A method for determining the optical axis angle precision of a UAV detector and related apparatus

Abstract

The application provides a method for determining the optical axis angle precision of a UAV detector and related devices, the method comprising: obtaining flight information of a UAV at different times, the flight information comprising one or more of position information, speed information and flight direction information. The flight information is input into a pre-constructed dynamic brain function network model to obtain the optical axis angle precision of the UAV detector. In this way, by constructing a dynamic brain function network model of the flight information of the UAV and the optical axis angle precision of the detector, the optical axis angle precision is directly output from the dynamic brain function network model according to the flight information of the UAV, greatly reducing the technical difficulty of determining the optical axis angle precision. The dynamic brain function network model is trained and has high accuracy. Compared with the optical target method which requires the use of precise measuring instruments and high-precision targets, the use of the dynamic brain function network model has lower cost and can realize low-cost and high-accuracy optical axis angle precision calculation.

Description

Technical Field

[0001] This invention relates to the field of computers, and in particular to a method and related apparatus for determining the optical axis angle accuracy of a drone detector. Background Technology

[0002] With the rapid development of technology, drones have been widely used in various fields, such as agriculture, geological exploration, and environmental monitoring. Drones can carry detection instruments, including various sensors, which can collect data from the target area for analysis. However, drones are affected by various factors during flight, such as airflow, wind direction, and terrain. These factors can cause errors in the data collected by the sensors, thus affecting the accuracy and reliability of the detection instruments.

[0003] Therefore, to improve the accuracy and reliability of the detector, it can be calibrated. One important calibration parameter is the optical axis angle accuracy. The optical axis angle refers to the angle between the center line of the sensor's field of view and the flight direction of the UAV. The size of the optical axis angle determines the sensor's field of view range and resolution, and also affects the measurement accuracy and reproducibility of the detector.

[0004] The commonly used method for measuring optical axis angle accuracy is the optical target method. This method utilizes a specially designed optical target to measure the center line of the field of view of the UAV detector at different angles, thereby calculating the optical axis angle accuracy. However, the optical target method requires precise measuring instruments and high-precision targets, thus incurring high costs and technical difficulties. Furthermore, the measurement accuracy of the optical target method is affected by light propagation, which can easily lead to inaccurate measurement results. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method and related device for determining the optical axis angle accuracy of a UAV detector, which can obtain a relatively accurate optical axis angle accuracy by using a dynamic brain functional network model. Compared with the optical target method for measuring optical axis angle accuracy, the method is less technically difficult and less costly.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] This application provides a method for determining the optical axis angle accuracy of a UAV detector, the method comprising:

[0008] The flight information of the UAV at different times is acquired, and the flight information includes one or more of the following: position information, speed information, and flight direction information;

[0009] The flight information is input into a pre-constructed dynamic brain function network model to obtain the optical axis angle accuracy of the UAV detector; the dynamic brain function network model is trained on a neural network model based on processing temporal relationships.

[0010] Optionally, the dynamic brain functional network model can be constructed as follows:

[0011] Training flight data of the UAV are obtained using at least one of a global positioning system, a barometer, an accelerometer, or a magnetometer;

[0012] The training optical axis angle data of the UAV detector is obtained by using at least one of an inertial measurement unit, a gyroscope, a camera, and a laser rangefinder;

[0013] The initial dynamic brain function network model is trained using the training flight data and the training optical axis angle data to obtain the dynamic brain function network model.

[0014] Optionally, the initial load prediction model includes an input layer, a hidden layer, and an output layer, wherein the hidden layer includes multiple nodes.

[0015] Optionally, the method further includes:

[0016] Identify the key nodes among the multiple nodes included in the hidden layer;

[0017] The flight information of the UAV is adjusted so that the optical axis angle accuracy output by the dynamic brain function network model is obtained using the key node.

[0018] Optionally, determining the key nodes among the multiple nodes included in the hidden layer includes:

[0019] Calculate the node degree of each node, and calculate the Rich-club coefficient of each node based on the node degree;

[0020] Nodes with a Rich-club coefficient greater than a threshold are identified as critical nodes.

[0021] Optionally, the method further includes:

[0022] A loss function is constructed based on the training optical axis angle data and the predicted optical axis angle data, wherein the predicted optical axis angle data is the output result of the initial dynamic brain function network model during the training process.

