Remote sensing image quantum identification method and device, storage medium and electronic device

Abstract

The application discloses a kind of remote sensing image quantum identification method, device, storage medium and electronic device, by receiving the remote sensing equipment to be identified remote sensing image photographed;The image classification model including quantum hollow convolutional neural network and classifier is input to the remote sensing image to be identified, the quantum hollow convolutional neural network is trained based on the pixel of remote sensing image to extract target object feature, the classifier is trained based on target object feature and predicts the kind of target object in remote sensing image;Receive the prediction data output by the image classification model, and provide the prediction data as output, the prediction data is used to characterize the kind of target object in the remote sensing image to be identified, using the high-speed computing capacity of quantum computation, the improvement of the identification speed of remote sensing image is realized.

Description

Technical Field

[0001] This invention belongs to the field of quantum computing technology, and in particular to a method, device, storage medium and electronic device for quantum recognition of remote sensing images. Background Technology

[0002] Extracting valuable knowledge from remote sensing images has become a research topic of great interest, and the identification of target objects within these images is an important method.

[0003] Faced with massive amounts of remote sensing image data, current methods relying on manual interpretation or classical computer processing are still not timely enough to quickly extract valuable data, failing to meet the objective needs of timely analysis and processing in modern life, such as environmental monitoring, disaster response, and urban management. Therefore, improving the recognition speed of remote sensing images is an urgent problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, storage medium, and electronic device for quantum recognition of remote sensing images, aiming to improve the recognition speed of remote sensing images.

[0005] One embodiment of the present invention provides a quantum recognition method for remote sensing images, the method comprising:

[0006] Receive remote sensing images to be identified captured by remote sensing equipment;

[0007] The remote sensing image to be identified is input into an image classification model that includes a quantum-dilated convolutional neural network and a classifier. The quantum-dilated convolutional neural network is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on the target object features.

[0008] The system receives the prediction data output by the image classification model and provides the prediction data as output, wherein the prediction data is used to characterize the type of target object in the remote sensing image to be identified.

[0009] Optionally, the quantum-hole convolutional neural network includes an encoding layer for encoding target pixels into an initial state of qubits. The target pixels are determined from the remote sensing image to be identified based on a preset dilation rate and convolution kernel. The initial state is used to extract target object features from the remote sensing image to be identified.

[0010] Optionally, the coding layer includes an H gate, an RY gate, and an RZ gate acting sequentially on the qubit. The rotation parameters of the RY gate are determined based on the inverse trigonometric function value of the target pixel, and the rotation parameters of the RZ gate are determined based on the inverse trigonometric function value of the square of the target pixel.

[0011] Optionally, the quantum hollow convolutional neural network further includes an entanglement layer and a measurement layer. The entanglement layer is used to evolve the initial state into an entangled state that includes target object feature information. The measurement layer is used to measure the qubits and extract the target object features of the remote sensing image to be identified from the entangled state.

[0012] Optionally, the entanglement layer includes CNOT gates acting sequentially on adjacent qubits, CNOT gates acting on qubits spaced apart, and parametric quantum logic gates acting on the qubits, wherein the parametric quantum logic gates are determined based on the loss function used to train the quantum-hole convolutional neural network model.

[0013] Optionally, the loss function is:

[0014]

[0015] Where Loss is the loss function, n is the number of remote sensing images used for training, and y i Let f be the true location of the target object in the i-th remote sensing image used for training. i Let p be the predicted location of the target object in the i-th remote sensing image used for training. i Let c be the true label probability distribution of the target object in the i-th remote sensing image used for training. i Let be the probability distribution of the predicted labels for the target objects in the i-th remote sensing image used for training.

[0016] Optionally, before inputting the remote sensing image to be identified into an image classification model including a quantum-dilated convolutional neural network and a classifier, the following steps are included:

[0017] The remote sensing image to be identified is subjected to size transformation and normalization processing to obtain the remote sensing image to be identified that is allowed to be input by the quantum dilated convolutional neural network.

