Unmanned aerial vehicle detection method, device and equipment based on 5G network and storage medium

Abstract

The invention relates to the technical field of signal detection, and discloses an unmanned aerial vehicle detection method, device and equipment based on a 5G network and a storage medium, the method is executed by a DAF responsible for data processing in an unmanned aerial vehicle detection system, the system further comprises an access and mobility management function AMF, and the method comprises the steps of receiving target sensor data corresponding to a target area, the target sensor data comprises a target RF signal and target image data corresponding to the target RF signal, processing the target RF signal through a random forest model to obtain a first recognition result, processing the target image data through an image recognition model to obtain a second recognition result, and outputting the second recognition result. And when both the first identification result and the second identification result indicate that the unmanned aerial vehicle exists in the target area, sending warning information to a third-party platform. According to the method, the random forest is combined with the image recognition model to perform anomaly analysis on the RF signal and the image data in the sensor data, so that high-precision real-time monitoring of the unmanned aerial vehicle is realized.

Description

Technical Field

[0001] The present invention relates to the field of signal detection technology, and in particular to a 5G network-based drone detection method, device, equipment and storage medium. Background Art

[0002] With the rapid development of drone technology, the application of drones in many fields such as industry, agriculture, logistics, etc. has become more and more common. However, this widespread use has also brought many security and regulatory challenges. The illegal intrusion or uncontrolled flight of drones may lead to serious safety hazards. Traditional drone detection methods mainly rely on technologies such as radar monitoring, visual recognition, and radio frequency RF signal analysis. However, these methods usually have problems such as insufficient detection accuracy, susceptibility to multipath interference, and slow response speed in complex environments, making it difficult to effectively deal with diverse drone threats. In addition, with the popularization of 5G networks, its high speed, low latency and wide coverage technical characteristics provide new technical support for building more efficient drone detection and monitoring systems, which makes the use of 5G networks for real-time drone detection and monitoring a hot topic in current research and application, and promotes the development of more accurate and real-time drone monitoring technology.

[0003] However, there are many shortcomings in drone detection and monitoring technology. For example, the detection accuracy of a single sensor is easily affected by environmental interference and equipment noise. The multi-sensor fusion method is too simple and fails to fully explore the complex relationship between different data sources, resulting in poor detection results. In addition, the existing system has a slow response speed when processing large-scale real-time data, which makes it difficult to meet the real-time monitoring needs in a highly dynamic environment. The generalization ability of the model is limited, and it is easy to overfit or fail to detect accurately in new environments or new data. Summary of the invention

[0004] In view of this, the present invention provides a drone detection method, device, equipment and storage medium based on a 5G network to solve the problems of low detection efficiency and limited model generalization ability of traditional drones due to rough data fusion solutions.

[0005] In a first aspect, the present invention provides a drone detection method based on a 5G network, the method is executed by a DAF responsible for data processing in a drone detection system, and the system also includes an access and mobility management function AMF; the method includes:

[0006] receiving target sensor data corresponding to the target area; the target sensor data including a target RF signal and target image data corresponding to the target RF signal;

[0007] Processing the target RF signal through a random forest model to obtain a first recognition result;

[0008] Processing the target image data through an image recognition model to obtain a second recognition result;

[0009] When both the first recognition result and the second recognition result indicate that a drone exists in the target area, a warning message is sent to the third-party platform.

[0010] The 5G network-based drone detection method provided in an embodiment of the present invention processes RF signals and image data respectively through a random forest model and an image recognition model. The RF signal can capture the radio activity of the drone, while the image data can provide intuitive visual information. The combination of the two can analyze the target sensor data of the target area from multiple angles and dimensions, thereby improving the accuracy of drone detection and effectively reducing false positives and false negatives in drone monitoring.

[0011] In an optional implementation, before receiving the target sensor data corresponding to the target area, the method further includes:

[0012] receiving sensor data sent by NWDAF; the sensor data includes RF signal and image data corresponding to the RF signal;

[0013] Fill missing values ​​in sensor data and perform data fusion to obtain valid sensor data;

[0014] Based on the effective sensor data, the random forest model and the image recognition model are trained respectively to obtain the target random forest model and the target image recognition model respectively.

