Unmanned aerial vehicle interference detection and classification method, system, equipment and medium

Abstract

The invention provides an unmanned aerial vehicle interference detection and classification method, system and device and a medium, and the method comprises the steps: obtaining a signal data set received by an unmanned aerial vehicle; analyzing the plurality of signal features corresponding to the signal data set to obtain the importance of the plurality of signal features in interference detection; based on the importance, performing dynamic grading processing on the plurality of signal features to obtain multi-level signal features; and analyzing and fusing the multi-level signal features by using a convolutional neural network to obtain an interference detection classification result of the unmanned aerial vehicle. According to the invention, the accuracy and robustness of interference detection and classification of the unmanned aerial vehicle can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV communication security, and particularly relates to a method, system, device and medium for UAV interference detection and classification. Background Art

[0002] With the rapid development of UAV technology, the security of UAV communication systems has attracted increasing attention. When UAVs are performing tasks, they may encounter various forms of radio signal interference. These interferences not only affect the communication quality of UAVs, but may also pose a threat to the safe flight of UAVs and even cause UAVs to lose control. Therefore, how to effectively detect and classify the interference signals encountered by UAVs has become one of the key issues in ensuring UAV communication security. Summary of the Invention

[0003] To solve the problems of the prior art, the present invention proposes a method, system, device and medium for UAV interference detection and classification, aiming to improve the accuracy and robustness of UAV interference detection and classification.

[0004] The object of the present invention is achieved by the following technical solutions:

[0005] On the one hand, an embodiment of the present invention provides a method for UAV interference detection and classification, the method comprising:

[0006] Obtaining a signal data set received by the UAV;

[0007] Analyzing a plurality of signal features corresponding to the signal data set to obtain the importance degrees of the plurality of signal features in interference detection;

[0008] Based on the importance degrees, performing dynamic hierarchical processing on the plurality of signal features to obtain multi-level signal features;

[0009] Using a convolutional neural network to analyze and fuse the multi-level signal features to obtain the UAV interference detection and classification result.

[0010] Optionally, the importance degree of each signal feature in interference detection is determined based on the influence value of each signal feature on the uniform distribution degree of the signal data set.

[0011] Optionally, the method for obtaining the influence value of each signal feature on the uniform distribution degree of the signal data set includes:

[0012] Obtaining the original uniformity index of the signal data set;

[0013] Obtain the partition uniformity index of the partitioned data set corresponding to each of the signal features; wherein, the partitioned data set corresponding to each of the signal features is obtained by partitioning the signal data set using each of the signal features;

[0014] Based on each of the signal features, subtract the original uniformity index from the partition uniformity index of the partitioned data set corresponding to the signal feature to obtain the influence value of each of the signal features on the uniformity of the signal data set.

[0015] Optionally, the obtaining of the original uniformity index of the signal data set includes:

[0016] Obtain the data volume of the signal data set and the number of subsets of the signal data of different categories in the signal data set;

[0017] Calculate the ratio between the number of subsets of the signal data of each category in the signal data set and the data volume to obtain a first ratio set;

[0018] Subtract a specific value from the square of each first ratio in the first ratio set to obtain the original uniformity index.

[0019] Optionally, the partitioned data set includes: a plurality of first subsets; the obtaining of the partition uniformity index of the partitioned data set corresponding to each of the signal features includes:

[0020] Determine the uniformity of each of the first subsets;

[0021] Calculate the ratio between the sub - data volume of each of the first subsets and the data volume of the signal data set to obtain the quantity ratio of each of the first subsets;

[0022] Multiply the quantity ratio of each of the first subsets by the uniformity to obtain an intermediate value of each of the first subsets;

[0023] Fuse the intermediate values of each of the first subsets to obtain the partition uniformity index.

[0024] Optionally, the signal data set includes: a plurality of second subsets; the analysis of the multiple signal features corresponding to the signal data set to obtain the importance of the multiple signal features in interference detection includes:

[0025] Based on each of the signal features, accumulate the influence values of the signal feature on the uniform distribution degree of each of the second subsets to obtain the total sum of the uniformity improvement amounts of each of the signal features;

[0026] For each of the said signal features, the ratio between the total sum of the improvement in the uniformity of the signal feature and the number corresponding to the multiple second sub-datasets is the importance of the corresponding signal feature in interference detection.

[0027] Optionally, the convolutional neural network includes: an independent-level feature convolutional neural network, a multi-level feature convolutional neural network; the number of levels of the multi-level feature convolutional neural network is determined by the number of levels in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network;

[0028] Using the convolutional neural network to analyze and fuse the multi-level signal features to obtain the interference detection classification result of the UAV includes:

[0029] Inputting the multi-level signal features into the multi-level feature convolutional neural network in the order of levels to obtain the preprocessed feature data output by each level feature convolutional neural network in the multi-level feature convolutional neural network;

[0030] Inputting the preprocessed feature data output by each level feature convolutional neural network into the independent-level feature convolutional neural network for fusion to obtain the interference detection classification result.

[0031] Optionally, the independent-level feature convolutional neural network sequentially includes: an input layer, a first convolutional layer, a second convolutional layer, a flattening layer, a fully connected layer, a merging layer introducing a self-attention mechanism, and a final fully connected layer.

[0032] Optionally, the construction process of the convolutional neural network includes:

[0033] Obtaining multiple historical signal features corresponding to a historical signal dataset; wherein, the historical signal dataset is the signal received by the UAV in a historical time period;

[0034] Based on the importance of the multiple historical signal features in interference detection, performing dynamic grading processing on the multiple historical signal features to obtain multi-level historical signal features;

[0035] Using the multi-level historical signal features to perform iterative training and performance evaluation on an initial neural network to obtain the convolutional neural network.

[0036] Optionally, before analyzing the multiple signal features corresponding to the signal dataset to obtain the importance of the multiple signal features in interference detection, the method further includes:

[0037] Performing normalization processing on the signal dataset to obtain a normalized signal set;

[0038] Perform signal analysis on the normalized signal set to obtain the multiple signal features.

[0039] On the other hand, an embodiment of the present invention provides a UAV interference detection and classification system, and the system includes:

[0040] A hierarchical feature processor, configured to obtain a signal data set received by the UAV; analyze multiple signal features corresponding to the signal data set to obtain the importance of the multiple signal features in interference detection; based on the importance, perform dynamic classification processing on the multiple signal features to obtain multi-level signal features;

[0041] An interference detection and classification module, configured to analyze and fuse the multi-level signal features by using a convolutional neural network to obtain an interference detection and classification result of the UAV.