[0018] Another embodiment of the present invention provides a quantum recognition device for remote sensing images, the device comprising:

[0019] The communication unit is used to receive remote sensing images to be identified captured by the remote sensing equipment.

[0020] The processing unit is used to input the remote sensing image to be identified into an image classification model including a quantum dilated convolutional neural network and a classifier. The quantum dilated convolutional neural network is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on the target object features.

[0021] The communication unit is also used to receive the prediction data output by the image classification model and provide the prediction data as output, wherein the prediction data is used to characterize the type of target object in the remote sensing image to be identified.

[0022] Optionally, the quantum-hole convolutional neural network includes an encoding layer for encoding target pixels into an initial state of qubits. The target pixels are determined from the remote sensing image to be identified based on a preset dilation rate and convolution kernel. The initial state is used to extract target object features from the remote sensing image to be identified.

[0023] The optional coding layer includes an H gate, an RY gate, and an RZ gate acting sequentially on the qubit. The rotation parameters of the RY gate are determined based on the inverse trigonometric function value of the target pixel, and the rotation parameters of the RZ gate are determined based on the inverse trigonometric function value of the square of the target pixel.

[0024] Optionally, the quantum hollow convolutional neural network further includes an entanglement layer and a measurement layer. The entanglement layer is used to evolve the initial state into an entangled state that includes target object feature information. The measurement layer is used to measure the qubits and extract the target object features of the remote sensing image to be identified from the entangled state.

[0025] Optionally, the entanglement layer includes CNOT gates acting sequentially on adjacent qubits, CNOT gates acting on qubits spaced apart, and parametric quantum logic gates acting on the qubits, wherein the parametric quantum logic gates are determined based on the loss function used to train the quantum-hole convolutional neural network model.

[0026] Optionally, the loss function is:

[0027]

[0028] Where Loss is the loss function, n is the number of remote sensing images used for training, and y i Let f be the true location of the target object in the i-th remote sensing image used for training. i Let p be the predicted location of the target object in the i-th remote sensing image used for training. i Let c be the true label probability distribution of the target object in the i-th remote sensing image used for training. i Let be the probability distribution of the predicted labels for the target objects in the i-th remote sensing image used for training.

[0029] Optionally, before inputting the remote sensing image to be identified into an image classification model including a quantum-dilated convolutional neural network and a classifier, the following steps are included:

[0030] The remote sensing image to be identified is subjected to size transformation and normalization processing to obtain the remote sensing image to be identified that is allowed to be input by the quantum dilated convolutional neural network.

[0031] Another embodiment of the present invention provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the above embodiments when running.

[0032] Another embodiment of the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the methods described in any of the above embodiments.

[0033] Compared with existing technologies, the present invention provides a remote sensing image recognition method, apparatus, storage medium, and electronic device. This method involves receiving a remote sensing image to be recognized captured by a remote sensing device; inputting the image into an image classification model comprising a quantum diffusing convolutional neural network (QDN) and a classifier; wherein the QDN is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on these features; receiving the prediction data output by the image classification model and providing the prediction data as output; and utilizing the high-speed computing power of quantum computing to improve the speed and accuracy of remote sensing image recognition. Attached Figure Description

[0034] Figure 1 Hardware structure block diagram of a computer terminal for a quantum recognition method for remote sensing images provided in an embodiment of the present invention;

[0035] Figure 2 A flowchart illustrating a quantum recognition method for remote sensing images provided in an embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the structure of a quantum circuit corresponding to an encoding layer provided in an embodiment of the present invention;

[0037] Figure 4 This is a schematic diagram of the structure of a quantum circuit corresponding to a quantum-hole convolutional neural network provided in an embodiment of the present invention;

[0038] Figure 5 This is a schematic diagram of the structure of a remote sensing image quantum recognition device provided in an embodiment of the present invention. Detailed Implementation

[0039] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0040] This invention first provides a quantum recognition method for remote sensing images, which can be applied to electronic devices, such as computer terminals, specifically ordinary computers, quantum computers, etc.

