Ultraviolet photomemristor inspired by bee eyes, its fabrication method and visual recognition system

Abstract

This invention belongs to the field of memristors and provides a bee-eye-inspired ultraviolet photoelectric memristor, its fabrication method, and a visual recognition system. The bee-eye-inspired ultraviolet photoelectric memristor includes, from bottom to top, a bottom electrode Pt, an HfO2 charge trapping layer, and an NbO layer. x UV photosensitive layer and top electrode ITO; NbO x The ultraviolet photosensitive layer absorbs ultraviolet light of a set wavelength and, under the synergistic effect of a bias voltage source, generates photogenerated carriers; the HfO2 charge-trapping layer provides defect states and functional sites for participating in conductance modulation; the HfO2 charge-trapping layer and NbO... x The ultraviolet photosensitive layer forms a heterojunction interface, which, through band matching and interface barrier distribution, influences the separation, transport, and accumulation behavior of photogenerated carriers. This effectively avoids the high energy consumption and large latency caused by frequent handling of visual data, providing crucial hardware support for ultra-low power edge intelligent vision.

Description

Technical Field

[0001] This invention belongs to the field of memristors, and particularly relates to an ultraviolet photoelectric memristor inspired by bee eyes, its preparation method, and a visual recognition system. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Traditional electronic vision systems follow the von Neumann architecture, which separates sensing, storage, and computation. This separation of "perception-transmission-processing" leads to the frequent transfer of massive amounts of unstructured visual data between sensors and memory, resulting in severe bandwidth limitations, high energy consumption, and large latency. In addition, traditional RGB cameras are susceptible to interference from visible light reflection in strong light environments, making it difficult to support the real-time, low-power edge vision detection requirements in resource-constrained environments.

[0004] To overcome these limitations, the "optoelectronic sensing-memristor-computing integrated" architecture, which mimics the efficient information processing mechanisms of biological vision, has become an important development direction. Among these, opto-memristors, as emerging neuromorphic vision hardware, can combine photoelectric sensitivity with synaptic plasticity, simultaneously achieving optical signal sensing, information storage, and in-situ computation at a single device level. This fundamentally avoids ineffective data transfer, significantly reducing system power consumption and response latency, and possesses significant innovation and technological advantages. However, existing biomimetic optoelectronic systems mostly rely on complex multi-terminal structures or spatially separated functional modules. Deep physical integration of specific spectral sensing, memory, and computation functions is difficult to achieve in simple two-terminal device structures, failing to fully reproduce the closed loop of biological vision computation from spectral feature sensing and dynamic memory to endogenous preprocessing, thus restricting their practical application in low-power edge vision systems. Summary of the Invention

[0005] To address the technical problems mentioned above, this invention provides an ultraviolet photomemristor inspired by bee eyes, its fabrication method, and a visual recognition system. This system effectively avoids the high energy consumption and large latency caused by frequent handling of visual data, providing key hardware support for ultra-low power edge intelligent vision.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides an ultraviolet photomemristor inspired by bee eyes.

[0007] A bee-eye-inspired ultraviolet photomemristor comprises, from bottom to top, a bottom electrode Pt, an HfO2 charge trapping layer, and an NbO layer. x The ultraviolet photosensitive layer and the top electrode ITO; The bottom electrode Pt and the top electrode ITO are connected to a bias voltage source; the NbOx The ultraviolet photosensitive layer is used to absorb ultraviolet light of a set wavelength and generate photogenerated carriers under the synergistic effect of a bias voltage source; the HfO2 charge trapping layer is used to provide defect states and functional sites for participating in conductivity modulation; the HfO2 charge trapping layer and NbO x The ultraviolet photosensitive layer constitutes the heterojunction interface, which affects the separation, transport and accumulation behavior of photogenerated carriers through band matching and interface barrier distribution.

[0008] In one implementation, under single-pulse or intermittent double-pulse ultraviolet light stimulation, the bee-eye inspired ultraviolet photomemristor exhibits short-term plasticity behaviors such as excitatory postsynaptic currents and double-pulse facilitation.

[0009] In one implementation, under continuous multi-pulse, high-frequency or wide-pulse ultraviolet light stimulation, the bee-eye inspired ultraviolet photomemristor exhibits a dynamic evolution and transformation from short-term plasticity to long-term plasticity through a charge accumulation and trapping mechanism.

