideband polarizer arranged sequentially along the incident polarized light path.

[0011] Furthermore, the first polarizer is a metal wire grid polarizer.

[0012] Furthermore, the contact sensing unit also includes a protective layer, which is spin-coated on the side of the reflective layer opposite to the elastic layer.

[0013] Furthermore, this application also provides a visual-tactile stress distribution sensing method for obtaining the stress distribution at the load contact position between a sample and the visual-tactile stress distribution sensing device using the aforementioned visual-tactile stress distribution sensing device, comprising the following steps: Acquire a first polarization image of the polarization-based visual-tactile sensing device when no sample load is applied; Acquire polarization images at different discrete angles when the same load is applied to several polarization-based visual-tactile sensing devices; Based on the first polarization image and the polarization images at different discrete angles when the same load is applied, the Stokes parameters of the sensed outgoing light are obtained. Based on the Stokes parameters of the incident polarized light and the sensed outgoing light, the stress distribution at the load contact position between the sample and the polarization-based visual-tactile sensing device is calculated, and the tactile information of the contact sensing interface is decoupled through the stress distribution information.

[0014] Furthermore, the method also includes the following steps: changing the load magnitude to obtain a continuously changing stress distribution, and performing load motion discrimination and early warning based on the continuously changing stress distribution.

[0015] Furthermore, it also includes the following steps: Calculate the principal stress difference using the following formula. σ: Where d is the thickness of the elastic layer; C is the stress optical constant of the elastic layer; and D is the optical path difference between the incident polarized light and the sensing outgoing light. Tactile data in the spatial domain are obtained based on the stress distribution; Based on the data from the spatial domain, the principal stress difference σ is subjected to a two-dimensional Fourier transform to obtain the periodic texture of the spatial frequency distribution; The main direction of the periodic texture is obtained based on the included angle β, and the texture features are extracted. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall structure of the polarization-based visual-tactile sensing device of this application; Figure 2 This is a cross-sectional view of the polarization-based visual-tactile sensing device of this application; Figure 3 This is a schematic diagram of the internal structure of the polarization-based visual-tactile sensing device of this application; Figure 4 This is a schematic diagram of the contact sensing unit of this application; Figure 5 This is a schematic diagram of the polarization-based visual-tactile perception method of this application under a monochromatic light source. Figure 6 This is a schematic diagram of the process of the polarization-based visual-tactile perception method of this application under a multispectral light source. Figure 7 In one embodiment of this application, a polarization-based force contact stress distribution map is obtained using a polarization-based visual-tactile sensing device and method; Figure 8 (a) and (b) are schematic diagrams of polarization-based pressure maps and three-dimensional point cloud views obtained using a polarization-based visual-tactile sensing device and method in one embodiment of this application; Figure 9 A schematic diagram of the polarization-based stress vector direction obtained using a polarization-based visual-tactile sensing device and method, as shown in one embodiment of this application; Figure 10 This is a texture contour result in one embodiment of this application.

[0017] Reference numerals: 10, polarized illumination unit; 11, incoherent monochromatic light source; 12, first polarizer; 13, quarter-wave plate; 14, multispectral light source; 15, achromatic broadband waveplate; 16, broadband polarizer; 20, beam splitter; 30, contact sensing unit; 31, elastic layer; 32, reflective layer; 33, protective layer; 40, detection and imaging unit; 41, phase adjuster; 42, second polarizer; 43, imaging camera; 50, housing. Detailed Implementation

[0018] Traditional polarization-based visual-tactile sensing methods and devices suffer from difficulties in detecting and sensing stress distribution in materials without fixed shapes, and are highly dependent on the integrity of markers. To address these issues, the applicant provides a polarization-based visual-tactile sensing device and method. This device uses reflective polarization imaging to read the stress-induced birefringence information of the photoelastic material in contact with the sample, enabling the sensing output of stress-related distribution in the contact area. This improves the distinguishability, robustness, and spatial resolution of contact states, and can effectively identify minute deformations. Furthermore, the use of polarization imaging avoids the reliance on surface texture or internal markers to identify stress distribution changes. By calculating the polarization information of the reflected polarized light, the stress field distribution on the contact surface can be directly obtained, thus expanding the detection scope beyond materials with fixed shapes. The following are some specific embodiments of this application: Example 1: Monochromatic Light Source Please refer to the following: Figures 1 to 3 A polarization-based visual-tactile sensing device includes: a polarization illumination unit 10, a beam splitter 20, a contact sensing unit 30, a detection imaging unit 40, and a housing 50.