[0015] The drone detection method based on 5G network provided by the embodiment of the present invention can ensure the quality and integrity of the training data by filling missing values ​​and fusing data of the sensor data provided by NWDAF. At the same time, by fusing and training data from different sources, the model can be better adapted to complex and changing environmental conditions. After preprocessing the data, the random forest model and image recognition model can be trained with valid sensor data, which can ensure that the model learning is more accurate and comprehensive, and help improve the accuracy and robustness of the model in drone detection tasks.

[0016] In an optional implementation, a random forest model and an image recognition model are trained based on valid sensor data to obtain a target random forest model and a target image recognition model, respectively, including:

[0017] Perform random forest model training based on RF signals in valid sensor data to obtain a target random forest model;

[0018] The image recognition model is trained based on the image data in the effective sensor data to obtain a target image recognition model.

[0019] The drone detection method based on 5G network provided by the embodiment of the present invention performs special training on effective sensor data to ensure that the random forest model and the image recognition model optimize the recognition of the characteristics of RF signals and image data respectively, and detect the drone from the perspective of RF signals and image data signals respectively. Finally, the recognition results of the two are comprehensively judged, which can significantly reduce the false positives and false negatives that may be generated by a single model. This multi-model fusion method can analyze sensor data more comprehensively, improve the accuracy and reliability of drone detection, and also improve the generalization ability and adaptability of the model.

[0020] In an optional implementation, performing missing value filling processing on sensor data and performing data fusion to obtain valid sensor data includes:

[0021] Perform median interpolation filling on the sensor data and remove noise to obtain smooth sensor data;

[0022] The smoothed sensor data is normalized and time-aligned to obtain standard sensor data;

[0023] The RF signal and image data at the same time in the standard sensor data are fused to obtain the comprehensive sensor data;

[0024] The integrated sensor data that meets the preset value range is confirmed as valid sensor data.

[0025] The 5G network-based drone detection method provided by the embodiment of the present invention can reduce random fluctuations and outliers in the data by filling missing values ​​and removing noise on the sensor data, making the data more authentic and reliable; at the same time, standardization is performed so that data from different sources and dimensions can be unified on the same scale for comparison and analysis, thereby improving the comparability and consistency of the data. In addition, time alignment is performed on the sensor data to ensure that the RF signal and image data at the same time point can correspond one-to-one, which facilitates the fusion of the two to obtain comprehensive sensor data, thereby realizing multi-angle and multi-dimensional drone detection and analysis.

[0026] In an optional implementation, a random forest model and an image recognition model are trained based on valid sensor data to obtain a target random forest model and a target image recognition model, respectively, including:

[0027] The valid sensor data is divided into training set data, validation set data and test set data according to the preset ratio;

[0028] Based on the training set data, the random forest model and the image recognition model are trained respectively to obtain the target random forest model and the target image recognition model respectively.

[0029] The 5G network-based drone detection method provided by an embodiment of the present invention can ensure that the model can be fully trained and verified at different stages by dividing the effective sensor data into different sets, thereby improving the generalization ability of the model. At the same time, through the participation of the verification set data, the model can be discovered and adjusted in time to prevent overfitting.

[0030] In an optional embodiment, the method further includes:

[0031] If the first recognition result indicates that there is no drone in the target area, and the second recognition result indicates that there is a drone in the target area, a first cross-validation is performed using an infrared sensor;

[0032] If the first cross-validation result still has anomalies, the detection threshold of the image recognition model is adjusted based on the second recognition result and the cross-validation result, and sent to NWDAF for storage.

[0033] The 5G network-based drone detection method provided by an embodiment of the present invention can reconfirm the target from the infrared radiation dimension by introducing an infrared sensor for cross-validation, thereby increasing the reliability and accuracy of detection. When the cross-validation result also has an abnormality, the detection threshold of the image recognition model is adjusted according to the second recognition result and the cross-validation result, and the parameter setting of the model can be optimized to make it more adaptable to the current environment and the characteristics of the target, thereby improving the accuracy of detection.

[0034] In an optional embodiment, the method further includes:

[0035] If the first recognition result indicates that a drone exists in the target area, and the second recognition result indicates that a drone does not exist in the target area, a spectrum analysis is performed to obtain a spectrum analysis result;

[0036] The spectrum analysis result and the RF signal information corresponding to the first recognition result are sent to the NWDAF for storage.

[0037] The drone detection method based on 5G network provided by the embodiment of the present invention can perform more in-depth processing of RF signals through spectrum analysis to verify the preliminary recognition results of the random forest model. This double verification mechanism helps to reduce false alarms and improve the accuracy of drone detection. At the same time, the spectrum analysis results and RF signal information are stored in NWDAF, providing a rich data source for subsequent data analysis and mining.

