Reconfigurable neuromorphic device, neuromorphic vision sensor array, and control method

Abstract

The application relates to the field of optoelectronic technology, in particular to a reconfigurable neuromorphic device, a neuromorphic visual sensor array and a control method, wherein the reconfigurable neuromorphic device comprises a ferroelectric bottom gate layer, a conductive floating gate layer, a tunneling dielectric layer, a two-dimensional semiconductor channel layer and a source-drain electrode which are stacked in sequence. The ferroelectric bottom gate applies electrostatic regulation to the channel through ferroelectric polarization, the floating gate is used for capturing and storing electric charges, the channel adopts a two-dimensional material with bipolar transport, wide spectrum photoelectric response and in-plane anisotropy, and the conductance state of the channel is modulated by the polarization intensity of the bottom gate. The structure not only realizes high-precision, low-power and reconfigurable conductance state regulation, but also enables a single device to cover the spectral range from visible light to near-infrared, effectively supports stable perception in weak light and night environments, the polarization-sensitive characteristics of the device support environmental target polarization information analysis, and the all-weather working capability in complex scenes such as variable illumination and reflection is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of optoelectronic technology, specifically to a reconfigurable neuromorphic device, a neuromorphic vision sensor array, and a pixel configuration method. Background Technology

[0002] All-weather perception capability refers to the system's ability to continuously and multi-dimensionally capture and visually fuse information about its surroundings at all times and in all environments. All-weather perception ensures that sensors can accurately and quickly detect moving objects even in adverse weather conditions or complex road conditions, enabling the system to make reasonable and safe driving decisions. Therefore, all-weather perception capability is crucial in fields such as autonomous driving, drones, and intelligent transportation.

[0003] Currently, solutions for all-weather sensing can be broadly categorized into two types. One involves increasing the sensing dimension through multi-module splicing, thereby improving the sensor's environmental adaptability. However, this modular approach, which physically integrates infrared sensors, cameras, and lidar, suffers from core drawbacks such as low spatial integration, high energy consumption, and difficulty in information fusion, hindering the large-scale application of all-weather sensing capabilities. Current research focuses more on materials innovation, specifically optimizing sensor performance based on the intrinsic properties of novel materials, such as two-dimensional materials. Two-dimensional materials, with their atomic-level thickness, broad spectral response, strong light-matter interaction, and unique physical properties, have become core materials for enhancing sensor sensing capabilities. However, existing research generally suffers from dimensional limitations, focusing only on the detection of single-dimensional optical information, such as visible light intensity or infrared light intensity, making it difficult to collaboratively sense multiple optical information within the same device to meet all-weather sensing requirements. Summary of the Invention

[0004] This invention provides a reconfigurable neuromorphic device, a neuromorphic vision sensor array, and a pixel configuration method to address the problem that related technologies struggle to collaboratively sense multidimensional optical information in the same device to achieve all-weather sensing.

[0005] In a first aspect, the present invention provides a reconfigurable neuromorphic device, comprising a bottom gate layer, a floating gate layer, a tunneling dielectric layer, a channel layer, and source / drain electrodes electrically connected to the channel layer, which are stacked sequentially; wherein the bottom gate layer is made of a ferroelectric material with polarization retention capability and is configured to apply electrostatic modulation to the channel layer through its polarization state; the floating gate layer is made of a conductive material and is configured to capture and store charge; the channel layer is made of a two-dimensional semiconductor material with bipolar electrical transport behavior and broad-spectrum photoelectric response characteristics, and its conductivity state is modulated by the polarization intensity of the bottom gate layer.

[0006] The reconfigurable neuromorphic device provided by this invention, through the synergistic modulation of the polarization of the ferroelectric bottom gate and the charge storage capability of the floating gate, controls a two-dimensional channel with bipolar transport and a wide-spectrum photoelectric response. This not only achieves high-precision, low-power, and reconfigurable conductance state control, but also enables a single device to cover the spectral range from visible light to near-infrared, effectively supporting stable sensing in low-light and nighttime environments. The device's polarization-sensitive characteristics support the analysis of environmental target polarization information, significantly improving its all-weather working capability in complex scenarios such as variable lighting conditions and reflections.

[0007] In some alternative implementations, the channel layer also has in-plane anisotropy.

[0008] The channel layer has in-plane anisotropy, which makes reconfigurable neuromorphic devices sensitive to the polarization direction of incident light. Without relying on external polarization optical elements, it simplifies the resolution process of polarization angle and degree of polarization of linearly polarized light, effectively suppresses specular reflection interference caused by rain, snow, fog or smooth surfaces, and enhances image contrast and target contour recognition capabilities.

[0009] In some optional embodiments, the ferroelectric material of the bottom gate layer is selected from at least one of barium titanate, barium strontium titanate, hafnium zirconium oxide, polyvinylidene fluoride (PVDF) and its copolymers, and copper indium phosphide sulfide crystal; the conductive material of the floating gate layer is selected from at least one of metal, graphene, carbon nanotubes, and conductive polymers; and the two-dimensional semiconductor material of the channel layer is selected from at least one of palladium diselenide, rhenium diselenide, black phosphorus, and black arsenic phosphorus two-dimensional materials.

[0010] Ferroelectric materials possess strong ferroelectric polarization and high durability, enabling stable and well-maintained electrostatic control. Conductive floating gate layers exhibit excellent charge trapping and storage capabilities, significantly extending device state retention time. Meanwhile, the two-dimensional semiconductor materials of the channel layer not only display a broad-spectrum photoelectric response but also possess bipolar transport behavior and significant in-plane anisotropy. The synergy of these three elements enables reconfigurable neuromorphic devices to not only support reliable imaging of neuromorphic vision sensor arrays in low-light and nighttime environments but also effectively suppress optical interference caused by adverse weather conditions such as rain, snow, and fog, significantly improving the reliability of visual perception across all times and environments.

[0011] In some alternative embodiments, the tunneling dielectric layer is composed of a high dielectric constant material or a two-dimensional insulating material, such as hafnium oxide, aluminum oxide, tantalum oxide, zirconium oxide, hexagonal boron nitride, two-dimensional molybdenum oxide, or two-dimensional tungsten oxide.

[0012] The tunneling dielectric layer has a high dielectric constant and good insulation properties, which can suppress leakage current while ensuring effective charge tunneling, thereby improving the switching ratio, control accuracy and operating stability of the device.

[0013] Secondly, the present invention also provides a neuromorphic vision sensor array, comprising a plurality of reconfigurable neuromorphic devices as described in any embodiment of the first method or the first aspect, and a control circuit coupled to the reconfigurable neuromorphic devices; wherein the plurality of reconfigurable neuromorphic devices are organized into dual-band dynamic pixel units; each dual-band dynamic pixel unit consists of two parallel reconfigurable neuromorphic devices; the control circuit is configured to: during the initialization phase, apply voltage signals of opposite polarities to the split bottom gates of the two reconfigurable neuromorphic devices in the dual-band dynamic pixel unit, so as to configure one of the reconfigurable neuromorphic devices as an NP-type operating state and the other reconfigurable neuromorphic device as a PN-type operating state, and maintain the operating state unchanged during subsequent consecutive frame exposures.

[0014] The neuromorphic vision sensor array provided by this invention organizes multiple reconfigurable neuromorphic devices into dual-band dynamic pixel units. During the initialization phase, a control circuit applies voltage signals of opposite polarities to two parallel devices within the unit, stabilizing them in NP-type and PN-type operating states that remain unchanged in subsequent consecutive frames. This enables complementary responses to light intensity changes in adjacent frames at the same spatial location. The PN-type device outputs a negative photocurrent representing the "off" event, while the NP-type device outputs a positive photocurrent representing the "on" event. By summing the two and performing inter-frame difference operations, background redundancy signals can be effectively canceled in static scenes (output ≈ 0), while light intensity changes caused by moving targets can be accurately preserved in dynamic scenes (output > 0). This design fully leverages the non-volatility, wide spectral response, and high sensitivity of reconfigurable neuromorphic devices. Without the need for mechanical shutters or complex timing control, it can achieve low power consumption and high signal-to-noise ratio across the entire time dimension, including dynamic visual perception under complex lighting conditions such as low light, nighttime, and sudden changes in brightness. This significantly improves the system's real-time performance, robustness, and environmental adaptability in scenarios such as autonomous driving, security monitoring, and robot vision.