[0041] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal for a quantum recognition method for remote sensing images provided in an embodiment of the present invention. (See diagram below.) Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing quantum recognition methods for remote sensing images are also shown. Optionally, the computer terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer terminal described above. For example, the computer terminal may also include components that are more complex than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0042] The memory 104 can be used to store software programs and modules for application software, such as the program instructions / modules corresponding to the quantum recognition method for remote sensing images in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0043] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the computer terminal. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0044] It's important to note that a true quantum computer has a hybrid structure, comprising two main parts: a classical computer responsible for performing classical computations and control, and a quantum device responsible for running quantum programs to achieve quantum computation. A quantum program is a sequence of instructions written in a quantum language such as QRunes that can run on a quantum computer, supporting operations on quantum logic gates and ultimately enabling quantum computing. Specifically, a quantum program is a sequence of instructions that operates on quantum logic gates according to a specific timing order.

[0045] In practical applications, due to limitations in the development of quantum device hardware, quantum computing simulations are often required to verify quantum algorithms, quantum applications, and so on. Quantum computing simulation is the process of simulating the execution of a quantum program corresponding to a specific problem using a virtual architecture (i.e., a quantum virtual machine) built with the resources of a regular computer. Typically, it is necessary to construct a quantum program corresponding to a specific problem. The quantum program referred to in this embodiment of the invention is a program written in a classical language that represents qubits and their evolution, wherein qubits, quantum logic gates, etc., related to quantum computing all have corresponding classical code representations.

[0046] Quantum circuits, also known as quantum logic circuits, are a manifestation of quantum programming and are the most commonly used general-purpose quantum computing model. They represent circuits that operate on qubits under an abstract concept. They consist of qubits, circuits (timelines), and various quantum logic gates. Finally, the results are often read out through quantum measurement operations.

[0047] Unlike traditional circuits that use metal wires to transmit voltage or current signals, in quantum circuits, the circuits can be seen as being connected by time. That is, the state of a quantum bit evolves naturally over time, following the instructions of the Hamiltonian operator until it encounters a logic gate and is operated on.

[0048] A quantum program corresponds to a single quantum circuit. The quantum program described in this invention refers to this single quantum circuit, where the total number of qubits in the single quantum circuit is the same as the total number of qubits in the quantum program. This can be understood as follows: a quantum program can consist of a quantum circuit, measurement operations on the qubits within the quantum circuit, registers storing the measurement results, and control flow nodes (jump instructions). A single quantum circuit can contain dozens, hundreds, or even thousands of quantum logic gate operations. The execution of a quantum program is the process of executing all the quantum logic gates in a specific timing order. It should be noted that the timing order refers to the chronological sequence in which individual quantum logic gates are executed.

[0049] It's important to note that in classical computing, the most basic unit is the bit, and the most fundamental control mode is the logic gate. Circuit control can be achieved through combinations of logic gates. Similarly, the way to process qubits is through quantum logic gates. Quantum logic gates enable the evolution of quantum states and are the foundation of quantum circuits. Quantum logic gates include single-qubit gates, such as Hadamard gates (H-gates), Pauli-X gates (X-gates), Pauli-Y gates (Y-gates), Pauli-Z gates (Z-gates), RX gates, RY gates, RZ gates, etc.; and multi-qubit gates, such as CNOT gates, CR gates, iSWAP gates, Tofoli gates, etc. Quantum logic gates are generally represented using unitary matrices, which are not only matrix forms but also operations and transformations. Generally, the effect of a quantum logic gate on a quantum state is calculated by left-multiplying the unitary matrix by the matrix corresponding to the right vector of the quantum state.