[0038] In an optional implementation, the random forest model and the image recognition model are trained based on the valid sensor data to obtain the target random forest model and the target image recognition model, respectively, and further include:

[0039] A second cross-validation is performed on the target random forest model and the target image recognition model, and parameters of the target random forest model and the target image recognition model are adjusted based on the second cross-validation results.

[0040] The 5G network-based drone detection method provided in an embodiment of the present invention divides the data set into different subsets, and repeatedly trains the model and evaluates the performance through the second cross-validation, so as to more comprehensively evaluate the performance of the model on different data sets. For example, the number of decision trees in the random forest model, the feature randomness, and the learning rate and number of iterations in the image recognition model can be adjusted to further optimize the performance of the model and make it more adaptable to the needs of practical applications, which helps to reduce the overfitting of the model to specific training data and improve the generalization performance of the model.

[0041] In a second aspect, the present invention provides a drone detection device based on a 5G network, the device comprising:

[0042] A receiving module, used for receiving target sensor data corresponding to the target area; the target sensor data includes a target RF signal and target image data corresponding to the target RF signal;

[0043] A first recognition module, used to process the target RF signal through a random forest model to obtain a first recognition result;

[0044] A second recognition module is used to process the target image data through an image recognition model to obtain a second recognition result;

[0045] The warning sending module is used to send a warning message to a third-party platform when the first recognition result and the second recognition result both indicate that a drone exists in the target area.

[0046] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the 5G network-based drone detection method of the above-mentioned first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0047] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the 5G network-based drone detection method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0048] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions, which are used to enable a computer to execute the 5G network-based drone detection method of the above-mentioned first aspect or any corresponding embodiment. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 is a system architecture diagram of a 5G network-based drone detection system according to an embodiment of the present invention;

[0051] Figure 2 is a flow chart of a 5G network-based drone detection method according to an embodiment of the present invention;

[0052] Figure 3 is an interactive schematic diagram of a drone detection method based on a 5G network according to an embodiment of the present invention;

[0053] Figure 4 is an interactive schematic diagram of another drone detection method based on a 5G network according to an embodiment of the present invention;

[0054] Figure 5 is a structural block diagram of a drone detection device based on a 5G network according to an embodiment of the present invention;

[0055] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0057] The more common drone detection and monitoring methods currently face some problems that cannot be ignored in practice. For example, when using a single sensor, the detection accuracy is easily affected by environmental interference and equipment noise; and the multi-sensor fusion method is too simple and fails to fully explore the complex relationship between different data sources, resulting in poor detection results. In addition, the existing system has a slow response speed when processing large-scale real-time data, which makes it difficult to meet the real-time monitoring needs in a highly dynamic environment. The generalization ability of the model is limited, and it is easy to overfit or fail to detect accurately in new environments or new data.

[0058] According to an embodiment of the present invention, an embodiment of a drone detection method based on a 5G network is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0059] This method is applied to the drone detection system based on 5G network, please refer to Figure 1 , Figure 1 The overall architecture of the system includes: data preparation and preprocessing module, model training module, model verification module, real-time monitoring and judgment module, exception handling and result storage module. Specifically:

[0060] Data preprocessing module: This module is responsible for collecting historical data from the sensor network and cleaning, converting and integrating the data to prepare for subsequent model training. It includes a data collection submodule, which is used to obtain RF signals and image data from the sensor network as the basic data for model training. The data cleaning and conversion submodule is used to handle missing values, remove noise, and standardize and time synchronize the data to ensure data consistency and quality. The data integration and partitioning submodule is used to integrate different types of data into comprehensive data points, check data quality, and partition the data into training sets, validation sets, and test sets.

[0061] Model training module: This module is mainly responsible for training models using preprocessed data, including random forest and image recognition models. Preferably, the image recognition model is a CNN model for subsequent real-time monitoring and judgment. It includes a random forest model training submodule, which is used to initialize the parameters of the random forest model, generate multiple decision trees, and determine the final prediction results through ensemble learning. The CNN model training submodule is used to define the convolutional neural network structure, pass image data layer by layer, and optimize the network parameters through multiple iterations.