[0023] The step of training the initial dynamic brain function network model using the training flight data and the training optical axis angle data to obtain the dynamic brain function network model includes:

[0024] Using the training flight data, the training optical axis angle data, and the loss function, the initial dynamic brain function network model is iteratively trained to obtain the dynamic brain function network model.

[0025] Optionally, the loss function is a mean squared error function constructed based on the training optical axis angle data and the predicted optical axis angle data.

[0026] This application also provides a device for determining the optical axis angle accuracy of a drone detector, the device comprising:

[0027] The flight information acquisition unit is used to acquire flight information of the UAV at different times, wherein the flight information includes one or more of position information, speed information and flight direction information;

[0028] The optical axis angle accuracy determination unit is used to input the flight information into a pre-constructed dynamic brain function network model to obtain the optical axis angle accuracy of the UAV detector; the dynamic brain function network model is trained on the basis of a neural network model based on processing temporal relationships.

[0029] This application also provides a device for determining the optical axis angle accuracy of a drone detector, including: a processor, a memory, and a system bus;

[0030] The processor and the memory are connected via the system bus;

[0031] The memory is used to store one or more programs, the one or more programs including instructions, which, when executed by the processor, cause the processor to perform any of the above-described methods for determining the optical axis angle accuracy of the UAV detector.

[0032] This application also provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform any of the above-described methods for determining the optical axis angle accuracy of the UAV detector.

[0033] This application provides a method for determining the optical axis angle accuracy of a drone detector. The method includes: since the drone's flight information affects the subsequent calibration of the detector using the optical axis angle accuracy, flight information of the drone at different times can be obtained. The flight information includes one or more of position information, velocity information, and flight direction information. The flight information is then input into a pre-constructed dynamic brain function network model to obtain the optical axis angle accuracy of the drone detector. The dynamic brain function network model is trained based on a neural network model that processes temporal relationships. In other words, by constructing a dynamic brain function network model of the drone's flight information and the detector's optical axis angle accuracy, the model directly outputs the optical axis angle accuracy based on the drone's flight information, greatly reducing the technical difficulty of determining the optical axis angle accuracy. Furthermore, the dynamic brain function network model is trained and has high accuracy. Compared to the optical target method, which requires precise measuring instruments and high-precision targets, this application utilizes a low-cost dynamic brain function network model, enabling low-cost and high-accuracy optical axis angle accuracy calculation. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart illustrating a method for determining the optical axis angle accuracy of a drone detector according to an embodiment of this application is shown.

[0036] Figure 2 A schematic diagram of the structure of a device for determining the optical axis angle accuracy of a drone detector provided in an embodiment of this application is shown. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] Currently, methods for measuring optical axis angle accuracy include static and dynamic measurement methods. Dynamic measurement methods involve collecting data during the actual flight of the UAV and calculating the optical axis angle accuracy by analyzing the sensor measurement results. Static measurement methods primarily use dedicated calibration devices to calibrate the detector, calculating the optical axis angle accuracy by measuring the relative positional relationship between the calibration device and the sensor.

[0039] The commonly used static measurement method for optical axis angle accuracy is the optical target method. Before using this method, a special target needs to be prepared. The target typically consists of multiple specific patterns, providing multiple different measurement points, and a corresponding coordinate system needs to be marked on it. During measurement, the UAV detector needs to be accurately installed at the center of the target, ensuring its optical axis is perpendicular to the target plane. Then, using a specific instrument, such as a goniometer or telescope, the center line of the detector's field of view at different angles is measured along the coordinate axes on the target, and the corresponding coordinates and angle data are recorded. Finally, the optical axis angle accuracy is calculated by processing and analyzing the measurement data. Specifically, the method for calculating the optical axis angle accuracy involves first associating the measurement points with the target coordinate system, then calculating the theoretical position of each measurement point, and comparing the theoretical position with the actual measured position to calculate the optical axis angle accuracy. For higher precision measurements, more complex algorithms such as multi-point calibration or nonlinear fitting can be used.

[0040] However, the optical target method requires sophisticated measuring instruments and high-precision targets, thus incurring high costs and technical difficulties. Furthermore, the manufacturing and use of the target may introduce errors, reducing the accuracy and reliability of the measurement results.