[0019] The polarization illumination unit 10 includes an incoherent monochromatic light source 11, a first polarizer 12, and a quarter-wave plate 13 arranged sequentially along the optical path.

[0020] The incoherent monochromatic light source 11 employs an incoherent monochromatic LED array, which, compared to using a laser to generate polarized light, fundamentally avoids laser speckle and ensures image clarity. Preferably, the incoherent monochromatic light source 11 uses red light with a center wavelength of 600nm, or red light with a center wavelength of 633nm, or green light with a center wavelength of 540nm. If the sensor used in the subsequent detection imaging unit 40 is a silicon-based CMOS sensor, it has a more sensitive response to this wavelength band. Furthermore, during polarization imaging, using monochromatic light can eliminate the chromatic aberration effect of the lens, improving image quality. The first polarizer 12 converts the natural light generated by the incoherent monochromatic light source 11 into polarized light, including linearly polarized light and circularly polarized light. Preferably, the first polarizer 12 is a metal wire grid polarizer. Compared to traditional iodine-based polarizers, the high polarization degree and high transmittance of the metal wire grid polarizer make it more durable, heat-resistant, humidity-resistant, and more stable in performance. The first polarizer 12 is set to be tightly bonded to the incoherent monochromatic light source 11. The first polarizer 12 and the incoherent monochromatic light source 11 can be bonded by optical adhesive or by vacuum bonding to reduce interface reflection loss and ensure light efficiency.

[0021] The polarized light formed at the first polarizer 12 enters the quarter-wave plate 13 at a fixed angle and generates circularly polarized light with a known polarization state as the incident polarized light.

[0022] Beam splitter 20 divides the incident polarized light into two beams: one is the imaging beam, which is incident on the detection imaging unit 40, and the other is the sensing beam, which is incident on the contact sensing unit 30. The optical axes of the imaging beam and the sensing beam are at an angle of 180°, and their optical axes are parallel to each other along the Z-axis. In this embodiment, the mirror surface of beam splitter 20 makes an angle of 45° with the optical axis of the incident polarized light. After being split by beam splitter 20, the horizontal incident polarized light is projected vertically upward onto the contact sensing unit 30. The reflected light, carrying sample stress information and with its optical axis parallel to the optical axis of the imaging beam, passes vertically downward again through beam splitter 20 and directly enters the detection imaging unit 40. This arrangement allows the imaging beam and the sensing light to overlap along the Z-axis, greatly saving space, avoiding shadows and uneven brightness caused by side illumination, providing optimal illumination conditions for orthophoto imaging, and featuring high uniformity, no speckle noise, high polarization degree, compact overall structure, and strong resistance to environmental interference.

[0023] Contact sensing unit 30 is used to contact the sample, so that the sample applies a load to it and obtains the stress distribution. Please refer to [link to relevant documentation]. Figure 4 The unit includes an elastic layer 31, a reflective layer 32, and a protective layer 33 stacked sequentially from the inside out. The elastic layer 31, the reflective layer 32, and the protective layer 33 are surrounded by a housing 50, with only the side of the protective layer 33 facing away from the elastic layer 31 facing outward, so that the housing 50 can protect the contact sensing unit 30.

[0024] The elastic layer 31 can be made of polydimethylsiloxane (PDMS) or polyurethane (PU) as the matrix. PDMS has excellent optical transparency and chemical stability, while polyurethane typically has higher toughness and abrasion resistance. PDMS can be cured using different formulations, and these different formulations can affect its Young's modulus and optical stress constant. Preferably, the modulus of the elastic layer 31 is set between 0.1 MPa and 1 MPa, and the optical stress constant is 1.2 × 10⁻⁶. - ¹² This hardness range simulates the mechanical properties of human fingertip soft tissue, providing sufficient deformation to ensure sensitivity while also offering enough support for rigid gripping. Furthermore, the elastic layer 31 is designed as a smooth plane without any surface texture or internal markings. The smooth surface reduces frictional interference during contact and avoids interference from textures on optical measurements.

