Guided quantum visual image recognition method and device

Abstract

The application provides a guided quantum visual image recognition method and device, the method comprising: designing a guided paradigm based on a visual high-low frequency complementary mechanism; constructing an enhanced quantum-classical hybrid model based on the guided paradigm for quantum image recognition. The enhanced quantum-classical hybrid model comprises: an early visual cortex module for extracting low spatial frequency (LSF) information of an input image; an orbitofrontal cortex module for generating an initial prediction based on the LSF information and locating a potential high spatial frequency (HSF) region based on an attention mechanism; a ventral visual pathway module for identifying the HSF region based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and a fusiform gyrus module for combining the HSF region obtained by the orbitofrontal cortex module and the ventral visual pathway module and identifying high spatial frequency (HSF) information by using a quantum re-uploading circuit. The application can improve the efficiency and accuracy of image classification.

Description

Technical Field

[0001] This invention relates to the fields of quantum machine learning and computer vision technology, and in particular to a guided quantum vision image recognition method and apparatus. Background Technology

[0002] Quantum Machine Learning (QML) is a discipline that combines quantum computing with machine learning algorithms to leverage the advantages of quantum computing to improve the performance and efficiency of machine learning models. Image classification, a key problem in classical machine learning, is a core task for validating the effectiveness of QML. Currently, QML research has expanded from simple research examples to more complex real-world tasks. The growing enthusiasm for QML has accelerated its application in computer vision (CV) tasks.

[0003] Figure 1 This paper showcases current popular QML paradigms, including the pure quantum paradigm (a), the parallel paradigm (b), and the serial paradigm (c). Early QML research focused primarily on specific quantum algorithms, gradually forming three main directions: Quantum Kernel Methods, which apply quantum kernels to classical kernel machine learning algorithms, such as Support Vector Machines (SVM); Quantum Convolutional Neural Networks (QCNN), such as quantum convolutions with fewer trainable parameters than classical CNNs achieved using amplitude embedding and unitary operations; and Quantum Neural Networks (QNN), based on variable parameter quantum circuits and trained via gradient descent. Representative work includes data re-upload, which bypasses the no-cloning restriction by repeatedly uploading classical data.

[0004] However, these quantum algorithms face limitations in the number of qubits when applied. For example... Figure 1 In quantum computing, the pure quantum paradigm (a) is typically limited to input sizes of 8×8 or smaller. To extend the applicability of QML to high-resolution images, researchers have proposed a parallel paradigm (b), but when the input size exceeds 100x100, the increase in the number of quantum convolution operations leads to significant time consumption. Another serial paradigm (c) fails to directly integrate the quantum part with the input data and introduces preprocessing steps such as convolution, pooling, fully connected layers, and even pre-trained models, making the mechanism of quantum advantage ambiguous. Specifically, it is difficult to demonstrate quantum advantage through methods other than performance metrics. Therefore, exploring new paradigms that can reasonably leverage the advantages of quantum circuits is crucial for QML and the entire field of quantum computing.

[0005] The human visual system exhibits a "Forest Before Trees" characteristic (focusing on the big picture before details). Recent research on human brain cognition has found that this characteristic stems from the complementary architecture of high spatial frequency (HSF) and low spatial frequency (LSF) signals during visual processing. For example... Figure 2 As shown, visual input is first processed in the early visual cortex (EVC) to generate a low-frequency signal containing the general outline of the image. Subsequently, the fronto-orbital cortex (OFC) receives the low-frequency signal, generates a preliminary prediction, and feeds it back to the fusiform gyrus to reduce the burden of subsequent analysis. Then, detailed information rich in high-frequency information is transmitted to the fusiform gyrus through the ventral visual pathway (VVS). Finally, the fusiform gyrus integrates the output of the OFC to delve into high-frequency details such as edges and textures to confirm the target recognition result.

[0006] Inspired by this, the present invention designs a novel and lightweight high- and low-frequency complementary QML architecture to jointly learn the low-frequency and high-frequency features of images in order to improve recognition speed and accuracy.

[0007] Furthermore, quantum data re-uploading is a QML technique that loads classical data multiple times during computation to improve the expressive power of quantum circuits. Compared to linear quantum models and quantum kernel methods, quantum re-uploading requires an exponential advantage in the number of qubits. Studies have shown that quantum re-uploaded circuits can fit Fourier series of arbitrary frequencies, exhibiting advantages in high-frequency representation. Subsequently, several studies have extended quantum re-uploading to image tasks, but the experimental datasets are relatively simple, making it difficult to apply to more complex tasks.