[0041] Furthermore, the optical target method is a method for measuring the shape and size of an object's surface using optical equipment. This method primarily involves projecting light onto the object being measured and then calculating the object's size and shape by observing the reflected or refracted light. Therefore, the measurement accuracy of the optical target method is affected by light propagation, such as by physical effects like light diffusion and refraction. These physical effects can cause deviations in light propagation, thus affecting the accuracy of the measurement results. The optical target method also presents problems when measuring irregularly shaped objects: because the surfaces of irregularly shaped objects lack obvious geometric features, it may be impossible to determine the measurement position and direction when using the optical target method, leading to inaccurate results. The optical target method also requires attaching special targets to the surface of the object being measured. These targets not only increase measurement costs but may also damage or contaminate the object's surface, thus affecting the reliability of the measurement results.

[0042] In other words, using the optical target method to measure the accuracy of the optical axis angle can easily lead to inaccurate measurement results.

[0043] Based on this, this application provides a method for determining the optical axis angle accuracy of a UAV detector. By constructing a dynamic brain function network model of the UAV's flight information and the detector's optical axis angle accuracy, the method directly outputs the optical axis angle accuracy based on the UAV's flight information using this dynamic brain function network model. This greatly reduces the technical difficulty of determining the optical axis angle accuracy. Furthermore, the dynamic brain function network model is trained and has high accuracy. Compared to the optical target method, which requires precise measuring instruments and high-precision targets, this application uses a dynamic brain function network model at a lower cost, enabling low-cost and high-accuracy optical axis angle accuracy calculation.

[0044] To better understand the technical solution and effects of this application, the specific embodiments will be described in detail below with reference to the accompanying drawings.

[0045] First Embodiment

[0046] refer to Figure 1 The diagram shows a flowchart of a method for determining the optical axis angle accuracy of a UAV detector according to an embodiment of this application. The method includes the following steps:

[0047] S101 acquires flight information of the drone at different times.

[0048] In the embodiments of this application, since the detector is mounted on a drone, the drone's flight information will affect the subsequent calibration of the detector. Therefore, in order to obtain calibration parameters for calibrating the detector, the drone's flight information at different times can be acquired. The detector includes a calibrator, and the calibration parameters include the calibrator's optical axis angle accuracy. The flight information may include one or more of position information, velocity information, and flight direction information. The flight information may also include altitude information, rotation angle information, angular velocity information, acceleration information, and gravity information, etc.

[0049] Specifically, multiple sensors can be mounted on the drone to collect its flight information. For example, at least one of the following sensors can be mounted on the drone: a Global Positioning System (GPS), a barometer, an accelerometer, or a magnetometer, to obtain flight information.

[0050] As an example, a barometer can be used to collect altitude and speed information of a drone.

[0051] As another example, accelerometers can be used to collect acceleration and gravity information of drones.

[0052] As another example, magnetometers can be used to collect the direction and heading information of drones.

[0053] As another example, GPS can be used to collect the location, speed, and flight direction information of drones.

[0054] In other words, multiple sensors can collect various types of information from the drone, and the collected data overlaps. This overlapping data can then be used to cross-verify the drone's flight information, thereby improving the accuracy of the flight information.

[0055] It should be noted that other sensors can also be mounted on the drone, and the optical axis angle accuracy of the detector can be calculated using the data collected by these sensors. For example, other sensors may include at least one of an inertial measurement unit (IMU), a gyroscope, a camera, and a laser rangefinder.

[0056] As an example, IMUs can be used to collect acceleration and angular velocity information of drones.

[0057] As another example, gyroscopes can be used to collect the speed and rotation angle information of a drone.

[0058] As another example, cameras and laser rangefinders can be used to collect information on the positional and directional differences between the drone and the detector.

[0059] The optical axis angle accuracy of the detector can be calculated from the data collected by the aforementioned sensors.

[0060] It should be noted that not all sensors used to calculate the optical axis angle accuracy may be equipped in actual drone flights, which could lead to the inability to calculate the optical axis angle accuracy. Furthermore, the calculation of the optical axis angle accuracy based on the collected data has a significant lag, which cannot meet the timely feedback requirements of the optical axis angle accuracy during actual drone flights.

[0061] In practical applications, the data collected by sensors can be cleaned and processed to improve its usability.

[0062] S102 inputs flight information into a pre-built dynamic brain function network model to obtain the optical axis angle accuracy of the UAV detector.