[0050] See Figure 2 , Figure 2 This is a flowchart illustrating a quantum recognition method for remote sensing images provided in an embodiment of the present invention. The method includes the following steps:

[0051] Step 201: Receive the remote sensing image to be identified captured by the remote sensing equipment;

[0052] Remote sensing equipment can include, for example, monitoring satellites, multi-angle imaging radiometers, long-wave thermal imagers, and thermal infrared imagers.

[0053] Step 202: Input the remote sensing image to be identified into an image classification model including a quantum dilated convolutional neural network and a classifier. The quantum dilated convolutional neural network is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on the target object features.

[0054] In a classic convolutional neural network, the convolution kernel is used in image processing to calculate the weighted average of pixels in a small region of the input image, which then becomes the corresponding pixel in the output image. The weights are defined by a function called the convolution kernel, also known as a filter.

[0055] Dilated convolution, also known as atrous deconvolution, is a novel convolutional approach proposed to address the image resolution reduction and information loss issues caused by downsampling in image semantic segmentation. By introducing a dilation rate parameter, dilated convolution allows a kernel of the same size to achieve a larger receptive field. Consequently, it also results in a smaller number of parameters compared to ordinary convolution for the same receptive field size. Therefore, it has wide applications in semantic segmentation, object detection, and speech synthesis.

[0056] Quantum dilated convolutional neural networks are constructed from quantum circuits, and the convolution calculation is performed by quantum circuits, unlike classical convolutional neural networks, which calculate the weighted average of pixels in the input image and weights in the convolution kernel.

[0057] The classifier can be either classical or quantum. Classical classifiers include fully connected layers in classical neural networks, while quantum classifiers include classifiers constructed from quantum circuits. No specific limitation is made here.

[0058] Step 203: Receive the prediction data output by the image classification model and provide the prediction data as output. The prediction data is used to characterize the type of target object in the remote sensing image to be identified.

[0059] The predicted data can be provided as output, either by sending it directly to the electronic devices of the associated monitoring personnel or by displaying it on the screen of the monitoring system. Alternatively, the predicted data can be further processed to determine whether further warnings and displays are needed based on the type and model of the target object.

[0060] Compared with existing technologies, the present invention provides a remote sensing image recognition method, apparatus, storage medium, and electronic device. This method involves receiving a remote sensing image to be recognized captured by a remote sensing device; inputting the image into an image classification model including a quantum dilated convolutional neural network (QDN) and a classifier. The QDN is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on these features; receiving the prediction data output by the image classification model and providing the prediction data as output. The prediction data characterizes the type of target object in the remote sensing image. By utilizing the high-speed computing power of quantum computing, the recognition speed of remote sensing images is improved.

[0061] Optionally, the quantum-hole convolutional neural network includes an encoding layer for encoding target pixels into an initial state of qubits. The target pixels are determined from the remote sensing image to be identified based on a preset dilation rate and convolution kernel. The initial state is used to extract target object features from the remote sensing image to be identified.

[0062] Assuming the kernel size is k and the dilation rate is d, the receptive field of the dilated convolution kernel is: n = k + (k-1)·(d-1). For example, if d = 2 and k = 2, then n = 3. For a 4×4 remote sensing image to be identified, if the original target pixels are the 1st, 2nd, 5th, and 6th pixels, then the target pixels in the dilated convolution are the 1st, 3rd, 9th, and 11th pixels. Although the specific target pixels are different, their number remains the same.

[0063] As can be seen, the difference from classical dilated convolution is that in quantum dilated convolutional neural networks, the dilation rate and convolution kernel are used to determine the target pixel from the remote sensing image to be identified. The target pixel is not calculated by weighted averaging with the weights in the convolution kernel. The target pixel is classical data to be encoded into a quantum state and is the input of the quantum circuit. The specific calculation is reflected in the quantum circuit. Therefore, only the size of the convolution kernel is needed, and the weights in the convolution kernel are not required.

[0064] Furthermore, the encoding method used by the encoding layer may be, for example, angle encoding, amplitude encoding, ground state encoding, etc.