[0062] Model validation module: This module is used to evaluate the performance of the training model and ensure the generalization ability and stability of the model on different data sets. It includes the following submodules: Cross-validation submodule, which is used to perform k-fold cross-validation using the validation set to evaluate the generalization performance of the model and avoid overfitting. Performance evaluation submodule, which evaluates the actual performance of the model through evaluation indicators (such as accuracy, stability, etc.) to ensure its suitability for real-time monitoring.

[0063] Real-time monitoring and judgment module: This module is responsible for processing real-time sensor service requests and using trained models to make abnormal judgments on real-time data. It includes a service request processing submodule, which is used to receive sensor service requests sent by third-party platforms, and AMF receives and forwards them to gNB to ensure the execution of sensor operations. The sensor operation and data collection submodule is used for the sensor network to perform sensor operations according to requests, collect real-time data and encrypt and return it. The real-time judgment submodule is used to decrypt real-time data and use random forest and CNN models to make abnormal judgments on sensor data.

[0064] The exception handling and result storage module is used to handle detected exceptions, record and analyze exception information, feedback warnings and store results to ensure continuous monitoring and optimization of the system. It includes an exception recording and analysis submodule, which is used to record specific information of abnormal data, analyze and eliminate environmental interference, and optimize the model when necessary. The warning information feedback submodule is used to feedback major abnormal information to the third-party platform to trigger the corresponding warning mechanism. The data storage and monitoring submodule is used to store the exception handling results in the database, continue to monitor the system, and update the monitoring status when there is no abnormality.

[0065] In this embodiment, a drone detection method based on a 5G network is provided. Figure 2 is a flow chart of a drone detection method based on a 5G network according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0066] Step S201, receiving target sensor data corresponding to the target area.

[0067] Specifically, the target sensor data includes the target RF signal and the target image data corresponding to the target RF signal. Therefore, it is necessary to reasonably arrange sensors that can capture RF signals and image data in or around the target area, including radio frequency scanners for capturing RF signals and high-definition cameras for capturing image data. It is also necessary to ensure that all sensors are connected to the drone detection system through the 5G network.

[0068] Furthermore, AMF receives the sensing service request from the third-party platform. The third-party platform encrypts the sensing service request, which generally includes the sensing area, sensing time, sensing frequency and duration, using the AES-256 algorithm, randomly generates a key and an initial vector, and then groups the data to be transmitted, and finally encrypts each group of data separately. The encrypted data, key and initial vector are sent to the AMF of the 5G core network.

[0069] Further, AMF forwards the sensor request to gNB. After receiving the encrypted sensor data, gNB performs decryption operation, merges relevant data, and transmits it to the sensor network. The sensor network starts the sensing operation according to the parameters it contains, activates the RF sensor and image capture device, and starts to continuously or periodically scan the radio spectrum in the target area, captures the RF signal that may come from the drone and captures the image data of the target area, and transmits the collected target sensor data to gNB. After gNB encrypts the target sensor data with AES-256, it sends it to AMF and then to DAF.

[0070] Step S202: Process the target RF signal through a random forest model to obtain a first recognition result.

[0071] Specifically, DAF receives and decrypts the target RF signal data from the target sensor data of gNB. Then, DAF requests and loads the trained random forest model from NWDAF, which is trained based on a large amount of drone RF signal data.

[0072] Furthermore, the preprocessed RF signal features are input into the random forest model for inference calculation. The random forest model classifies or regresses the input features through the voting mechanism of multiple decision trees to determine whether there is an RF signal of a drone. The random forest model outputs the first recognition result, that is, the judgment of whether there is a drone in the target area.

[0073] Step S203: Process the target image data through an image recognition model to obtain a second recognition result.

[0074] Specifically, DAF receives and decrypts the target image data from the target sensor data of gNB. Then, DAF requests and loads the trained image recognition model from NWDAF.

[0075] Furthermore, the preprocessed target image data is input into the image recognition model, which will automatically extract key features from the image, such as the shape, color, and texture of the drone. Using the extracted features, the image recognition model will classify the image to determine whether it contains a drone. Based on the classification result of the model, the second recognition result is output, that is, the determination of whether there is a drone in the target area.

[0076] Step S204: When both the first recognition result and the second recognition result indicate that there is a drone in the target area, a warning message is sent to the third-party platform.

[0077] Specifically, DAF first collects the first recognition result and the second recognition result obtained by the random forest model and the image recognition model in steps S202 and S203. According to the preset logical judgment rules, if both recognition results reach or exceed the preset threshold, that is, both indicate that there is a drone in the target area, DAF will generate a warning message and send it to the third-party platform.