[0063] In the embodiments of this application, after obtaining the flight information of the UAV at different times through S101, the flight information can be input into the pre-constructed dynamic brain function network model. The dynamic brain function network model outputs the optical axis angle accuracy of the detector, thereby realizing the direct acquisition of the optical axis angle accuracy using the flight information and the dynamic brain function network model, which greatly reduces the technical difficulty of determining the optical axis angle accuracy.

[0064] It should be noted that the Dynamic Brain Functional Network (DNB) model is a network model describing the interactions between neurons in the human brain and can be applied to dynamic problems in machine learning. The DNB model can transform time-series data into a graphical structure, allowing for data analysis using graphical analytics techniques. Flight information, being spatiotemporal relational data and also time-series data, can therefore be analyzed using the DNB model. Furthermore, the optical axis angle accuracy is affected by the UAV's flight information, making it also related to spatiotemporal relationships. Therefore, the DNB method can be used to construct a spatiotemporal relationship model between the UAV and the detection device, thus describing the complex spatiotemporal interactions between them.

[0065] To enable the use of a dynamic brain functional network model to describe the complex spatiotemporal interaction between the UAV and the detector, a dynamic brain functional network model can be trained based on a neural network model that processes temporal relationships. The specific training process will be described in the second embodiment.

[0066] In summary, the method for determining the optical axis angle accuracy of a UAV detector provided in this application constructs a dynamic brain function network model of the UAV's flight information and the detector's optical axis angle accuracy. This model directly outputs the optical axis angle accuracy based on the UAV's flight information, greatly reducing the technical difficulty of determining the optical axis angle accuracy. The dynamic brain function network model is trained and has high accuracy. Compared to the optical target method, which requires precise measuring instruments and high-precision targets, this application uses a low-cost dynamic brain function network model, enabling low-cost and high-accuracy optical axis angle accuracy calculation.

[0067] Second Embodiment

[0068] This embodiment will describe the construction process of the dynamic brain functional network model mentioned in the above embodiments. Specifically, it may include the following steps S201-S203:

[0069] S201: Acquire training flight data of the UAV using at least one of a global positioning system, a barometer, an accelerometer, or a magnetometer.

[0070] S202, using at least one of an inertial measurement unit, a gyroscope, a camera, and a laser rangefinder to acquire the training optical axis angle data of the UAV detector.

[0071] In the embodiments of this application, in order to construct a dynamic brain function network model, it is necessary to collect a large amount of training sample data, that is, to collect training flight data of drones and training optical axis angle data of detectors, so as to use the training flight data and training optical axis angle data of detectors to train and obtain a dynamic brain function network model.

[0072] Specifically, as in the first embodiment, at least one of the following sensors can be used to acquire flight information: a global positioning system, a barometer, an accelerometer, or a magnetometer. At least one of the following sensors can be used to acquire optical axis angle accuracy: an inertial measurement unit, a gyroscope, a camera, and a laser rangefinder. Therefore, at least one of the following sensors can be used to acquire training flight data, and at least one of the following sensors can be used to acquire training optical axis angle data.

[0073] While collecting training flight data and training optical axis angle data, an initial dynamic brain function network model can also be constructed and the model parameters initialized.

[0074] As an example, the initial dynamic brain functional network model can be represented as:

[0075]

[0076] in, Let f(y(t),t) represent the rate of change at time t, and let f(y(t),t) represent the change of state y(t) at time t.

[0077] In the embodiments of this application, the initial dynamic brain function network model can be trained using training flight data and training optical axis angle data. During training, the training optical axis angle data is used as the rate of change, the training flight data is used as the state, and time is used as the change quantity.

[0078] It should be noted that the initial dynamic brain function network model can be a multi-layer neural network model, where each layer represents a time step, and each time step can consist of an input layer, a hidden layer, and an output layer, where the hidden layer includes multiple nodes.

[0079] The input layer takes the drone's flight information as input, and the hidden layer learns from the training flight data and training optical axis angle data to capture the long-term dependencies between these two types of time-series data. The output layer outputs the optical axis angle accuracy, thereby obtaining the optical axis angle accuracy of the drone detection calibrator.

[0080] The specific structure of the hidden layer can be a recurrent neural network (RNN) or a long short-term memory network (LSTM). These two structures can adaptively process data with temporal relationships and capture long-term dependencies in time series data.

