A quantum-classical feature fusion identification method and system for high-altitude operation safety monitoring
By using a quantum classical feature fusion identification method, the problem of feature fracture and hidden deformation caused by steel frame obstruction at high-altitude operation sites has been solved, enabling efficient monitoring and early warning of risks in high-altitude operations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHENGDU ANSWER INFORMATION TECH CO LTD
- Filing Date
- 2026-05-26
- Publication Date
- 2026-06-26
AI Technical Summary
At high-altitude work sites, the image features of critical safety nodes are broken due to the obstruction of steel structure components. Traditional methods are prone to feature semantic drift when dealing with severe obstruction, and conventional time series models are insufficient in their ability to warn of hidden deformations.
A quantum classical feature fusion recognition method is adopted. By acquiring images of high-altitude operation sites, the spatial coordinates and local image features of safety nodes are extracted, a spatial constraint matrix is constructed for sparsity constraint, quantum final state is generated and drift signal is measured, and a probability distribution decision of high-altitude operation risk is generated by combining an adaptive fusion strategy.
It effectively reduces the effective dimension of the parameter space, improves the ability to monitor hidden risks under complex working conditions, and significantly enhances the early warning capability for progressive deformation.
Smart Images
Figure CN122289879A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of quantum computing and computer vision technology, and more specifically, to a quantum classical feature fusion recognition method and system for high-altitude operation safety monitoring. Background Technology
[0002] High-altitude work sites often suffer from obstructions caused by steel structural components. These obstructions can lead to fragmented image features at critical safety points. Traditional methods based on classical neural networks are prone to semantic drift when dealing with severe obstruction. Furthermore, subtle deformations such as slow slippage of safety buckles and shifts in the body's center of gravity occur with minimal changes in pixel space, making conventional time-series models insufficient for warning of such progressive risks. While recent research has attempted to incorporate quantum machine learning, existing methods often use fixed-topology quantum circuits, failing to integrate the constraints of real-world physical distances into the entangled structure design. Moreover, the fusion of quantum and classical branches is relatively simple, often involving linear splicing, resulting in limited robustness under complex conditions. Therefore, there is an urgent need in this field for a risk identification method that can adapt to physical obstruction and capture minute deformations. Summary of the Invention
[0003] The purpose of this invention is to provide a quantum classical feature fusion identification method and system for high-altitude operation safety monitoring, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:
[0004] Firstly, this application provides a quantum classical feature fusion and identification method for high-altitude operation safety monitoring, including:
[0005] Acquire images of high-altitude operations, extract the spatial coordinates and corresponding local image features of multiple safety nodes from the images, and generate dynamic entanglement parameters based on the local image features;
[0006] The local image features are normalized, and the normalized local image features are encoded into initial qubits;
[0007] A spatial constraint matrix is constructed based on the distance between each spatial coordinate, and the spatial constraint matrix is used to perform sparsification constraints on the dynamic entanglement parameters.
[0008] The initial qubits and the entanglement parameters after sparsification are input into the quantum circuit. The entanglement gates in the quantum circuit are controlled by the spatial constraint matrix to drive the quantum circuit to undergo unitary evolution that is isomorphic to the spatial positional relationship of each safe node, thereby generating the quantum final state.
[0009] The quantum final state is measured to obtain a quantum characteristic vector, and a drift signal is generated based on the difference of the quantum characteristic vector at different times;
[0010] The quantum feature vector is adaptively fused with the local image features, and the fused features and the drift signal are input into the evaluation network to generate a probability distribution decision vector for high-altitude operation risks.
[0011] Secondly, this application also provides a quantum classical feature fusion identification system for high-altitude operation safety monitoring, including:
[0012] The image acquisition module is used to acquire images of the high-altitude operation site, extract the spatial coordinates of multiple safety nodes and corresponding local image features from the images, and generate dynamic entanglement parameters based on the local image features;
[0013] A quantum state preparation module is used to normalize the local image features and encode the normalized local image features into initial qubits;
[0014] The constraint construction module is used to construct a spatial constraint matrix based on the distance between each spatial coordinate, and to use the spatial constraint matrix to perform sparsification constraints on the dynamic entanglement parameters.