[0078] Optionally, the warning message contains detailed information such as the drone’s location, detection time, and detected target sensor data.

[0079] In summary, the drone detection method based on 5G network provided by the embodiment of the present invention is as follows: Figure 3 As shown in the figure, the third-party platform encrypts the sensor service request containing specific requirements to ensure the security of the data during transmission. The encrypted data is sent to the access and mobility management function AMF of the 5G core network. As a key component of the core network, AMF is responsible for receiving and forwarding this data to the base station gNB. After receiving the encrypted sensor data, gNB performs decryption, merges and organizes the decrypted data, and then passes it to the sensor network.

[0080] The sensor network then starts to collect key data such as video, images, and RF signals of the target area using various sensors previously deployed, including radio frequency scanners for capturing RF signals and high-definition cameras for capturing image data. After the collection is completed, the sensor network transmits the data back to the gNB, which again encrypts the data with AES-256 to ensure the security of the data during transmission to the AMF. The AMF acts as a transit station and forwards the encrypted data to the data processing function DAF.

[0081] As the core processing unit of the drone detection system, DAF decrypts and preprocesses the data, and then requests the trained random forest and image processing models, such as the convolutional neural network CNN model, VGG model or LeNet-5 model, from the network data analysis function NWDAF. After obtaining the model, it is loaded on DAF, and the preprocessed target sensor data is used for real-time judgment. The random forest model analyzes the RF signal and identifies the radio activity characteristics of the drone; while the image recognition model processes the image data and captures the visual features of the drone, and obtains the first recognition result and the second recognition result respectively. When the recognition results of the two models both point to the existence of a drone in the target area, DAF will immediately trigger the warning mechanism and send a warning message to the third-party platform through a secure communication channel so that timely countermeasures can be taken.

[0082] In some optional implementations, before receiving the target sensor data corresponding to the target area, the method further includes:

[0083] receiving sensor data sent by NWDAF; the sensor data includes RF signal and image data corresponding to the RF signal;

[0084] Fill missing values ​​in sensor data and perform data fusion to obtain valid sensor data;

[0085] Based on the effective sensor data, the random forest model and the image recognition model are trained respectively to obtain the target random forest model and the target image recognition model respectively.

[0086] like Figure 4 As shown, the system starts to continuously collect sensor data for the past month from various previously deployed sensors, including radio frequency scanners for capturing RF signals and high-definition cameras for capturing image data. After collection, the data is transmitted to the NWDAF database for storage and passed to the data processing and analysis function DAF in real time.

[0087] After receiving the data, DAF fills in missing values ​​and performs preprocessing for data fusion to ensure data consistency and integrity, and then transmits the preprocessed valid data back to NWDAF for storage to facilitate subsequent management and traceability.

[0088] Next, DAF uses the random forest algorithm to train the model. By initializing the model parameters, multiple decision trees are constructed and trained. Each tree learns independently based on different data subsets. Finally, the prediction results of each decision tree are aggregated and voted to obtain the target random forest model. At the same time, DAF deploys image recognition models, such as CNN models, to train image recognition models. By designing the network architecture, such as the stacking of convolutional layers, pooling layers, and fully connected layers, the model can automatically extract high-level features from image data, and after multiple rounds of iterative optimization, the target image recognition model is obtained to achieve accurate recognition of drone images. Finally, the trained model is returned to NWDAF.

[0089] In some optional implementations, the random forest model and the image recognition model are trained based on the valid sensor data to obtain the target random forest model and the target image recognition model, respectively, including:

[0090] Perform random forest model training based on RF signals in valid sensor data to obtain a target random forest model;

[0091] The image recognition model is trained based on the image data in the effective sensor data to obtain a target image recognition model.

[0092] DAF trains a random forest model based on the RF signal in the effective sensor data. First, initialize the model parameters, such as the number of trees, the depth of the tree, etc.; then, use the training set data to generate multiple decision trees, and train them through decision tree algorithms such as CART, ID3, etc.; finally, integrate the prediction results of each decision tree, such as through a voting mechanism, to determine the final prediction result. After multiple rounds of iteration and optimization, the target random forest model is obtained. At the same time, DAF trains the target image recognition model based on the image data in the effective sensor data. Taking the CNN model as an example, first, define the CNN network structure suitable for drone recognition, including convolution layer, pooling layer, activation function and fully connected layer; then, use the target image data to train the CNN, and adjust the network weights through the back propagation algorithm so that the model can learn the feature representation of the drone image; finally, tune the model based on the training results to avoid overfitting. After sufficient training, the target image recognition model is obtained.