[0081] As one possible implementation, the hidden layer adopts a recurrent neural network structure. During the training process, the hidden state of the recurrent neural network structure needs to be calculated using training flight data and training optical axis angle data.

[0082] As an example, the hidden states of a recurrent neural network structure can be calculated using the following formula:

[0083] h t =f(W hh h t-1 +W xh x t +b h )

[0084] Where, x t This represents the input at time t, h t H represents the hidden state at time t. t-1 W represents the hidden state at time t-1. hh W represents the weight matrix between hidden states. xh b represents the weight matrix between the input layer and the hidden layer. h This represents the bias term of the hidden layer. In other words, the hidden state at time t is affected by the hidden state at time t-1.

[0085] In the embodiments of this application, when training an initial dynamic brain function network model using training flight data and training optical axis angle data to obtain a dynamic brain function network model, a loss function can be constructed to guide the training of the initial dynamic brain function network model, thereby improving the output accuracy of the final dynamic brain function network model.

[0086] Backpropagation can be used to train an initial dynamic brain function network model to minimize the error between the predicted and actual outputs. The predicted output is the predicted optical axis angle data, which represents the output of the initial dynamic brain function network model during training. The actual output is the training optical axis angle data, which is the data used to train the initial dynamic brain function network model. In other words, the loss function can be constructed based on the training and predicted optical axis angle data. Therefore, using the training flight data, the training optical axis angle data, and the loss function, the initial dynamic brain function network model can be iteratively trained to obtain the dynamic brain function network model. The output of the dynamic brain function network model trained using this loss function has high accuracy.

[0087] As an example, the mean squared error function, constructed from the training and predicted optical axis angle data, can be used as the loss function. The formula for the mean squared error function is as follows:

[0088]

[0089] Where N represents the number of samples, y i Indicates the actual output. This indicates the predicted output.

[0090] Third Embodiment

[0091] This embodiment will detail how to improve the accuracy of optical axis angle precision using a dynamic brain functional network model, thereby improving the accuracy of the detector. Specifically, it may include the following steps S301-S303:

[0092] S301, Identify the key nodes among the multiple nodes included in the hidden layer.

[0093] In the embodiments of this application, the Rich-club analysis method is a method for evaluating the importance and connection strength of nodes in complex networks. Rich-club analysis can be used to discover rich clubs within a network, each comprising multiple key nodes, and to assess the influence and control these key nodes have over the entire network. In dynamic brain functional networks, Rich-club analysis can identify which nodes in the hidden layer have a significant impact on UAV flight and detector operation, guiding subsequent optimization and improvements.

[0094] In other words, the Rich-club analysis method can be used to identify key nodes among the multiple nodes included in the hidden layer, and these key nodes can then be used to optimize the structure of the dynamic brain functional network model and assist in actual flight.

[0095] Specifically, the node degree of each node in the hidden layer can be calculated. Node degree refers to the number of connections each node has with other nodes. Node degree is directly proportional to the number of connections; that is, the higher the number of connections, the greater the node degree. Nodes with higher node degrees have stronger connections with other nodes and greater control over the network.

[0096] After calculating the degree of each node, the Rich-club coefficient of each node can be calculated. The Rich-club coefficient can be used to identify key nodes among multiple nodes. The Rich-club coefficient represents the strength and influence of connections between nodes with high degrees in a network. Typically, node degree and the Rich-club coefficient are directly proportional; that is, the Rich-club coefficient increases with increasing node degree. Rich-club analysis provides a deeper understanding of the network structure and relationships. It considers the growth trend of node degree and identifies concentrated connections among nodes with high degrees. Using the Rich-club coefficient helps determine which nodes not only have high degrees but also form important "rich clubs" with more significant control and influence over the entire network. In other words, Rich-club analysis provides a more comprehensive network insight to guide optimization and improvement.

[0097] After obtaining the Rich-club coefficient, nodes with higher Rich-club coefficients can be identified as critical nodes. Specifically, nodes with Rich-club coefficients greater than a threshold can be identified as critical nodes, where the threshold can be determined based on the actual situation.

[0098] S302, adjust the UAV's flight information so that the optical axis angle accuracy of the dynamic brain function network model output is obtained using key nodes.

[0099] In the embodiments of this application, nodes with high Rich-club coefficients are considered the most important nodes in the network through Rich-club analysis. These nodes are identified as key nodes, and their influence and control over the entire network are evaluated. Key nodes can help optimize the network structure and operation methods, improve the optical axis angle accuracy of the UAV detector calibrator, and thus better meet the needs of practical applications.