[0025] A reflective layer 32 covers the side of the elastic layer 31 facing the protective layer 33, reflecting the incident polarized light back into the detection and imaging unit 40. The reflective layer 32 can be fabricated using a coating method, where an extremely thin metal film is deposited on the outside of the elastic layer 31 by magnetron sputtering or vapor deposition, thus forming a reflective surface with high flatness and good optical performance. In this embodiment, the coated metal film is an aluminum film, which has a reflectivity of approximately 92%, exhibiting high reflectivity. Furthermore, when incident near normal, the influence of the aluminum film on the polarization state of the incident polarized light can be compensated for through calibration, improving accuracy. The sensing beam passes through the elastic layer 31, is reflected by the reflective layer 32, returns along its original optical path, and is incident on the imaging unit 40.

[0026] The protective layer 33 is spin-coated on the side of the reflective layer 32 facing away from the elastic layer 31. Polydimethylsiloxane can be used for spin-coating, with a thickness of about 0.1 mm. The thin protective layer and curing process make its birefringence effect negligible or can be compensated by baseline subtraction.

[0027] The detection imaging unit 40 includes a phase adjuster 41, a second polarizer 42, and an imaging camera 43 arranged sequentially along the optical axis of the imaging beam away from the beam splitter 20. The phase adjuster 41 can employ a quarter-wave plate or a variable phaser platform as used in the prior art. The phase adjuster 41 can rotate about an axis parallel to the Z-axis to detect the polarization phase using a four-step or eight-step phase-shifting method as used in the prior art. The sensor-emerged light carrying stress information passes through the phase adjuster 41 and is then incident on the second polarizer 42 in a set direction, before entering the imaging camera 43. The imaging camera 43 captures multiple frames of polarized images as the phase adjuster 41 rotates. By analyzing and calculating these multiple frames of polarized images, the load distribution on the sample can be obtained. For example, under the combined action of the phase adjuster 41 and the second polarizer 42, the horizontal component, vertical component, and ±45° component of the polarized light are measured at rotation angles of 0°, 45°, 90°, and 135°, thereby obtaining the Stokes parameters S0, S1, and S2.

[0028] Based on the structure of the aforementioned polarization-based visual-tactile sensing device, this application also provides a polarization-based visual-tactile sensing method. Please refer to [link to relevant documentation]. Figure 5 This includes the following steps: Step S10: Obtain the first polarization image of the sample when it is unloaded using the polarization-based visual-tactile sensing device described above.

[0029] When there is no non-contact sample in the contact sensing unit 30, the polarization illumination unit 10 emits light, and the detection imaging unit 40 captures an image, acquiring a set of baseline polarization images as the first polarization image, denoted as... .

[0030] Step S20: Make the contact sensing unit 30 of the polarization-based visual-tactile sensing device come into contact with the sample, and acquire polarization images of several samples at different discrete angles when the sample is loaded with load.

[0031] After the contact sensing unit 30 comes into contact with the sample, it is subjected to a load. Under the applied load, a set of load polarization images is acquired as the second polarization image, denoted as... k is the number of images according to the polarization modulation method, such as the 4-step or 8-step method in the prior art.

[0032] Then, the discrete angle of the phase adjuster 41 is adjusted to obtain polarization images at different discrete angles θ_k under the same load, and one frame of image is acquired at each angle:

[0033] Step S30: Calculate and obtain the Stokes parameter S' of the sensor output light based on the first polarization image under no load and the polarization images under different discrete angles under load.

[0034] The image intensity sequences obtained in steps S10 and S20 are calculated using the following formula through orthogonal expansion:

[0035]

[0036]

[0037]

[0038] Thus, the complete Stokes parameters are obtained. .

[0039] Since different wavelengths have different phase delays, multi-wavelength information can be used for joint calculation to eliminate the 2π integer multiple phase ambiguity problem that may occur under single wavelength, thereby expanding the measurable stress range. The final stress distribution can be obtained by weighted averaging of the results from each channel.