[0093] In some optional implementations, missing value filling processing is performed on sensor data and data fusion is performed to obtain valid sensor data, including:

[0094] Perform median interpolation filling on the sensor data and remove noise to obtain smooth sensor data;

[0095] The smoothed sensor data is normalized and time-aligned to obtain standard sensor data;

[0096] The RF signal and image data at the same time in the standard sensor data are fused to obtain the comprehensive sensor data;

[0097] The integrated sensor data that meets the preset value range is confirmed as valid sensor data.

[0098] The system begins to continuously collect sensor data from the past month from various previously deployed sensors, including radio frequency scanners for capturing RF signals and high-definition cameras for capturing image data. After collection, the data is transmitted to the NWDAF database for storage and passed to the data processing and analysis function DAF in real time.

[0099] DAF preprocesses the data, including: the first step is data cleaning. Specifically, the missing values ​​are processed by filling the missing values ​​with the median filling method; noise is removed by using the weighted moving average method, which means that different weights are given to data at different time points when calculating the average value, and the most recent data is given a higher weight to smooth the data. The second step is data conversion, which is to first perform standard normalization to standardize the data to the same scale. Then time synchronization is performed, and the data at different time points are aligned to a unified time axis using linear interpolation. The third step is data integration. The RF signal and image features at the same time point are fused into a comprehensive data point. Check whether the fused signal strength, image features and radar data are within the expected range (RF signal strength is -100dBm~0dBm, frequency is 300MHz~6GHz, image features are 0~1, and the latitude and longitude range of the position coordinates). If they are not within the expected range, they are deleted and valid sensor data is finally obtained.

[0100] In some optional implementations, the random forest model and the image recognition model are trained based on the valid sensor data to obtain the target random forest model and the target image recognition model, respectively, including:

[0101] The valid sensor data is divided into training set data, validation set data and test set data according to the preset ratio;

[0102] Based on the training set data, the random forest model and the image recognition model are trained respectively to obtain the target random forest model and the target image recognition model respectively.

[0103] The valid sensor data obtained in the above steps are divided into a training set (30%), a validation set (15%) and a test set (15%), and then the processed data is transmitted back to NWDAF for storage.

[0104] Therefore, when DAF performs random forest model training, the model parameters are initialized first, and the number of decision trees (n_estimators = 100), maximum depth (max_depth = null), and minimum number of sample splits (min_samples_split = 2) are set; then the decision tree is generated, and 100 sub-sample sets are randomly sampled from the training data set, and each sub-sample set is used to train a decision tree; the decision tree is trained, and at each node of each tree, a feature subset is randomly selected for splitting, and the best feature points and splitting points are selected to construct the decision tree. The decision trees are then integrated. The prediction results of all trees are voted to determine the final prediction result.

[0105] When training image recognition models in DNF, especially CNN models, it is necessary to divide the training data into small batches, train the models batch by batch, perform multiple iterations, and gradually optimize the model parameters.

[0106] In some optional embodiments, the method further comprises:

[0107] If the first recognition result indicates that there is no drone in the target area, and the second recognition result indicates that there is a drone in the target area, a first cross-validation is performed using an infrared sensor;

[0108] If the first cross-validation result still has anomalies, the detection threshold of the image recognition model is adjusted based on the second recognition result and the cross-validation result, and sent to NWDAF for storage.

[0109] If the first recognition result indicates that there is no drone in the target area, and the second recognition result indicates that there is a drone in the target area, the specific information of the current detection image of NWDAF is first returned, including the image frame, monitoring area and time. Then, the infrared sensor is used for cross-validation. According to the abnormal detection situation, the detection threshold of the CNN model is returned to NWDAF to optimize the model. Finally, the results are stored in the NWDAF database.

[0110] In some optional embodiments, the method further comprises:

[0111] If the first recognition result indicates that a drone exists in the target area, and the second recognition result indicates that a drone does not exist in the target area, a spectrum analysis is performed to obtain a spectrum analysis result;

[0112] The spectrum analysis result and the RF signal information corresponding to the first recognition result are sent to the NWDAF for storage.