[0010] A second aspect of the present invention provides a method for fabricating an ultraviolet photomemristor inspired by bee eyes.

[0011] A method for fabricating an ultraviolet photomemristor inspired by bee eyes, comprising: Preparation of bottom electrode Pt; An HfO2 charge trapping layer was prepared on the bottom electrode Pt using atomic layer deposition. NbO was prepared on an HfO2 charge-trapping layer using reactive magnetron sputtering. x Ultraviolet photosensitive layer; In NbO x On the ultraviolet photosensitive layer, a top electrode pattern orthogonal to the bottom electrode Pt is defined by photolithography, and the top electrode ITO is fabricated to form an ultraviolet photomemristor inspired by bee eyes.

[0012] In one implementation, a substrate with a thermally oxidized layer is ultrasonically cleaned sequentially in acetone, anhydrous ethanol, and deionized water, followed by oxygen plasma activation treatment to improve surface adhesion. A negative photoresist is spin-coated using photolithography and developed to define the bottom electrode pattern. Subsequently, a Ti thin film is deposited as an adhesion layer using DC magnetron sputtering technology, followed by in-situ deposition of a Pt thin film. After deposition, ultrasonic stripping is performed in acetone, followed by rinsing with deionized water and drying with nitrogen to obtain the bottom electrode Pt.

[0013] In one implementation, the bottom electrode Pt is transferred to an atomic layer deposition system, using tetramethylaminohafnium as the hafnium source precursor and deionized water as the oxygen source precursor. Under constant chamber temperature, multiple alternating deposition cycles are performed with a standard cycle of 0.1 seconds of tetramethylaminohafnium pulse, 10 seconds of nitrogen purging, 0.1 seconds of deionized water pulse, and 10 seconds of nitrogen purging. Finally, a HfO2 thin film of a predetermined thickness is uniformly grown on the bottom electrode to obtain an HfO2 charge trapping layer.

[0014] In one implementation, the HfO2 charge-trapping layer sample is transferred to a magnetron sputtering cavity using a high-purity metallic Nb target. Reactive sputtering is performed in a mixed atmosphere of argon and oxygen, with the gas flow rate ratio controlled to ensure the formation of NbO with the target defect states and exhibiting substoichiometry. x Thin film; NbO was deposited under set room temperature, working pressure and RF power conditions. x Photosensitive layer.

[0015] As one implementation method, the mixed gas flow ratio is controlled as Ar:O2 = 30:3, the working gas pressure is 0.5Pa, and the radio frequency power is 120W.

[0016] As one implementation method, the bee-eye inspired ultraviolet photomemristor is finally annealed to improve the transmittance and conductivity of the top electrode ITO and promote lattice relaxation and stable distribution of oxygen vacancies at the heterojunction interface.

[0017] A third aspect of the present invention provides a visual recognition system.

[0018] A visual recognition system, comprising: The sensing and computing module includes an ultraviolet photomemristor inspired by bee eyes, as described above, for in-situ sensing of ultraviolet visual signals; the photomemristor exhibits dynamic synaptic plasticity under ultraviolet light pulse stimulation, and uses the charge trapping dynamics of the heterojunction interface to sense the target ultraviolet image and perform feature extraction preprocessing to obtain the original two-dimensional visual image. The signal conditioning and encoding module, which is connected to the sensing and computing unit, is used to preprocess the original two-dimensional visual image according to the pixel intensity and discretely encode it into an ultraviolet light pulse sequence with a corresponding spatiotemporal distribution. The neuromorphic computing module integrates a physical reservoir computing architecture and a lightweight convolutional neural network. It receives ultraviolet light pulse sequences and performs ultraviolet image compression and ultraviolet image classification and recognition based on the high-dimensional mapping calculated by the physical reservoir computing architecture and the semantic aggregation of the lightweight convolutional neural network.

[0019] The beneficial effects of this invention are: (1) This invention highly integrates ultraviolet light sensing, information storage and feature processing into a single HfO2 / NbO x In heterojunction photomemristors, NbO is used x The ultraviolet photosensitive layer absorbs ultraviolet light of a set wavelength and generates photogenerated carriers under the synergistic effect of a bias voltage source. The HfO2 charge trapping layer provides defect states and functional sites for participating in conductivity modulation. Through band matching and interface barrier distribution, the separation, transport and accumulation behavior of photogenerated carriers are jointly affected. This breaks through the architectural bottleneck of the physical separation of "sensing-transmission-processing" in traditional vision systems. This integrated design at the underlying physical level effectively avoids the high energy consumption and large latency caused by frequent handling of visual data, and provides key hardware support for ultra-low power edge intelligent vision.