[0040] Step S40: Calculate the stress distribution at the load contact position based on the Stokes parameters of the incident polarized light and the sensed outgoing light.

[0041] When incident polarized light is perpendicularly incident on an elastic layer 31 with thickness d and stress optical constant C, according to the birefringence effect, the principal stress difference... The relationship between σ and the optical path difference D generated after passing through the sample with the load is as follows:

[0042] Based on two-way reflection, the relationship between optical path difference and phase retardation is as follows: , Therefore, the principal stress difference and phase delay can be expressed as:

[0043] When there is contact between the contact sensing unit 30 and the sample generating the load, a sensing beam with Stokes parameter S is incident perpendicularly on the contact sensing unit 30. The pressing area will generate a phase delay δ on the incident light, and the angle between the principal axis of its birefringence fast axis and the X-axis of the image space coordinate system (i.e., the X-axis on the pressing surface) is β. The transfer matrix of the sample is M(β,δ). Then, according to the polarization transmission theory of the sensing outgoing light, the Stokes parameter S' of the sensing outgoing light can be expressed as:

[0044] The matrix form is as follows:

[0045] Solving this matrix yields:

[0046] Therefore, by obtaining the Stokes parameters of the incident polarized light and the sensed outgoing light, the phase delay at the contact position can be obtained, and the stress distribution change at the contact position can be obtained. Stress correlation diagrams related to the stress distribution change can be obtained, including force contact stress distribution diagrams, pressure diagrams, three-dimensional point cloud views, and stress vector direction diagrams.

[0047] The obtained stress distribution map can yield relevant parameters such as peak stress, contact area, and shear stress that can be used for feedback control. These features can be used to achieve applications such as normal force estimation, shear / slip warning, and gripping force feedback adjustment. Figures 6 to 8 This is a stress correlation diagram obtained after using the above-mentioned polarization-based visual-tactile sensing device and method in actual testing.

[0048] Furthermore, due to the microscopic protrusions and depressions of the sample surface texture, a spatially non-uniform stress distribution is generated in the elastic layer 31, and the obtained force contact stress distribution map itself is a mapping of the surface texture. Therefore, the texture features can be further extracted according to the following steps: Step S41: Obtain spatial domain data based on the force contact stress distribution diagram.

[0049] Local statistics can be calculated from the force contact stress distribution map using existing technologies. These local statistics include the mean, variance, gradient magnitude, and gray-level co-occurrence matrix.

[0050] Step S42: Based on the data from the spatial domain, analyze the principal stress differences. σ is subjected to a two-dimensional Fourier transform to obtain the periodic texture of the spatial frequency distribution.

[0051] Since periodic textures exhibit significant frequency peaks in the spectrum, they can be obtained through periodic textures.

[0052] Step S43: Obtain the main direction of the periodic texture based on the included angle β, thereby realizing the automatic identification and classification of different texture types of the sample and extracting texture features.

[0053] The texture recognition described above does not require modification of the optical model or addition of hardware; it only requires adding spatial feature extraction and classification steps based on the stress distribution.

[0054] Step S50: Repeat steps S20 to S50, changing the load magnitude, obtaining the continuously changing stress distribution and its corresponding stress correlation diagram, and performing load motion discrimination and early warning. Figure 9As shown, motion discrimination and early warning are achieved by acquiring stress correlation maps of consecutive frames and setting edge shear thresholds based on stress change peak values, time-domain stress thresholds, or spatial distribution. Besides thresholding, discrimination can also be achieved through statistical feature methods, pattern recognition methods, or machine learning.

[0055] Example 2: Multispectral Light Source This embodiment has a structure that is basically the same as that of the polarization-based visual-tactile sensing device in Embodiment 1, except that the polarization illumination unit 10 uses a multispectral light source capable of emitting beams with at least two different center wavelengths. This multispectral light source includes a multispectral light source 14, an achromatic broadband waveplate 15, and a broadband polarizer 16 arranged sequentially along the optical path. The multispectral light source 14 can be a multi-wavelength LED array. Correspondingly, when the polarization illumination unit 10 is a multispectral light source, the phase adjuster 41 can use an achromatic quarter-wave plate or a liquid crystal variable retarder, as used in the prior art.