[0042] Optionally, the importance of each signal feature in interference detection is determined based on the influence value of each signal feature on the uniform distribution degree of the signal data set.

[0043] Optionally, the hierarchical feature processor is specifically configured to obtain an original uniformity index of the signal data set; obtain a partition uniformity index of a partition data set corresponding to each signal feature; wherein, the partition data set corresponding to each signal feature is obtained by partitioning the signal data set by using each signal feature; based on each signal feature, subtract the partition uniformity index of the partition data set corresponding to the signal feature from the original uniformity index to obtain the influence value of each signal feature on the uniform distribution degree of the signal data set.

[0044] Optionally, the hierarchical feature processor is specifically configured to obtain the data volume of the signal data set and the number of subsets of signal data of different categories in the signal data set; calculate the ratio between the number of subsets of signal data of each category in the signal data set and the data volume to obtain a first ratio set; subtract the square of each first ratio in the first ratio set from a specific value to obtain the original uniformity index.

[0045] Optionally, the partition data set includes: a plurality of first sub-data sets;

[0046] The hierarchical feature processor is specifically configured to determine the uniformity of each of the first sub-datasets; calculate the ratio between the amount of sub-data in each of the first sub-datasets and the amount of data in the signal dataset to obtain the quantity ratio of each of the first sub-datasets; multiply the quantity ratio of each of the first sub-datasets by the uniformity to obtain the intermediate value of each of the first sub-datasets; and fuse the intermediate values of each of the first sub-datasets to obtain the partition uniformity index.

[0047] Optionally, the signal dataset includes: a plurality of second sub-datasets;

[0048] The hierarchical feature processor is specifically configured to accumulate the influence values of each signal feature on the uniform distribution degree of each second sub-dataset based on each signal feature to obtain the total sum of the uniformity improvement amounts of each signal feature; for each signal feature, the ratio between the total sum of the uniformity improvement amounts of the signal feature and the quantity corresponding to the plurality of second sub-datasets is the importance of the corresponding signal feature in interference detection.

[0049] Optionally, the convolutional neural network includes: an independent hierarchical feature convolutional neural network, a multi-level feature convolutional neural network; the number of levels of the multi-level feature convolutional neural network is determined by the number of levels in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network;

[0050] The interference detection classification module is further configured to input the multi-level signal features into the multi-level feature convolutional neural network in order of levels to obtain the preprocessed feature data output by each level feature convolutional neural network in the multi-level feature convolutional neural network;

[0051] Input the preprocessed feature data output by each level feature convolutional neural network into the independent hierarchical feature convolutional neural network for fusion to obtain the interference detection classification result.

[0052] Optionally, the independent hierarchical feature convolutional neural network sequentially includes: an input layer, a first convolutional layer, a second convolutional layer, a flattening layer, a fully connected layer, a merging layer introducing self-attention mechanism, and a final fully connected layer.

[0053] Optionally, the system further includes:

[0054] A building block of a convolutional neural network for obtaining multiple historical signal features corresponding to a historical signal dataset; wherein the historical signal dataset is the signals received by the drone during a historical time period; dynamically grading the multiple historical signal features based on their importance in interference detection to obtain multi-level historical signal features; and using the multi-level historical signal features to iteratively train and evaluate the performance of an initial neural network to obtain the convolutional neural network.

[0055] Optionally, the hierarchical feature processor includes:

[0056] A preprocessing unit for normalizing the signal dataset to obtain a normalized signal set; and parsing the normalized signal set to obtain the multiple signal features.

[0057] On the other hand, an embodiment of the present invention also provides an electronic device, including: at least one processor and a memory; the memory and the processor are connected by a bus;

[0058] The memory is used to store one or more programs;

[0059] When the one or more programs are executed by the at least one processor, the drone interference detection and classification method as described above is implemented.

[0060] Correspondingly, an embodiment of the present invention also provides a readable storage medium with an execution program stored thereon. When the execution program is executed, the drone interference detection and classification method as described above is implemented.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] An embodiment of the present invention provides a drone interference detection and classification method, system, device, and medium. During the execution of this method, first, analyze multiple signal features corresponding to the signal dataset received by the drone to obtain the importance of the multiple signal features in interference detection; then, based on this importance, dynamically grade the multiple signal features to obtain multi-level signal features; finally, use a convolutional neural network to analyze and fuse the multi-level signal features, thereby obtaining the interference detection and classification results of the drone. In this way, on the one hand, by dynamically grading multiple signal features of the drone, the importance of signal features in interference detection can be accurately identified. On the other hand, combining a convolutional neural network for the analysis and deep fusion of multi-level signal features can improve the accuracy and robustness of drone interference detection and classification, so that the drone can better meet the application requirements in a complex electromagnetic environment.

[0063] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, rather than limiting the technical solutions provided by the embodiments of the present invention. Brief Description of the Drawings

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings, where:

[0065] Figure 1 is a schematic flowchart of a method for detecting and classifying UAV interference provided by an embodiment of the present invention;

[0066] Figure 2 is a schematic framework diagram of a method for detecting and classifying UAV interference based on a deep fusion network of dynamic hierarchical features provided by an embodiment of the present invention;

[0067] Figure 3 is a schematic flowchart of a method for detecting and classifying UAV interference based on a deep fusion network of dynamic hierarchical features provided by an embodiment of the present invention;

[0068] Figure 4 is a schematic composition diagram of a UAV interference detection and classification system provided by an embodiment of the present invention;

[0069] Figure 5 is a schematic composition diagram of an electronic device provided by an embodiment of the present invention. Detailed Embodiments

[0070] The following will illustrate the embodiments of the present invention with reference to the drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, rather than for limiting the protection scope of the present invention.

[0071] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0072] In the following description, the terms "first / second / third" only distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first / second / third" can be interchanged with a specific order or sequence when permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the embodiments of the present invention belong. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the embodiments of the present invention.

[0074] With the rapid development of unmanned aerial vehicle (UAV) technology, the security of UAV communication systems has been increasingly emphasized. When performing tasks, UAVs may encounter various forms of radio signal interference, which not only affects the communication quality of UAVs but also may pose a threat to the safe flight of UAVs and even lead to out-of-control of UAVs. Therefore, how to effectively detect and classify the interference signals encountered by UAVs has become one of the key issues in ensuring UAV communication security.