[0015] In some alternative implementations, the multiple reconfigurable neuromorphic devices are further organized into polarization static pixel units, each consisting of four reconfigurable neuromorphic devices arranged regularly with crystal orientations of 0°, 45°, 90°, and 135°. The control circuit is configured to apply voltage signals of corresponding polarities to the four reconfigurable neuromorphic devices in the polarization static pixel unit during the initialization phase, thereby configuring two reconfigurable neuromorphic devices with crystal orientations of 0° and 45° into a first operating state, and configuring two reconfigurable neuromorphic devices with crystal orientations of 90° and 135° into a second operating state opposite to the first operating state, wherein the first operating state is NP type or PN type.

[0016] This implementation arranges multiple reconfigurable neuromorphic devices with in-plane anisotropy in a crystal orientation of 0°, 45°, 90°, and 135° to form a polarization static pixel unit. During the initialization phase, these units are configured into two complementary working states, enabling the synchronous acquisition of photocurrent response signals in four polarization directions within the same frame. Thanks to the intrinsic sensitivity of the two-dimensional semiconductor material in the channel layer to polarized light, the polarization information of the incident light can be acquired without the need for an external polarizer or optical modulator. This effectively suppresses specular reflection interference caused by rain, snow, fog, water surfaces, or smooth surfaces such as glass, significantly enhancing image contrast, edge sharpness, and target contour recognition capabilities. It also greatly improves the system's environmental perception reliability and all-weather operation capability in harsh weather, strong glare, or low-texture scenes.

[0017] Thirdly, the present invention also provides a control method for a neuromorphic visual sensor array, comprising: during the exposure of the current frame, reading a first photocurrent signal output by a reconfigurable neuromorphic device in an NP-type operating state within a dual-band dynamic pixel unit of the neuromorphic visual sensor array of the second aspect; during the exposure of the next frame, reading a second photocurrent signal output by a reconfigurable neuromorphic device in a PN-type operating state within a dual-band dynamic pixel unit; and adding the first photocurrent signal and the second photocurrent signal to obtain an inter-frame differential output signal.

[0018] The control method for a neuromorphic vision sensor array provided by this invention reads the complementary photocurrent signals output by two reconfigurable neuromorphic devices, fixed in NP-type and PN-type operating states in a dual-band dynamic pixel unit, respectively, in adjacent frames, and adds them together to generate an inter-frame differential output signal. This method effectively suppresses static background information while highly sensitively preserving light intensity changes caused by moving targets. This method requires no complex timing switching or external reference frames, and utilizes the inherent non-volatile operating states and opposite-polarity photoelectric response characteristics of the devices to achieve low-power, high signal-to-noise ratio, and high temporal resolution dynamic visual perception.

[0019] In some optional embodiments, the multiple reconfigurable neuromorphic devices are further organized into polarization static pixel units, each consisting of four reconfigurable neuromorphic devices arranged in a regular crystal orientation of 0°, 45°, 90°, and 135°. The control circuit is configured to, during the initialization phase, apply voltage signals of corresponding polarities to the four reconfigurable neuromorphic devices in the polarization static pixel unit to configure two reconfigurable neuromorphic devices with 0° and 45° crystal orientations as a first operating state, and the two reconfigurable neuromorphic devices with 90° and 135° crystal orientations as a second operating state opposite to the first operating state. The control method for the neuromorphic vision sensor array, which is of NP or PN type, further includes the following steps: during the same frame exposure, the photocurrent signals of four reconfigurable neuromorphic devices in the polarization static pixel unit are read to obtain four polarization-resolved response signals; the sum of horizontal-vertical linear polarization components, the sum of diagonal linear polarization components, and the difference between horizontal-vertical linear polarization components are determined according to the four polarization-resolved response signals; the polarization angle of each pixel position is calculated according to the ratio of the sum of horizontal-vertical linear polarization components to the sum of diagonal linear polarization components; and the degree of polarization is calculated according to the difference between horizontal-vertical linear polarization components, the sum of horizontal-vertical linear polarization components, and the sum of diagonal polarization components.

[0020] This implementation method synchronously reads four polarization-resolved response signals output by reconfigurable neuromorphic devices with crystal orientations of 0°, 45°, 90°, and 135° within the same frame. Based on the precise calculation of the sum and difference of horizontal-vertical polarization components and the sum of polarization components in the diagonal direction, it fully utilizes the in-plane anisotropy and reconfigurable characteristics of the two-dimensional semiconductor material in the channel layer. High-precision polarization information reconstruction can be achieved without external polarization optical elements. The obtained polarization information can effectively distinguish between diffuse reflection and specular reflection, significantly suppress glare and stray interference in scenes such as rain, snow, fog, water surfaces, or glass, enhance image contrast, edge sharpness, and target contour clarity, and greatly improve the system's perception reliability, anti-interference ability, and all-weather working performance in complex lighting, harsh weather, or low-texture environments.

[0021] In some optional implementations, the four polarization-resolved response signals include a first polarization-resolved response signal, a second polarization-resolved response signal, a third polarization-resolved response signal, and a fourth polarization-resolved response signal. The first polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 0° crystal orientation and in a first operating state; the second polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 45° crystal orientation and in a first operating state; the third polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 90° crystal orientation and in a second operating state; and the fourth polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 135° crystal orientation and in a second operating state. Determining the sum of horizontal-vertical linear polarization components, the sum of diagonal polarization components, and the difference between horizontal-vertical linear polarization components based on the four polarization-resolved response signals includes: determining the sum of horizontal-vertical linear polarization components using the sum of the first and third polarization-resolved response signals; determining the sum of diagonal polarization components using the sum of the second and fourth polarization-resolved response signals; and determining the difference between horizontal-vertical linear polarization components using the difference between the first and third polarization-resolved response signals.

[0022] This implementation method enables precise extraction of polarization information in the horizontal-vertical and diagonal directions through specific signal combinations. The polarization angle and degree of polarization can be calculated efficiently and accurately within a single frame, effectively suppressing specular reflection interference and improving imaging contrast.

[0023] Fourthly, the present invention also provides an all-weather perception visual sensing system, including a neuromorphic visual sensor array and a control circuit coupled to the array; the neuromorphic visual sensor array is configured to perform a control method for the neuromorphic visual sensor array as described in the third aspect and any embodiment of the third aspect under the control of the control circuit.

