A 3D pose real-time recognition system and method based on brain-inspired metamaterials

Abstract

The present invention discloses a three-dimensional attitude real-time recognition system and method based on a brain-inspired metamaterial. The system includes: a high-directivity lens antenna, a brain-inspired metamaterial, a plurality of detection components, and a server; wherein, the brain-inspired metamaterial is located between the high-directivity lens antenna and the plurality of detection components, and the plurality of detection components are electrically connected to the server; wherein, the manufacturing data of the brain-inspired metamaterial is generated based on phase arrangement information, and the phase arrangement information is generated after training a diffraction neural network. Since the present application obtains phase arrangement information after training a diffraction neural network, manufactures a brain-inspired metamaterial according to the phase arrangement information, and performs three-dimensional attitude real-time recognition based on the brain-inspired metamaterial, the efficiency of the object recognition process is improved, a light-speed-level processing speed is achieved, and the dynamic data of the object can be processed in real time.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and particularly to a three-dimensional pose real-time recognition system and method based on brain-like metamaterials. Background Art

[0002] Object recognition is a computer technology related to image processing and computer vision, and is often used for the detection of digital images, classification, and labeling of semantic objects. Object recognition is closely related to our lives, such as video surveillance, target tracking, and annotation and segmentation of images. To achieve object recognition, it is usually divided into two steps: one is to obtain an image data set through a camera or the like; the other is to perform data processing by a computer combining deep learning and other pattern recognition algorithms. However, with the rapid development of big data and the Internet of Things, the above steps are difficult to meet the requirements of light-speed parallel data acquisition and processing.

[0003] At the beginning of this century, the emergence of metamaterials and photonics has prompted researchers to consider object recognition technology from a new optical perspective. Metamaterials are artificially designed and have characteristic parameters that materials existing in nature do not have. Since Professor Walser of the University of Texas at Austin in the United States named it "metamaterial", it has been widely adopted by international scholars. It is usually translated as "metamaterial" by domestic scholars, and is also translated as "left-handed material", "negative refractive index material", "anisotropic medium", "artificial electromagnetic material", "meta-structure material", etc. Among them, the method of designing brain-like metamaterials through neural network algorithms can design any more accurate metamaterial characteristic parameters required compared with the traditional artificial design method. As an important part of metamaterials, it has attracted extensive attention from domestic and foreign scholars in recent years; compared with the implementation method based on electrical computing, optical computing has great potential in achieving high throughput, specific scenarios, and real-time tasks, and at the same time has the remarkable advantages of high-speed operation, low energy consumption, and strong parallel capabilities. So far, what has been achieved by optical computing includes mathematical operators, logical operators, etc. In terms of optical imaging, there have also been related progresses, such as computational imaging, edge detection, and non-line-of-sight imaging. These pioneering methods have effectively completed the step of image acquisition in object recognition. However, if the image processing step of the object recognition task continues to be performed, digital calculations such as dimensionality reduction and feature extraction are still indispensable.

[0004] On the other hand, object recognition (e.g., handwritten digits) is often used as an example to demonstrate various optical neural network architectures such as nanophotonic deep learning circuits and hybrid optoelectronic convolutional neural networks. These architectures essentially utilize a series of existing datasets to improve or replace digital computing, completing the second step: image processing. For field applications, light-to-electron / electron-to-light conversion is used for real-time communication between electronic image sensors and optical processing components. The operating speed of the entire system is ultimately limited by the electronic input / output speed. Given the above factors, it is of great significance to achieve three-dimensional (3D) optical object recognition in the real world. Summary of the Invention

[0005] Embodiments of the present application provide a three-dimensional pose real-time recognition system and method based on brain-inspired metamaterials. To provide a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary part is not a general review, nor is it intended to identify key / important constituent elements or delineate the protection scope of these embodiments. Its sole purpose is to present some concepts in a simple form as a prelude to the detailed description that follows.

[0006] In a first aspect, embodiments of the present application provide a three-dimensional pose real-time recognition system based on brain-inspired metamaterials, the system comprising:

[0007] a high-directivity lens antenna, brain-inspired metamaterials, a plurality of detection components, and a server; wherein,

[0008] the brain-inspired metamaterials are located between the high-directivity lens antenna and the plurality of detection components, and the plurality of detection components are electrically connected to the server; wherein,

[0009] wherein, the manufacturing data of the brain-inspired metamaterials is generated based on phase arrangement information, and the phase arrangement information is generated after training a diffraction neural network.