[0075] Traditional interference detection methods mainly rely on spectrum analysis technology, which determines the presence of interference by monitoring the change in signal strength within a specific frequency range. However, this method has problems such as low detection accuracy and slow response speed, making it difficult to meet the requirements of modern UAVs for rapid and accurate interference detection. In addition, traditional methods often ignore the correlation between signal features and cannot make full use of signal feature information for efficient interference classification.

[0076] In recent years, deep learning technology has achieved remarkable results in fields such as image recognition and natural language processing due to its powerful feature extraction and pattern recognition capabilities. Methods based on deep learning have also been attempted to be applied to UAV interference detection, and by constructing models such as convolutional neural networks to automatically learn signal features, the efficiency and accuracy of interference detection have been improved. However, most of the existing deep learning-based interference detection methods adopt static feature extraction methods and do not fully consider the importance differences between different features, resulting in insufficiently refined feature extraction and affecting the effects of interference detection and classification.

[0077] Based on the above problems, the present invention proposes a method, system, device, and medium for UAV interference detection and classification. On the one hand, by dynamically grading multiple signal features of UAVs, it can accurately identify the importance of signal features in interference detection. On the other hand, by combining convolutional neural networks for multi-level signal feature analysis and deep fusion, it can improve the accuracy and robustness of UAV interference detection and classification, so that UAVs can better adapt to the application requirements in complex electromagnetic environments.

[0078] Example 1:

[0079] An embodiment of the present invention provides a method for detecting and classifying UAV interference, as Figure 1 shown, which is a schematic flow chart of a method for detecting and classifying UAV interference provided by an embodiment of the present invention; among them, the following description is made in combination with Figure 1 shown in:

[0080] Step 101, obtain a signal data set received by the UAV.

[0081] In some embodiments of the present invention, the UAV may refer to an unmanned aircraft that is controlled by a radio remote control device and a self-prepared program control device, or is completely or intermittently autonomously controlled by an on-board computer. Among them, the UAV may be a multi-rotor UAV or a fixed-wing UAV, etc., and the present invention does not make any limitation thereto.

[0082] In some embodiments of the present invention, the signal data set received by the UAV includes, but is not limited to: signal data captured by the receiving module and / or antenna of the UAV; among them, the signal data set includes, but is not limited to: possible radio frequency interference, background noise, and other potential disturbance signals. For example, when the UAV is performing a task, it may receive co-frequency interference from other wireless devices or accidental signal blocking, etc.

[0083] Step 102, analyze multiple signal features corresponding to the signal data set to obtain the importance of the multiple signal features in interference detection.

[0084] In some embodiments of the present invention, the multiple signal features corresponding to the signal data set include, but are not limited to: subcarrier spacing, meta time, subcarrier, cyclic prefix (CP) length, received power, signal-to-noise ratio, signal power, noise power, etc.

[0085] In some embodiments of the present invention, the signal data in the signal data set can be first subjected to feature extraction by a hierarchical feature processor to obtain multiple signal features, and then the hierarchical feature processor is continued to analyze the multiple signal features obtained by extraction, such as: calculating the importance of each signal feature in the multiple signal features in interference detection.

[0086] In some embodiments of the present invention, before executing step 102, the following steps A1 and A2 can also be executed:

[0087] Step A1, perform normalization processing on the signal data set to obtain a normalized signal set.

[0088] In some embodiments of the present invention, the signal dataset may be normalized first to obtain a normalized signal set. Here, the signals in the normalized signal set may be further preprocessed, including but not limited to denoising and filtering, to enhance the signal quality in the obtained normalized signal set.

[0089] Step A2: Perform signal analysis on the normalized signal set to obtain the multiple signal features.

[0090] In some embodiments of the present invention, the normalized signal set may be subjected to signal analysis to obtain the signal features of each normalized signal in the normalized signal set, thereby obtaining multiple signal features.

[0091] It should be noted that the signal features corresponding to different normalized signals in the normalized signal set may be the same or different; at the same time, analyzing each normalized signal can obtain at least one signal feature.

[0092] In this way, by first normalizing the received signal dataset and then performing analysis, more accurate multiple signal features can be obtained, thereby providing more accurate parameter support for subsequent UAV interference detection and classification.

[0093] In some embodiments of the present invention, a hierarchical feature extractor may be used to determine the importance of signal features in interference detection; among them, determining the importance of signal features in interference detection can be determined by the influence value of the signal feature on the set uniformity index of its corresponding signal dataset; among them, the set uniformity index of the signal dataset can be used to characterize the uniform distribution degree of the signal data in the signal dataset.

[0094] In some embodiments of the present invention, the importance of each signal feature in interference detection is determined based on the influence value of each signal feature on the uniform distribution degree of the signal dataset.

[0095] Here, the influence value of the signal feature on the uniform distribution degree of the signal dataset can be directly determined as the importance of the signal feature in interference detection, or alternatively, the influence value of the signal feature on the uniform distribution degree of the signal dataset can be analyzed and processed to obtain the importance of the signal feature in interference detection.

[0096] Exemplarily, the importance of signal feature 1 in interference detection is the influence value of signal feature 1 on the uniform distribution degree of the signal dataset. At the same time, the importance of signal feature 2 in interference detection is obtained after analyzing the influence value of signal feature 2 on the uniform distribution degree of the signal dataset, and it is not necessarily the influence value of signal feature 2 on the uniform distribution degree of the signal dataset.

[0097] In some embodiments of the present invention, the influence values of different signal features on the uniform distribution degree of the signal dataset may be the same or different, and the present invention does not make any limitation thereto.

[0098] Among them, the influence value of the signal feature on the uniform distribution degree of the signal dataset can be directly represented by a numerical value.

[0099] It should be noted that the uniform distribution degree of the signal dataset can be used to characterize the uniformity of the signal data in the signal dataset distributed according to a specific law.

[0100] In this way, based on the influence value of each obtained signal feature on the uniform distribution degree of the signal dataset, determining the importance of the corresponding signal feature in interference detection can improve the convenience and accuracy of determining the recognition importance.

[0101] In some embodiments of the present invention, the obtaining manner of the influence value of each of the signal features on the uniform distribution degree of the signal dataset can be implemented by the following steps A1 to A3 ( Figure 1 not shown):

[0102] Step A1: Obtain the original uniformity index of the signal dataset.