[0056] Please see Figure 6 The steps of the polarization-based visual-tactile sensing method corresponding to the polarization-based visual-tactile sensing device in this embodiment are basically the same as those in Embodiment 1, except that: In step S10, N discrete wavelength channels provided by the multispectral light source are obtained. Each wavelength channel λ i The following set of polarization image sequences is represented as: I k (λ i ), k=1,2,…,K; i=1,2,…,N, where K is the number of phase shift steps and N is the number of wavelength channels.

[0057] The acquisition methods of multispectral light sources include, but are not limited to, the following two: First, the time-sequence switching method, that is, lighting up LEDs of each wavelength in sequence, completing a set of phase-shift acquisitions during the lighting of each wavelength, and obtaining polarization images corresponding to each wavelength; Second, the simultaneous illumination method, that is, lighting up LEDs of multiple wavelengths at the same time, and the imaging camera 43 adopts a color CMOS sensor with a Bayer color filter array, and extracts polarization images corresponding to each wavelength according to the R, G, and B channels respectively.

[0058] In step S20, each wavelength channel λ is acquired independently. i Polarization images of several samples under different discrete angles when loaded.

[0059] In step S30, for each wavelength channel λ i The Stokes parameters S'(λ) for each wavelength channel were calculated using the same orthogonal expansion method as in Example 1. i ).

[0060]

[0061]

[0062]

[0063]

[0064] After performing step S30, proceed to step S60: for each wavelength channel λ i The phase delay δ(λ) is calculated according to the following formula. i ) and the direction of the principal stress axis β(λ) i ) .

[0065]

[0066]

[0067] Wherein, S(λ) i S'(λ) is the Stokes parameter of the incident polarized light. i ) represents the Stokes parameter of the emitted light.

[0068] Step S70:

[0069] The unique advantage of using a multispectral light source lies in its ability to eliminate the phase ambiguity inherent in single-wavelength schemes. In single-wavelength schemes, the range limitation of the inverse cosine function leads to folding ambiguity when the phase delay exceeds π, thus restricting the measurable stress range. In multispectral schemes, for example, two wavelengths λ1 and λ2 can be used to form an equivalent wavelength Λ. 12 =(λ1·λ2) / (|λ2 By utilizing the phase difference information of the two wavelength channels, the phase order is determined, thereby expanding the unambiguous stress measurement range from λ / (4Cd) to Λ. 12 / (4Cd), specifically: For single-wavelength phase ambiguity, the unambiguous measurement range for a single wavelength (δ=π) is:

[0070] For dual-wavelength deblurring, taking dual-wavelength as an example, the phase delays produced by the same stress Δσ at the same location for the two wavelengths λ1 and λ2 (λ1 < λ2) are as follows: The difference between the two is:

[0071] Equivalent wavelength:

[0072] The phase difference can then be expressed as:

[0073] Due to Λ 12 λ1, λ2, the unambiguous stress measurement range corresponding to the equivalent wavelength (Δδ=π):

[0074] Taking λ1=540nm and λ2=633nm as an example, we get:

[0075] Therefore, the unambiguous measurement range is expanded to multiples of a single wavelength:

[0076] As an example, the results detected using the polarization-based visual-tactile sensing device and method of this application can be found in [reference needed]. Figures 7 to 10 Compared with existing technologies, the polarization-based visual-tactile sensing device and method of this application have the following advantages: 1. This method directly utilizes the photoelastic effect to convert stress measurement into the measurement of optical polarization signals. Compared to indirect measurement schemes that map deformation to a stress field, this is a direct measurement method. Through physical calculations of polarization information, the contact stress distribution can be directly, fully, and quantitatively observed and measured. Therefore, the stress field can be used to sense and distinguish more minute tactile information, and the stress vector relationship and stress distribution can be used to detect and provide early warnings of micro-slippage. Most current multi-axis force detection technologies calculate the displacement of the marker and the actual stress magnitude, then use statistical methods such as machine learning to establish a relationship to deduce the stress magnitude—a statistical result. This application, however, uses a physical mechanism to directly calculate the stress magnitude—a physical calculation result.