[0020] (2) This invention cleverly draws on the biological visual processing mechanism of the compound eye of bees, utilizing NbO x The intrinsic optical properties of the thin film enable highly specific sensing of the ultraviolet band, effectively shielding interference such as specular reflection of visible light in natural environments. Simultaneously, by combining the dynamic migration mechanism of charge-trapping defect states distributed within the dense HfO2 layer deposited by ALD with interfacial oxygen vacancies, the dynamic evolution from short-term to long-term plasticity of biological synapses is perfectly replicated, providing an excellent physical hardware foundation for processing dynamic spatiotemporal visual signals.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1 This is a flowchart of visual information perception and neural network simulation recognition based on a single photomemristor, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the bee-eye inspired ultraviolet photomemristor according to an embodiment of the present invention; Figure 3 NbO in this embodiment of the invention x A schematic diagram of the crystal structure of HfO2; Figure 4 XPS plot of Nb element in an embodiment of the present invention; Figure 5 XPS plot of Hf element in an embodiment of the present invention; Figure 6 The image shows the single-ultraviolet pulse collision response recovery time curve of the device in this embodiment of the invention. Figure 7 The curves showing the synaptic double-pulse facilitation behavior of the device in this embodiment of the invention under ultraviolet light; Figure 8 The present invention provides test curves of the device under different numbers of pulses under ultraviolet light in embodiments thereof; Figure 9 The present invention provides test curves of the device under different pulse frequencies under ultraviolet light in embodiments thereof; Figure 10 The present invention provides test curves of the device under ultraviolet light with different pulse widths according to embodiments of the invention. Figure 11 The device in this embodiment of the invention is subjected to binary test curves under ultraviolet light; Figure 12 This is a schematic diagram of the visual recognition workflow based on a pool computing-lightweight convolutional neural network collaborative architecture according to an embodiment of the present invention. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0025] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0026] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0027] With the deep integration of artificial intelligence, the Internet of Things, and edge computing technologies, machine vision is rapidly developing towards lower power consumption, miniaturization, and high real-time performance. Against this backdrop, biomimetic visual perception and pattern recognition targeting ultraviolet light have attracted significant attention due to their enormous potential in fields such as precision agriculture and ecological monitoring.

[0028] To achieve efficient and accurate visual pattern recognition in resource-constrained environments, there is an urgent need to develop a new generation of optoelectronic sensing-in-memory resistor integrated vision systems. This system should be based on a heterojunction HfO2 / NbOx opto-memristor with ultraviolet selective response, emphasizing its innovative architecture that integrates ultraviolet light sensing, dynamic synaptic weight storage, and in-situ feature extraction at a single device level, fundamentally breaking through the traditional paradigm of separating visual sensing and computation. Simultaneously, this underlying physical device and the upper-level neuromorphic algorithm work efficiently together to achieve efficient compression and accurate classification of target ultraviolet visual images. By integrating the inherent sensing-in-memory resistor's advantages with a lightweight algorithm architecture, this invention aims to construct a visual recognition system that combines high energy efficiency, high recognition accuracy, and biomimetic intelligence, providing innovative technical solutions for precision agriculture, UAV automated inspection, and edge intelligent vision chips, and promoting the integrated development of low-power brain-like intelligence and machine vision.

[0029] This invention aims to solve the problems of high power consumption and large latency in existing machine vision systems due to their adherence to the von Neumann architecture, as well as the susceptibility of conventional RGB cameras to interference under complex lighting conditions and their inability to support edge vision detection in resource-constrained environments. It provides a bee-eye inspired visual recognition system that integrates ultraviolet light sensing, dynamic storage, and neuromorphic computing.