[0008] Therefore, this invention delves into the high-frequency expressive power of data re-upload methods and combines them with the low-frequency complementarity of classic networks to find suitable practical application scenarios in image classification tasks. Summary of the Invention

[0009] To address the aforementioned problems, the present invention aims to provide a guided quantum vision image recognition method and device. By innovating the paradigm, it breaks through the limitations of the existing QML paradigm in high-dimensional data processing. Based on the high- and low-frequency complementary mechanism of the human visual system, it fully utilizes the high-frequency expressive power of the quantum reloading method and the low-frequency complementarity of the classical network to achieve efficient fusion of the quantum and classical parts, thereby improving the efficiency and accuracy of image classification.

[0010] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0011] On the one hand, a guided quantum vision image recognition method is provided, the method comprising the following steps:

[0012] S1: Design a guiding paradigm based on the complementary mechanism of visual high and low frequencies;

[0013] S2: Construct an enhanced quantum-classical hybrid model based on the aforementioned guiding paradigm for quantum image recognition; the enhanced quantum-classical hybrid model includes: an early visual cortex module, an orbitofrontal cortex module, a ventral visual pathway module, and a fusiform gyrus module.

[0014] Specifically, the early visual cortex module extracts low spatial frequency (LSF) information from the input image; the orbitofrontal cortex module generates an initial prediction based on the LSF information and locates potential high spatial frequency (HSF) regions based on an attention mechanism; the ventral visual pathway module identifies HSF regions based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and the fusiform gyrus module combines the HSF regions obtained by the orbitofrontal cortex module and the ventral visual pathway module, and uses a quantum reload circuit to identify high spatial frequency (HSF) information.

[0015] Optionally, the guidance paradigm based on the visual high- and low-frequency complementary mechanism includes:

[0016] A classical convolutional neural network is used to process global low spatial frequency (LSF) information, and the quantum reload circuit is guided to focus on the high spatial frequency (HSF) region to process high spatial frequency HSF information.

[0017] Optionally, the early visual cortex module employs a classic convolutional neural network structure, including two convolutional layers and two pooling layers, as expressed in the following formula:

[0018] P1 = MaxPool1(W1*X + b1)

[0019] Z Lsr =MaxPool2(W2*P1+b2)

[0020] Where X is the input image, W1 is the weight of the first convolutional layer, b1 is the bias of the first convolutional layer, and MaxPool... l This represents the first pooling layer, P1 is the output after passing through the first convolutional layer and the first pooling layer, W2 is the weight of the second convolutional layer, b2 is the bias of the second convolutional layer, MaxPool2 represents the second pooling layer, and Z represents the weight of the second pooling layer. LSF It is the output after the second convolutional layer and the second pooling layer, namely the LSF information extracted by the early visual cortex module.

[0021] Optionally, in the orbitofrontal cortex module, the initial prediction is generated as follows:

[0022] h Classical =ReLU(W OFC Z LsF+b OFC )

[0023] Among them, W OFC b is the weight function of the initial prediction network. OFC h is the bias function of the initial prediction network, ReLU is the activation function of the initial prediction network, and h is the bias function. Classical This is the generated initial prediction result;

[0024] The potential HSF region is located based on the attention mechanism, as shown below:

[0025] A HSF =(W HSF *Z LSF +b HsF )

[0026]

[0027] Among them, W HSF These are the weights of a convolutional network based on an attention mechanism, b HSF It is the bias of a convolutional network based on the attention mechanism, A HSF It is the attention map output by a convolutional network based on the attention mechanism; (i,j) represents attention map A. HSF The coordinates of the location in the graph; the value of each location (i,j) is the attention score of the region where the location is located, corresponding to the degree of human interest in the HSF; (i*,j*) is the A HSF The coordinates corresponding to the maximum value in the middle; (i orig j orig ) represents the position coordinates of the HSF region in the original image; X represents the original input image, which is a square grayscale image with a side length of H, and h is the side length of the square HSF region extracted by the attention mechanism of the current module; This indicates the high spatial frequency (HSF) region for the localization of the orbitofrontal cortex module.

[0028] Optionally, in the ventral visual pathway module, the high-frequency complexity index based on the two-dimensional discrete Fourier transform is calculated as follows:

[0029]

[0030] Among them, R i,j This represents a segment of the original input image X into non-overlapping regions of size h*h, where i and j index the regions in the height and width directions, respectively; (u,v) represents each region R. i,j The coordinates of the point in C u,v Indicates R i,j The frequency coefficients f obtained by performing a two-dimensional discrete Fourier transform u,v E represents the calculated frequency amplitude of the region. HighE Low R respectively i,j The sum of the high-frequency and low-frequency components at all locations, r Ri,j Indicates a high-frequency complexity index;

[0031] HSF region selection is as follows:

[0032]

[0033] in, This indicates the HSF region identified by the ventral visual pathway module.