[0077] 2. A fundamental breakthrough in perception: Direct measurement of internal stress distribution. Existing visual-tactile technologies can only indirectly infer surface contact force, while this invention enables direct and visual observation of the full-field stress distribution within the sensing medium. By directly converting stress into optical signals through the photoelastic effect, it avoids the complex process and accumulated errors of traditional methods that "calculate force from deformation," providing more fundamental and richer mechanical information. Monochrome LEDs are used to suppress ambient light, thereby improving the signal-to-noise ratio. A coaxial beam splitter facilitates compact optical path integration, and a two-way reflective design enhances the phase delay signal of contact information. Contact information based on the stress field enables spatial perception of the full-field force distribution, and combined with the stress vector direction, it enables temporal perception and early warning of slippage.

[0078] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A polarization-based visual-tactile sensing device, characterized in that, include: A polarization illumination unit is used to generate incident polarized light; A beam splitter is used to split the incident polarized light into an imaging beam and a sensing beam, wherein the optical axis of the imaging beam and the sensing beam are at an angle of 180°. A contact sensing unit is used to contact the sample so that the sample loads the contact sensing unit. It includes an elastic layer and a reflective layer. After the sensing beam passes through the elastic layer, it is reflected by the reflective layer and then passes through the beam splitter to form a sensing output beam parallel to the optical axis of the imaging beam. The detection imaging unit includes a phase adjuster, a second polarizer, and an imaging camera arranged sequentially along the optical axis of the imaging beam toward the direction away from the beam splitter. The phase adjuster can rotate about an axis parallel to the optical axis of the imaging beam to change the polarization phase angle. The imaging camera captures several frames of polarized images with different polarization phase angles as the phase adjuster rotates, so that the imaging beam and the sensing output light enter the detection imaging unit in the same direction. The detection imaging unit can capture and acquire several frames of polarized images with different polarization phase angles.

2. The polarization-based visual-tactile sensing device according to claim 1, characterized in that: The polarization illumination unit includes an incoherent monochromatic light source, a first polarizer, and a quarter-wave plate arranged sequentially along the incident polarized light path.

3. The polarization-based visual-tactile sensing device according to claim 1, characterized in that: The polarization illumination unit includes a multispectral light source, an achromatic wideband waveplate, and a wideband polarizer arranged sequentially along the incident polarized light path.

4. The polarization-based visual-tactile sensing device according to claim 2, characterized in that: The first polarizer is a metal wire grid polarizer.

5. The polarization-based visual-tactile sensing device according to claim 1, characterized in that: The contact sensing unit further includes a protective layer, which is spin-coated on the side of the reflective layer opposite to the elastic layer.

6. A polarization-based visual-tactile sensing method, using the polarization-based visual-tactile sensing device of claim 1, to obtain the stress distribution at the load contact position between the sample and the polarization-based visual-tactile sensing device, characterized in that, Includes the following steps: Acquire a first polarization image of the polarization-based visual-tactile sensing device when no sample load is applied; Acquire polarization images at different discrete angles when the same load is applied to several polarization-based visual-tactile sensing devices; Based on the first polarization image and the polarization images at different discrete angles when the same load is applied, the Stokes parameters of the sensed outgoing light are obtained. Based on the Stokes parameters of the incident polarized light and the sensed outgoing light, the stress distribution at the load contact position between the sample and the polarization-based visual-tactile sensing device is calculated, and the tactile information of the contact sensing interface is decoupled through the stress distribution information.

7. The polarization-based visual-tactile perception method according to claim 6, characterized in that, It also includes the following steps: By changing the load magnitude, a continuously changing stress distribution is obtained, and load movement is judged and warned based on the continuously changing stress distribution.

8. The polarization-based visual-tactile perception method according to claim 6, characterized in that, It also includes the following steps: Calculate the principal stress difference using the following formula. σ: Where d is the thickness of the elastic layer; C is the stress optical constant of the elastic layer; and D is the optical path difference between the incident polarized light and the sensing outgoing light. Tactile data in the spatial domain are obtained based on the stress distribution; Based on the data from the spatial domain, the principal stress difference σ is subjected to a two-dimensional Fourier transform to obtain the periodic texture of the spatial frequency distribution; The main direction of the periodic texture is obtained based on the included angle β, and the texture features are extracted.

Images