[0103] In some embodiments of the present invention, the original uniformity index of the signal dataset can also be directly represented by a numerical value. Here, the original uniformity index can be used to characterize the data distribution uniformity of the signal dataset in the initial distribution state, that is, the original uniformity index can be used to evaluate the uniformity degree of the signal dataset.

[0104] It should be noted that for a signal dataset containing multiple signal categories (normal signal, pulse interference signal, blocking interference signal), the original uniformity index (i.e., the set uniformity index) can be represented by the following formula (1):

[0105]

[0106] Among them, U(S) represents the original uniformity index of the signal dataset S, and the smaller the value of U(S), the higher the uniformity degree of the signal dataset S. Here, the signal dataset S represents a signal data set containing multiple categories, and K represents the total number of categories in the signal dataset S. p k represents the proportion of the signal data of category k in the signal dataset S, that is, p k = the number of elements of category k (the number of sub-signal data) / the total number of elements (data volume) of the signal dataset S.

[0107] In other words, the above step A1 can be implemented through the following process:

[0108] First step, obtain the data volume of the signal data set and the sub - quantities of the signal data of different categories in the signal data set.

[0109] In some embodiments of the present invention, the data volume of the signal data set S is the total number of elements included in the signal data set S, which is the sum of the sub - quantities of the signal data of different categories in the signal data set S.

[0110] Exemplarily, consider a scenario of UAV interference detection, and a signal data set S is received; among them, the signal data in the signal data set S is divided into several categories, such as: "normal signal", "pulse interference signal", "blocking interference signal", etc. Among them, the data volume N of the signal data set S is 100, which includes: 50 normal signals, 30 pulse interference signals, and 20 blocking interference signals.

[0111] Second step, calculate the ratio between the sub - quantity of the signal data of each category in the signal data set and the data volume to obtain a first ratio set.

[0112] In some embodiments of the present invention, calculating the ratio between the sub - quantity of the signal data of each category in the signal data set S and the data volume is to determine the proportion of the signal data of each category in the signal data set S, so as to obtain a first ratio set.

[0113] Here, continuing to refer to the above description, the ratio between the sub - quantity of the normal signal and the data volume is: 50 / 100; the ratio between the sub - quantity of the pulse interference signal and the data volume is: 30 / 100; the ratio between the sub - quantity of the blocking interference signal and the data volume is: 20 / 100; thus, the first ratio set includes: 50 / 100, 30 / 100, 20 / 100.

[0114] Third step, subtract a specific value from the square of each first ratio in the first ratio set to obtain the original uniformity index.

[0115] In some embodiments of the present invention, the specific value can be 1. Subtract 1 from the square of each ratio in the first ratio set to obtain the original uniformity index of the signal data set.

[0116] In some embodiments of the present invention, the specific value 1 and the parameters of the first ratio set (50 / 100, 30 / 100, 20 / 100) described above can be substituted into formula (1) to obtain the original uniformity index U(S) of the signal data set S = 1 - [(50 / 100) 2 +(30 / 100) 2 +(20 / 100) 2=0.62. Here, the value of the original uniformity index of the signal data set S is 0.62, indicating that the class distribution of the signals in the signal data set S is not very uniform.

[0117] It should be noted that in the actual application of UAV interference detection, it is usually desired that the signal characteristics of different interference classes in the signal data set are more evenly distributed in the set, that is, the higher the value of the original uniformity index of the signal data set, the more conducive it is to providing parameter support for improving the accuracy and efficiency of UAV interference detection subsequently.

[0118] In some embodiments of the present invention, the original uniformity index of the signal data set can also be used to evaluate the performance of the hierarchical feature processor when identifying signal features at different interference levels.

[0119] In this way, by performing simple logical operations on the data volume of the signal data set and the number of sub-signal data of different classes therein, the value representing the original uniformity index of the signal data set can be obtained more accurately, thereby providing more accurate parameter support for determining the influence value of the uniformity of each signal feature corresponding to the signal data set with reference to the original uniformity subsequently.

[0120] Step A2: Obtain the partition uniformity index of the partition data set corresponding to each of the signal features.

[0121] Wherein, the partition data set corresponding to each of the signal features is obtained by partitioning the signal data set using each of the signal features.

[0122] In some embodiments of the present invention, the partition uniformity indexes of the partition data sets corresponding to different signal features may be the same or different, and the present invention does not make any limitation thereto.

[0123] It should be noted that the partition data set corresponding to the signal feature is the data set obtained by partitioning the signal data in the signal data set using the signal feature. Here, the number of the partition data sets corresponding to the signal feature is usually at least two.

[0124] In some embodiments of the present invention, the partition data set includes: a plurality of first sub-data sets, that is, the partition data set corresponding to the signal feature includes: a plurality of first sub-data sets; correspondingly, the above step 10212 can be implemented through the following process:

[0125] The first step is to determine the uniformity of each of the first sub-data sets.

[0126] In some embodiments of the present invention, the uniformity of each first sub-data set can be determined first; wherein, the uniformities of different first sub-data sets may be the same or different.

[0127] It should be noted that the uniformity of the first sub - dataset, which characterizes the degree of uniform distribution of the data within the first sub - dataset, can also be represented by a numerical value.

[0128] In the second step, calculate the ratio between the sub - data volume of each of the first sub - datasets and the data volume of the signal dataset to obtain the quantity ratio of each of the first sub - datasets.

[0129] In some embodiments of the present invention, calculate the ratio between the sub - data volume and the data volume of each first sub - dataset to obtain the quantity ratio of each sub - dataset.

[0130] It should be noted that the sub - data volume of the first sub - dataset is the number of elements within the first sub - dataset. Correspondingly, the data volume of the signal dataset is as described above: the number of elements within the signal dataset.

[0131] In some embodiments of the present invention, the quantity ratios of different first sub - datasets can be the same or different. Exemplarily, the quantity ratio of the first sub - dataset 1 is: 10 / 100; the quantity ratio of the first sub - dataset 2 is: 30 / 100; the quantity ratio of the second sub - dataset 3 is: 60 / 100.

[0132] In the third step, multiply the quantity ratio of each of the first sub - datasets by the uniformity to obtain the intermediate value of each of the first sub - datasets.