[0065] In an angle encoding method provided by the present invention, the encoding layer includes an H gate, an RY gate, and an RZ gate acting sequentially on the qubit. The rotation parameters of the RY gate are determined based on the inverse trigonometric function value of the target pixel, and the rotation parameters of the RZ gate are determined based on the inverse trigonometric function value of the square of the target pixel.

[0066] like Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a quantum circuit corresponding to an encoding layer provided in an embodiment of the present invention. The quantum circuit corresponding to the encoding layer includes 4 qubits, each of which is sequentially acted upon by an H gate, an RY gate, and an RZ gate, and the qubits are initially in the 0 state.

[0067] For a 4×4 remote sensing image to be identified, four target pixels x1, x2, x3, and x4 are determined. The rotation angles of the RY gate and RZ gate acting on the first qubit are arctan(x1) and arctan(x2) respectively. 2 The rotation angles of the RY gate and RZ gate acting on the second qubit are arctan(x2) and arctan(x2), respectively. 2The rotation angles of the RY gate and RZ gate acting on the third qubit are arctan(x3) and arctan(x3), respectively. 2 The rotation angles of the RY gate and RZ gate acting on the fourth qubit are arctan(x4) and arctan(x4), respectively. 2 ).

[0068] Optionally, the quantum hollow convolutional neural network further includes an entanglement layer and a measurement layer. The entanglement layer is used to evolve the initial state into an entangled state that includes target object feature information. The measurement layer is used to measure the qubits and extract the target object features of the remote sensing image to be identified from the entangled state.

[0069] Here, the entangled layer functions similarly to that in a classic convolutional neural network, calculating a weighted average of the target pixel and the weights in the convolution kernel to obtain the image features to be identified. Unlike classic dilated convolution operations, the quantum dilated convolution operation in this embodiment is achieved through the evolution of quantum states in a quantum circuit.

[0070] Secondly, a classic 2×2 convolution kernel includes 4 weights, which require at least 4 classical bits to store. At the same time, 4 target pixels also require at least 4 classical bits to store. If the weights are decimals or fractions, or if the target pixels have more than just binary values ​​of 1 or 0, then even more classical bits will be required. In contrast, a quantum circuit only requires 4 qubits for storage and computation. The bit resources required are far less than those of a classic convolution operation, which can save computer internal resources and improve the utilization rate of computer internal resources.

[0071] Finally, based on the superposition and entanglement characteristics of quantum computing, parallel computing can be achieved, thus improving the recognition speed of remote sensing images.

[0072] In a specific embodiment provided by the present invention, the entanglement layer includes CNOT gates acting sequentially on adjacent qubits, CNOT gates acting on qubits spaced apart, and parametric quantum logic gates acting on the qubits, wherein the parametric quantum logic gates are determined based on the loss function used to train the quantum void convolutional neural network model.

[0073] like Figure 4 As shown, Figure 4 This is a schematic diagram of the quantum circuit structure corresponding to a quantum-hole convolutional neural network provided in an embodiment of the present invention. The entanglement layer includes CNOT gates acting sequentially on adjacent qubits, CNOT gates acting on qubits spaced apart, and parametric quantum logic gates acting on the qubits.

[0074] The CNOT gates acting on adjacent qubits include: CNOT gates acting on the first and second qubits, CNOT gates acting on the second and third qubits, CNOT gates acting on the third and fourth qubits, and CNOT gates acting on the fourth and first qubits.

[0075] The CNOT gates acting on the qubits spaced apart include: CNOT gates acting on the first and third qubits, CNOT gates acting on the second and fourth qubits, CNOT gates acting on the third and first qubits, and CNOT gates acting on the fourth and second qubits.