[0010] Optionally, the high-directivity lens antenna is configured to emit electromagnetic waves to the object to be recognized;

[0011] the brain-inspired metamaterials are configured to process the electromagnetic waves into the electric field intensities of different poses of the object to be recognized;

[0012] the plurality of detection components are configured to detect the electric field intensities of different poses of the object to be recognized and send the electric field intensities to the server;

[0013] the server is configured to perform three-dimensional pose recognition based on the received pose data.

[0014] Optionally, the manufacturing data of the brain-inspired metamaterials is generated according to the following steps, including:

[0015] Input the size, pose, rotation angle of the object to be recognized, and the distance from the metamaterial into the simulation software for simulation, and use the MATLAB-CST joint simulation method to generate network training samples;

[0016] Input the network training samples into the diffraction neural network for training, and output the loss value of the network;

[0017] When the loss value reaches the minimum, generate a trained diffraction neural network;

[0018] Determine the network parameters in the trained diffraction neural network as the phase arrangement information of the brain-like metamaterial;

[0019] Generate manufacturing data of the brain-like metamaterial based on the phase arrangement information.

[0020] Optionally, when the loss value reaches the minimum, output multiple phase arrangement information, including:

[0021] When the loss value does not reach the minimum, adjust the parameters of the diffraction neural network and continue to execute the step of inputting the sample data set into the diffraction neural network for training until the loss value reaches the minimum and stop training.

[0022] Optionally, the server includes a signal processing module and an attitude recognition module; among them,

[0023] The signal processing module and the attitude recognition module are electrically connected.

[0024] Optionally, the signal processing module is used to collect and preprocess the electric field intensities of the object to be recognized in different postures on multiple detection components;

[0025] The attitude recognition module is used to generate the corresponding three-dimensional attitude of the object to be recognized after analyzing the preprocessed electric field intensities of the object to be recognized in different postures.

[0026] Optionally, the signal processing module at least includes a radio frequency switch, a broadband amplifier, a mixer, and an FPGA; among them, the radio frequency switch, the broadband amplifier, the mixer, and the FPGA are electrically connected in sequence.

[0027] Optionally, the radio frequency switch collects the electric field intensities of the object to be recognized in different postures from multiple detection components at a microsecond speed, and the electric field intensities in different postures are processed by the radio frequency switch, the broadband amplifier, the mixer, and the FPGA in sequence to generate the preprocessed electric field intensities of the object to be recognized in different postures.

[0028] Optionally, the brain-like metamaterial is composed of a dense array of sub-wavelength neural units; among them,

[0029] The brain-like metamaterial includes two layers of metamaterials, and the size of the unit structure on each layer of metamaterial is consistent with the phase arrangement information.

[0030] In a second aspect, an embodiment of the present application provides a method for real-time three-dimensional pose recognition based on brain-inspired metamaterials. The method includes:

[0031] A high-directivity lens antenna emits electromagnetic waves to the object to be recognized;

[0032] The brain-inspired metamaterials process the electromagnetic waves into the electric field strengths of different poses of the object to be recognized;

[0033] Multiple detection components detect the electric field strengths of different poses of the object to be recognized and send the electric field strengths to the server;

[0034] The server performs three-dimensional pose recognition based on the received pose data.

[0035] The technical solution provided by the embodiment of the present application may include the following beneficial effects:

[0036] In the embodiment of the present application, a high-directivity lens antenna emits electromagnetic waves to the object to be recognized, the brain-inspired metamaterials process the electromagnetic waves into the electric field strengths of different poses of the object to be recognized, multiple detection components detect the electric field strengths of different poses of the object to be recognized and send the electric field strengths to the server, and the server performs three-dimensional pose recognition based on the received pose data. Since the present application obtains the phase arrangement information after training the diffraction neural network, manufactures the brain-inspired metamaterials according to the phase arrangement information, and performs real-time three-dimensional pose recognition based on the brain-inspired metamaterials, the efficiency of the object recognition process is improved, the processing speed at the speed of light is achieved, and the dynamic data of the object can be processed in real time.

[0037] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present invention and used together with the specification to explain the principles of the present invention.