[0133] In some embodiments of the present invention, the quantity ratio of each first sub - dataset can be correspondingly multiplied by its uniformity to obtain the intermediate value of each first sub - dataset. Exemplarily, multiply the quantity ratio of the first sub - dataset 1 by the uniformity of the first sub - dataset 1 to obtain the intermediate value of the first sub - dataset 1; multiply the quantity ratio of the first sub - dataset 2 by the uniformity of the first sub - dataset 2 to obtain the intermediate value of the first sub - dataset 2.

[0134] In the fourth step, fuse the intermediate values of each of the first sub - datasets to obtain the partition uniformity index.

[0135] In some embodiments of the present invention, fusing the intermediate values of each first sub - dataset can refer to adding the intermediate values of each sub - dataset to obtain a numerical value, which can be used to characterize the partition uniformity index of the partitioned dataset corresponding to the signal feature.

[0136] In this way, by analyzing the uniformity and sub - data volume of multiple first sub - datasets corresponding to the signal feature and combining with the data volume of the signal dataset, the partition uniformity index of the partitioned dataset corresponding to the signal feature can be obtained more conveniently.

[0137] Step A3: Based on each of the signal features, subtract the original uniformity index from the partition uniformity index of the partition data set corresponding to the signal feature to obtain the influence value of each signal feature on the uniformity of the signal data set.

[0138] In some embodiments of the present invention, taking signal feature 1 as an example, it may be to subtract the original uniformity index from the partition uniformity index of the partition data set corresponding to signal feature 1 to obtain the influence value of signal feature 1 on the uniformity of the signal data set.

[0139] Referring to the above description, the original uniformity index of the signal data set S is U(S). The partition data set corresponding to signal feature 1 includes: the first sub-data set S firsthalf and the first sub-data set S latterhalf . Here, the combination of the first sub-data set S firsthalf and the first sub-data set S latterhalf can obtain the signal data set S. Then, the influence value of this signal feature 1 on the uniformity of the signal data set S can refer to the following formula (2):

[0140]

[0141] where ΔU represents the influence value of signal feature 1 on the uniformity of the signal data set S, that is, the amount of decrease in the uniformity of this signal feature 1 in the signal data set S; U(S firsthalf ) represents the uniformity of the first sub-data set S firsthalf , U(S latterhalf ) represents the uniformity of the first sub-data set S latterhalf , N firsthalf represents the amount of sub-data in the first sub-data set S firsthalf (the number of elements it contains), N latterhalf represents the amount of sub-data in the first sub-data set S latterhalf (the number of elements it contains); N represents the amount of data in the signal data set S (the number of elements it contains); correspondingly, that is, it represents the partition uniformity index of the partition data set corresponding to signal feature 1.

[0142] In some embodiments of the present invention, for a signal data set S of unmanned aerial vehicles containing multiple categories, if we want to evaluate the importance of the signal feature of "signal strength", before considering the "signal strength" feature, the original uniformity index U(S) of the signal data set S may be relatively high. Then it may be because signals of different categories are mixed together in the entire signal data set S. However, when the signal data set S is partitioned according to the "signal strength" feature, if the uniformity of the signal data set S is significantly improved (ΔU is significantly less than U(S)), then it can be considered that "signal strength" is an important feature.

[0143] It should be noted that if there is a signal data set S containing two categories (interference signals and non-interference signals), its original uniformity index is U(S). When the signal data set S is divided into two subsets (the multiple first subsets mentioned above) according to the "signal strength" feature, the uniformity of each subset is U(S firsthalf ) and U(S latterhalf ). If the sum of the uniformities of these two subsets is greater than the uniformity of the original data set, that is, U(S firsthalf ) + U(S latterhalf ) > U(S), then it can be considered that the "signal strength" feature makes a significant contribution to improving the uniformity of the signal data set S, and thus is an important signal feature.

[0144] That is to say, the influence value of the signal feature on the uniform distribution degree of the signal data set is calculated by evaluating the contribution of the signal feature to the decrease in the value corresponding to the original uniformity index of the signal data set.

[0145] In this way, by directly subtracting the original uniformity index of the signal data set from the partition uniformity index of the partition data set corresponding to each signal feature, the influence value of each signal feature on the uniformity of the signal data set is obtained, providing parameter support for determining the importance of the signal feature in interference detection.

[0146] In some embodiments of the present invention, multiple signal features can be represented by a signal feature set, such as: X = {x1, x2,..., x n}; where n is the number of signal features. Here, the importance degree, that is, the importance, of each signal feature in interference detection can be obtained through a hierarchical feature processor.

[0147] Here, before determining the importance of each signal feature in interference detection among multiple signal features, the multiple signal features can also be processed such as normalized processing first to eliminate the differences in dimension and value range among the multiple signal features, so that the corresponding different indicators or features are comparable.

[0148] In some embodiments of the present invention, the importance set of multiple signal features in interference detection can be represented by I = {i1, i2,..., i n}; where i n represents the importance of the nth signal feature in interference detection, and I can represent the importance set of signal features obtained through the hierarchical feature processor.

[0149] In some embodiments of the present invention, if the influence value of the signal feature on the uniform distribution degree of the signal data set is directly regarded as the importance of the signal feature in interference detection, then this importance can also be represented by a numerical value.

[0150] In some embodiments of the present invention, when the signal data set includes a plurality of second sub-data sets, the above step 102 can be implemented in the following ways of step 1021 and step 1022 ( Figure 1 not shown in the figure):

[0151] Step 1021: Based on each of the signal features, accumulate the influence values of the signal features on the uniform distribution degree of each of the second sub-data sets to obtain the total sum of the uniform degree improvement amounts of each of the signal features.

[0152] In some embodiments of the present invention, assume there is an evaluation unit that can be used to evaluate the relevant parameters of a plurality of signal features corresponding to a signal data set. By using this evaluation unit, it can be evaluated, that is, it may be found that: for the signal feature X (for example, signal strength), in a plurality of second sub-data sets included in the signal data set, in a certain specific set, it significantly increases the distribution uniformity of the data within the set; exemplarily, such as: this signal feature X can help distinguish normal communication signals and interfered signals. Or, by using this evaluation unit, it can be evaluated, that is, it is found that for the signal feature Y (for example, signal frequency offset), in another specific set of a plurality of second sub-data sets included in the signal data set, the improvement amount of the uniformity of the set is relatively large, because this signal feature Y can effectively distinguish signals with frequency interference and signals with normal frequency. Then it is necessary to use the evaluation unit to evaluate the influence value of the signal feature a on the uniform distribution degree of all second sub-data sets and accumulate them to obtain the total sum of the uniform degree improvement amounts of the signal feature a, so as to better determine the importance of the signal feature a in interference detection with the help of the total sum of the uniform degree improvement amounts of the signal feature a.