[0015] The quantum evolution module is used to input the initial qubit and the entanglement parameters after sparsification constraint into the quantum circuit, and use the spatial constraint matrix to control the entanglement gate in the quantum circuit to drive the quantum circuit to perform unitary evolution isomorphic to the spatial position relationship of each safe node, and generate the quantum final state;
[0016] The measurement module is used to measure the quantum final state, obtain the quantum feature vector, and generate a drift signal based on the difference of the quantum feature vector at different times;
[0017] The fusion evaluation module is used to adaptively fuse the quantum feature vector with the local image features, and input the fused features and the drift signal into the evaluation network to generate a probability distribution decision vector for the risks of high-altitude operations.
[0018] Thirdly, this application also provides a quantum classical feature fusion identification device for high-altitude operation safety monitoring, comprising:
[0019] Memory, used to store computer programs;
[0020] A processor is used to implement the steps of the quantum classical feature fusion recognition method for high-altitude operation safety monitoring when executing the computer program.
[0021] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described quantum classical feature fusion identification method for high-altitude operation safety monitoring.
[0022] The beneficial effects of this invention are as follows:
[0023] 1. This invention constructs a spatial constraint matrix by extracting the spatial coordinates of secure nodes and sparses the dynamic entanglement parameters, so that the activation state of the entanglement gate in the quantum circuit and the actual spatial position relationship of each secure node form an isomorphic mapping. This dynamically generates a sparse entangled structure consistent with the physical topology of the field in a high-dimensional Hilbert space, effectively reducing the effective dimension of the parameter space, alleviating the quantum gradient vanishing problem, and reducing noise.
[0024] 2. This invention obtains quantum feature vectors by measuring the quantum final state, and generates drift signals by utilizing the difference in expected values of quantum feature vectors at adjacent times, thus transforming the gradual deformation that is not easily perceptible to the naked eye during high-altitude operations into quantifiable drift signals.
[0025] 3. This invention adopts an adaptive fusion strategy to dynamically adjust the fusion weights of quantum features and classical image features. Under difficult working conditions such as steel frame obstruction, it can adaptively enhance the contribution of quantum topological features and input the fused features and drift signals into the evaluation network, which significantly improves the early warning capability of hidden risks.
[0026] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description
[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0028] Figure 1 This is a schematic diagram of the quantum classical feature fusion and recognition method for high-altitude operation safety monitoring as described in this embodiment of the invention;
[0029] Figure 2 This is a schematic diagram of the quantum classical feature fusion and recognition system for high-altitude operation safety monitoring as described in this embodiment of the invention.
[0030] Figure 3 This is a schematic diagram of the quantum classical feature fusion identification device for high-altitude operation safety monitoring as described in an embodiment of the present invention.
[0031] Marked in the image:
[0032] 800. Quantum-classical feature fusion identification device for safety monitoring of high-altitude operations; 801. Processor; 802. Memory; 803. Multimedia component; 804. I / O interface; 805. Communication component. Detailed Implementation
[0033] 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.
[0034] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0035] Example 1:
[0036] This embodiment provides a quantum classical feature fusion recognition method for safety monitoring of high-altitude operations, applicable to high-altitude operation sites where monitoring cameras face severe obstruction by steel structural components.
[0037] See Figure 1 The figure shows that this method includes:
[0038] S1. Acquire images of the high-altitude operation site, extract the spatial coordinates of multiple safety nodes and their corresponding local image features from the images, and generate dynamic entanglement parameters based on the local image features;
[0039] This embodiment uses high-altitude maintenance of power transmission towers as an application scenario. The monitoring camera collects video streams from the work site in real time and performs the following operations on the image frames at the current moment:
[0040] First, a classic feature extraction network is used to process the image frame at the current moment to locate and identify N key safety nodes. Specifically, the safety nodes include at least one of the following: the safety belt hook point, the center of gravity of the worker's skeleton, and the connection point of the fall arrestor. Preferably, the classic feature extraction network employs a lightweight attention mechanism convolutional neural network.