[0076] Parametric quantum logic gates include: R(α1, β1, γ1) acting on the first qubit, R(α2, β2, γ2) acting on the second qubit, R(α3, β3, γ3) acting on the third qubit, and R(α4, β4, γ4) acting on the fourth qubit. α i ,β i γ i The angles of rotation around the x-axis, y-axis, and z-axis are respectively determined based on the loss function used to train the quantum dilated convolutional neural network model, where i = 1, 2, 3, 4.

[0077] Can Figure 4 The first module consists of four CNOT gates acting on adjacent qubits, the second module consists of four CNOT gates acting on qubits spaced apart, and the third module consists of four parametric quantum logic gates. Figure 4 This is just one specific embodiment of the present invention, in which there is only one of each module. In other embodiments, the number of each module can be set according to the requirements, and there can be one or more.

[0078] Different loss functions can be selected to train the quantum dilated convolutional neural network and the classifier separately, or the same loss function can be selected to train the quantum dilated convolutional neural network and the classifier together; no limitation is made here.

[0079] In a specific embodiment provided by the present invention, the loss function is:

[0080]

[0081] Where Loss is the loss function, n is the number of remote sensing images used for training, and y i Let f be the true location of the target object in the i-th remote sensing image used for training. i Let p be the predicted location of the target object in the i-th remote sensing image used for training. iLet c be the true label probability distribution of the target object in the i-th remote sensing image used for training. i Let be the probability distribution of the predicted labels for the target objects in the i-th remote sensing image used for training.

[0082] The actual location and the predicted location can be represented by coordinates. For example, if the remote sensing images are placed in the same coordinate system, the target objects in the remote sensing images can be represented by a set of coordinates, or by the centroid coordinates obtained from that set of coordinates.

[0083] Furthermore, before identifying the target object in the remote sensing image or inputting the remote sensing image to be identified into an image classification model including a quantum-dilated convolutional neural network and a classifier, the method further includes: performing size transformation and normalization processing on the remote sensing image to be identified or the remote sensing image used for training to obtain a remote sensing image that the quantum-dilated convolutional neural network can input. This maintains the uniformity of the remote sensing images, facilitating subsequent identification and training.

[0084] It should be noted that the prediction data can be predictions for major categories or for minor categories, such as each subcategory under each major category, or predictions for specific models within each subcategory. There are no restrictions on this.

[0085] See Figure 5 , Figure 5 A schematic diagram of a remote sensing image quantum recognition device provided in an embodiment of the present invention. The device includes:

[0086] The communication unit 501 is used to receive remote sensing images to be identified captured by the remote sensing equipment;

[0087] The processing unit 502 is used to input the remote sensing image to be identified into an image classification model including a quantum dilated convolutional neural network and a classifier. The quantum dilated convolutional neural network is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on the target object features.

[0088] The communication unit 503 is also used to receive the prediction data output by the image classification model and provide the prediction data as output, wherein the prediction data is used to characterize the type of target object in the remote sensing image to be identified.

[0089] Optionally, the quantum-hole convolutional neural network includes an encoding layer for encoding target pixels into an initial state of qubits. The target pixels are determined from the remote sensing image to be identified based on a preset dilation rate and convolution kernel. The initial state is used to extract target object features from the remote sensing image to be identified.

[0090] Optionally, the coding layer includes an H gate, an RY gate, and an RZ gate acting sequentially on the qubit. The rotation parameters of the RY gate are determined based on the inverse trigonometric function value of the target pixel, and the rotation parameters of the RZ gate are determined based on the inverse trigonometric function value of the square of the target pixel.

[0091] Optionally, the quantum hollow convolutional neural network further includes an entanglement layer and a measurement layer. The entanglement layer is used to evolve the initial state into an entangled state that includes target object feature information. The measurement layer is used to measure the qubits and extract the target object features of the remote sensing image to be identified from the entangled state.

[0092] Optionally, the entanglement layer includes CNOT gates acting sequentially on adjacent qubits, CNOT gates acting on qubits spaced apart, and parametric quantum logic gates acting on the qubits, wherein the parametric quantum logic gates are determined based on the loss function used to train the quantum-hole convolutional neural network model.