[0039] Figure 1 is a schematic structural diagram of a real-time three-dimensional pose recognition system based on brain-inspired metamaterials provided by an embodiment of the present application;

[0040] Figure 2 is a schematic diagram of the structure of a brain-inspired metamaterial unit provided by an embodiment of the present application;

[0041] Figure 3 is a schematic diagram of the overall structure of a brain-inspired metamaterial provided by an embodiment of the present application;

[0042] Figure 4 is a schematic diagram of the influence of the radius of the unit structure on the phase provided by an embodiment of the present application;

[0043] Figure 5 It is a schematic diagram of the effect of identifying rabbits in different postures provided by an embodiment of the present application;

[0044] Figure 6 It is a schematic flow diagram of a three-dimensional posture real-time recognition method based on brain-like metamaterials provided by an embodiment of the present application;

[0045] Figure 7 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application;

[0046] Figure 8 It is a schematic diagram of a storage medium provided by an embodiment of the present application. Detailed implementation manners

[0047] The following description and the accompanying drawings fully disclose specific implementation manners of the present invention, enabling those skilled in the art to practice them.

[0048] It should be clear that the described embodiments are only a part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0049] When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present invention. On the contrary, they are only examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0050] In the description of the present invention, it should be understood that terms such as "first", "second", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In addition, in the description of the present invention, unless otherwise specified, "a plurality of" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects before and after.

[0051] The present application provides a method, device, storage medium, and terminal for real-time three-dimensional pose recognition based on brain-inspired metamaterials to solve the problems existing in the above-mentioned related technical problems. In the technical solution provided by the present application, after training the diffraction neural network, phase arrangement information is obtained, and based on the phase arrangement information, brain-inspired metamaterials are manufactured, and real-time three-dimensional pose recognition is performed based on the brain-inspired metamaterials, which improves the efficiency of the object recognition process, achieves the processing speed of the light speed level, and can process the dynamic data of the object in real time. The following uses exemplary embodiments for detailed description.

[0052] Please refer to Figure 1 , which is a schematic flowchart of a system for real-time three-dimensional pose recognition based on brain-inspired metamaterials provided by an embodiment of the present application. As Figure 1 shown, the system of the embodiment of the present application includes: a high-directivity lens antenna 1, a brain-inspired metamaterial 2, a plurality of detection components, and a server; wherein, the brain-inspired metamaterial 2 is located between the high-directivity lens antenna 1 and the plurality of detection components, and the plurality of detection components are electrically connected to the server; wherein, the manufacturing data of the brain-inspired metamaterial 2 is generated based on the phase arrangement information, and the phase arrangement information is generated after training the diffraction neural network.

[0053] In this embodiment, each of the plurality of detection components includes a detector 3, a probe 4, and a connection line 5; wherein, the detector 3, the probe 4, and the connection line 5 are electrically connected.

[0054] In this embodiment, the high-directivity lens antenna is used to emit electromagnetic waves to the object to be recognized; the brain-inspired metamaterial is used to process the electromagnetic waves into the electric field intensities of different postures of the object to be recognized; the plurality of detection components are used to detect the electric field intensities of different postures of the object to be recognized and send the electric field intensities to the server; the server is used to perform three-dimensional pose recognition according to the received pose data.

[0055] For example, place a freely moving rabbit in front of the brain-inspired metamaterial. The spatial electromagnetic field scattered by the rabbit is secondarily scattered by the metamaterial and then processed into different state data representing the lying, running, and standing postures of the rabbit. Therefore, by using reverse design technology to design the brain-inspired metamaterial, the user's requirements are transformed into the spatial metamaterial structure, and the scattered field data is automatically analyzed and processed.

[0056] Furthermore, the brain-inspired metamaterial can further achieve abnormal dynamic optical illusions and can convert the dynamic movement sequence of the rabbit into a holographic video of a giraffe or other objects. Designing the brain-inspired metamaterial mainly includes three stages: data collection, optimization of the scattering matrix, and pose recognition.

[0057] In this embodiment, Figure 1The application scenario of the present invention is shown. When the electromagnetic wave emitted by the high-directivity lens antenna 1 propagates to the target rabbit to be measured, a strong scattered electromagnetic field will be generated. After the scattered electromagnetic field passes through the brain-like metamaterial structure 2 designed by the present invention, electromagnetic field responses will be generated at different positions of the three detectors 3 at the back end. A rabbit moves freely in front of the neural network metamaterial 2 under the incidence of transverse electromagnetic waves, i.e., TE waves. At the same time, the three detectors 3, i.e., small single-polarization antennas, are connected to a series of complex components to sense the electromagnetic wave intensity, i.e., the electric field amplitude, on the output plane. To quantify the test results, we define the variance of the electric field amplitude and the incident electric field amplitude, i.e., E / E 0 as the test index.