[0030] Figure 2 A schematic diagram of a specific structure of an ultraviolet photomemristor inspired by bee eyes, according to an embodiment of the present invention, is provided. Figure 2 In the process, the ultraviolet photomemristor inspired by bee eyes consists of: from bottom to top, a bottom electrode Pt, an HfO2 charge trapping layer, and an NbO layer. x The ultraviolet photosensitive layer and the top electrode ITO; The bottom electrode Pt and the top electrode ITO are connected to a bias voltage source; NbO x The ultraviolet photosensitive layer absorbs ultraviolet light of a set wavelength and, under the synergistic effect of a bias voltage source, generates photogenerated carriers; the HfO2 charge-trapping layer provides defect states and functional sites for participating in conductance modulation; the HfO2 charge-trapping layer and NbO... x The ultraviolet photosensitive layer constitutes the heterojunction interface, which affects the separation, transport and accumulation behavior of photogenerated carriers through band matching and interface barrier distribution.

[0031] Under the combined effect of ultraviolet light stimulation and applied bias voltage, photogenerated carriers undergo directional separation and cross-interface transport at the interface, causing dynamic changes in local charge distribution, interface barrier, and defect state occupancy. At the same time, the reversible migration and redistribution of defects such as oxygen vacancies near the interface further modulates the formation of local conductive paths and electrodynamic processes.

[0032] Inspired by bee eyes, ultraviolet photoelectromerists possess dynamic synaptic plasticity that depends on the history of light stimulation: under ultraviolet light stimulation with different parameters, the device can dynamically evolve from short-term plasticity behaviors such as excitatory postsynaptic currents and double-pulse facilitation induced by single pulses or intermittent double pulses to long-term plasticity induced by continuous multi-pulse, high-frequency or wide-pulse induced by the internal charge accumulation and trapping mechanism.

[0033] Under single-pulse or intermittent double-pulse ultraviolet light stimulation, the bee-eye inspired ultraviolet photomemristor exhibits short-term plasticity behaviors such as excitatory postsynaptic currents and double-pulse facilitation.

[0034] Under continuous multi-pulse, high-frequency, or wide-pulse ultraviolet light stimulation, the bee-eye inspired ultraviolet photomemristor exhibits a dynamic evolution and transformation from short-term plasticity to long-term plasticity through a charge accumulation and trapping mechanism.

[0035] like Figure 1 As shown, the method for fabricating a bee-eye-inspired ultraviolet photomemristor according to an embodiment of the present invention includes: Step 1: Prepare the bottom electrode Pt.

[0036] A SiO2 / Si substrate with a 300nm thermally oxidized layer was sequentially ultrasonically cleaned in acetone, anhydrous ethanol, and deionized water, followed by oxygen plasma activation to improve surface adhesion. A negative photoresist was spin-coated using photolithography and developed to define the bottom electrode pattern. Subsequently, a 10nm thick Ti film was deposited as an adhesion layer using DC magnetron sputtering, followed by in-situ deposition of a Pt film. After deposition, ultrasonic peeling was performed in acetone, followed by rinsing with deionized water and drying with nitrogen to obtain a strip-shaped Pt bottom electrode. In some specific implementations, when preparing Pt thin films by DC magnetron sputtering, the deposition time is precisely controlled to 400 seconds to obtain a bottom electrode Pt with a thickness of 50 nm.

[0037] Step 2: An HfO2 charge trapping layer is prepared on the bottom electrode Pt using atomic layer deposition.

[0038] The sample obtained in step 1 was transferred to an atomic layer deposition system, using tetramethylaminohafnium as the hafnium source precursor and deionized water as the oxygen source precursor. Under a constant chamber temperature, multiple alternating deposition cycles were performed, consisting of a 0.1-second tetramethylaminohafnium pulse, a 10-second nitrogen purge, a 0.1-second deionized water pulse, and a 10-second nitrogen purge, to uniformly grow a HfO2 film of a predetermined thickness on the bottom electrode. For example, the constant chamber temperature for atomic layer deposition was set to 250°C, and 90 standard cycles were precisely performed to obtain a 10 nm thick HfO2 charge trapping layer.

[0039] Step 3: Prepare NbO on the HfO2 charge-trapping layer using reactive magnetron sputtering. x Ultraviolet photosensitive layer.

[0040] The sample with the deposited HfO2 layer was transferred to a magnetron sputtering chamber using a high-purity metallic Nb target. Reactive sputtering was performed in a mixed atmosphere of argon and oxygen, with precise control of the gas flow ratio to ensure the formation of NbO with the target defect states (oxygen vacancies) in a substoichiometric manner. x Thin film. NbO was deposited under specified room temperature, working pressure, and RF power conditions. x Photosensitive layer.