[0093] Optionally, the loss function is:

[0094]

[0095] Where Loss is the loss function, n is the number of remote sensing images used for training, and y i Let f be the true location of the target object in the i-th remote sensing image used for training. i Let p be the predicted location of the target object in the i-th remote sensing image used for training. i Let c be the true label probability distribution of the target object in the i-th remote sensing image used for training. i Let be the probability distribution of the predicted labels for the target objects in the i-th remote sensing image used for training.

[0096] Optionally, before inputting the remote sensing image to be identified into an image classification model including a quantum-dilated convolutional neural network and a classifier, the following steps are included:

[0097] The remote sensing image to be identified is subjected to size transformation and normalization processing to obtain the remote sensing image to be identified that is allowed to be input by the quantum dilated convolutional neural network.

[0098] Another embodiment of the present invention provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the method embodiments above when running.

[0099] Specifically, in this embodiment, the storage medium can be configured to store a computer program for performing the following steps:

[0100] Receive remote sensing images to be identified captured by remote sensing equipment;

[0101] The remote sensing image to be identified is input into an image classification model that includes a quantum-dilated convolutional neural network and a classifier. The quantum-dilated convolutional neural network is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on the target object features.

[0102] The system receives the prediction data output by the image classification model and provides the prediction data as output, wherein the prediction data is used to characterize the type of target object in the remote sensing image to be identified.

[0103] Specifically, in this embodiment, the storage medium may include, but is not limited to, USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks, and other media capable of storing computer programs.

[0104] Another embodiment of the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the method embodiments described above.

[0105] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0106] Specifically, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0107] Receive remote sensing images to be identified captured by remote sensing equipment;

[0108] The remote sensing image to be identified is input into an image classification model that includes a quantum-dilated convolutional neural network and a classifier. The quantum-dilated convolutional neural network is trained to extract target object features based on the pixels of the remote sensing image, and the classifier is trained to predict the type of target object in the remote sensing image based on the target object features.

[0109] The system receives the prediction data output by the image classification model and provides the prediction data as output, wherein the prediction data is used to characterize the type of target object in the remote sensing image to be identified.

[0110] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A remote sensing image quantum recognition method, characterized in that, The method comprises: receiving a remote sensing image to be identified photographed by a remote sensing device; inputting the remote sensing image to be identified into an image classification model comprising a quantum cavity convolutional neural network and a classifier, the quantum cavity convolutional neural network being trained to extract target object features based on pixels of the remote sensing image, and the classifier being trained to predict the category of the target object in the remote sensing image based on the target object features; the loss function for training the quantum cavity convolutional neural network model is: wherein Loss is a loss function, n is the number of remote sensing images used for training, is a true position of the object in the i-th remote sensing image used for training, is a predicted position of the object in the i-th remote sensing image used for training, is a true label probability distribution of the object in the i-th remote sensing image used for training, is a predicted label probability distribution of the object in the i-th remote sensing image used for training; receiving prediction data output by the image classification model, and providing the prediction data as output, the prediction data being used to represent the category of the target object in the remote sensing image to be identified; wherein the quantum cavity convolutional neural network comprises an entanglement layer, the entanglement layer comprising a CNOT gate acting on adjacent quantum bits in sequence, a CNOT gate acting on quantum bits at intervals, and a parametric quantum logic gate acting on the quantum bits, the rotation angle of the parametric quantum logic gate being determined based on the loss function for training the quantum cavity convolutional neural network model; wherein adjacent quantum bits include: from the first bit quantum bit to the second last bit quantum bit, in sequence, the next bit quantum bit is adjacent to the quantum bit, and the last bit quantum bit is also adjacent to the first bit quantum bit; the quantum bits at intervals include: from the first bit quantum bit to the second last bit quantum bit, in sequence, the quantum bits at intervals of 1 bit are adjacent to the quantum bits; wherein each group of quantum bits at intervals is acted on twice by the CONT gate, and the control bits of the two quantum bits in the same group are different when they are acted on twice by the CONT gate; wherein the parameters of the parametric quantum logic gate are used to indicate the angles of rotation around the x-axis, y-axis and z-axis.