[0058] In this embodiment, manufacturing data needs to be pre-generated when manufacturing the brain-like metamaterial. When generating the manufacturing data, first, the size, pose, rotation angle of the object to be recognized, and the distance from the metamaterial are input into the simulation software for simulation, and the network training samples are generated by the MATLAB-CST joint simulation method. Then, the network training samples are input into the diffraction neural network for training, and the loss value of the network is output. When the loss value reaches the minimum, the trained diffraction neural network is generated. Secondly, the network parameters in the trained diffraction neural network are determined as the phase arrangement information of the brain-like metamaterial. Finally, the manufacturing data of the brain-like metamaterial is generated based on the phase arrangement information.

[0059] Furthermore, when the loss value does not reach the minimum, adjust the parameters of the diffraction neural network and continue to execute the step of inputting the sample data set into the diffraction neural network for training until the loss value reaches the minimum and then stop training.

[0060] Specifically, in the data acquisition stage, we considered four parameters of the rabbit to generate simulation data, i.e., the size, posture, and rotation angle of the rabbit, and its distance from the brain-like metamaterial. We imported these models into the commercial software package CST Microwave Studio and continuously generated data using the MATLAB-CST joint simulation method. The simulation data was shuffled, and 80% was randomly selected as the training set, and the remaining 20% was used to test the performance of the brain-like metamaterial.

[0061] For example, three rabbits of different body sizes (large, medium, and small), three different postures (lying posture, running posture, and standing posture) were considered. The normal direction of the corresponding metamaterial ranged from -30° to 30° with a step of 10°, and the distance between the rabbit and the brain-like metamaterial ranged from 10 mm to 100 mm with a step of 10 mm. Using the commercial simulation software CST Microwave Studio, we simulated and obtained hundreds of data sets containing all the above scenarios.

[0062] Specifically, during the network training phase, Python version 3.5.0 was used for training. The TensorFlow framework version 1.10.0 was on the server (GeForce 249 10GTX TITAN X 227GPU and Intel(R) Xeon(R) CPU X5570 @ 2.93GHz with 48GB RAM, running the Linux 250 operation 228 system). Our neural network took dozens of minutes to fit.

[0063] For example, the diffractive neural network algorithm was adopted to optimize the scattering matrix. Of course, topology optimization and genetic algorithms are also good choices. Each hidden layer of the neural network contains 30×40 cells with a size of 13×13mm 2 , the axial distance between layers was set to 300mm (an optimizable parameter), and the designed metamaterial operates at 8.6GHz. The simulated data was shuffled and input into the neural network to accelerate the convergence of the algorithm. 80% of the data in the dataset was used for the training of the neural network, 20% of the data was used for testing, and it was assumed that the transmission amplitude of the cells was uniform. As the loss value decreased, the accuracy of the training set reached 99%, and the accuracy of the test set reached 98%, which proves that the neural network is reliable and there is almost no overfitting.

[0064] Specifically, several samples were randomly selected from the test set to observe the results of numerical simulation. Three sub-regions with a radius less than one wavelength were selected, corresponding to three different postures respectively. The classification basis for different postures is the maximum value of the signals of the left, middle, and right single-polarization detectors. Finally, all the results were correctly focused on the regions corresponding to the training results.

[0065] Specifically, the target located in the electromagnetic field will excite strong scattered waves, which will finally be focused on three predefined different detection points after passing through the metamaterial structure. Mathematically, the above process can be expressed as U l+1 =F -1 (F(S l ·U l )·H l ), where H l is the transmission equation of the free space by the Rayleigh-Sommerfeld diffraction method, F represents the operator of the Fourier transform, and S l represents the scattering matrix; where l ranges from 1 to L, and L is the number of layers of the metamaterial; U 1 is the initial electromagnetic field generated by the target to be measured, and U l+1 is the electromagnetic field on the output plane. The optimization goal is to minimize the loss function Among them, N is the total number of samples, and G is the base value. To achieve the goal of reducing the loss function value, we can use methods based on gradient descent (such as topology optimization) or heuristic algorithms (such as simulated annealing algorithm) to optimize the scattering matrix. To simplify the design complexity, we divide the brain-inspired metamaterial structure into several sub-regions, each sub-region corresponding to a specific part of the target rabbit. According to the electric field intensity at different detection positions, we can quickly judge the posture of the target rabbit.