[0024] The all-weather perception visual sensing system of the present invention integrates a neuromorphic visual sensor array with a control circuit, enabling it to complete the recognition of light intensity changes and the calculation of polarization parameters without relying on an external processor. This not only significantly reduces data transmission bandwidth and system power consumption, but also significantly improves the real-time perception, anti-interference ability, and multi-dimensional information fusion efficiency in complex environments such as darkness, rain, snow, and fog. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of the structure of a reconfigurable neuromorphic device according to an embodiment of the present invention; Figure 2 This is a three-dimensional structural schematic diagram of an example of a reconfigurable neuromorphic device according to an embodiment of the present invention; Figure 3 These are optical mirror images of a reconfigurable neuromorphic device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the transfer curve when the same voltage is applied simultaneously to the split bottom gate; Figure 5 These are schematic diagrams of four configurations under different electrostatic doping conditions; Figure 6 This is a schematic diagram of the output curve of an NP junction; Figure 7 This is a schematic diagram of the output curve of a PN junction; Figure 8 It is the I under different light intensities for devices at 1064nm in NP and PN configurations. ds -V ds Schematic diagram of the curve; Figure 9 It is I at 1064nm sc A schematic diagram illustrating the linear relationship between light intensity and light intensity; Figure 10 It is the I under different light intensities for the device at 532nm in NP and PN configurations. ds -V ds Schematic diagram of the curve; Figure 11 It is I at 532nm sc A schematic diagram illustrating the linear relationship between light intensity and light intensity; Figure 12 This is a schematic diagram of the photoelectric response state retention test; Figure 13 This is a schematic diagram illustrating the polarization dependence of the device in a 1064nm NP configuration with zero bias voltage; Figure 14 This is a schematic diagram illustrating the polarization dependence of the device in a 1064nm PN configuration under zero bias voltage; Figure 15 This is a schematic diagram illustrating the polarization dependence of the device in a 532nm NP configuration with zero bias voltage. Figure 16 This is a schematic diagram illustrating the polarization dependence of the device in a 532nm PN configuration at zero bias voltage; Figure 17 This is a schematic diagram of a neuromorphic vision sensor array and its working principle; Figure 18 This is a schematic diagram of the first type of control method for a neuromorphic visual sensor array according to an embodiment of the present invention; Figure 19 This is a schematic diagram of motion detection results for a small car as a moving object in both bright and dark environments. Figure 20 It is a schematic diagram of pixel brightness distribution in original, static and dynamic images in visible light and dark environments; Figure 21 This is a schematic diagram of the polarization imaging results with a small car as the moving object in a reflective scene. Figure 22 This is a schematic diagram of the second process of the neuromorphic visual sensor array control method according to an embodiment of the present invention; Among them, 1. Source; 2. Channel layer; 3. Drain; 4. Tunneling dielectric layer; 5. Floating gate layer; 6. Bottom gate layer; 7. Bottom electrode; 8. Bottom electrode; 9. Substrate; 11. Neuromorphic device NP unit; 12. Neuromorphic device PN unit; 13. Sub-region; 14. 45° dual-band dynamic pixel unit; 15. 0° dual-band dynamic pixel unit; 16. 90° dual-band dynamic pixel unit; 17. 135° dual-band dynamic pixel unit; 18. First polarization static pixel unit; 19. Second polarization static pixel unit. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0029] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0030] This invention provides a reconfigurable neuromorphic device, such as... Figure 1The reconfigurable neuromorphic device shown includes a bottom gate layer 6, a floating gate layer 5, a tunneling dielectric layer 4, a channel layer 2, and source / drain electrodes electrically connected to the channel layer 2, stacked sequentially. The bottom gate layer 6 is made of a ferroelectric material with polarization retention capability and is configured to apply electrostatic modulation to the channel layer 2 through its polarization state. The floating gate layer 5 is made of a conductive material and is configured to capture and store charge. The channel layer 2 is made of a two-dimensional semiconductor material with bipolar electrical transport behavior and broad-spectrum photoelectric response characteristics, and its conductivity state is modulated by the polarization intensity of the bottom gate layer 6. The source / drain electrodes include a source electrode 1 and a drain electrode 3.

[0031] Among them, ferroelectric materials have the ability to maintain polarization.

[0032] Specifically, the optical response band of the two-dimensional semiconductor material covers the visible to near-infrared range.

[0033] The reconfigurable neuromorphic device provided in this embodiment uses the polarization of the ferroelectric bottom gate and the charge storage capability of the floating gate to synergistically regulate a two-dimensional channel with bipolar transport and a wide-spectrum photoelectric response. This not only achieves high-precision, low-power, and reconfigurable conductance state regulation, but also enables a single device to cover the spectral range from visible light to near-infrared, effectively supporting stable sensing in low-light and nighttime environments. The device's polarization-sensitive characteristics support the analysis of environmental target polarization information, significantly improving its all-weather working capability in complex scenarios such as variable lighting conditions and reflections.

[0034] In some alternative implementations, the channel layer 2 also exhibits in-plane anisotropy. The in-plane anisotropy of the channel layer 2 enables the reconfigurable neuromorphic device to be sensitive to the polarization direction of incident light, simplifying the resolution process of the polarization angle and degree of polarization of linearly polarized light without relying on external polarization optical elements. This effectively suppresses specular reflection interference caused by rain, snow, fog, or smooth surfaces, enhancing image contrast and target contour recognition capabilities.

[0035] In some optional embodiments, the ferroelectric material is selected from at least one of barium titanate, barium strontium titanate, hafnium zirconium oxide, polyvinylidene fluoride (PVDF) and its copolymers, and copper indium phosphide sulfide crystal; the conductive material is selected from at least one of metal, graphene, carbon nanotubes, and conductive polymers; and the two-dimensional semiconductor material is selected from at least one of palladium diselenide, rhenium diselenide, black phosphorus, and black arsenic phosphorus two-dimensional materials.

[0036] For example, the bottom gate layer 6 is made of barium titanate, the floating gate layer 5 is made of metal or graphene, and the channel layer 2 is made of palladium diselenide or rhenium diselenide. Among them, barium titanate has strong ferroelectric polarization and high durability, which can provide stable and good polarization maintenance electrostatic control; the metal floating gate or graphene floating gate has excellent charge trapping and storage capabilities, which can significantly extend the device state retention time; while palladium diselenide or rhenium diselenide not only exhibits a wide-spectrum photoelectric response, but also has bipolar transport behavior and significant in-plane anisotropy; the synergy of the three enables the reconfigurable neuromorphic device to not only support reliable imaging in low light and nighttime environments, but also effectively suppress optical interference caused by adverse weather conditions such as rain, snow, and fog, significantly improving the reliability of visual perception in all weather and all scenarios.

[0037] In some optional embodiments, the tunneling dielectric layer 4 is composed of a high dielectric constant material or a two-dimensional insulating material. The high dielectric constant material refers to an insulating material with a dielectric constant significantly higher than that of conventional silicon dioxide, and possessing low leakage current and good film-forming properties. Specifically, the high dielectric constant material is selected from at least one of hafnium oxide, aluminum oxide, tantalum oxide, and zirconium oxide; the two-dimensional insulating material is selected from at least one of hexagonal boron nitride, two-dimensional molybdenum oxide, and two-dimensional tungsten oxide.

[0038] The tunneling dielectric layer 4 is made of a high dielectric constant material or a two-dimensional insulating material, which can suppress leakage current while ensuring effective charge tunneling, thereby improving the device's switching ratio, control accuracy, and operational stability. For example, the tunneling dielectric layer 4 can be made of alumina, boron nitride, hafnium oxide, etc.

[0039] like Figure 2 As shown, the reconfigurable neuromorphic device includes, from top to bottom, a metal Sb / Au source / drain electrode, a palladium diselenide thin film, a hafnium oxide thin film as a tunneling layer 4, a metal Au thin film as a floating gate layer 5, a barium titanate thin film as a bottom gate layer 6, a metal Cr / Au bottom electrode 7, a metal Cr / Au bottom electrode 8, and a silicon dioxide / silicon substrate 9. The metal Sb / Au source / drain electrode is deposited on the palladium diselenide thin film and the hafnium oxide thin film, the metal Au thin film is located between the hafnium oxide thin film and the barium titanate thin film, and the metal Cr / Au bottom electrode 7 and the metal Cr / Au bottom electrode 8 are deposited on the silicon dioxide / silicon substrate 9 and covered by the barium titanate thin film.

[0040] Figure 2The reconfigurable neuromorphic device shown uses palladium diselenide as the channel material, enabling bipolar electrical transport behavior and polarization sensitivity. A split-gate electrostatic modulation of the PN and NP junctions is then employed to obtain bidirectional photocurrent. Barium titanate, a ferroelectric material with good polarization and durability, is used as the bottom gate, Au, a stable material, as the floating gate, and hafnium oxide, a material with good dielectric properties, is used as the tunneling layer 4. The device operates using a composite mechanism of a ferroelectric gate and a floating gate. Barium titanate provides ferroelectric polarization, and the floating gate layer 5 assists in charge trapping, thereby increasing the current retention time after an applied electrical pulse.