[0034] Optionally, the processing procedure of the spindle-shaped return module is as follows:

[0035] Sobel operator edge enhancement:

[0036]

[0037] Among them, K x K y These represent the horizontal and vertical convolution kernels of the Sobel operator, respectively. These represent the horizontal and vertical gradients after human-like attention localization based on the orbitofrontal cortex module, used to capture key features of objects in vision; E represents the edge gradients of the image in the horizontal and vertical directions after applying the Sobel operator to the HSF region; Humans and E Metrics These represent the edge intensity maps of the processed image, used to enhance the outline and details of objects;

[0038] Feature connection and flattening:

[0039] E Combined =Concat(E Human E Metrie )

[0040] z HSF =Flatten(E Combined )

[0041] Where Concat represents feature concatenation, and Flatten represents feature flattening; E Cobined Z represents the features after the connection. HSF Indicates HSF information;

[0042] Quantum re-upload circuit processing:

[0043] h Quantum =f Quantum ((z HSF )

[0044]

[0045] Among them, f Quantum This represents a quantum reload circuit with multiple parameterized quantum gates, h Quantum L represents the output vector of the circuit; L represents the number of nonlinear quantum blocks stacked in the circuit, and l represents the index of each specific nonlinear quantum block. This represents the classical eigenvector of the nth input data in the l-th layer. In quantum circuits, these eigenvectors are processed by R... z The rotating gate is encoded as the rotation angle of a quantum state; θ represents the set of all trainable parameters in the quantum circuit. U represents the k-th parameterized quantum gate module in the l-th layer. SE The parameters, This represents the parameters of the additional parameterized quantum gate module at the end of the l-th layer, used to adjust the quantum state before the output; U (l) Represents the overall quantum operation of the l-th layer, including L repeated modular operations and the final entanglement operation before the output at the end; U SE R represents a parameterized strongly entangled layer with a CZ entanglement gate. z Represents a single-qubit rotation gate around the Z-axis; z (l) Z represents the classical output vector of the l-th layer. j Let d represent the Pauli-Z operator acting on the j-th qubit, d represent the number of qubits in the quantum circuit, i.e., the system dimension, and ψ represent the number of qubits. (l) f represents the quantum state after the l-th layer has been processed. M This represents the Mth classical post-processing function, responsible for processing the z-values ​​output by the quantum circuit. (l) Further mapping to the final result h Quantum .

[0046] On the other hand, a guided quantum vision image recognition device is provided for implementing the method described in any of the above embodiments. The device includes an enhanced quantum classical hybrid model constructed based on a guided paradigm, wherein the guided paradigm is based on a visual high-low frequency complementary mechanism;

[0047] The enhanced quantum classical hybrid model includes: an early visual cortex module, an orbitofrontal cortex module, a ventral visual pathway module, and a fusiform gyrus module;

[0048] The early visual cortex module is used to extract low spatial frequency (LSF) information from the input image; the orbitofrontal cortex module is used to generate initial predictions based on LSF information and locate potential high spatial frequency (HSF) regions based on an attention mechanism; the ventral visual pathway module is used to identify HSF regions based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and the fusiform gyrus module is used to combine the HSF regions obtained by the orbitofrontal cortex module and the ventral visual pathway module and use a quantum reload circuit to identify high spatial frequency (HSF) information.

[0049] On the other hand, an electronic device is provided, the electronic device comprising:

[0050] processor;

[0051] A memory storing computer-readable instructions, which, when loaded and executed by the processor, implement the steps of the guided quantum vision image recognition method described above.

[0052] On the other hand, a computer-readable storage medium is provided, wherein program code is stored in the computer-readable storage medium, and the program code can be called by a processor to execute the steps of the guided quantum vision image recognition method described above.

[0053] The beneficial effects of the technical solution provided by this invention include at least the following:

[0054] (1) This invention proposes a quantum-classical hybrid paradigm inspired by human cognition, namely the guided paradigm. This guided paradigm processes the complete low-frequency information of an image through a classical network, while simultaneously guiding the quantum re-upload circuit to focus on the high-frequency complex regions in the image, thereby effectively alleviating the limitations of existing QML paradigms in high-dimensional data processing. The unique design of the guided paradigm achieves efficient synergy between the classical and quantum components, ensuring that the advantages of quantum computing are fully utilized.

[0055] (2) This invention designs a high- and low-frequency dual-channel complementary architecture. The classical channel is responsible for processing the low-frequency global structure of the image, while the quantum channel focuses on modeling high-frequency details. This design not only reduces the processing burden of the quantum part, but also makes up for the shortcomings of the classical part in high-frequency representation, thus achieving fast and accurate image classification.

[0056] (3) This invention proposes a high-frequency complexity index based on two-dimensional discrete Fourier transform (2D-DFT). This index is used to identify complex high-frequency regions in images that are difficult for classical networks to represent effectively, guiding quantum resources to focus on processing these key regions. The introduction of this high-frequency complexity index enhances the ability of the quantum part to represent high-frequency features, further improving the classification performance of the overall model.