[0041] Furthermore, the spatial coordinates of each security node in the pixel coordinate system are extracted sequentially. ,in Centered on each critical security node, local image feature vectors are extracted using the classic feature extraction network. The local image feature vectors of N security nodes Constructing a local image feature set .
[0042] Furthermore, the statistical distribution of local features in the current image frame is extracted as scene context features. And mapped to dynamic entanglement parameters. In this embodiment, the dynamic entanglement parameters for The matrix, where the first... The element represents the first element. The qubit and the first The original entanglement strength between qubits.
[0043] Based on the above embodiments, this method further includes:
[0044] S2. Perform L2 norm normalization on the local image features, and encode the normalized local image features into initial qubits;
[0045] Specifically, for the first Local image feature vectors of each security node Its normalized eigenvector is :
[0046] ;
[0047] Furthermore, amplitude encoding is used to prepare the initial qubits of the normalized local image features. .
[0048] Based on the above embodiments, this method further includes:
[0049] S3. Construct a spatial constraint matrix based on the distance between each spatial coordinate, and use the spatial constraint matrix to perform sparsification constraints on the dynamic entanglement parameters;
[0050] Specifically, step S3 includes:
[0051] S31. Calculate the Euclidean distance between the spatial coordinates of any two safe nodes:
[0052] S32. Perform an exponential decay transformation on each of the Euclidean distances to obtain the corresponding constraint weights, and arrange all constraint weights in the order of safe node pairing to form a spatial constraint matrix:
[0053] ;
[0054] In the formula, The attenuation coefficient is... , The first , The spatial coordinates of each secure node For constraint weights.
[0055] It can be seen that the closer two safe nodes are in the image space, the better. The value approaches 1 when the two safe nodes are far apart in the image space or when there is a steel frame occlusion. Approaching 0.
[0056] By all constitute a Spatial constraint matrix diagonal elements .
[0057] Furthermore, using the aforementioned spatial constraint matrix For the dynamic entanglement parameters Apply Hadamard product sparsity constraints:
[0058] ;
[0059] In the formula, This represents the entanglement parameters after sparsification constraints.
[0060] According to the Hadamard product sparsity constraint, when two safe nodes are too far apart or are obstructed by a steel frame, When it approaches 0, the corresponding entanglement parameter It was also forcibly compressed to near zero.
[0061] Based on the above embodiments, this method further includes:
[0062] S4. Input the initial qubit and the entanglement parameters after sparsification constraint into the quantum circuit, and use the spatial constraint matrix to control the entanglement gate in the quantum circuit to drive the quantum circuit to perform unitary evolution isomorphic to the spatial position relationship of each safe node, and generate the quantum final state;
[0063] In this embodiment, the quantum circuit adopts a variable quantum circuit structure comprising L layers, each layer consisting of alternating single-bit rotation layers and controlled entanglement layers, wherein the number of qubits in the circuit is equal to the number of secure nodes N;
[0064] Specifically, the single-qubit rotation layer is configured with a Pauli rotation gate for each qubit, and the rotation gate is configured with globally learnable angle parameters. It is used to perform fundamental feature rotations on the initial qubits.
[0065] Specifically, the controlled entanglement layer utilizes a controlled phase gate (CPHASE) as the entanglement gate for any two qubits. The rotation angle of the controlled phase gate is determined by the entanglement parameters after sparsification constraints. Decide.
[0066] Specifically, step S4 includes:
[0067] S41. Obtain the matrix elements in the spatial constraint matrix corresponding to any two safe nodes;
[0068] S42. Determine whether the matrix element is less than a preset threshold:
[0069] If so, then set the rotation angle of the controlled phase gate between the corresponding two qubits in the quantum circuit to 0, so that the entanglement gate between the corresponding two qubits is decoupled;
[0070] In this embodiment, the preset threshold The noise level can be dynamically adjusted based on the actual depolarization noise level of the hardware. Preferably, .