[0041] Specifically, Figure 2 The reconfigurable neuromorphic device shown can be fabricated using the following methods: (1) Peeling of multilayer palladium diselenide two-dimensional material: Place the two-dimensional palladium diselenide block material on a blue transparent tape for mechanical peeling, and gently press the material surface with the blade of a scalpel. Use tweezers to gently separate the block material from the blue tape, leaving a relatively thin and uniform two-dimensional material on the blue tape. Fold the blue tape in half 3 to 5 times until the material is evenly distributed on the tape.

[0042] (2) Sample preparation of multilayer palladium diselenide two-dimensional material: Cut PDMS into small pieces with a scalpel and stick them onto a clean glass slide. Gently attach the blue transparent tape with the material attached to the PDMS and use a cotton swab to gently press the tape to ensure it adheres completely to the PDMS. Quickly peel the tape off the PDMS to attach the desired material. Select a two-dimensional material of appropriate thickness and size under an optical microscope. Use the blue tape to remove the surrounding material of the desired material to obtain the required two-dimensional material sample.

[0043] (3) Preparation of the split gate electrode: Polymethyl methacrylate (PMMA) A8 adhesive was coated onto a silica / silicon substrate using a spin coater. The substrate was heated at 150°C for 3 minutes and then cooled. The required electrode pattern was lithographically patterned using an electron beam lithography machine. After exposure, the substrate was placed in a developer for 20 seconds and then fixed in a fixer for 40 seconds. Then, Cr / Au gate electrodes with 5 nm Cr and 10 nm Au metal were deposited by vapor deposition. After vapor deposition, the adhesive was removed to obtain a sample containing only the gate electrode.

[0044] (4) Preparation of multilayer barium titanate two-dimensional material: Barium titanate and strontium aluminate were prepared on a strontium titanate substrate by pulsed laser deposition at 740℃ and 700℃. PDMS was attached to one side of the barium titanate substrate and immersed in deionized water for 6 hours to dissolve the strontium aluminate layer and separate the strontium titanate substrate. At this point, only the multilayer barium titanate two-dimensional material remained on the PDMS. The sample was precisely transferred using a two-dimensional material transfer platform. The PDMS with the target material was attached to a glass slide, and both the glass slide and the silicon wafer were fixed on the transfer platform. The left, right, front, and back knobs were adjusted to align the material with the target area. The lifting knob was adjusted to slowly lower the glass slide until the target material on the PDMS was completely attached to the silicon wafer. The slide was heated at 60℃ for 5 minutes. After heating, the glass slide was slowly lifted to detach the PDMS from the silicon wafer, and the target material was transferred to the silicon wafer.

[0045] (5) Fabrication of floating gate layer: The specific steps are the same as in 3. An Au floating gate layer with a spacing of 1.3 μm and a thickness of 30 nm is fabricated on a silicon wafer.

[0046] (6) Fabrication of source and drain electrodes: The specific steps are the same as in 3. A pair of 20nm Sb metal and 30nm Au metal electrodes with a spacing of 7.5μm are fabricated on the silicon wafer as the source and drain electrodes.

[0047] (7) Transfer of multilayer barium titanate two-dimensional material: The prepared two-dimensional material sample was precisely transferred using a transfer platform. The barium titanate film from step 4 was picked up from the substrate using PVB, attached to PDMS, and transferred to the split bottom gate using a two-dimensional transfer system. It was placed in ethanol and heated at 50°C for 2 hours to remove PVB. The transfer steps were the same as in step 4.

[0048] (8) Transfer of floating gate layer: The specific steps are the same as in step 7. Align the Au floating gate layer with the split bottom gate below and cover it on barium titanate. Use ethanol again to remove PVB.

[0049] (9) Preparation of hafnium oxide layer: 25 nm hafnium oxide was prepared at 250 °C by atomic layer deposition.

[0050] (10) Transfer of palladium diselenide channel: The specific steps are the same as in step 4. Transfer palladium diselenide to hafnium oxide and align it with the bottom electrode so that it is in the middle part of the bottom electrode.

[0051] (11) Transfer of source and drain electrodes: Transfer and align the source and drain electrode materials obtained in step 6 to complete the device fabrication. See step 7 for details of the transfer.

[0052] Figure 3The optical microscope images of the reconfigurable neuromorphic device clearly show the physical layout and spatial assembly relationship of each functional layer. The images are labeled with the source, drain, palladium diselenide channel layer, hafnium oxide tunneling dielectric layer, gold floating gate layer, barium titanate bottom gate layer, and two independently set split bottom gates on the left and right. The functional regions of the PN junction and NP junction are clearly marked, and a 20μm scale bar is also marked to characterize the device's microscopic size. The functional layers are vertically stacked, with the source and drain electrodes symmetrically distributed at both ends of the palladium diselenide channel layer. The split bottom gates are arranged in parallel below the bottom gate layer. The overall structure is aligned and regular, with no obvious interlayer misalignment or physical defects. This visually verifies the consistency between the actual fabrication structure and the design scheme of the device, and provides reliable physical structural support for independently applying voltages of opposite polarities to the split bottom gates to achieve PN / NP state switching and thus achieve bipolar photocurrent output.

[0053] Figure 4 By obtaining the transfer curves when the same gate voltage is applied simultaneously to the split bottom gate, the electrical properties of palladium diselenide under full gate control can be obtained. The counterclockwise hysteresis loop indicates that ferroelectric modulation dominates under gate voltage control, and the hysteresis window preliminarily suggests that the palladium diselenide channel exhibits non-volatile electrical behavior. Because the two gates can be controlled independently and configured with different voltages, the channel can exhibit four different polarities. Figure 5 This indicates that when positive (negative) gate voltages are applied simultaneously to the gates, the channel exhibits hole (electron) doping, and the device displays linear PP (NN) electrical properties. When opposite but equal gate voltages are applied to both gates, the device exhibits NP (PN) electrical properties. Furthermore, the effect of different voltage pulses on the channel current is investigated. Theoretically, barium titanate can be manipulated into a neutral phase-difference mixed configuration. By applying opposite voltage pulses of equal absolute value to the split bottom gates, the polarization state of barium titanate can be controlled. The polarization intensity increases with the gradual application of the pulse amplitude, increasing the internal electric field, which in turn increases the accumulation of holes and electrons, resulting in an increasing trend in the channel currents of both the PN and NP junctions. Figure 6 and Figure 7 As shown, the split bottom gate is configured with voltage pulses of the same absolute value but opposite values, with a pulse intensity interval of 0.1V. When reading the source-drain current at zero bias, the current not only shows an increasing trend, but the device also maintains a stable current when the gate pulse is removed. This proves that the device can not only be finely controlled, but also has non-volatile behavior, which is helpful for subsequent frame difference calculation.

[0054] To investigate the photoresponse characteristics of this neuromorphic device, a scanning photocurrent testing system was used to verify whether its photoresponse characteristics in the visible and near-infrared bands were stable and controllable. Theoretically, electron-hole pairs generated under illumination separate under the influence of the built-in electric field of a PN or NP junction, generating an electromotive force inside the semiconductor. At zero bias, this will produce a short-circuit photocurrent. This is the photovoltaic effect. During retinal signal transmission, the visual signal is split into ON and OFF signals. This physiological mechanism matches the bipolar photocurrent characteristics output by this device in NP / PN states, thus simulating the dynamic light intensity response of biological vision. When the incident laser wavelength is 532nm or 1064nm, an opposite voltage pulse (±4V, 50ms) is applied to the split bottom gate, such as... Figure 8 and Figure 10 As shown, the PN junction exhibits a negative photocurrent under zero bias, while the NP junction exhibits a positive photocurrent under zero bias. To detect the light intensity sensitivity of this sensor, the light intensity was gradually increased, and the current showed an increasing trend. From the measured... - Extracting different light intensities from the curve Fitting is performed to obtain the PN and NP photodiodes generated It has a linear relationship with the incident light intensity, such as Figure 9 and Figure 11 As shown. Fitting R under 1064nm PN or NP junction state. 2 The fitted R values ​​are 0.98447 and 0.98448 for the 532nm laser state, respectively. 2 The values ​​are 0.9415 and 0.9605 respectively, both close to 1, indicating a good fit. This not only shows that the photocurrent generation originates from the photovoltaic effect but also reflects the stability of the device's photoelectric performance. This linear relationship applies not only to dynamic changes caused by variations in light intensity but also to the implementation of internal computational applications within the sensor. Figure 12 As shown, after removing the gate voltage, the photoresponse exhibits 10 when the source-drain voltage is zero. 3 The hold time of s demonstrates the long-term stability of the positive and negative photocurrent states, further indicating that the device exhibits non-volatile behavior, which is beneficial for reducing power consumption.