[0113] If the first recognition result indicates that there is a drone in the target area, it will first return to NWDAF to record the specific information of the current abnormal signal, including time, signal strength and frequency. Then DAF will use MATLAB to perform spectrum analysis to eliminate environmental interference, and finally return the analysis results and records to the historical database in NWDAF for storage.

[0114] In some optional implementations, the random forest model and the image recognition model are trained based on the valid sensor data to obtain the target random forest model and the target image recognition model, respectively, further comprising:

[0115] A second cross-validation is performed on the target random forest model and the target image recognition model, and parameters of the target random forest model and the target image recognition model are adjusted based on the second cross-validation results.

[0116] Use the validation set to perform k-fold cross validation to evaluate the generalization performance of the model and ensure the stability of the model on different data. Specifically, when the accuracy of the training set is higher than 90% and the accuracy of the validation set is lower than 80%, the target model can be considered overfitting. At this time, the model complexity can be reduced, and the maximum depth of the random forest model can be reduced or the minimum number of sample splits can be increased to prevent the model from overfitting the data.

[0117] When the accuracy of the training set is lower than 70% and the accuracy of the validation set is lower than 70%, the target model can be considered underfitting. At this time, it is necessary to increase the complexity of the model, increase the number of decision trees or increase the maximum depth of the random forest model to improve the expressive power of the model or rebuild the network results, increase the number of network layers or neurons for the CNN model, and appropriately reduce the learning rate.

[0118] In this embodiment, a drone detection device based on a 5G network is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the term "module" may be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0119] This embodiment provides a drone detection device based on a 5G network, such as Figure 5 As shown, including:

[0120] The receiving module 501 is used to receive target sensor data corresponding to the target area; the target sensor data includes a target RF signal and target image data corresponding to the target RF signal.

[0121] The first recognition module 502 is used to process the target RF signal through a random forest model to obtain a first recognition result.

[0122] The second recognition module 503 is used to process the target image data through the image recognition model to obtain a second recognition result.

[0123] The warning sending module 504 is used to send a warning message to a third-party platform when the first recognition result and the second recognition result both indicate that there is a drone in the target area.

[0124] In some optional embodiments, the device further comprises:

[0125] The data receiving module is used to receive the sensor data sent by the NWDAF; the sensor data includes the RF signal and the image data corresponding to the RF signal.

[0126] The preprocessing module is used to fill missing values ​​in sensor data and perform data fusion to obtain valid sensor data.

[0127] The training module is used to train the random forest model and the image recognition model based on the effective sensor data, and obtain the target random forest model and the target image recognition model respectively.

[0128] In some optional embodiments, the training module includes:

[0129] The first model training submodule is used to perform random forest model training based on the RF signal in the valid sensor data to obtain a target random forest model.

[0130] The second model training submodule is used to perform image recognition model training based on the image data in the effective sensor data to obtain a target image recognition model.

[0131] In some optional embodiments, the preprocessing module includes:

[0132] The denoising submodule is used to perform median interpolation filling on the sensor data and remove noise to obtain smooth sensor data.

[0133] The standardization submodule is used to standardize the smoothed sensor data and perform time alignment to obtain standard sensor data.

[0134] The fusion submodule is used to fuse the RF signal and image data at the same time in the standard sensor data to obtain the comprehensive sensor data.

[0135] The judgment range submodule is used to confirm the comprehensive sensor data that meets the preset value range as valid sensor data.

[0136] In some optional implementations, the training module further includes:

[0137] The grouping submodule is used to divide the valid sensor data into training set data, validation set data and test set data according to a preset ratio.

[0138] The training model submodule is used to train the random forest model and the image recognition model based on the training set data to obtain the target random forest model and the target image recognition model respectively.

[0139] In some optional embodiments, the device further comprises:

[0140] The infrared verification module is used to perform a first cross-verification using an infrared sensor if the first recognition result indicates that there is no drone in the target area and the second recognition result indicates that there is a drone in the target area.

[0141] The detection threshold module is used to adjust the detection threshold of the image recognition model based on the second recognition result and the cross-validation result if there is still an abnormality in the first cross-validation result, and send it to NWDAF for storage.

[0142] In some optional embodiments, the device further comprises:

[0143] The spectrum analysis module is used to perform spectrum analysis to obtain a spectrum analysis result if the first recognition result indicates that a drone exists in the target area and the second recognition result indicates that no drone exists in the target area.