[0153] For example, if a plurality of second sub-data sets all have the corresponding signal feature X, then the total sum of the uniform degree improvement amounts of the signal feature X is the sum of the uniform degree improvement amounts in these sets (a plurality of second sub-data sets). Here, if the signal feature X can significantly improve the uniformity in most of the plurality of second sub-data sets, then its cumulative importance value, that is, the total sum of the uniform degree improvement amounts will be very high, indicating that the signal feature X plays an important role in detecting UAV communication signal interference.

[0154] In some embodiments of the present invention, through this signal feature importance accumulation strategy, that is, the total sum of the uniform degree improvement amounts of each signal feature, not only can the signal features that are most critical for UAV interference detection be identified, but also the relative importance of these signal features in the interference detection task can be quantified.

[0155] It should be noted that the sum of the improvement amounts of the uniformity of each signal feature is obtained by accumulating the improvement amounts of the uniformity on all sets, that is, multiple second sub-datasets, so as to obtain the total improvement amount of the uniformity of the signal feature in the entire evaluation unit. That is to say, all the second sub-datasets are traversed, and the influence value of each signal feature on the uniform distribution degree of each second sub-dataset is determined and accumulated.

[0156] Step 1022: For each of the signal features, the ratio between the sum of the improvement amounts of the uniformity of the signal feature and the number corresponding to the multiple second sub-datasets is the importance of the corresponding signal feature in interference detection.

[0157] In some embodiments of the present invention, the influence value of the signal feature j on the uniform distribution degree of each second sub-dataset can be set as Then the sum of the improvement amounts of the uniformity of the signal feature j in all the second sub-datasets (the signal dataset includes U second sub-datasets) is U j , and can refer to the following formula (3):

[0158]

[0159] where U j characterizes the importance of the signal feature j in all evaluation units, that is, all the second sub-datasets, and U represents the total number of evaluation units, that is, the second sub-datasets.

[0160] In some embodiments of the present invention, the above steps 1021 and 1022 indicate that a series of analyses need to be performed on the influence value of the signal feature on the uniform distribution degree of the signal dataset to obtain the importance of the signal feature in interference detection.

[0161] In this way, in the case where the signal dataset includes multiple second sub-datasets, the influence value of the signal feature on the uniform distribution degrees of the multiple second sub-datasets can also be accumulated for analysis to more accurately determine the importance of the signal feature in interference detection.

[0162] In some embodiments of the present invention, a hierarchical feature processor is used to identify the influence values of multiple signal features, that is, signal features of different interference categories, on the uniform distribution of the data in the signal dataset, so as to use the identified influence values as important parameters for evaluating the importance of the multiple signal features in interference detection.

[0163] Step 103: Based on the importance, perform dynamic grading processing on the multiple signal features to obtain multi-level signal features.

[0164] In some embodiments of the present invention, a hierarchical feature processor may continue to be employed to perform clustering analysis on multiple signal features based on their importance in interference detection, that is, to achieve dynamic hierarchical (stratified) processing of multiple signal features, thereby obtaining multi-level signal features. That is to say, the hierarchical feature processor performs clustering processing on signal features with similar importance through a clustering algorithm to achieve hierarchical classification of multiple signal features.

[0165] It should be noted that the core of dynamically classifying multiple signal features based on importance to obtain multi-level signal features is to identify which signal features among multiple signal features are the most critical for interference detection. Exemplarily, if it is found that signal frequency is very crucial in the interference detection process, then signal feature b that is highly correlated with frequency changes can be determined as a key feature, that is, it is considered that the importance of signal feature b will be relatively high, and thus signal feature b can be classified into the signal feature set corresponding to the higher level in the multi-level signal features.

[0166] In some embodiments of the present invention, a hierarchical feature processor is employed to divide signal features with similar importance in interference detection into one level through a clustering algorithm. For example, if the importance levels of multiple signal features are 0.96, 0.43, 0.89, 0.74, 0.88, 0.71 in sequence, then the hierarchical feature processor will divide them into multiple signal features corresponding to level 1, and the importance levels of these multiple signal features in interference detection are [0.96, 0.89, 0.88]; multiple signal features corresponding to level 2, and the importance levels of these multiple signal features in interference detection are [0.74, 0.71]; the signal feature corresponding to level 3, and the importance level of this signal feature in interference detection is [0.43]; among them, level 1 has the highest level.

[0167] It should be noted that the higher the level in the multi-level signal features, the greater the influence of the signal features corresponding to that level on the tag (subsequent interference detection), and the greater the corresponding obtained weight.

[0168] Step 104: Use a convolutional neural network to analyze and fuse the multi-level signal features to obtain the interference detection classification result of the UAV.

[0169] In some embodiments of the present invention, the convolutional neural network is a deep learning model. Here, the convolutional neural network is a pre-trained (or pre-constructed) convolutional neural network.

[0170] In some embodiments of the present invention, the construction process of the convolutional neural network includes:

[0171] The first step is to obtain multiple historical signal features corresponding to the historical signal dataset.

[0172] Among them, the historical signal dataset is the signal received by the UAV in the historical time period.

[0173] In some embodiments of the present invention, the acquired historical signal dataset can be preprocessed and parsed first to obtain its corresponding multiple historical signal features. Here, the multiple historical signal features include but are not limited to: subcarrier spacing, meta-time, subcarriers, CP length, received power, signal-to-noise ratio, signal power, and noise power.

[0174] Here, the historical signal dataset includes but is not limited to: various types of radio frequency interference signals, background noise, and other potential disturbance signals received by the UAV through its internal receiving module and / or antenna in the historical time period.

[0175] It should be noted that the historical time period can refer to any historical duration, for example: one month before the current moment, one week before the current moment, etc.

[0176] In the second step, based on the importance of the multiple historical signal features in interference detection, dynamic hierarchical processing is performed on the multiple historical signal features to obtain multi-level historical signal features.

[0177] In some embodiments of the present invention, a hierarchical feature processor can also be used to first analyze the multiple historical signal features to obtain the importance of the multiple historical signal features in interference detection, and then based on the importance of the multiple historical signal features in interference detection, dynamic hierarchical processing is performed on the multiple historical signal features to obtain multi-level historical signal features.