[0055] Polarization, besides intensity, is another dimension of information reflecting the inherent optical properties of an object. The angle of linear polarization (AoLP) clearly reflects the direction of linearly polarized light and can be used to enhance image contrast and target recognition. The degree of linear polarization (DoLP) represents the proportion of linearly polarized light intensity to the total light intensity, and can intuitively reflect the degree of light polarization. Analyzing DoLP can remove reflections in images, improve image quality, and aid in target recognition and classification. Currently, the measurement of AoLP and DoLP mainly relies on the mature four-pixel method, including measurements in four directions: 0°, 45°, 90°, and 135°. This device is not only sensitive to changes in light intensity but also to light polarization. To investigate the polarization characteristics of this device, a polarized light system was used to test the photocurrent at different polarization angles under 1064nm and 532nm illumination. We used a polarized photoelectric system to test the angle-resolved photocurrent under 1064nm and 532nm illumination. The polarization angle of the incident light can be changed by rotating the angle of the half-wave plate. The initial position is such that the incident polarized light is perpendicular to the sawtooth direction of PdSe2. Sampling is performed every 15° to measure the change of photocurrent with polarization angle. Figure 13 Taking the NP state at 1064nm as an example, an opposite gate voltage pulse is applied to the split-bottom gate voltage to put the device in the NP state. The channel is then illuminated with incident light of a certain power, and the half-wave plate is automatically rotated at a certain angle. The photoresponse of the device at different incident angles is obtained under zero bias. Under different incident light powers, the photocurrent of the NP junction exhibits good 360° periodicity, and the fitting results with the sine function are good. For example, Figure 13 and Figure 14 As shown, at 1064 nm, the polarization ratio of the NP junction state is mostly around 4.8, while that of the PN junction state is mostly around 3.8. This device also exhibits good optical response in the visible light band. From... Figure 15 and Figure 16 It can be seen that, under the NP and PN configurations at 532nm, the relationship between the photocurrent value and the incident light angle under different incident light powers still exhibits a good 360° periodicity. The polarization ratio of the NP junction is mostly around 1.4, while that of the PN junction is mostly around 1.25. This indicates that the device can still accurately adapt to the dynamic changes in light intensity within the visible light band, and it is also very likely that it can successfully analyze the polarization information of objects, such as AoLP and DoLP, through sensor calculations, thereby enhancing the contrast of difficult-to-distinguish target objects.

[0056] This invention also provides a neuromorphic vision sensor array, including multiple reconfigurable neuromorphic devices and a control circuit coupled to the reconfigurable neuromorphic devices; wherein the multiple reconfigurable neuromorphic devices are organized into dual-band dynamic pixel units; each dual-band dynamic pixel unit consists of two parallel reconfigurable neuromorphic devices; the control circuit is configured to: during the initialization phase, apply voltage signals of opposite polarities to the split bottom gates of the two reconfigurable neuromorphic devices in the dual-band dynamic pixel unit, so as to configure one of the reconfigurable neuromorphic devices as an NP-type operating state and the other reconfigurable neuromorphic device as a PN-type operating state, and maintain the operating state unchanged during subsequent consecutive frame exposures.

[0057] The aforementioned neuromorphic vision sensor array organizes multiple reconfigurable neuromorphic devices into dual-band dynamic pixel units. During the initialization phase, the control circuit applies voltage signals of opposite polarities to two parallel devices within the unit, stabilizing them in NP-type and PN-type operating states that remain unchanged in subsequent consecutive frames. This enables complementary responses to light intensity changes in adjacent frames at the same spatial location. The PN-type device outputs a negative photocurrent representing the "light removal (OFF)" event, while the NP-type device outputs a positive photocurrent representing the "light supply (ON)" event. By summing the two and performing inter-frame difference operations, background redundancy signals can be effectively canceled in static scenes (output ≈ 0), while light intensity changes caused by moving targets can be accurately preserved in dynamic scenes (output > 0). This design fully leverages the non-volatility, wide spectral response, and high sensitivity of reconfigurable neuromorphic devices, enabling low-power, high signal-to-noise ratio all-weather dynamic visual perception, including complex lighting conditions such as low light, nighttime, and sudden changes in brightness, without the need for mechanical shutters or complex timing control. This significantly improves the system's real-time performance, robustness, and environmental adaptability in scenarios such as autonomous driving, security monitoring, and robot vision.

[0058] Furthermore, multiple reconfigurable neuromorphic devices are organized into polarization static pixel units, each consisting of four reconfigurable neuromorphic devices arranged regularly with crystal orientations of 0°, 45°, 90°, and 135°. The control circuit is configured to apply voltage signals of corresponding polarities to the four reconfigurable neuromorphic devices in the polarization static pixel unit during the initialization phase, thereby configuring two reconfigurable neuromorphic devices with crystal orientations of 0° and 45° into a first operating state, and configuring two reconfigurable neuromorphic devices with crystal orientations of 90° and 135° into a second operating state opposite to the first operating state, wherein the first operating state is NP type or PN type.

[0059] For example, the polarization static pixel units after the working state configuration are divided into two types: the first type consists of PN-type reconfigurable neuromorphic devices with 0° and 45° crystal orientation and NP-type reconfigurable neuromorphic devices with 90° and 135° crystal orientation; the second type consists of PN-type reconfigurable neuromorphic devices with 90° and 135° crystal orientation and NP-type reconfigurable neuromorphic devices with 0° and 45° crystal orientation.

[0060] In other words, the neuromorphic vision sensor array arranges multiple reconfigurable neuromorphic devices with in-plane anisotropy according to the crystal orientation rules of 0°, 45°, 90° and 135° to form polarization static pixel units. During the initialization phase, these units are configured into two sets of complementary working states, so that photocurrent response signals in four polarization directions can be acquired simultaneously within the same frame. Thanks to the intrinsic sensitivity of the two-dimensional semiconductor material of the channel layer to polarized light, the polarization information of the incident light can be resolved without external polarizers or optical modulators. This effectively suppresses specular reflection interference caused by rain, snow, fog, water surfaces or smooth surfaces such as glass, significantly enhances image contrast, edge sharpness and target contour recognition capabilities, and greatly improves the system's environmental perception robustness and all-weather operation capability in harsh weather, strong glare or low-texture scenes.