[0057] (4) This invention achieves an integrated design of the Early Visual Cortex (EVC) module, the Orbitofrontal Cortex (OFC) module, the Ventral Visual Pathway (VVS) module, and the Fusiform module. These modules are respectively responsible for low-frequency information extraction, initial prediction generation and high-frequency region localization, high-frequency region identification based on image high-frequency complexity index, and detailed identification of high-frequency information through a quantum re-upload circuit. The integration method and collaborative working mechanism in this invention further enhance the performance and interpretability of the overall system.

[0058] (5) This invention achieves efficient integration of quantum and classical components through paradigm innovation, complementary design of high and low frequency dual channels and innovation of high frequency complexity index, thereby improving the efficiency and accuracy of image classification. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a schematic diagram of the currently popular QML paradigm;

[0061] Figure 2 This is a schematic diagram of the high- and low-frequency complementary mechanism of human vision;

[0062] Figure 3 This is a flowchart of a guided quantum vision image recognition method provided in an embodiment of the present invention;

[0063] Figure 4 This is a schematic diagram of the guiding paradigm provided in the embodiments of the present invention;

[0064] Figure 5 This is a schematic diagram of the EQC model provided in an embodiment of the present invention. Detailed Implementation

[0065] 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0066] In embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the term "exemplary" is intended to present the concept in a specific manner.

[0067] This invention provides a guided quantum vision image recognition method, which can be implemented by an electronic device, such as a terminal or a server. Figure 3 As shown, the processing flow of this method may include the following steps:

[0068] S1: Design a guiding paradigm based on the complementary mechanism of visual high and low frequencies.

[0069] Humans tend to perceive the overall outline of a scene (forest) before focusing on local details (trees) during visual processing; this is known as the "Forest Before Trees" effect. This mechanism not only effectively reduces cognitive load but also ensures perceptual accuracy while maintaining processing efficiency.

[0070] Inspired by this, the present invention proposes an innovative guiding paradigm, such as Figure 4 As shown, the classical network processes the complete forest and guides the quantum circuit to the complex trees. This design has a dual advantage: the classical network undertakes feature extraction of the complete input, breaking through the input dimensionality limitation; the quantum circuit directly processes the raw input, ensuring that the unique advantages of quantum computing are fully utilized.

[0071] Specifically, in this invention, the guidance paradigm based on the visual high- and low-frequency complementary mechanism includes:

[0072] A classical convolutional neural network is used to process global low spatial frequency (LSF) information, and the quantum reload circuit is guided to focus on the high spatial frequency (HSF) region to process high spatial frequency HSF information.

[0073] The visual high- and low-frequency complementary mechanism has the following advantages:

[0074] 1) Rapid preliminary judgment: Preliminary judgment is made by quickly transmitting LSF information, saving time for subsequent processing. 2) Reduced cognitive load: Relevant information is processed centrally based on preliminary predictions, reducing cognitive load. 3) Improved recognition accuracy and speed: Integrating HSF and LSF information improves the accuracy and speed of object recognition.

[0075] S2: Constructing an Enhanced Quantum-Classical Hybrid (EQC) model based on the guided paradigm for quantum image recognition. For example... Figure 5As shown, the enhanced quantum classical hybrid model (EQC model) includes: the early visual cortex module (EVC), the orbitofrontal cortex module (OFC), the ventral visual pathway module (VVS), and the fusiform module.

[0076] The system consists of several modules: the early visual cortex module extracts low spatial frequency (LSF) information from the input image; the orbitofrontal cortex module generates an initial prediction based on the LSF information and locates potential high spatial frequency (HSF) regions based on an attention mechanism; the ventral visual pathway module identifies HSF regions based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and the fusiform gyrus module combines the HSF regions obtained by the orbitofrontal cortex module and the ventral visual pathway module and uses a quantum reload circuit to identify high spatial frequency (HSF) information.

[0077] The EQC model enhances interpretability through a modular design aligned with the various stages of human vision and draws inspiration from biological processes to solve image classification tasks more naturally. By thoroughly integrating classical and quantum channels, the EQC model fully leverages the strengths of both paradigms to capture rich information from images.

[0078] Specifically, the classical channel is responsible for low-frequency processing and high-frequency region discovery, while the quantum channel focuses on modeling local high frequencies. This complementary design of high and low frequency dual channels reduces the processing burden of the quantum part while compensating for the shortcomings of the classical part in high-frequency representation, enabling fast and accurate image classification and alleviating the limitations of QML in terms of input dimensions.