[0178] It should be noted that the specific execution logic here is similar to the execution logic of steps 102 and 103 above, and the corresponding description can refer to the description of steps 102 and 103 above, which will not be repeated here.

[0179] In the third step, the multi-level historical signal features are used to iteratively train and perform performance evaluation on the initial neural network to obtain the convolutional neural network.

[0180] In some embodiments of the present invention, the initial neural network can be a pre-set network model, and its internal model parameters can be determined according to actual requirements.

[0181] In some embodiments of the present invention, multi-level historical signal features are fed into an initial neural network for training so that the initial neural network can better learn the essence of each level of historical signal features in the multi-level historical signal features. At the same time, in the process of constructing the initial neural network to obtain a convolutional neural network, an adaptive learning rate adjustment mechanism can also be adopted, that is, according to the change of the loss function value during the training process, the learning rate is automatically adjusted to ensure that the initial neural network converges quickly in the initial stage of training and is finely tuned in the later stage of training to avoid overfitting, so as to obtain a convolutional neural network with high generalization ability and detection accuracy.

[0182] It should be noted that if the convolutional neural network includes: an independent-level feature convolutional neural network (for fusing the input data), a multi-level feature convolutional neural network (for analyzing multi-level signal features); the number of levels of the multi-level feature convolutional neural network is determined by the number of levels in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network; correspondingly, the initial neural network may include: an initial independent-level feature convolutional neural network, an initial multi-level feature convolutional neural network; the number of levels of the initial multi-level feature convolutional neural network is determined by the number of levels in the multi-level historical signal features, and the network structure complexity of the high-level feature convolutional neural network in the initial multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network. Here, the multi-level historical signal features can be input into the initial multi-level feature convolutional neural network for analysis according to different levels, and different historical level features are correspondingly input. Based on the analysis results, the initial multi-level feature convolutional neural network is trained. At the same time, the analysis results can also be input into the initial independent-level feature convolutional neural network for splicing or fusion, and based on the splicing or fusion results, the initial independent-level feature convolutional neural network is trained.

[0183] In some embodiments of the present invention, if the initial neural network includes: an initial independent-level feature convolutional neural network and an initial multi-level feature convolutional neural network, then the initial independent-level feature convolutional neural network and the initial multi-level feature convolutional neural network are simultaneously iteratively trained and performance evaluated. The execution logic is similar to that corresponding to the iterative training and performance evaluation of one of the initial independent-level feature convolutional neural network and the initial multi-level feature convolutional neural network, and will not be elaborated here.

[0184] In this way, through the multiple historical signal features corresponding to the historical signal dataset, the initial neural network is iteratively trained and performance evaluated, and a convolutional neural network with a more accurate output interference detection result can be obtained.

[0185] In some embodiments of the present invention, the convolutional neural network includes: an independent-level feature convolutional neural network, a multi-level feature convolutional neural network; the number of levels of the multi-level feature convolutional neural network is determined by the number of levels in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network; correspondingly, step 104 above can be implemented in the following manner of step 1041 and step 1042 (not shown in the figure):

[0186] Step 1041: Input the multi-level signal features into the multi-level feature convolutional neural network in order of levels, and obtain the preprocessed feature data output by each level feature convolutional neural network in the multi-level feature convolutional neural network.

[0187] In some embodiments of the present invention, the signal features corresponding to different levels in the multi-level signal features can be sent to the corresponding multi-level feature convolutional neural networks for analysis; that is, for multiple signal features with higher levels in the multi-level signal features, they need to be input into the higher-level feature convolutional neural network in the multi-level feature convolutional neural network, and correspondingly, for multiple signal features with lower levels in the multi-level signal features, they need to be input into the lower-level feature convolutional neural network in the multi-level feature convolutional neural network; exemplarily, if for "highly correlated" signal features, their importance in interference detection is very high, then they will be classified into high-level signal features, and thus need to be input into a deeper or more complex neural network in the multi-level feature convolutional neural network for analysis to capture their subtle feature data.

[0188] It should be noted that the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network; among them, the network structure can refer to the structure composed of the functions corresponding to its input layer, output layer, and hidden layer.

[0189] Step 1042: Input the preprocessed feature data output by each level feature convolutional neural network into the independent-level feature convolutional neural network for fusion to obtain the interference detection classification result.

[0190] In some embodiments of the present invention, an independent-level feature convolutional neural network is used to splice or fuse the preprocessed feature data output by each level feature convolutional neural network, that is, each level feature convolutional neural network, to obtain the corresponding detection result.

[0191] In some embodiments of the present invention, an independent hierarchical feature convolutional neural network may be employed to merge the outputs of each hierarchical feature convolutional neural network in a multi-level feature convolutional neural network, and a fully connected layer is used for interference type classification to output an interference detection classification result.

[0192] Here, the interference detection classification result can be represented by the presence of interference (pulse interference, blocking interference, noise interference, etc.) or the absence of interference.

[0193] It should be noted that the independent hierarchical feature convolutional neural network is a fusion network that is responsible for integrating information from different levels to obtain the final detection result. In this way, the information of each hierarchical feature can be fully utilized to improve the accuracy and robustness of UAV interference detection.

[0194] In some embodiments of the present invention, the independent hierarchical feature convolutional neural network can adopt an attention mechanism to enhance the attention to key hierarchical features, reduce the dependence on redundant or irrelevant features, and thus improve the generalization ability and detection accuracy of the independent hierarchical feature convolutional neural network.

[0195] In some embodiments of the present invention, the independent hierarchical feature convolutional neural network sequentially includes: an input layer, a first convolutional layer, a second convolutional layer, a flattening layer, a fully connected layer, a merging layer with a self-attention mechanism, and a final fully connected layer.

[0196] In some embodiments of the present invention, in the independent hierarchical feature convolutional neural network, the input layer is used to receive the signal feature data preprocessed by each hierarchical feature convolutional neural network in the multi-level feature convolutional neural network; the first convolutional layer and the second convolutional layer sequentially extract primary features and high-level features, that is, the first convolutional layer is used to perform a convolutional operation on the input signal data (the output of the input layer) to extract primary features; the second convolutional layer is used to perform a further convolutional operation on the output of the first convolutional layer to extract higher-level features; the flattening layer is used to flatten the output tensor of the second convolutional layer into a one-dimensional vector (i.e., one-dimensionalize the feature vector) for input into the fully connected layer; the fully connected layer is used to perform a linear transformation on the flattened feature vector to extract a higher-level feature representation; the merging layer is used to merge the outputs of multiple fully connected layers to form a higher-dimensional feature representation, and at the same time, a self-attention mechanism is introduced into the merging layer. By calculating the correlation between features, a weight is assigned to each feature (implemented through the self-attention mechanism), making the independent hierarchical feature convolutional neural network pay more attention to features with high importance; the final connection layer is used to perform a linear transformation on the weighted features to generate an interference detection classification result.