[0041] The mixed gas flow ratio is precisely controlled at Ar:O2 = 30:3 (sccm), the working gas pressure is 0.5Pa, and the radio frequency power is 120W.

[0042] As an alternative implementation method, the deposition time is controlled at 1250 seconds to obtain NbO with a thickness of 20 nm. x layer.

[0043] Step 4: In NbO x On the ultraviolet photosensitive layer, a top electrode pattern orthogonal to the bottom electrode Pt is defined by photolithography, and the top electrode ITO is fabricated to form an ultraviolet photomemristor inspired by bee eyes.

[0044] The top electrode pattern, orthogonal to the bottom electrode, was defined again using photolithography. A transparent, conductive ITO film was deposited using radio frequency magnetron sputtering in a pure argon atmosphere. Following an acetone ultrasonic lift-off process, a Pt / HfO2 / NbO film was formed. x An ITO heterojunction photoresist was then placed in a rapid annealing furnace for annealing to improve the transmittance and conductivity of the ITO electrode and promote lattice relaxation and stable distribution of oxygen vacancies at the heterojunction interface.

[0045] As an alternative implementation, the radio frequency magnetron sputtering conditions for the ITO thin film are: pure argon gas flow rate of 30 sccm, working gas pressure of 0.5 Pa, radio frequency power of 80 W, and deposition for 500 seconds to obtain an ITO thin film with a thickness of 50 nm.

[0046] As an alternative implementation method, the annealing process is carried out in an argon atmosphere at a temperature of 350°C for 30 minutes.

[0047] After the above steps, a Pt / HfO2 / NbO2-based product is obtained. x The ITO heterojunction photoresistor consists of, from bottom to top, a bottom electrode Pt, a charge trapping layer HfO2, and an ultraviolet photosensitive layer NbO. x And the top electrode ITO, its device structure diagram is as follows Figure 2 As shown.

[0048] In this embodiment, in order to clarify the atomic-level interface characteristics of the core heterojunction material, Figure 3 NbO was demonstrated x A schematic diagram of the crystal structure of Pt / HfO2. This is to demonstrate the properties of Pt / HfO2 / NbO. x The chemical composition and interfacial properties of the / ITO heterojunction composite material were characterized by XPS testing. The XPS plot of Nb elemental composition is shown below. Figure 4 As shown, the XPS plot of the Hf element is as follows: Figure 5 As shown. Specifically, the Nb 3d energy spectrum reveals NbO x The multivalent state distribution of Nb in HfO2 is consistent with the coexistence of different niobium oxide phases; while the Hf4f energy spectrum confirms the characteristic valence state of Hf in HfO2.

[0049] Based on Pt / HfO2 / NbO x The single-pulse collision response test experiment of the ITO opto-memristor was conducted under the specific test conditions of applying a single ultraviolet light pulse with a wavelength of 350 nm. For example... Figure 6 The device shown exhibits a single ultraviolet pulsed postsynaptic response recovery time curve. Pulse widths ranging from 100 to 900 ms can induce varying degrees of excitatory postsynaptic currents, demonstrating typical short-term plasticity characteristics.

[0050] The experiment on the detection of dual-pulse facilitated behavior based on photomemristors involved applying two consecutive ultraviolet light pulses. Figure 7 As shown, the excitatory postsynaptic current induced by the second pulse is significantly higher than that induced by the first pulse. The synaptic double-pulse facilitation behavior curve of the device under ultraviolet light exhibits typical biological synaptic characteristics that decay with increasing pulse interval time, confirming its memory enhancement potential in processing spatiotemporal optical signals.

[0051] To verify the evolutionary mechanism of short-term memory to long-term memory under multiple pulse stimulation, tests were conducted with different numbers of pulses. For example... Figure 8 As shown, with the increase of the number of ultraviolet light pulses (from 1 to 50), the postsynaptic current showed a cumulative growth trend and the decay time was significantly prolonged, indicating that repetitive stimulation promoted the transition from short-term plasticity to long-term plasticity, which is highly consistent with the memory consolidation mechanism in biological nervous systems.

[0052] The experimental results based on different frequencies and pulse width modulations of the opto-memristor are as follows: Figure 9 and Figure 10 As shown. In Figure 9In the test curves of the device under ultraviolet light with different pulse frequencies, high-frequency (5 Hz) stimulation can induce a slowly decaying current and directly enter the long-term memory state, while low-frequency (0.1 Hz) only maintains the short-term memory characteristics. Figure 10 The test curves of the device under ultraviolet light with different pulse widths further show that a wider light pulse can significantly improve the retention time of synaptic weight and the peak value of postsynaptic current.