[0071] Specifically, if Then two qubits Rotation angle of the CPHASE door between The value is forcibly set to 0. Since a rotation angle of 0 for the CPHASE gate is equivalent to an identity transformation, the entanglement gate is effectively decoupled, and entanglement is no longer generated between the two qubits. Then keep That is, the original value of the entanglement parameter after the sparsification constraint is used as the rotation angle.
[0072] In this embodiment, the entangled topology in the quantum circuit is dynamically pruned using the aforementioned control method: only qubits corresponding to spatially close and unobstructed safe nodes retain strong entanglement connections; while entanglement between distant or obstructed node pairs is severed. This makes the topology of the entangled network is isomorphic to the physical topology of the work site.
[0073] S43. Configure the rotation angles of all the entangled gates after adjustment into the quantum circuit, drive the quantum circuit to perform unitary evolution, so as to dynamically generate a sparse entangled network isomorphic to the physical topology of the work site in quantum space, and obtain the quantum final state. ;
[0074] Specifically, the rotation angle of each CPHASE gate is configured to the corresponding position in the variable quantum circuit. Then, the initial qubits are... The input quantum circuit sequentially passes through L layers of single-qubit rotation and controlled entanglement. In each layer, a single-qubit rotation gate rotates the quantum state according to the current global learning parameters, and then a CPHASE gate applies entanglement between the corresponding qubit pairs according to the configured rotation angle. After L layers of evolution, the circuit outputs the final quantum state. .
[0075] Based on the above embodiments, this method further includes:
[0076] S5. Measure the quantum final state to obtain a quantum characteristic vector, and generate a drift signal based on the difference of the quantum characteristic vector at different times;
[0077] Specifically, step S5 includes:
[0078] S51. Measure the single-bit expected value of each qubit in the quantum final state along the PauliZ operator:
[0079] ;
[0080] In the formula, Indicates the first The expected value of a single qubit. This represents the Pauli Z operator.
[0081] S52. Arrange all individual expected values in the same order as the secure nodes to obtain the quantum eigenvector. ;
[0082] S53. Calculate the average value of all components of the quantum eigenvector at the current moment, as the expected value at the current moment;
[0083] In this embodiment, the time corresponding to the current video frame is ,but The quantum eigenvector at time t is ,calculate The arithmetic mean of all components yields the expected value at the current moment. ;
[0084] The expected value This reflects the overall magnetization of the entire quantum system along the Pauli Z direction at the current moment. When the work site is under normal conditions... When the value stabilizes near the baseline, and the safety buckle slowly slips off, an abnormal evolution occurs in the global topological implicit association, leading to... Drifting has occurred.
[0085] S54. Obtain the expected value of the quantum feature vector at the adjacent previous time step, calculate the absolute value of the difference between the expected value at the current time step and the expected value at the previous time step, and obtain the drift signal;
[0086] Specifically, reading Expected value at time If the current frame is the first frame, then initialize. .
[0087] for The drift signal is obtained by calculating the absolute value of the difference between the expected value at the current time and the expected value at the previous time. .
[0088] Based on the above embodiments, this method further includes:
[0089] S6. Adaptively fuse the quantum feature vector with the local image features, and input the fused features and the drift signal into the evaluation network to generate a probability distribution decision vector for high-altitude operation risks;
[0090] Specifically, S6 includes:
[0091] S61. Transfer the local image features Mapping to spatial dimensions The first mapping feature is obtained. The first mapping feature The dimension is ,in, This refers to the number of secure nodes.
[0092] S62. Transfer the quantum eigenvector Mapping to spatial dimensions The second mapping feature is obtained. ;
[0093] S63. Calculate the inner product correlation between the first mapping feature and the second mapping feature, and generate dynamic fusion weights based on the inner product correlation:
[0094] ;
[0095] In the formula, Indicates dynamic fusion weights. Represents a sigmoid function. For dimension parameters, The weight matrix is a learnable bilinear weight matrix. This is for gating bias.