[0100] The dynamic brain functional network model and Rich-club analysis method can be applied to practical operations. Through spatiotemporal observation and feedback adjustments, the optical axis angle accuracy of the UAV detector calibrator can be continuously improved.

[0101] As one possible approach, by using a dynamic brain functional network model and the Rich-club analysis method, it is possible to identify which nodes have a significant impact on UAV flight and the operation of the detection instrument, and these nodes are designated as key nodes. Using these key nodes, the UAV's flight information can be adjusted so that the optical axis angle accuracy output by the dynamic brain functional network model is the same as that obtained using the key nodes. This further improves the accuracy of the output optical axis angle, thereby enhancing the accuracy of the detection instrument.

[0102] As another possible approach, by using a dynamic brain functional network model and the Rich-club analysis method, it is possible to identify which nodes have a significant impact on the optical axis angle accuracy and designate these nodes as key nodes. The position and attitude of the detector relative to the UAV can then be adjusted, specifically the positional and directional differences between the detector and the UAV, so that the optical axis angle accuracy output by the dynamic brain functional network model is obtained using the key nodes.

[0103] As another possible implementation method, the Rich-club analysis method can also optimize the data acquisition and processing process, remove data noise and redundant information, improve data quality and accuracy, and optimize the structure of dynamic brain functional network models, thereby improving the efficiency and accuracy of output optical axis angle precision.

[0104] In summary, based on the dynamic brain functional network model and the Rich-club analysis method, not only can the accuracy of the detector be improved, but it can also guide the flight and operation of UAVs, thereby improving the performance and efficiency of UAVs.

[0105] Fourth embodiment

[0106] Based on the method for determining the optical axis angle accuracy of the UAV detector provided in the above embodiments, this application also provides a device for determining the optical axis angle accuracy of the UAV detector.

[0107] See Figure 2 This is a schematic diagram of the structure of a device for determining the optical axis angle accuracy of a drone detector provided in this embodiment. The device 200 includes:

[0108] The flight information acquisition unit 210 is used to acquire flight information of the UAV at different times, wherein the flight information includes one or more of position information, speed information and flight direction information;

[0109] The optical axis angle accuracy determination unit 220 is used to input the flight information into a pre-constructed dynamic brain function network model to obtain the optical axis angle accuracy of the UAV detector; the dynamic brain function network model is trained on the basis of a neural network model based on processing temporal relationships.

[0110] In one implementation of this embodiment, the device 200 further includes a model building unit, which is used for:

[0111] Training flight data of the UAV are obtained using at least one of a global positioning system, a barometer, an accelerometer, or a magnetometer;

[0112] The training optical axis angle data of the UAV detector is obtained by using at least one of an inertial measurement unit, a gyroscope, a camera, and a laser rangefinder;

[0113] The initial dynamic brain function network model is trained using the training flight data and the training optical axis angle data to obtain the dynamic brain function network model.

[0114] In one implementation of this embodiment, the initial load prediction model includes an input layer, a hidden layer, and an output layer, wherein the hidden layer includes multiple nodes.

[0115] In one implementation of this embodiment, the device 200 further includes an adjustment unit, the adjustment unit being configured to:

[0116] Identify the key nodes among the multiple nodes included in the hidden layer;

[0117] The flight information of the UAV is adjusted so that the optical axis angle accuracy output by the dynamic brain function network model is obtained using the key node.

[0118] In one implementation of this embodiment, the adjustment unit is used for:

[0119] Calculate the node degree of each node, and calculate the Rich-club coefficient of each node based on the node degree;

[0120] Nodes with a Rich-club coefficient greater than a threshold are identified as critical nodes.

[0121] In one implementation of this embodiment, the device 200 further includes a loss function construction unit, which is configured to:

[0122] A loss function is constructed based on the training optical axis angle data and the predicted optical axis angle data, wherein the predicted optical axis angle data is the output result of the initial dynamic brain function network model during the training process.

[0123] The model building unit is used for:

[0124] Using the training flight data, the training optical axis angle data, and the loss function, the initial dynamic brain function network model is iteratively trained to obtain the dynamic brain function network model.

[0125] In one implementation of this embodiment, the loss function is a mean squared error function constructed based on the training optical axis angle data and the predicted optical axis angle data.