[0061] Figure 17This paper demonstrates a neuromorphic vision sensor array constructed from two-dimensional materials and its working principle. The neuromorphic vision sensor array includes neuromorphic device NP units 11 and neuromorphic device PN units 12 arranged at 0°, 45°, 90°, and 135°. The NP units 11 and PN units 12 are reconfigurable neuromorphic devices obtained by applying opposite voltages to two split bottom gates. A sub-region 13 is formed by four NP units 11 and four PN units 12, as shown in the figure. Multiple neuromorphic device units can be recombined into pixel units through pixel reconfiguration. The pixel units are of two types: one is four sets of dual-band dynamic pixel units at different angles, namely, 45° dual-band dynamic pixel unit 14, 0° dual-band dynamic pixel unit 15, 90° dual-band dynamic pixel unit 16, and 135° dual-band dynamic pixel unit 17. Each of these groups consists of two neuromorphic device NP units 11 and two neuromorphic device PN units 12 with the same orientation; the other group consists of a first polarization static pixel unit 18 and a second polarization static pixel unit 19, both of which are composed of two neuromorphic device NP units 11 and two neuromorphic device PN units 12. The first polarization static pixel unit 18 is composed of a 0° neuromorphic device PN unit 12, a 45° neuromorphic device PN unit 12, a 90° neuromorphic device NP unit 11, and a 135° neuromorphic device NP unit 11. The second polarization static pixel unit 19 is composed of a 90° neuromorphic device PN unit 12, a 135° neuromorphic device PN unit 12, a 0° neuromorphic device NP unit 11, and a 45° neuromorphic device NP unit 11.

[0062] This invention also provides a control method for a neuromorphic vision sensor array, utilizing the control circuitry within the array. Specifically, the neuromorphic vision sensor array includes multiple reconfigurable neuromorphic devices and a control circuit coupled to them; wherein the multiple reconfigurable neuromorphic devices are organized into dual-band dynamic pixel units; each dual-band dynamic pixel unit consists of two parallel reconfigurable neuromorphic devices; the control circuitry is configured to: during the initialization phase, apply voltage signals of opposite polarities to the split bottom gates of the two reconfigurable neuromorphic devices in the dual-band dynamic pixel unit, thereby configuring one reconfigurable neuromorphic device in an NP-type operating state and the other in a PN-type operating state, maintaining these operating states unchanged during subsequent consecutive frame exposures.

[0063] For example, the working state configuration process of the reconfigurable neuromorphic device in the dual-band dynamic pixel unit of the neuromorphic vision sensor array is as follows: Different polarity voltage signals are applied to the split gate of the two reconfigurable neuromorphic devices in the dual-band dynamic pixel unit through a control circuit, configuring the first reconfigurable neuromorphic device as an NP-type working state and the second reconfigurable neuromorphic device as a PN-type working state. Specifically, a first polarity voltage signal (voltage amplitude ±3V~±5V, pulse duration 30ms~100ms) is applied to the first reconfigurable neuromorphic device, utilizing the synergistic effect of the polarization of the ferroelectric material in the gate layer and the charge storage of the floating gate layer to electrostatically dope the two-dimensional semiconductor channel layer, forming an NP junction structure. A second polarity voltage signal (with the same amplitude and pulse duration as the first polarity voltage signal) is applied to the second reconfigurable neuromorphic device, reversing the polarity of the split gate voltage to reconstruct the doping type of the two-dimensional semiconductor channel layer, forming a PN junction structure (with the built-in electric field direction opposite to that of the NP junction).

[0064] Figure 18 This is a schematic flowchart of the first embodiment of the neuromorphic visual sensor array control method according to the present invention, as shown below. Figure 18 As shown, the control method for the neuromorphic vision sensor array includes the following steps: Step S1801: During the exposure of the current frame, read the first photocurrent signal output by the reconfigurable neuromorphic device in the NP-type operating state within the dual-band dynamic pixel unit of the neuromorphic vision sensor array.

[0065] For example, during the current frame exposure, the first photocurrent signal I output by the NP-type operating device under illumination is acquired. NP The illumination is visible light (405nm~532nm) or near-infrared light (780nm~1064nm). The first photocurrent signal is generated by the separation of photogenerated carriers driven by the built-in electric field of the NP junction, which corresponds to the "ON" response of the scene light intensity and characterizes the light intensity information of the current frame.

[0066] Step S1802: During the next frame exposure, read the second photocurrent signal output by the reconfigurable neuromorphic device in the PN-type operating state within the dual-band dynamic pixel unit.

[0067] For example, during the next frame exposure, while keeping the operating configuration of both devices unchanged, the second photocurrent signal I output by the PN-type operating device under the same illumination conditions is acquired. PN The second photocurrent signal is generated by the reverse separation of photogenerated carriers driven by the built-in electric field of the PN junction, corresponding to the "off" response of the scene light intensity, and characterizing the light intensity information of the next frame.

[0068] Step S1803: Add the first photocurrent signal and the second photocurrent signal to obtain the inter-frame differential output signal.

[0069] For example, the control circuit synchronously acquires the first photocurrent signal I. NP With the second photocurrent signal I PN The sum of the two is used as the inter-frame differential output signal, i.e., G = I NP + I PN If there are no moving targets in the scene, and the light intensity information in the two frames is consistent, I NP with I PN When the absolute values ​​of the two signals are close and their polarities are opposite, the output signal G≈0, achieving suppression of static background redundancy information; if there is a moving target in the scene, the light intensity information at the target location changes in adjacent frames, I NP with I PN The amplitude balance is broken, the output signal G>0, and only the dynamic information of the moving target is retained, realizing all-weather (bright / dark environment) dynamic detection.

[0070] Therefore, it can be seen that neuromorphic vision sensor arrays can achieve positive and negative photocurrents through static configuration, and then mimic the human eye's ability to detect moving objects through bipolar cells and retinal neurons. Dynamic video can be segmented into a time-evolving image stream ( Figure 19 a& Figure 19 d) The dynamic detection function is mainly based on the bidirectional photocurrent characteristics. A diagram illustrating its working principle is shown below. Figure 19 b. Compared to detection only in visible light, this device exhibits optical response in both the visible light (532nm) and near-infrared (1064nm) bands, and can simulate the dynamic detection of objects across the entire time dimension. It can detect car motion using visible spectral radiation in bright environments and near-infrared spectral radiation in dark environments. Figure 19 c& Figure 19 e). Figure 20 a and Figure 20 b shows the results of simulated object motion detection in bright and dark environments. The pixel brightness of the original motion image, after brightness normalization, is irregularly distributed between 0 and 1 (grayscale), containing both moving object information and static redundant information. If the information between two frames is static, then the brightness distribution of almost all output pixels is close to zero (blue). If there is a moving object between two frames, and static information is removed, the brightness of most pixels is close to zero, while the brightness of pixels associated with the moving object is greater than zero (red). Outputting only the result frame containing motion information significantly reduces the power consumption of data transmission and processing static information.

[0071] The control method for a neuromorphic vision sensor array provided by this invention reads the complementary photocurrent signals output by two reconfigurable neuromorphic devices, fixed in NP-type and PN-type operating states in a dual-band dynamic pixel unit, respectively, in adjacent frames, and adds them together to generate an inter-frame differential output signal. This method effectively suppresses static background information while highly sensitively preserving light intensity changes caused by moving targets. This method requires no complex timing switching or external reference frames, and utilizes the inherent non-volatile operating states and opposite-polarity photoelectric response characteristics of the devices to achieve low-power, high signal-to-noise ratio, and high temporal resolution dynamic visual perception.

[0072] This invention also provides a control method for a neuromorphic vision sensor array, applied to a control circuit in a neuromorphic vision sensor array. The neuromorphic vision sensor array includes multiple reconfigurable neuromorphic devices and a control circuit coupled to the reconfigurable neuromorphic devices. The multiple reconfigurable neuromorphic devices are organized into dual-band dynamic pixel units and polarization static pixel units. Each dual-band dynamic pixel unit consists of two parallel reconfigurable neuromorphic devices, and each polarization static pixel unit consists of four reconfigurable neuromorphic devices arranged regularly with crystal orientations of 0°, 45°, 90°, and 135°. The control circuit is configured to: during the initialization phase, apply voltage signals of opposite polarities to the split bottom gates of the two reconfigurable neuromorphic devices in the dual-band dynamic pixel unit to configure one of the reconfigurable neuromorphic devices as an NP-type operating state and the other as a PN-type operating state, and maintain the operating state unchanged during subsequent consecutive frame exposures; apply voltage signals of corresponding polarities to the four reconfigurable neuromorphic devices in the polarization static pixel unit to configure the two reconfigurable neuromorphic devices with crystal orientations of 0° and 45° as a first operating state, and the two reconfigurable neuromorphic devices with crystal orientations of 90° and 135° as a second operating state opposite to the first operating state, wherein the first operating state is either NP-type or PN-type.