[0144] The storage result module is used to send the spectrum analysis result and the RF signal information corresponding to the first recognition result to the NWDAF for storage.

[0145] In some optional implementations, the training module further includes:

[0146] The second cross-validation submodule is used to perform a second cross-validation on the target random forest model and the target image recognition model, and adjust the parameters of the target random forest model and the target image recognition model based on the second cross-validation results.

[0147] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0148] The 5G network-based drone detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0149] The embodiment of the present invention also provides a computer device having the above Figure 5 The drone detection device based on 5G network is shown.

[0150] See also Figure 6 , Figure 6 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.

[0151] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0152] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0153] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0154] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0155] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0156] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0157] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0158] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0159] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A drone detection method based on 5G network, characterized in that: The method is performed by a DAF responsible for data processing in a drone detection system, and the system also includes an access and mobility management function AMF; the method includes: receiving target sensor data corresponding to the target area; the target sensor data including a target RF signal and target image data corresponding to the target RF signal; Processing the target RF signal through a random forest model to obtain a first recognition result; Processing the target image data through an image recognition model to obtain a second recognition result; When both the first recognition result and the second recognition result indicate that a drone exists in the target area, a warning message is sent to a third-party platform.

2. The method according to claim 1, characterized in that Before receiving the target sensor data corresponding to the target area, the method further includes: receiving sensor data sent by NWDAF; the sensor data including RF signal and image data corresponding to the RF signal; Performing missing value filling processing on the sensor data and performing data fusion to obtain valid sensor data; Based on the effective sensor data, a random forest model and an image recognition model are trained respectively to obtain a target random forest model and a target image recognition model respectively.

3. The method according to claim 2, characterized in that The performing of random forest model and image recognition model training based on the effective sensor data to obtain a target random forest model and a target image recognition model respectively includes: Performing random forest model training based on the RF signal in the valid sensor data to obtain a target random forest model; An image recognition model is trained based on the image data in the effective sensor data to obtain a target image recognition model.

4. The method according to claim 3, characterized in that The performing missing value filling processing on the sensor data and data fusion to obtain valid sensor data includes: Performing median interpolation filling on the sensor data and removing noise to obtain smoothed sensor data; The smoothed sensor data is standardized and time-aligned to obtain standard sensor data; fusing the RF signal and the image data at the same time in the standard sensor data to obtain comprehensive sensor data; The integrated sensor data that meets a preset value range is confirmed as the valid sensor data.

5. The method according to claim 4, characterized in that The performing of random forest model and image recognition model training based on the effective sensor data to obtain a target random forest model and a target image recognition model respectively includes: Dividing the valid sensor data into training set data, validation set data and test set data according to a preset ratio; Based on the training set data, a random forest model and an image recognition model are trained respectively to obtain a target random forest model and a target image recognition model respectively.

6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: If the first recognition result indicates that there is no drone in the target area, and the second recognition result indicates that there is a drone in the target area, performing a first cross-validation using an infrared sensor; If there is still an abnormality in the first cross-validation result, the detection threshold of the image recognition model is adjusted based on the second recognition result and the cross-validation result, and sent to NWDAF for storage.

7. The method according to any one of claims 6, characterized in that The method further comprises: If the first recognition result indicates that there is a drone in the target area, and the second recognition result indicates that there is no drone in the target area, performing spectrum analysis to obtain a spectrum analysis result; The spectrum analysis result and the RF signal information corresponding to the first recognition result are sent to NWDAF for storage.

8. The method according to claim 4, characterized in that The training of the random forest model and the image recognition model based on the valid sensor data to obtain the target random forest model and the target image recognition model respectively further includes: A second cross-validation is performed on the target random forest model and the target image recognition model, and parameters of the target random forest model and the target image recognition model are adjusted based on the second cross-validation result.

9. A drone detection device based on 5G network, characterized in that: The device comprises: A receiving module, configured to receive target sensor data corresponding to a target area; the target sensor data includes a target RF signal and target image data corresponding to the target RF signal; A first recognition module, configured to process the target RF signal through a random forest model to obtain a first recognition result; A second recognition module, used to process the target image data through an image recognition model to obtain a second recognition result; The warning sending module is used to send a warning message to a third-party platform when the first recognition result and the second recognition result both indicate that there is a drone in the target area.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the 5G network-based drone detection method described in any one of claims 1 to 8 by executing the computer instructions.

Images