[0096] S64. The first mapping feature and the second mapping feature are weighted and summed according to the dynamic fusion weights to obtain the fused features:
[0097] ;
[0098] In the formula, Indicates the characteristics after fusion. This is element-wise multiplication.
[0099] Classic features become unreliable when severe occlusion occurs. When the quantum spectral density approaches 1, the fusion features are mainly contributed by quantum features; when the classical features are clear, The value approaches 0, indicating that the fusion features are mainly contributed by classical features.
[0100] Specifically, step S6 further includes:
[0101] S65. The fused features are concatenated with the drift signal in the time dimension to form a time-series input feature;
[0102] Specifically, the features after fusing the current frame Flattened into a one-dimensional vector, and compared with the drift signal of the current frame. The input features are concatenated to form the input features for a single time step. For multiple consecutive frames, the input features are stacked in chronological order to obtain temporal input features.
[0103] S66. Input the temporal input features into the long short-term memory unit to obtain the hidden state vector;
[0104] In this embodiment, the evaluation network employs Long Short-Term Memory (LSTM) units. The LSM units capture long-term dependencies in temporal features through their internal forget gate, input gate, and output gate structures. After iterations through all time steps, the hidden state vector for the last time step is output.
[0105] S67. Input the hidden state vector into the fully connected layer, and output the probability distribution decision vector through the classification function. Each component of the probability distribution decision vector corresponds to the probability of a risk category.
[0106] Based on the above embodiments, this method further includes;
[0107] S7. Joint training via backpropagation;
[0108] Specifically, step S7 includes:
[0109] S71. Construct a total loss function, and embed orthogonal constraint loss terms into the total loss function;
[0110] In this embodiment, the total loss function comprises two terms: the first term is the task loss, used to measure the difference between the risk probability distribution output by the evaluation network and the true label, which can be achieved using cross-entropy loss; the second term is the orthogonal constraint loss term. The design goal of the orthogonal constraint loss term is to measure the distance between local image features and quantum feature vectors. By embedding the orthogonal constraint loss term into the total loss function, the accuracy of risk classification is optimized while ensuring that the two types of features are kept far apart in the feature space.
[0111] S72. Utilize the orthogonal constraint loss term to increase the distance metric between the local image features and the quantum feature vector;
[0112] Specifically, during training, the gradient of the total loss function with respect to the learnable parameters of each layer is calculated using the backpropagation algorithm, and the distance between the two features in the feature space gradually increases.
[0113] S73. Calculate the gradient based on the total loss function, and update the network parameters through backpropagation to constrain the local image features to be separated from the quantum feature vector in the feature space.
[0114] In summary, this invention significantly reduces the dimensionality of the parameter space and improves the reliability of hidden risk monitoring under complex working conditions by deeply coupling physical space and quantum topology.
[0115] Example 2:
[0116] like Figure 2 As shown, this embodiment provides a quantum classical feature fusion recognition system for high-altitude operation safety monitoring, the system comprising:
[0117] The image acquisition module is used to acquire images of the high-altitude operation site, extract the spatial coordinates of multiple safety nodes and corresponding local image features from the images, and generate dynamic entanglement parameters based on the local image features;
[0118] A quantum state preparation module is used to normalize the local image features and encode the normalized local image features into initial qubits;
[0119] The constraint construction module is used to construct a spatial constraint matrix based on the distance between each spatial coordinate, and to use the spatial constraint matrix to perform sparsification constraints on the dynamic entanglement parameters.
[0120] The quantum evolution module is used to input the initial qubit and the entanglement parameters after sparsification constraint into the quantum circuit, and use the spatial constraint matrix to control the entanglement gate in the quantum circuit to drive the quantum circuit to perform unitary evolution isomorphic to the spatial position relationship of each safe node, and generate the quantum final state;
[0121] The measurement module is used to measure the quantum final state, obtain the quantum feature vector, and generate a drift signal based on the difference of the quantum feature vector at different times;
[0122] The fusion evaluation module is used to adaptively fuse the quantum feature vector with the local image features, and input the fused features and the drift signal into the evaluation network to generate a probability distribution decision vector for the risks of high-altitude operations.