[0079] Furthermore, the Early Visual Cortex (EVC) module employs a classic convolutional neural network structure, including two convolutional layers and two pooling layers, with the aim of rapidly processing the input image and extracting coarse-grained low-frequency information.

[0080] The formula is expressed as follows:

[0081] P1 = MaxPoo1 l (W1*X+b1)

[0082] Z LsF =MaxPool2(W2*P1+b2)

[0083] Where X is the input image, W1 is the weight of the first convolutional layer, b1 is the bias of the first convolutional layer, MaxPool1 represents the first pooling layer, P1 is the output after the first convolutional layer and the first pooling layer, W2 is the weight of the second convolutional layer, b2 is the bias of the second convolutional layer, MaxPool2 represents the second pooling layer, and Z... LSF It is the output after the second convolutional layer and the second pooling layer, namely the LSF information extracted by the early visual cortex module.

[0084] Furthermore, in the orbitofrontal cortex module (OFC), the initial prediction is generated as follows:

[0085] h Classical =ReLU(W OFC Z LSF +b OFC )

[0086] Among them, W OFC b is the weight function of the initial prediction network. OFC h is the bias function of the initial prediction network, ReLU is the activation function of the initial prediction network, and h is the bias function. Classical This is the generated initial prediction result;

[0087] The potential HSF region is located based on the attention mechanism, as shown below:

[0088] A HsF =(W HsF *Z LsF +b HsF )

[0089]

[0090] Among them, W HSF These are the weights of a convolutional network based on an attention mechanism, b HSF It is the bias of a convolutional network based on the attention mechanism, A HSF It is the attention map output by a convolutional network based on the attention mechanism; (i, j) represents the attention map A. HSF The coordinates of the location in the graph; the value of each location (i, j) is the attention score of the region where the location is located, corresponding to the degree of human interest in the HSF; (i*, j*) is the A HSF The coordinates corresponding to the maximum value in the middle; (i orig j orig ) represents the position coordinates of the HSF region in the original image; X represents the original input image, which is a square grayscale image with a side length of H, and h is the side length of the square HSF region extracted by the attention mechanism of the current module; This indicates the high spatial frequency (HSF) region for the localization of the orbitofrontal cortex module.

[0091] Furthermore, in the ventral visual pathway module (VVS), the high-frequency complexity index based on two-dimensional discrete Fourier transform (2D-DFT) is calculated as follows:

[0092]

[0093] Among them, R i,jThis represents a segment of the original input image X into non-overlapping regions of size h*h, where i and j index the regions in the height and width directions, respectively; (u,v) represents each region R. i,j The coordinates of the point in C u,v Indicates R i,j The frequency coefficients f obtained by performing a two-dimensional discrete Fourier transform u,v E represents the calculated frequency amplitude of the region. High E Low R respectively i,j The sum of the high-frequency and low-frequency components at all locations, r Ri,j Indicates a high-frequency complexity index;

[0094] HSF region selection is as follows:

[0095]

[0096] in, This indicates the HSF region identified by the ventral visual pathway module.

[0097] Furthermore, the processing procedure of the fusiform module is as follows:

[0098] Sobel operator edge enhancement:

[0099]

[0100] Among them, K x K y These represent the horizontal and vertical convolution kernels of the Sobel operator, respectively. These represent the horizontal and vertical gradients after human-like attention localization based on the orbitofrontal cortex module, used to capture key features of objects in vision; E represents the edge gradients of the image in the horizontal and vertical directions after applying the Sobel operator to the HSF region; Humans and E Metrics These represent the edge intensity maps of the processed image, used to enhance the outline and details of objects;

[0101] Feature connection and flattening:

[0102] E Combined =Concat(E Human E Metric )

[0103] z HSF =Flatten(E) Combined )

[0104] Where Concat represents feature concatenation, and Flatten represents feature flattening; E Cobined Z represents the features after the connection. HSF This indicates HSF information.

[0105] Quantum re-upload circuit processing:

[0106] h Quantum =f Quantum ((z HSF )

[0107]

[0108] Among them, f Quantum This represents a quantum reload circuit with multiple parameterized quantum gates, h Quantum L represents the output vector of the circuit; L represents the number of nonlinear quantum blocks stacked in the circuit, and l represents the index of each specific nonlinear quantum block. This represents the classical eigenvector of the nth input data in the l-th layer. In quantum circuits, these eigenvectors are processed by R... z The rotating gate is encoded as the rotation angle of a quantum state; θ represents the set of all trainable parameters in the quantum circuit. U represents the k-th parameterized quantum gate module in the l-th layer. SE The parameters, This represents the parameters of the additional parameterized quantum gate module at the end of the l-th layer, used to adjust the quantum state before the output; U (l) Represents the overall quantum operation of the l-th layer, including L repeated modular operations and the final entanglement operation before the output at the end; U SE R represents a parameterized strongly entangled layer with a CZ entanglement gate. z Represents a single-qubit rotation gate around the Z-axis; z (l) Z represents the classical output vector of the l-th layer. j Let d represent the Pauli-Z operator acting on the j-th qubit, d represent the number of qubits in the quantum circuit, i.e., the system dimension, and ψ represent the number of qubits. (l) f represents the quantum state after the l-th layer has been processed. M This represents the Mth classical post-processing function, responsible for processing the z-values ​​output by the quantum circuit. (l) Further mapping to the final result h Quantum .