[0066] Furthermore, after the network is trained, the parameters in the network can be determined as the phase arrangement information, and the manufacturing data of the brain-inspired metamaterial is generated based on the phase arrangement information. Finally, the brain-inspired metamaterial is manufactured according to the manufacturing data.

[0067] Specifically, in the physical design stage of the brain-inspired metamaterial, different from the existing metamaterials designed by neural network training, such as non-uniform body metamaterials and transmissive metasurfaces (regarded as layered metamaterials), we consider the method of stacking layered metamaterials. The brain-inspired metamaterial consists of a dense array of sub-wavelength neural units, each of which can act as an independent neural unit and is connected to the neural units in the next layer at the same time. Using a properly designed metamaterial, once the training is completed, the neural network is universal in both the microwave band and the visible light band. In the microwave band, we designed a layered composite unit structure, and the transmission phase of each unit structure covers the range of 0-2π. It should be noted that due to the sharp change of the transmission phase, Figure 4 the amplitude in the shaded area will drop sharply, and they are very sensitive to the dielectric constant of the substrate and the geometric dimensions of the metal discs in the assembly. This problem can be alleviated to a certain extent by increasing the number of stacking layers, but at the same time, it will reduce the robustness of the neural network and cause stronger mutual coupling. Therefore, we only adopt the working area outside the shaded area in the dispersion relation diagram.

[0068] In this embodiment, the server includes a signal processing module and an attitude recognition module; among them, the signal processing module and the attitude recognition module are electrically connected.

[0069] In this embodiment, the signal processing module is used to collect and preprocess the electric field intensities of different postures of the object to be recognized on multiple detection components; the attitude recognition module is used to generate the three-dimensional attitude corresponding to the object to be recognized after analyzing the preprocessed electric field intensities of different postures of the object to be recognized.

[0070] In this embodiment, the signal processing module at least includes a radio frequency switch, a broadband amplifier, a mixer, and an FPGA; among them, the radio frequency switch, the broadband amplifier, the mixer, and the FPGA are electrically connected in sequence.

[0071] In this embodiment, the RF switch collects the electric field strengths of the object to be recognized in different postures from multiple detection components at the microsecond speed, and the electric field strengths in different postures are sequentially processed by the RF switch, broadband amplifier, mixer, and FPGA to generate the preprocessed electric field strengths of the object to be recognized in different postures.

[0072] In this embodiment, the brain-inspired metamaterial consists of a dense array of sub-wavelength neural units; wherein, the brain-inspired metamaterial includes two layers of metamaterials, and the sizes and phase arrangement information of the unit structures on each layer of metamaterial are the same. As Figure 3 shown, the size of the unit structure on each layer of metamaterial is determined by the obtained phase distribution. 9 is the substrate of the single-layer brain-inspired metamaterial structure; 10 is the metal sheet on the metamaterial unit structure; 11 is the hole connecting the brain-inspired metamaterial structure. The unit structure of the brain-inspired metamaterial is as Figure 2 shown, the unit structure includes the substrate 6 of the unit structure, the metal disc 7 with variable radius on the designed unit structure, and the metal frame 8 on the unit structure. The radius of the metal disc 7 on the unit structure will cause the change of the metamaterial phase, as Figure 4 shown.

[0073] Specifically, in the measurement stage, the high-directivity lens antenna centered on the brain-inspired metamaterial is used as the excitation source. Three monopole probes we manufactured ourselves are connected to an RF switch (HMC641ALC4) to collect amplitude signals from three ports at the microsecond speed. The received signals are amplified by a broadband amplifier and mixed at 6 GHz. Then, we use AD9361 as the RF processor, which includes a low-noise amplifier (LNA), mixer, and other components, and use Xilinx ZYNQ for data processing, accelerated by the field-programmable gate array (FPGA).

[0074] For example Figure 5 shown, 12 is a rabbit in a lying posture; 13 is the position of the electrical signal response on the detector after the electromagnetic wave passes through the rabbit in the lying posture and the metasurface; 14 is a rabbit in a running posture; 15 is the position of the electrical signal response on the detector after the electromagnetic wave passes through the rabbit in the running posture and the metasurface; 16 is a rabbit in a standing posture; 17 is the position of the electrical signal response on the detector after the electromagnetic wave passes through the rabbit in the standing posture and the metasurface.