[0073] Figure 22 This is a schematic diagram of a second flowchart of the neuromorphic visual sensor array control method according to an embodiment of the present invention, as shown below. Figure 22 As shown, the control method for the neuromorphic vision sensor array includes the following steps: Step S2201: During the exposure of the current frame, read the first photocurrent signal output by the reconfigurable neuromorphic device in the NP-type operating state within the dual-band dynamic pixel unit of the neuromorphic vision sensor array.

[0074] Step S2202: During the exposure of the current frame, read the photocurrent signals of the four reconfigurable neuromorphic devices in the polarization static pixel unit to obtain four polarization-resolved response signals.

[0075] Specifically, the four polarization-resolved response signals include a first polarization-resolved response signal, a second polarization-resolved response signal, a third polarization-resolved response signal, and a fourth polarization-resolved response signal. The first polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 0° crystal orientation and in an NP-type operating state; the second polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 45° crystal orientation and in an NP-type operating state; the third polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 90° crystal orientation and in a PN-type operating state; and the fourth polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 135° crystal orientation and in a PN-type operating state.

[0076] Step S2203: Determine the sum of the horizontal-vertical linear polarization components, the sum of the diagonal linear polarization components, and the difference between the horizontal-vertical linear polarization components based on the four polarization resolution response signals.

[0077] Specifically, determining the sum of horizontal-vertical linear polarization components, the sum of diagonal polarization components, and the difference between horizontal-vertical linear polarization components based on the four polarization resolution response signals includes: determining the sum of horizontal-vertical linear polarization components using the sum of the first and third polarization resolution response signals; determining the sum of diagonal polarization components using the sum of the second and fourth polarization resolution response signals; and determining the difference between horizontal-vertical linear polarization components using the difference between the first and third polarization resolution response signals.

[0078] Step S2204: Calculate the polarization angle of each pixel position based on the ratio of the sum of the horizontal-vertical linear polarization components to the sum of the diagonal linear polarization components and the formula.

[0079] Specifically, the polarization angle of each pixel position is calculated using the following formula 1, based on the horizontal-vertical linear polarization components and the diagonal linear polarization components.

[0080] Formula 1 In Formula 1, S1 represents the sum of horizontal and vertical linear polarization components, and S2 represents the sum of diagonal linear polarization components.

[0081] Step S2205: Calculate the degree of polarization based on the difference between the horizontal and vertical linear polarization components, the sum of the horizontal and vertical linear polarization components, and the sum of the diagonal linear polarization components.

[0082] Specifically, the degree of polarization is calculated using the following formula 2 based on the difference between the horizontal and vertical linear polarization components, the sum of the horizontal and vertical linear polarization components, and the sum of the diagonal linear polarization components.

[0083] Formula 2 In Formula 2, S1 represents the sum of horizontal and vertical linear polarization components, S2 represents the sum of diagonal linear polarization components, and S0 represents the difference between horizontal and vertical linear polarization components.

[0084] Step S2206: During the next frame exposure, read the second photocurrent signal output by the reconfigurable neuromorphic device in the PN-type operating state within the dual-band dynamic pixel unit.

[0085] Step S2207: During the next frame exposure, read the photocurrent signals of the four reconfigurable neuromorphic devices within the polarization static pixel unit to obtain four polarization-resolved response signals.

[0086] Step S2208: Determine the sum of the horizontal-vertical linear polarization components, the sum of the diagonal linear polarization components, and the difference between the horizontal-vertical linear polarization components based on the four polarization resolution response signals.

[0087] The method for calculating the sum of horizontal-vertical linear polarization components, the sum of diagonal linear polarization components, and the difference between horizontal-vertical linear polarization components is the same as in step S2203, and will not be repeated here.

[0088] Step S2209: Calculate the polarization angle of each pixel position based on the ratio of the sum of the horizontal-vertical linear polarization components to the sum of the diagonal linear polarization components.

[0089] The method for calculating the polarization angle is the same as in step S2204, and will not be repeated here.

[0090] Step S2210: Calculate the degree of polarization based on the difference between the horizontal and vertical linear polarization components, the sum of the horizontal and vertical linear polarization components, and the sum of the diagonal linear polarization components.

[0091] The method for calculating the polarization angle is the same as in step S2205, and will not be repeated here.

[0092] Step S2211: Add the first photocurrent signal and the second photocurrent signal to obtain the inter-frame differential output signal.

[0093] The simulation was achieved by capturing four static images at different polarization angles. Figure 21 a) and working principle ( Figure 21 (b) Perform the following steps. Calculate the linear polarization angle using Formula 1. The result should be within the range of 0 to 180 degrees. Assign color values, and finally output a color image containing the target polarization angle feature information. Figure 21 c). The linear polarization degree is calculated using Formula 2. The result should be within the range of 0 to 1. A color value is assigned, and finally, a color image containing the target polarization degree feature information is output. Figure 21d). Compared with the image in the reflection scene, the analyzed polarization angle and polarization degree image has a more obvious contrast with the target object, making it stand out more in the image and facilitating more accurate detection of the target object.

[0094] This embodiment analyzes polarization information through differences between frames at different angles and simple calculations, which can reduce the interference of complex environments on target information detection, highlight the limitations of the sensor in suppressing reflection interference, and enhance target detection and recognition capabilities. In summary, this polarization-sensitive reconfigurable neuromorphic sensor array holds promise for all-weather dynamic detection and static interference immunity. Furthermore, it is expected to further improve recognition accuracy by analyzing the polarization characteristics of light in complex light propagation scenarios such as fog, haze, and snow.

[0095] This embodiment also provides an all-weather perception visual sensing system, including a neuromorphic visual sensor array and a control circuit coupled to the array; the neuromorphic visual sensor array is configured to execute the control method described above under the control of the control circuit. By integrating the neuromorphic visual sensor array with the control circuit, the all-weather perception visual sensing system can complete the recognition of light intensity changes and the calculation of polarization parameters without relying on an external processor. This not only significantly reduces data transmission bandwidth and system power consumption but also significantly improves the real-time perception, anti-interference capability, and multi-dimensional information fusion efficiency in complex environments such as darkness, rain, snow, and fog.

[0096] In summary, this invention utilizes the intrinsic properties of two-dimensional semiconductor materials, such as a wide spectral response range, bipolar electrical transport behavior, and in-plane anisotropy, to fabricate an electrically programmable, reconfigurable polarization-sensitive neuromorphic vision device. Based on this, a novel arrangement is proposed, moving from modular splicing to an integrated design, and from single-dimensional detection to multi-dimensional perception. This novel structural design enables dynamic detection in both bright and dark environments and the reconstruction of static polarization information distribution in complex reflective scenes. It will demonstrate broad application prospects in multiple fields such as autonomous driving, drone operations, smart security, and industrial monitoring, promoting scientific and technological progress and industrial innovation in related fields.

[0097] In the field of autonomous driving, the wide-spectrum response (visible and near-infrared bands) and positive and negative photocurrent switching capability of this invention, combined with differential algorithms, can accurately detect the dynamic trajectories of vehicles, pedestrians, and obstacles in complex scenarios such as heavy rain, dense fog, and strong light at night. At the same time, it can capture the polarization information of static traffic signs to avoid misjudgments caused by reflection interference. The stable non-volatile photocurrent design under zero bias voltage greatly reduces power consumption, and the high integration design is suitable for small installation spaces, improving driving safety and system reliability.