[0109] Through the collaborative work of the above modules, the EQC model can effectively distinguish and integrate coarse and fine information, achieving efficient and accurate image classification.

[0110] The guided quantum vision image recognition method proposed in this invention brings about many beneficial effects through innovative technical solutions, including but not limited to the following:

[0111] 1. Break through the limitations of the existing QML paradigm;

[0112] Optimizing the Paradigm for Combining Quantum and Classical Algorithms: In the current era of Noisy Intermediate-Scale Quantum (NISQ), this invention proposes a guiding paradigm that effectively optimizes the combination of quantum and classical algorithms. By processing global low-frequency information through classical networks and guiding quantum circuits to focus on high-frequency complex regions, the application scope of quantum algorithms in high-dimensional image data processing is significantly expanded. This method not only alleviates the problem of the number of quantum feature extraction operations expanding with the input dimension but also improves the utilization efficiency of quantum computing resources, enabling QML models to process larger-scale and higher-resolution image data.

[0113] 2. Improve the applicability of quantum algorithms in image classification tasks;

[0114] Classification task optimization combining high-frequency Fourier analysis: This invention introduces a high-frequency complexity index based on two-dimensional discrete Fourier transform (2D-DFT) to accurately identify high-frequency regions suitable for quantum processing. This method combining high-frequency Fourier analysis allows quantum algorithms to focus on capturing subtle texture and edge information in images, thus exhibiting significant performance improvements in various complex image classification tasks (such as medical image classification and texture classification). Experimental results show that the EQC architecture based on this invention outperforms existing classical and quantum baseline models on multiple high-resolution and high-frequency complexity datasets, verifying its superiority in practical applications.

[0115] 3. QML paradigm innovation;

[0116] The Guiding Paradigm: This invention proposes for the first time a quantum-classical hybrid paradigm inspired by human cognition—the Guiding Paradigm. This paradigm guides quantum circuits through classical networks to the most suitable high-frequency complex regions for processing, effectively alleviating the dimensionality limitation problem of QML when processing high-dimensional data. This innovative paradigm design achieves efficient synergy between the classical and quantum components, ensuring the full utilization of the advantages of quantum computing, thereby improving the overall image recognition performance of the model.

[0117] 4. Complementary architecture design;

[0118] High- and low-frequency complementary dual-channel design: Drawing inspiration from the high- and low-frequency complementary mechanism in the human visual system, this invention designs a dual-channel architecture. The classical channel is responsible for processing the low-frequency global structure of the image, while the quantum channel focuses on modeling high-frequency details. This design not only reduces the processing burden of the quantum part but also compensates for the shortcomings of the classical part in high-frequency representation, achieving fast and accurate image classification. The high- and low-frequency complementary dual-channel design is the core innovation of this invention in the quantum-classical hybrid network architecture, significantly improving the classification efficiency and accuracy of the model.

[0119] 5. High-frequency region detection strategy;

[0120] Innovative High-Frequency Region Discovery Strategies: This invention proposes two efficient and novel high-frequency region discovery strategies: an HSF localization algorithm that mimics human attention and an HSF localization algorithm driven by a high-frequency complexity index based on 2D-DFT. These strategies can accurately identify complex high-frequency regions in images that are difficult for classical networks to represent effectively, guiding quantum resources to focus on processing these key regions and further enhancing the quantum part's ability to represent high-frequency features. Through these high-frequency region discovery strategies, this invention expands the application scope of quantum-classical hybrid algorithms, enabling them to adapt to more diverse and complex image classification tasks.

[0121] 6. Enhance the interpretability and robustness of the model;

[0122] Modular Design and Bioinspiration: The EQC model, through its modular design aligned with the various stages of human vision, not only enhances its interpretability but also improves its robustness under different noise environments. The Early Visual Cortex (EVC) module is responsible for low-frequency information extraction, the Orbitofrontal Cortex (OFC) module generates initial predictions and locates high-frequency regions, the Ventral Visual Pathway (VVS) module identifies high-frequency regions based on the Fourier high-frequency complexity index, and the Fusiform module performs detailed identification of high-frequency information through a quantum reload circuit. This collaborative working mechanism among the modules enables the EQC model to maintain efficient training and inference performance when processing high-resolution and high-frequency complex images, avoiding training difficulties such as gradient vanishing.