[0123] Based on the above embodiments, the safety node includes at least one of the following: safety belt hook point, worker's skeletal center of gravity point, and fall arrestor connection point.
[0124] Based on the above embodiments, the constraint construction module includes:
[0125] The distance calculation unit is used to calculate the Euclidean distance between the spatial coordinates of any two safe nodes;
[0126] The matrix construction unit is used to perform an exponential decay transformation on each of the Euclidean distances to obtain the corresponding constraint weights, and arrange all the constraint weights in the order of safe node pairing to form a spatial constraint matrix.
[0127] Based on the above embodiments, the quantum circuit adopts a variable quantum circuit structure, which includes a single-bit rotation layer and a controlled entanglement layer; wherein, the single-bit rotation layer uses a Pauli rotation gate to perform characteristic rotation on the initial quantum bit, and the controlled entanglement layer uses a controlled phase gate as the entanglement gate.
[0128] Based on the above embodiments, the quantum evolution module includes:
[0129] An element acquisition unit is used to acquire matrix elements in the spatial constraint matrix corresponding to any two safe nodes;
[0130] The decoupling judgment unit is used to determine whether the matrix element is less than a preset threshold. When the matrix element is less than the preset threshold, the rotation angle of the controlled phase gate between the corresponding two qubits in the quantum circuit is set to 0, so that the entanglement gate between the corresponding two qubits is decoupled.
[0131] An evolution execution unit is used to configure the rotation angles of all entangled gates after adjustment into the quantum circuit, drive the quantum circuit to perform unitary evolution, so as to dynamically generate a sparse entangled network that is isomorphic to the physical topology of the work site in quantum space, and obtain the quantum final state.
[0132] Based on the above embodiments, the measurement module includes:
[0133] A measurement unit is used to measure the individual expected value of each qubit in the quantum final state along the Pauli Z operator;
[0134] A feature-forming unit is used to arrange all individual expected values in the same order as the secure node to obtain the quantum feature vector;
[0135] The expectation calculation unit is used to calculate the average value of all components of the quantum feature vector at the current moment, as the expectation value at the current moment;
[0136] The drift calculation unit is used to obtain the expected value of the quantum feature vector at the adjacent previous time step, calculate the absolute value of the difference between the expected value at the current time step and the expected value at the previous time step, and obtain the drift signal.
[0137] Based on the above embodiments, the fusion evaluation module includes:
[0138] The first mapping unit is used to map the local image features to a preset dimension space to obtain the first mapped features;
[0139] The second mapping unit is used to map the quantum feature vector to the preset dimension space to obtain the second mapping feature;
[0140] The weight generation unit is used to calculate the inner product correlation between the first mapping feature and the second mapping feature, and generate dynamic fusion weights based on the inner product correlation.
[0141] The weighted fusion unit is used to perform a weighted summation of the first mapping feature and the second mapping feature according to the dynamic fusion weight to obtain the fused feature.
[0142] Based on the above embodiments, the evaluation network includes a long short-term memory unit; the fusion evaluation module further includes:
[0143] The temporal splicing unit is used to splice the fused features with the drift signal in the time dimension to form temporal input features;
[0144] A timing processing unit is used to input the timing input features into the long short-term memory unit to obtain a hidden state vector;
[0145] The decision output unit is used to input the hidden state vector into the fully connected layer and output the probability distribution decision vector through a classification function.
[0146] Based on the above embodiments, the system further includes a joint training module for performing joint training of backpropagation; the joint training module includes:
[0147] A loss construction unit is used to construct a total loss function, in which an orthogonal constraint loss term is embedded;
[0148] A distance constraint unit is used to increase the distance metric between the local image features and the quantum feature vector by utilizing the orthogonal constraint loss term.
[0149] The parameter update unit is used to calculate the gradient based on the total loss function and update the network parameters through backpropagation to constrain the local image features to be separated from the quantum feature vector in the feature space.
[0150] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.