[0098] In the field of drone operations, the self-powered characteristics of this invention can greatly reduce the energy consumption of the sensing module, thereby potentially extending the drone's flight time. Its wide spectrum and multi-dimensional sensing capabilities enable it to work stably in complex environments such as low light, nighttime, and strong reflectivity in mountainous areas. The highly integrated design can reduce the overall size and weight, broadening the application development of drones.

[0099] In the field of smart security, traditional security cameras have low perception accuracy in scenarios such as night, heavy fog, and backlight, and rely on external power supply. However, the all-weather dynamic detection capability of this invention can clearly identify the movement trajectory of people and vehicles in pitch black night and in bad weather. The static polarization information perception function effectively copes with interference scenarios such as glass reflection, improving the accuracy of monitoring. Its self-powered characteristics make it possible to deploy in remote areas with weak power supply conditions.

[0100] In the field of industrial monitoring, this invention can meet the long-term sensing needs of complex industrial scenarios and extreme environments. In industrial production, its multi-dimensional sensing capabilities can monitor and capture potential equipment malfunctions in real time, and its low-power self-powered design conforms to green and sustainable development.

[0101] In summary, this invention, through the design of a synergistic structure of two-dimensional palladium diselenide and a split-gate, a barium titanate / gold / hafnium oxide architecture and novel arrangement, and a differential algorithm, enables multi-dimensional all-weather sensing in the visible-near-infrared band, stable operation with zero bias and low power consumption, and highly integrated miniaturized deployment. This invention can be widely applied in multiple fields such as autonomous driving, drones, smart security, and industrial monitoring, solving the core problems of existing technologies such as low spatial integration, single sensing dimension, poor signal synchronization, high power consumption, and reliance on external power supply. It provides innovative solutions for intelligent and long-term sensing in various fields, promoting technological upgrading and large-scale implementation in related industries.

[0102] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A reconfigurable neuromorphic device, characterized in that, It includes a bottom gate layer, a floating gate layer, a tunneling dielectric layer, a channel layer, and source / drain electrodes electrically connected to the channel layer, which are stacked sequentially. The bottom gate layer is made of a ferroelectric material with polarization retention capability and is configured to apply electrostatic control to the channel layer through its polarization state. The floating gate layer is made of a conductive material and is configured to capture and store charge; The channel layer is composed of a two-dimensional semiconductor material with bipolar electrical transport behavior and broad-spectrum photoelectric response characteristics, and its conductivity state is modulated by the polarization intensity of the bottom gate layer.

2. The reconfigurable neuromorphic device according to claim 1, characterized in that, The channel layer also exhibits in-plane anisotropy.

3. The reconfigurable neuromorphic device according to claim 1 or 2, characterized in that, The ferroelectric material is selected from at least one of barium titanate, barium strontium titanate, hafnium zirconium oxide, polyvinylidene fluoride (PVDF) and its copolymers, and copper indium phosphide sulfide crystal; the conductive material is selected from at least one of metal, graphene, carbon nanotubes, and conductive polymers; and the two-dimensional semiconductor material is selected from at least one of palladium diselenide, rhenium diselenide, black phosphorus, and black arsenic phosphorus two-dimensional materials.

4. The reconfigurable neuromorphic device according to claim 1, characterized in that, The tunneling dielectric layer is composed of a high dielectric constant material or a two-dimensional insulating material.

5. A neuromorphic visual sensor array, characterized in that, include: Multiple reconfigurable neuromorphic devices as described in any one of claims 1 to 4, and control circuitry coupled to said reconfigurable neuromorphic devices; The plurality of reconfigurable neuromorphic devices are organized into dual-band dynamic pixel units; each dual-band dynamic pixel unit consists of two parallel reconfigurable neuromorphic devices. The control circuit is configured to: during the initialization phase, apply voltage signals of opposite polarity to the split bottom gates of the two reconfigurable neuromorphic devices in the dual-band dynamic pixel unit, so as to configure one of the reconfigurable neuromorphic devices as an NP-type operating state and the other as a PN-type operating state, and maintain the operating state unchanged during subsequent consecutive frame exposures.

6. The neuromorphic visual sensor array according to claim 5, characterized in that, The plurality of reconfigurable neuromorphic devices are further organized into polarization static pixel units, each consisting of four reconfigurable neuromorphic devices arranged regularly with crystal orientations of 0°, 45°, 90° and 135°. The control circuit is configured to: during the initialization phase, apply voltage signals of corresponding polarity to the four reconfigurable neuromorphic devices in the polarization static pixel unit to configure two of the reconfigurable neuromorphic devices with 0° and 45° crystal orientations as a first operating state, and configure two of the reconfigurable neuromorphic devices with 90° and 135° crystal orientations as a second operating state opposite to the first operating state, wherein the first operating state is NP type or PN type.

7. A control method for a neuromorphic visual sensor array, characterized in that, include: During the current frame exposure, the first photocurrent signal output by the reconfigurable neuromorphic device in the dual-band dynamic pixel unit of the neuromorphic vision sensor array as described in claim 5 is read. During the next frame exposure, the second photocurrent signal output by the reconfigurable neuromorphic device in the PN-type operating state within the dual-band dynamic pixel unit is read. The first photocurrent signal and the second photocurrent signal are added together to obtain the inter-frame differential output signal.

8. The control method for the neuromorphic visual sensor array according to claim 7, characterized in that, The plurality of reconfigurable neuromorphic devices are further organized into polarization static pixel units, each consisting of four reconfigurable neuromorphic devices arranged regularly with crystal orientations of 0°, 45°, 90°, and 135°. The control circuit is configured to: during the initialization phase, apply voltage signals of corresponding polarities to the four reconfigurable neuromorphic devices in the polarization static pixel unit to configure two of the reconfigurable neuromorphic devices with crystal orientations of 0° and 45° as a first operating state, and configure the two reconfigurable neuromorphic devices with crystal orientations of 90° and 135° as a second operating state opposite to the first operating state, wherein the first operating state is NP type or PN type. The control method of the neuromorphic vision sensor array further includes: during the same frame exposure, reading the photocurrent signals of the four reconfigurable neuromorphic devices in the polarization static pixel unit to obtain four polarization-resolved response signals. Based on the four polarization resolution response signals, determine the sum of the horizontal-vertical linear polarization components, the sum of the diagonal linear polarization components, and the difference between the horizontal-vertical linear polarization components, respectively. The polarization angle of each pixel position is calculated based on the ratio of the sum of the horizontal and vertical linear polarization components to the sum of the diagonal linear polarization components. The degree of polarization is calculated based on the difference between the horizontal and vertical linear polarization components, the sum of the horizontal and vertical linear polarization components, and the sum of the diagonal linear polarization components.

9. The control method for the neuromorphic visual sensor array according to claim 8, characterized in that, The four polarization-resolved response signals include a first polarization-resolved response signal, a second polarization-resolved response signal, a third polarization-resolved response signal, and a fourth polarization-resolved response signal. The first polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 0° crystal orientation and in a first operating state; the second polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 45° crystal orientation and in the first operating state; the third polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 90° crystal orientation and in a second operating state; and the fourth polarization-resolved response signal corresponds to a reconfigurable neuromorphic device with a 135° crystal orientation and in the second operating state. The step of determining the sum of horizontal-vertical linear polarization components, the sum of diagonal linear polarization components, and the difference between horizontal-vertical linear polarization components based on the four polarization resolution response signals includes: The sum of the first polarization resolution response signal and the third polarization resolution response signal is used to determine the horizontal-vertical linear polarization components. The diagonal polarization components are determined by the sum of the second polarization resolution response signal and the fourth polarization resolution response signal. The difference between the horizontal and vertical linear polarization components is determined by the difference between the first polarization resolution response signal and the third polarization resolution response signal.

10. An all-weather perception visual sensing system, characterized in that, The device includes a neuromorphic visual sensor array and a control circuit coupled to the neuromorphic visual sensor array; the neuromorphic visual sensor array is configured to perform a control method for the neuromorphic visual sensor array as described in any one of claims 7 to 9 under the control of the control circuit.

Images