[0053] Based on the binary test curve of the device under ultraviolet light, as shown in... Figure 11 As shown in the figure. In the test, up to 16 discrete current states (such as 0000 to 1111) were generated by mapping optical pulse sequences, forming clear and separate clusters of class features. This shows that the opto-memristor array has a strong ability to distinguish and encode the input optical pulse sequence patterns.

[0054] In one or more embodiments, a visual recognition system is also provided, comprising: The sensing and computing module includes an ultraviolet photomemristor inspired by bee eyes, as described above, for in-situ sensing of ultraviolet visual signals; the photomemristor exhibits dynamic synaptic plasticity under ultraviolet light pulse stimulation, and uses the charge trapping dynamics of the heterojunction interface to sense the target ultraviolet image and perform feature extraction preprocessing to obtain the original two-dimensional visual image. The signal conditioning and encoding module, which is connected to the sensing and computing unit, is used to preprocess the original two-dimensional visual image according to the pixel intensity and discretely encode it into an ultraviolet light pulse sequence with a corresponding spatiotemporal distribution. The neuromorphic computing module integrates a physical reservoir computing architecture and a lightweight convolutional neural network. It receives ultraviolet light pulse sequences and performs ultraviolet image compression and ultraviolet image classification and recognition based on the high-dimensional mapping calculated by the physical reservoir computing architecture and the semantic aggregation of the lightweight convolutional neural network.

[0055] To enhance the system's intelligence and accuracy in recognizing complex images, the dynamic response of the aforementioned device array was combined with reservoir computing and a lightweight convolutional neural network. A schematic diagram of its visual recognition workflow based on the RC-CNN collaborative architecture is shown below. Figure 12 As shown, the original image is discretized based on pixel intensity and encoded into corresponding ultraviolet light pulse sequences. These sequences are then subjected to high-dimensional feature mapping via an RC layer, followed by semantic extraction and classification matching via a CNN. This collaborative system successfully simulated the precise recognition capability of a bee's compound eye for ultraviolet "nectar source guidance" patterns, effectively extracting key discriminative information such as petal morphology, texture details, and color distribution. In simulated application tests, the system achieved an accuracy rate exceeding 93% in recognizing ultraviolet visual patterns of flowers.

[0056] This invention deeply integrates the nonlinear conductivity dynamics of the underlying hardware devices with the upper-level intelligent software algorithms at the system level, constructing a heterogeneous architecture that coordinates physical reservoir computing and lightweight convolutional neural networks. By directly mapping visual features using optical pulse sequences for encoding, the continuity and detail of the analog signal are preserved to the maximum extent. This successfully achieves efficient feature extraction and accurate classification of complex ultraviolet visual patterns of flowers (with a system recognition accuracy of over 93%). This hardware-software collaborative framework points to a highly promising technological direction for future applications in resource-constrained environments such as automated UAV inspections and precision agriculture.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A bee-eye-inspired ultraviolet photomemristor, characterized in that, include: From bottom to top: bottom electrode Pt, HfO2 charge trapping layer, NbO x The ultraviolet photosensitive layer and the top electrode ITO; The bottom electrode Pt and the top electrode ITO are connected to a bias voltage source; the NbO x The ultraviolet photosensitive layer is used to absorb ultraviolet light of a set wavelength and generate photogenerated carriers under the synergistic effect of a bias voltage source; the HfO2 charge trapping layer is used to provide defect states and functional sites for participating in conductivity modulation; the HfO2 charge trapping layer and NbO x The ultraviolet photosensitive layer constitutes the heterojunction interface, which affects the separation, transport and accumulation behavior of photogenerated carriers through band matching and interface barrier distribution.

2. The bee-eye-inspired ultraviolet photomemristor as described in claim 1, characterized in that, Under single-pulse or intermittent double-pulse ultraviolet light stimulation, the bee-eye inspired ultraviolet photomemristor exhibits short-term plasticity behaviors such as excitatory postsynaptic currents and double-pulse facilitation.