[0075] In the embodiments of the present application, a high-directivity lens antenna emits electromagnetic waves to the object to be recognized. The brain-inspired metamaterial processes the electromagnetic waves into the electric field intensities of different postures of the object to be recognized. A plurality of detection components detect the electric field intensities of different postures of the object to be recognized and send the electric field intensities to the server. The server performs three-dimensional posture recognition based on the received posture data. Since the present application obtains the phase arrangement information after training the diffraction neural network, manufactures the brain-inspired metamaterial according to the phase arrangement information, and performs real-time three-dimensional posture recognition based on the brain-inspired metamaterial, the efficiency of the object recognition process is improved, the processing speed of the light speed level is achieved, and the dynamic data of the object can be processed in real time.

[0076] Please refer to Figure 6 , which is a schematic flowchart of a three-dimensional posture real-time recognition system based on a brain-inspired metamaterial provided by an embodiment of the present application. As Figure 6 shown, the system of the embodiment of the present application includes:

[0077] S101, a high-directivity lens antenna emits electromagnetic waves to the object to be recognized;

[0078] S102, the brain-inspired metamaterial processes the electromagnetic waves into the electric field intensities of different postures of the object to be recognized;

[0079] S103, a plurality of detection components detect the electric field intensities of different postures of the object to be recognized and send the electric field intensities to the server;

[0080] S104, the server performs three-dimensional posture recognition based on the received posture data.

[0081] In the embodiments of the present application, a high-directivity lens antenna emits electromagnetic waves to the object to be recognized. The brain-inspired metamaterial processes the electromagnetic waves into the electric field intensities of different postures of the object to be recognized. A plurality of detection components detect the electric field intensities of different postures of the object to be recognized and send the electric field intensities to the server. The server performs three-dimensional posture recognition based on the received posture data. Since the present application obtains the phase arrangement information after training the diffraction neural network, manufactures the brain-inspired metamaterial according to the phase arrangement information, and performs real-time three-dimensional posture recognition based on the brain-inspired metamaterial, the efficiency of the object recognition process is improved, the processing speed of the light speed level is achieved, and the dynamic data of the object can be processed in real time.

[0082] The embodiment of the present application also provides an electronic device to execute the three-dimensional posture real-time recognition method of the above-mentioned electronic device. Please refer to Figure 7 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. As Figure 7As shown, the electronic device 4 includes: a processor 400, a memory 401, a bus 402, and a communication interface 403. The processor 400, the communication interface 403, and the memory 401 are connected through the bus 402. A computer program that can run on the processor 400 is stored in the memory 401. When the processor 400 runs the computer program, it executes the three-dimensional attitude real-time recognition method of the electronic device provided in any of the foregoing embodiments of the present application.

[0083] Among them, the memory 401 may include a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 403 (which can be wired or wireless), a communication connection is realized between this device network element and at least one other network element, and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0084] The bus 402 may be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Among them, the memory 401 is used to store programs. After receiving an execution instruction, the processor 400 executes the program. The three-dimensional attitude real-time recognition method of the electronic device disclosed in any of the foregoing embodiments of the present application can be applied to the processor 400 or implemented by the processor 400.

[0085] The processor 400 may be an integrated circuit chip with the ability to process signals. In the implementation process, each step of the above method may be completed by the integrated logic circuit of the hardware in the processor 400 or the instructions in the form of software. The above-mentioned processor 400 may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application may be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 401, and the processor 400 reads the information in the memory 401 and combines its hardware to complete the steps of the above method.

[0086] The electronic device provided by the embodiment of the present application and the three-dimensional attitude real-time recognition method provided by the electronic device provided by the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the method adopted, run or implemented by it.

[0087] The embodiment of the present application also provides a computer-readable storage medium corresponding to the three-dimensional attitude real-time recognition method provided by the foregoing embodiment. Please refer to Figure 8 , which shows that the computer-readable storage medium is an optical disc 30, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will execute the three-dimensional attitude real-time recognition method provided by any of the foregoing embodiments.

[0088] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here one by one.