[0151] Example 3:
[0152] Corresponding to the above method embodiments, this embodiment also provides a quantum classical feature fusion identification device for high-altitude operation safety monitoring. The quantum classical feature fusion identification device for high-altitude operation safety monitoring described below and the quantum classical feature fusion identification method for high-altitude operation safety monitoring described above can be referred to in correspondence with each other.
[0153] Figure 3 This is a block diagram illustrating a quantum classical feature fusion identification device 800 for high-altitude operation safety monitoring, according to an exemplary embodiment. Figure 3 As shown, the quantum classical feature fusion identification device 800 for high-altitude operation safety monitoring may include: a processor 801 and a memory 802. The quantum classical feature fusion identification device 800 for high-altitude operation safety monitoring may also include one or more of the following: a multimedia component 803, an I / O interface 804, and a communication component 805.
[0154] The processor 801 controls the overall operation of the quantum classical feature fusion identification device 800 for high-altitude work safety monitoring, to complete all or part of the steps in the aforementioned quantum classical feature fusion identification method for high-altitude work safety monitoring. The memory 802 stores various types of data to support the operation of the quantum classical feature fusion identification device 800 for high-altitude work safety monitoring. This data may include, for example, instructions for any application or method operating on the quantum classical feature fusion identification device 800 for high-altitude work safety monitoring, as well as application-related data, such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the quantum classical feature fusion identification device 800 for high-altitude operation safety monitoring and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, and an NFC module.
[0155] In an exemplary embodiment, the quantum classical feature fusion identification device 800 for high-altitude operation safety monitoring can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned quantum classical feature fusion identification method for high-altitude operation safety monitoring.
[0156] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the quantum classical feature fusion and recognition method for high-altitude work safety monitoring described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above. These program instructions may be executed by the processor 801 of the quantum classical feature fusion and recognition device 800 for high-altitude work safety monitoring to complete the quantum classical feature fusion and recognition method for high-altitude work safety monitoring described above.
[0157] Example 4:
[0158] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the quantum classical feature fusion identification method for high-altitude operation safety monitoring described above.
[0159] A readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of the quantum classical feature fusion and recognition method for safety monitoring of high-altitude operations as described in the above method embodiment are implemented.
[0160] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.
[0161] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0162] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A quantum-classical feature fusion and recognition method for safety monitoring of high-altitude operations, characterized in that, include: Acquire images of high-altitude operations, extract the spatial coordinates and corresponding local image features of multiple safety nodes from the images, and generate dynamic entanglement parameters based on the local image features; The local image features are normalized, and the normalized local image features are encoded into initial qubits; A spatial constraint matrix is constructed based on the distance between each spatial coordinate, and the spatial constraint matrix is used to perform sparsification constraints on the dynamic entanglement parameters. The initial qubits and the entanglement parameters after sparsification are input into the quantum circuit. The entanglement gates in the quantum circuit are controlled by the spatial constraint matrix to drive the quantum circuit to undergo unitary evolution that is isomorphic to the spatial positional relationship of each safe node, thereby generating the quantum final state. The quantum final state is measured to obtain a quantum characteristic vector, and a drift signal is generated based on the difference of the quantum characteristic vector at different times; The quantum feature vector is adaptively fused with the local image features, and the fused features and the drift signal are input into the evaluation network to generate a probability distribution decision vector for high-altitude operation risks.
2. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, The safety node includes at least one of the following: safety belt hook point, worker's skeletal center of gravity, and fall arrestor connection point.
3. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, A spatial constraint matrix is constructed based on the distances between each spatial coordinate, including: Calculate the Euclidean distance between the spatial coordinates of any two safe nodes; An exponential decay transformation is performed on each Euclidean distance to obtain the corresponding constraint weights. All constraint weights are then arranged in the order of safe node pairing to form a spatial constraint matrix.
4. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, The quantum circuit adopts a variable quantum circuit structure, which includes a single-bit rotation layer and a controlled entanglement layer; wherein, the single-bit rotation layer uses a Pauli rotation gate to perform characteristic rotation on the initial quantum bit, and the controlled entanglement layer uses a controlled phase gate as the entanglement gate.
5. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 4, characterized in that, The entanglement gates in the quantum circuit are controlled using the spatial constraint matrix to drive the quantum circuit to undergo unitary evolution isomorphic to the spatial positional relationship of each safe node, generating a quantum final state, including: Obtain the matrix elements in the spatial constraint matrix corresponding to any two safe nodes; Determine whether the matrix element is less than a preset threshold: If so, then set the rotation angle of the controlled phase gate between the corresponding two qubits in the quantum circuit to 0, so that the entanglement gate between the corresponding two qubits is decoupled; The rotation angles of all entangled gates after adjustment are configured into the quantum circuit, and the quantum circuit is driven to perform unitary evolution to dynamically generate a sparse entangled network that is isomorphic to the physical topology of the work site in quantum space, thereby obtaining the quantum final state.
6. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, Measuring the quantum final state to obtain a quantum eigenvector, and generating a drift signal based on the difference in the quantum eigenvector at different times, including: Measure the single-bit expected value of each qubit in the quantum final state along the Pauli Z operator; Arrange all the individual expected values in the same order as the secure nodes to obtain the quantum feature vector; Calculate the average value of all components of the quantum eigenvector at the current moment, and use it as the expected value at the current moment; The expected value of the quantum feature vector at the adjacent previous time step is obtained, and the absolute value of the difference between the expected value at the current time step and the expected value at the previous time step is calculated to obtain the drift signal.
7. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, Adaptive fusion of the quantum feature vector and the local image features includes: The local image features are mapped to a preset dimensional space to obtain the first mapped features; The quantum feature vector is mapped to the preset dimensional space to obtain the second mapped feature; Calculate the inner product correlation between the first mapping feature and the second mapping feature, and generate dynamic fusion weights based on the inner product correlation; The first mapping feature and the second mapping feature are weighted and summed according to the dynamic fusion weight to obtain the fused feature.
8. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, The evaluation network includes long short-term memory units; The fused features and the drift signal are input into the evaluation network to generate a probability distribution decision vector for high-altitude operation risks, including: The fused features are concatenated with the drift signal in the time dimension to form a time-series input feature; The temporal input features are input into the Long Short-Term Memory unit to obtain the hidden state vector; The hidden state vector is input into a fully connected layer, and the probability distribution decision vector is output through a classification function.
9. The quantum classical feature fusion and recognition method for high-altitude operation safety monitoring according to claim 1, characterized in that, It also includes joint training via backpropagation: Construct a total loss function and embed orthogonal constraint loss terms into the total loss function; The orthogonal constraint loss term is used to increase the distance metric between the local image features and the quantum feature vector; The gradient is calculated based on the total loss function, and the network parameters are updated through backpropagation to constrain the local image features to be separated from the quantum feature vector in the feature space.
10. A quantum-classical feature fusion recognition system for safety monitoring of high-altitude operations, characterized in that, include: The image acquisition module is used to acquire images of the high-altitude operation site, extract the spatial coordinates of multiple safety nodes and corresponding local image features from the images, and generate dynamic entanglement parameters based on the local image features; A quantum state preparation module is used to normalize the local image features and encode the normalized local image features into initial qubits; The constraint construction module is used to construct a spatial constraint matrix based on the distance between each spatial coordinate, and to use the spatial constraint matrix to perform sparsification constraints on the dynamic entanglement parameters. The quantum evolution module is used to input the initial qubit and the entanglement parameters after sparsification constraint into the quantum circuit, and use the spatial constraint matrix to control the entanglement gate in the quantum circuit to drive the quantum circuit to perform unitary evolution isomorphic to the spatial position relationship of each safe node, and generate the quantum final state; The measurement module is used to measure the quantum final state, obtain the quantum feature vector, and generate a drift signal based on the difference of the quantum feature vector at different times; The fusion evaluation module is used to adaptively fuse the quantum feature vector with the local image features, and input the fused features and the drift signal into the evaluation network to generate a probability distribution decision vector for the risks of high-altitude operations.