[0126] Furthermore, this application embodiment also provides a device for determining the optical axis angle accuracy of a drone detector, including: a processor, a memory, and a system bus;

[0127] The processor and the memory are connected via the system bus;

[0128] The memory is used to store one or more programs, the one or more programs including instructions, which, when executed by the processor, cause the processor to perform any of the above-described methods for determining the optical axis angle accuracy of the UAV detector.

[0129] Furthermore, this application embodiment also provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to execute any of the above-described methods for determining the optical axis angle accuracy of the UAV detector.

[0130] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0131] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0132] The above description is merely a preferred embodiment of this application. Although this application has disclosed preferred embodiments above, it is not intended to limit this application. Any person skilled in the art can make many possible variations and modifications to the technical solutions of this application using the methods and techniques disclosed above, or modify them into equivalent embodiments with equivalent changes, without departing from the scope of the technical solutions of this application. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of this application without departing from the content of the technical solutions of this application shall still fall within the protection scope of the technical solutions of this application.

Claims

1. A method for determining the optical axis angle accuracy of a UAV detector, characterized in that, The method includes: The flight information of the UAV at different times is acquired, and the flight information includes one or more of the following: position information, speed information, and flight direction information; The flight information is input into a pre-constructed dynamic brain function network model to obtain the optical axis angle accuracy of the UAV detector; the dynamic brain function network model is trained on a neural network model based on processing temporal relationships; the dynamic brain function network model includes an input layer, a hidden layer and an output layer, and the hidden layer includes multiple nodes; The method further includes: Identify the key nodes among the multiple nodes included in the hidden layer; The flight information of the UAV is adjusted so that the optical axis angle accuracy output by the dynamic brain function network model is obtained using the key node; The key nodes among the multiple nodes included in the hidden layer are: Calculate the node degree of each node, and calculate the Rich-club coefficient of each node based on the node degree; Nodes with a Rich-club coefficient greater than a threshold are identified as critical nodes.

2. The method according to claim 1, characterized in that, The dynamic brain functional network model is constructed as follows: Training flight data of the UAV are obtained using at least one of a global positioning system, a barometer, an accelerometer, or a magnetometer; The training optical axis angle data of the UAV detector is obtained by using at least one of an inertial measurement unit, a gyroscope, a camera, and a laser rangefinder; The initial dynamic brain function network model is trained using the training flight data and the training optical axis angle data to obtain the dynamic brain function network model.

3. The method according to claim 2, characterized in that, The method further includes: A loss function is constructed based on the training optical axis angle data and the predicted optical axis angle data, wherein the predicted optical axis angle data is the output result of the initial dynamic brain function network model during the training process. The step of training the initial dynamic brain function network model using the training flight data and the training optical axis angle data to obtain the dynamic brain function network model includes: Using the training flight data, the training optical axis angle data, and the loss function, the initial dynamic brain function network model is iteratively trained to obtain the dynamic brain function network model.

4. The method according to claim 3, characterized in that, The loss function is a mean squared error function constructed based on the training optical axis angle data and the predicted optical axis angle data.

5. A device for determining the optical axis angle accuracy of a drone detector, characterized in that, The device includes: The flight information acquisition unit is used to acquire flight information of the UAV at different times, wherein the flight information includes one or more of position information, speed information and flight direction information; The optical axis angle accuracy determination unit is used to input the flight information into a pre-constructed dynamic brain function network model to obtain the optical axis angle accuracy of the UAV detector; the dynamic brain function network model is trained on a neural network model based on processing temporal relationships; the dynamic brain function network model includes an input layer, a hidden layer and an output layer, and the hidden layer includes multiple nodes; The device further includes an adjustment unit, the adjustment unit being configured to: Identify the key nodes among the multiple nodes included in the hidden layer; The flight information of the UAV is adjusted so that the optical axis angle accuracy output by the dynamic brain function network model is obtained using the key node; The adjustment unit is specifically used for: Calculate the node degree of each node, and calculate the Rich-club coefficient of each node based on the node degree; Nodes with a Rich-club coefficient greater than a threshold are identified as critical nodes.

6. A device for determining the optical axis angle accuracy of a drone detector, characterized in that, include: Processor, memory, system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, the one or more programs including instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a terminal device, cause the terminal device to perform the method described in any one of claims 1-4.

Images