[0089] The computer-readable storage medium provided by the above embodiments of the present application and the three-dimensional pose real-time recognition method provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0090] The above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

[0091] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program for three-dimensional pose real-time recognition can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory or a random access memory, etc.

[0092] The above-disclosed are only the preferred embodiments of the present application. Of course, the scope of rights of the present application cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. A three-dimensional pose real-time recognition system based on a brain-inspired metamaterial, characterized in that, the system includes: a high-directivity lens antenna, a brain-inspired metamaterial, a plurality of detection components, and a server; wherein, the brain-inspired metamaterial is located between the high-directivity lens antenna and the plurality of detection components, and the plurality of detection components are electrically connected to the server; wherein, the manufacturing data of the brain-inspired metamaterial is generated based on the phase arrangement information, and the phase arrangement information is generated after training a diffraction neural network; wherein, the manufacturing data of the brain-inspired metamaterial is generated according to the following steps, including: inputting the size, pose, rotation angle of the object to be recognized, and the distance from the metamaterial into a simulation software for simulation, and generating network training samples by using the MATLAB-CST joint simulation method; inputting the network training samples into a diffraction neural network for training, and outputting the loss value of the network; when the loss value reaches the minimum, generating a trained diffraction neural network; determining the network parameters in the trained diffraction neural network as the phase arrangement information of the brain-inspired metamaterial; generating the manufacturing data of the brain-inspired metamaterial based on the phase arrangement information; wherein, the brain-inspired metamaterial is composed of a dense array of sub-wavelength neural units; wherein, the brain-inspired metamaterial includes two layers of metamaterials, and the size of the unit structure on each layer of the metamaterial is consistent with the phase arrangement information.

2. The three-dimensional pose real-time recognition system based on a brain-inspired metamaterial according to claim 1, characterized in that, the high-directivity lens antenna is used to emit electromagnetic waves to the object to be recognized; the brain-inspired metamaterial is used to process the electromagnetic waves into the electric field strengths of different poses of the object to be recognized; the plurality of detection components are used to detect the electric field strengths of different poses of the object to be recognized, and send the electric field strengths to the server; the server is used to perform three-dimensional pose recognition according to the received pose data.

3. The three-dimensional pose real-time recognition system based on a brain-inspired metamaterial according to claim 1, characterized in that, when the loss value reaches the minimum, outputting a plurality of phase arrangement information, including: when the loss value does not reach the minimum, adjusting the parameters of the diffraction neural network and continuing to execute the step of inputting the sample data set into the diffraction neural network for training until the loss value reaches the minimum and then stopping training.

4. The three-dimensional pose real-time recognition system based on a brain-inspired metamaterial according to claim 1, characterized in that, the server includes a signal processing module and a pose recognition module; wherein, the signal processing module and the pose recognition module are electrically connected.

5. The three-dimensional pose real-time recognition system based on a brain-inspired metamaterial according to claim 4, characterized in that, the signal processing module is used to collect and preprocess the electric field strengths of different poses of the object to be recognized on the plurality of detection components; the pose recognition module is used to generate the three-dimensional pose corresponding to the object to be recognized after analyzing the preprocessed electric field strengths of different poses of the object to be recognized.

6. A three-dimensional attitude real-time recognition system based on brain-inspired metamaterials according to claim 5, characterized in that, the signal processing module at least includes a radio frequency switch, a broadband amplifier, a mixer, and an FPGA; wherein, the radio frequency switch, the broadband amplifier, the mixer, and the FPGA are electrically connected in sequence.

7. A three-dimensional attitude real-time recognition system based on brain-inspired metamaterials according to claim 6, characterized in that, the radio frequency switch collects the electric field strengths of different attitudes of the object to be recognized from the multiple detection components at a microsecond speed, and the electric field strengths of different attitudes are processed by the radio frequency switch, the broadband amplifier, the mixer, and the FPGA in sequence to generate the preprocessed electric field strengths of different attitudes of the object to be recognized.

8. A three-dimensional attitude real-time recognition method based on brain-inspired metamaterials implemented by using claim 1, characterized in that, the method includes: a high-directivity lens antenna emits electromagnetic waves to the object to be recognized; the brain-inspired metamaterials process the electromagnetic waves into the electric field strengths of different attitudes of the object to be recognized; multiple detection components detect the electric field strengths of different attitudes of the object to be recognized and send the electric field strengths to the server; the server performs three-dimensional attitude recognition according to the received attitude data.

Images