[0123] Accordingly, embodiments of the present invention also provide a guided quantum vision image recognition device, the device comprising an enhanced quantum classical hybrid model (EQC model) constructed based on a guided paradigm, wherein the guided paradigm is a guided paradigm based on a visual high-low frequency complementary mechanism.

[0124] The enhanced quantum classical hybrid model includes: an early visual cortex module, an orbitofrontal cortex module, a ventral visual pathway module, and a fusiform gyrus module.

[0125] The early visual cortex module is used to extract low spatial frequency (LSF) information from the input image; the orbitofrontal cortex module is used to generate initial predictions based on LSF information and locate potential high spatial frequency (HSF) regions based on an attention mechanism; the ventral visual pathway module is used to identify HSF regions based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and the fusiform gyrus module is used to combine the HSF regions obtained by the orbitofrontal cortex module and the ventral visual pathway module and use a quantum reload circuit to identify high spatial frequency (HSF) information.

[0126] The apparatus in this embodiment can be used to execute the technical solutions of the above method embodiments. Its implementation principle and technical effects are similar, and will not be repeated here.

[0127] In summary, this invention achieves a significant performance improvement in image classification tasks through paradigm innovation, complementary high- and low-frequency dual-channel design, and innovative high-frequency complexity metrics. Its groundbreaking technical solution not only expands the application scope of QML in high-dimensional data processing but also improves the classification accuracy and training efficiency of the model.

[0128] In an exemplary embodiment, the present invention also provides an electronic device, the electronic device comprising:

[0129] processor;

[0130] A memory storing computer-readable instructions, which, when loaded and executed by the processor, implement the steps of the guided quantum vision image recognition method described above.

[0131] In an exemplary embodiment, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the steps of the guided quantum vision image recognition method described above. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0132] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0133] The use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0134] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0135] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0136] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0137] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0138] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0139] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0140] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0142] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A guided quantum vision image recognition method, characterized in that, Includes the following steps: S1: Design a guiding paradigm based on the complementary mechanism of visual high and low frequencies; S2: Construct an enhanced quantum-classical hybrid model based on the aforementioned guiding paradigm for quantum image recognition; The enhanced quantum classical hybrid model includes: an early visual cortex module, an orbitofrontal cortex module, a ventral visual pathway module, and a fusiform gyrus module; Specifically, the early visual cortex module extracts low spatial frequency (LSF) information from the input image; the orbitofrontal cortex module generates an initial prediction based on the LSF information and locates potential high spatial frequency (HSF) regions based on an attention mechanism; the ventral visual pathway module identifies HSF regions based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and the fusiform gyrus module combines the HSF regions obtained by the orbitofrontal cortex module and the ventral visual pathway module, and uses a quantum reload circuit to identify high spatial frequency (HSF) information. The processing procedure of the spindle-shaped return module is as follows: Sobel operator edge enhancement: Among them, K x K y These represent the horizontal and vertical convolution kernels of the Sobel operator, respectively. These represent the horizontal and vertical gradients after human-like attention localization based on the orbitofrontal cortex module, used to capture key features of objects in vision; E represents the edge gradients of the image in the horizontal and vertical directions after applying the Sobel operator to the HSF region; Humans and E Metrics These represent the edge intensity maps of the processed image, used to enhance the outline and details of objects; This indicates the high spatial frequency (HSF) region for the localization of the orbitofrontal cortex module. This indicates the HSF region identified by the ventral visual pathway module; Feature connection and flattening: AND Combined =Concat(E Human ,AND Metric ) z HSF =Flatten(E Combined ) Where Concat represents feature concatenation, and Flatten represents feature flattening; E Cobined Z represents the features after the connection. HSF Indicates HSF information; Quantum re-upload circuit processing: h Quantum =f Quantum (z HSF ) Among them, f Quantum This represents a quantum reload circuit with multiple parameterized quantum gates, h Quantum L represents the output vector of the circuit; L represents the number of nonlinear quantum blocks stacked in the circuit, and l represents the index of each specific nonlinear quantum block. This represents the classical eigenvector of the nth input data in the l-th layer. In quantum circuits, these eigenvectors are processed by R... z The rotating gate is encoded as the rotation angle of a quantum state; θ represents the set of all trainable parameters in the quantum circuit. U represents the k-th parameterized quantum gate module in the l-th layer. SE The parameters, This represents the parameters of the additional parameterized quantum gate module at the end of the l-th layer, used to adjust the quantum state before the output; U (l) Represents the overall quantum operation of the l-th layer, including L repeated modular operations and the final entanglement operation before the output at the end; U SE R represents a parameterized strongly entangled layer with a CZ entanglement gate. z Represents a single-qubit rotation gate around the Z-axis; z (l) Z represents the classical output vector of the l-th layer. j Let d represent the Pauli-Z operator acting on the j-th qubit, d represent the number of qubits in the quantum circuit, i.e., the system dimension, and ψ represent the number of qubits. (l) f represents the quantum state after the l-th layer has been processed. M This represents the Mth classical post-processing function, responsible for processing the z-values ​​output by the quantum circuit. (l) Further mapping to the final result h Quantum .