2. The method of claim 1, wherein, The quantum cavity convolutional neural network comprises an encoding layer for encoding a target pixel to an initial state of a quantum bit, the target pixel being determined from the remote sensing image to be identified based on a preset dilation rate and a convolution kernel, and the initial state being used to extract the target object features of the remote sensing image to be identified.

3. The method of claim 2, wherein, The encoding layer comprises H gates, RY gates and RZ gates acting on the quantum bits in sequence, the rotation parameter of the RY gate being determined based on the inverse trigonometric function value of the target pixel, and the rotation parameter of the RZ gate being determined based on the inverse trigonometric function value of the square of the target pixel.

4. The method of claim 3, wherein, The quantum cavity convolutional neural network further comprises an entanglement layer and a measurement layer, the entanglement layer being used to evolve the initial state to an entangled state comprising target object feature information, and the measurement layer being used to measure the quantum bits to extract the target object features of the remote sensing image to be identified from the entangled state.

5. The method according to any one of claims 1 to 4, wherein Before inputting the remote sensing image to be identified into the image classification model comprising the quantum cavity convolutional neural network and the classifier, the method comprises: performing size transformation and normalization processing on the remote sensing image to be identified to obtain a remote sensing image to be identified allowed to be input into the quantum cavity convolutional neural network.

6. A remote sensing image quantum recognition device, characterized in that, The device comprises: a communication unit for receiving a remote sensing image to be identified photographed by a remote sensing device; The processing unit is configured to input the remote sensing image to be identified into an image classification model including a quantum void convolutional neural network and a classifier, the quantum void convolutional neural network is trained to extract target object features based on pixels of the remote sensing image, and the classifier is trained to predict the category of the target object in the remote sensing image based on the target object features; and the loss function for training the quantum void convolutional neural network model is as follows: wherein Loss is a loss function, n is the number of remote sensing images used for training, is a true position of the object in the i-th remote sensing image used for training, is a predicted position of the object in the i-th remote sensing image used for training, is a true label probability distribution of the object in the i-th remote sensing image used for training, is a predicted label probability distribution of the object in the i-th remote sensing image used for training; The communication unit is further configured to receive the prediction data output by the image classification model and provide the prediction data as an output, the prediction data being used to represent the category of the target object in the remote sensing image to be identified. The quantum void convolutional neural network includes an entanglement layer, the entanglement layer includes a CNOT gate acting on adjacent quantum bits, a CNOT gate acting on quantum bits at intervals, and a parametric quantum logic gate acting on the quantum bits, and the rotation angle of the parametric quantum logic gate is determined based on the loss function for training the quantum void convolutional neural network model; wherein adjacent quantum bits include: from the first bit quantum bit to the second last bit quantum bit, in turn with the next bit quantum bit as adjacent quantum bits, and the last bit quantum bit and the first bit quantum bit are also adjacent quantum bits. The quantum bits at intervals include: from the first bit quantum bit to the second last bit quantum bit, in turn with the quantum bits at intervals of 1 as interval quantum bits. Each group of quantum bits at intervals is acted on twice by the CONT gate, and the control bits of the two quantum bits in the same group are different when the CONT gate acts on them twice. The parameters of the parametric quantum logic gate are used to indicate the angles of rotation around the x-axis, the y-axis and the z-axis.

7. A storage medium, characterized by The storage medium stores a computer program, and the computer program is configured to execute the method described in any one of claims 1-5 when running. 8.An electronic device comprising a memory and a processor, the electronic device comprising: The memory stores a computer program, and the processor is configured to execute the computer program to execute the method described in any one of claims 1-5.

Images