3. The bee-eye-inspired ultraviolet photomemristor as described in claim 1, characterized in that, Under continuous multi-pulse, high-frequency, or wide-pulse ultraviolet light stimulation, the bee-eye inspired ultraviolet photomemristor exhibits a dynamic evolution and transformation from short-term plasticity to long-term plasticity through a charge accumulation and trapping mechanism.

4. A method for fabricating a bee-eye-inspired ultraviolet photomemristor as described in any one of claims 1-3, characterized in that, include: Preparation of bottom electrode Pt; An HfO2 charge trapping layer was prepared on the bottom electrode Pt using atomic layer deposition. NbO was prepared on an HfO2 charge-trapping layer using reactive magnetron sputtering. x Ultraviolet photosensitive layer; In NbO x On the ultraviolet photosensitive layer, a top electrode pattern orthogonal to the bottom electrode Pt is defined by photolithography, and the top electrode ITO is fabricated to form an ultraviolet photomemristor inspired by bee eyes.

5. The method for fabricating an ultraviolet photomemristor inspired by bee eyes as described in claim 4, characterized in that, The substrate with the thermal oxide layer was ultrasonically cleaned in acetone, anhydrous ethanol and deionized water in sequence, and then activated by oxygen plasma to improve surface adhesion. The bottom electrode pattern was defined by spin-coating negative photoresist and developing. Subsequently, using DC magnetron sputtering technology, a Ti film was first deposited as an adhesion layer, and Pt film was then deposited in situ. After deposition, the film was ultrasonically peeled off in acetone, rinsed with deionized water, and dried with nitrogen to obtain the bottom electrode Pt.

6. The method for fabricating an ultraviolet photomemristor inspired by bee eyes as described in claim 4, characterized in that, The bottom electrode Pt was transferred to an atomic layer deposition system, using tetramethylaminohafnium as the hafnium source precursor and deionized water as the oxygen source precursor. Under constant chamber temperature, multiple alternating deposition cycles were performed with a standard cycle of 0.1 seconds of tetramethylaminohafnium pulse, 10 seconds of nitrogen purging, 0.1 seconds of deionized water pulse, and 10 seconds of nitrogen purging. Finally, a HfO2 thin film of a predetermined thickness was uniformly grown on the bottom electrode to obtain an HfO2 charge trapping layer.

7. The method for fabricating an ultraviolet photomemristor inspired by bee eyes as described in claim 4, characterized in that, The HfO2 charge-trapping layer sample was transferred to a magnetron sputtering chamber using a high-purity metallic Nb target. Reactive sputtering was performed in a mixed atmosphere of argon and oxygen, with the gas flow ratio controlled to ensure the formation of NbO with the target defect states and exhibiting substoichiometry. x Thin film; NbO was deposited under set room temperature, working pressure and RF power conditions. x Photosensitive layer.

8. The method for fabricating an ultraviolet photomemristor inspired by bee eyes as described in claim 7, characterized in that, The mixed gas flow ratio is controlled at Ar:O2 = 30:3, the working gas pressure is 0.5Pa, and the radio frequency power is 120W.

9. The method for fabricating an ultraviolet photomemristor inspired by bee eyes as described in claim 4, characterized in that, Inspired by the bee eye, the ultraviolet photomemristor is finally annealed to improve the transmittance and conductivity of the top electrode ITO and promote lattice relaxation and stable distribution of oxygen vacancies at the heterojunction interface.

10. A visual recognition system, characterized in that, include: The sensing and computing module includes a bee-eye inspired ultraviolet photomemristor as described in any one of claims 1-3, used for in-situ sensing of ultraviolet visual signals; the photomemristor exhibits dynamic synaptic plasticity under ultraviolet light pulse stimulation, and uses the charge trapping dynamics of the heterojunction interface to sense the target ultraviolet image and perform feature extraction preprocessing to obtain the original two-dimensional visual image. The signal conditioning and encoding module, which is connected to the sensing and computing unit, is used to preprocess the original two-dimensional visual image according to the pixel intensity and discretely encode it into an ultraviolet light pulse sequence with a corresponding spatiotemporal distribution. The neuromorphic computing module integrates a physical reservoir computing architecture and a lightweight convolutional neural network. It receives ultraviolet light pulse sequences and performs ultraviolet image compression and ultraviolet image classification and recognition based on the high-dimensional mapping calculated by the physical reservoir computing architecture and the semantic aggregation of the lightweight convolutional neural network.

Images