2. The guided quantum vision image recognition method according to claim 1, characterized in that, The guidance paradigm based on the visual high- and low-frequency complementary mechanism includes: A classical convolutional neural network is used to process global low spatial frequency (LSF) information, and the quantum reload circuit is guided to focus on the high spatial frequency (HSF) region to process high spatial frequency HSF information.

3. The guided quantum vision image recognition method according to claim 1, characterized in that, The early visual cortex module employs a classic convolutional neural network structure, including two convolutional layers and two pooling layers, as expressed in the following formula: P1 = MaxPool1(W1*X + b1) Z LSF =MaxPool2(W2*P1+b2) Where X is the input image, W1 is the weight of the first convolutional layer, b1 is the bias of the first convolutional layer, MaxPool1 represents the first pooling layer, P1 is the output after the first convolutional layer and the first pooling layer, W2 is the weight of the second convolutional layer, b2 is the bias of the second convolutional layer, MaxPool2 represents the second pooling layer, and Z... LSF It is the output after the second convolutional layer and the second pooling layer, namely the LSF information extracted by the early visual cortex module.

4. The guided quantum vision image recognition method according to claim 3, characterized in that, In the orbitofrontal cortex module, the initial prediction is generated as follows: h Classical =ReLU(W OFC WITH LSF +b OFC ) Among them, W OFC b is the weight function of the initial prediction network. OFC h is the bias function of the initial prediction network, ReLU is the activation function of the initial prediction network, and h is the bias function. Classical This is the generated initial prediction result; The potential HSF region is located based on the attention mechanism, as shown below: A HSF =(W HSF *Z LSF +b HSF ) Among them, W HSF These are the weights of a convolutional network based on an attention mechanism, b HSF It is the bias of a convolutional network based on the attention mechanism, A HSF It is the attention map output by a convolutional network based on the attention mechanism; (i, j) represents the attention map A. HSF The coordinates of the location in the graph; the value of each location (i, j) is the attention score of the region where the location is located, corresponding to the degree of human interest in the HSF; (i*, j*) is the A HSF The coordinates corresponding to the maximum value in the middle; (i orig j orig ) represents the position coordinates of the HSF region in the original image; X represents the original input image, which is a square grayscale image with a side length of H, and h is the side length of the square HSF region extracted by the attention mechanism of the current module; This indicates the high spatial frequency (HSF) region for the localization of the orbitofrontal cortex module.

5. The guided quantum vision image recognition method according to claim 4, characterized in that, In the ventral visual pathway module, the high-frequency complexity index based on the two-dimensional discrete Fourier transform is calculated as follows: Among them, R i,j This represents a segment of the original input image X into non-overlapping regions of size h*h, where i and j index the regions in the height and width directions, respectively; (u,v) represents each region R. i,j The coordinates of the point in C u,v Indicates R i,j The frequency coefficients f obtained by performing a two-dimensional discrete Fourier transform u,v E represents the calculated frequency amplitude of the region. High E Low R respectively i,j The sum of the high-frequency and low-frequency components at all locations, r Ri,j Indicates a high-frequency complexity index; HSF region selection is as follows: in, This indicates the HSF region identified by the ventral visual pathway module.

6. A guided quantum vision image recognition device, the device being used to implement the method as described in any one of claims 1 to 5, characterized in that, The device includes an enhanced quantum classical hybrid model constructed based on a guiding paradigm, wherein the guiding paradigm is a guiding paradigm based on a visual high-low frequency complementary mechanism; The enhanced quantum classical hybrid model includes: an early visual cortex module, an orbitofrontal cortex module, a ventral visual pathway module, and a fusiform gyrus module; The early visual cortex module is used to extract low spatial frequency (LSF) information from the input image; the orbitofrontal cortex module is used to generate initial predictions based on LSF information and locate potential high spatial frequency (HSF) regions based on an attention mechanism; the ventral visual pathway module is used to identify HSF regions based on a high-frequency complexity index of two-dimensional discrete Fourier transform; and the fusiform gyrus module is used to combine the HSF regions obtained by the orbitofrontal cortex module and the ventral visual pathway module and use a quantum reload circuit to identify high spatial frequency (HSF) information.

7. An electronic device, characterized in that, The electronic device includes: processor; A memory storing computer-readable instructions that, when loaded and executed by the processor, implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 5.

Images