[0197] That is to say, the merging layer introducing the self-attention mechanism evaluates its importance by accumulating the improvement in the uniformity of features across multiple data sets. Among them, for key signal features, such as signal strength, which can significantly improve the uniformity in the signal data set and has a relatively high cumulative importance value, it indicates that this signal feature plays an important role in detecting communication signal interference. Thus, the relative importance of each signal feature in the interference detection task is quantified through this evaluation strategy.

[0198] In this way, in the independent hierarchical feature convolutional neural network, by adopting the attention mechanism to enhance the attention to key hierarchical features, the dependence on redundant or irrelevant features can be reduced, thereby improving the generalization ability and detection accuracy of the interference detection classification results output by the independent hierarchical feature convolutional neural network.

[0199] In this way, first, through the multi-level feature convolutional neural network, the multi-level signal features are preprocessed level by level, and then the preprocessed feature data output by each level of feature convolutional neural network is fused through the independent hierarchical feature convolutional neural network, which can improve the accuracy and robustness of the interference detection and classification of the unmanned aerial vehicle, so that the unmanned aerial vehicle can better adapt to the application requirements in the complex electromagnetic environment.

[0200] Based on the above description, refer to Figure 2 As shown, it is a schematic framework diagram of a method for detecting and classifying interference of an unmanned aerial vehicle based on a dynamic hierarchical feature deep fusion network provided by an embodiment of the present invention. Among them, first, start at 201; second, execute 202 using a hierarchical feature processor, which specifically includes the following processes: 1. Obtain the signal data set received by the unmanned aerial vehicle receiving module and perform normalization processing on this signal data set; 2. Calculate the importance of each signal feature corresponding to the signal data set in interference detection; 3. Based on the importance calculated in 2, through a clustering algorithm, group the signal features with close importance to achieve the classification of multiple signal features, thereby obtaining multi-level signal features; then, execute 203 using an interference detection and classification module, which specifically includes the following processes: According to different levels, correspondingly input the multi-level signal features output by the hierarchical feature processor into hierarchical feature convolutional neural networks with different structural complexities for preprocessing to obtain preprocessed data, and input the preprocessed data output by each hierarchical feature convolutional neural network into an independent convolutional neural network for data splicing to obtain the interference detection and classification results of the unmanned aerial vehicle; finally, end at 204.

[0201] Correspondingly, refer to Figure 3 As shown, it is a schematic flow diagram of a method for detecting and classifying interference of an unmanned aerial vehicle based on a dynamic hierarchical feature deep fusion network provided by an embodiment of the present invention. The steps included are as follows:

[0202] 301. Obtain the signal characteristics corresponding to the data of the UAV receiving end, including: subcarrier spacing, meta-time, subcarriers, CP length, received power, signal-to-noise ratio, signal power, and noise power.

[0203] 302. Normalize multiple signal characteristics. Here, a hierarchical feature processor can be used to normalize multiple signal characteristics.

[0204] 303. The hierarchical feature processor performs hierarchical processing on multiple signal characteristics. Here, clustering analysis can be performed on multiple normalized signal characteristics to achieve hierarchical processing.

[0205] 304. Feed the signal characteristics of different levels into the corresponding hierarchical feature convolutional neural network. Here, the number of levels of the hierarchical feature convolutional neural network is the same as the number of levels corresponding to multiple signal characteristics; it can be to feed the signal characteristics of different levels into their respective matching hierarchical feature convolutional neural networks for preprocessing according to different levels.

[0206] 306. Through an independent hierarchical feature convolutional neural network, splice the results obtained by multiple hierarchical feature convolutional neural networks in the previous step to obtain the corresponding interference detection classification results.

[0207] The UAV interference detection and classification method provided by the embodiments of the present invention can support real-time or near-real-time interference detection and classification, and is applicable to the UAV flight safety monitoring system; on the one hand, by dynamically grading multiple signal characteristics of the UAV, it can accurately identify the importance of signal characteristics in interference detection. On the one hand, combining convolutional neural networks for multi-level signal feature analysis and deep fusion can improve the accuracy and robustness of UAV interference detection and classification, so that the UAV can better adapt to the application requirements in complex electromagnetic environments; in other words, the embodiments of the present invention provide a UAV interference detection and classification method based on a dynamic hierarchical feature deep fusion network, aiming to improve the anti-interference ability of the UAV communication system and ensure the normal operation of the UAV in complex electromagnetic environments.

[0208] Embodiment 2:

[0209] Based on the same inventive concept, the present invention also provides a UAV interference detection and classification system, as Figure 4 shown, which is a schematic diagram of the composition of a UAV interference detection and classification system provided by an embodiment of the present invention. Among them, the UAV interference detection and classification system 400 includes:

[0210] The hierarchical feature processor 401 is configured to obtain a signal data set received by the drone; analyze a plurality of signal features corresponding to the signal data set to obtain the importance of the plurality of signal features in interference detection; based on the importance, perform dynamic hierarchical processing on the plurality of signal features to obtain multi-level signal features.

[0211] The interference detection classification module 402 is configured to use a convolutional neural network to analyze and fuse the multi-level signal features to obtain an interference detection classification result of the drone.

[0212] Optionally, the importance of each signal feature in interference detection is determined based on the influence value of each signal feature on the uniform distribution degree of the signal data set.

[0213] Optionally, the hierarchical feature processor is specifically configured to obtain an original uniformity index of the signal data set; obtain a partition uniformity index of a partition data set corresponding to each signal feature, where the partition data set corresponding to each signal feature is obtained by partitioning the signal data set using each signal feature; based on each signal feature, subtract the partition uniformity index of the partition data set corresponding to the signal feature from the original uniformity index to obtain the influence value of each signal feature on the uniform distribution degree of the signal data set.

[0214] Optionally, the hierarchical feature processor is specifically configured to obtain the data volume of the signal data set and the number of sub-signal data of different categories in the signal data set; calculate the ratio between the number of sub-signal data of each category in the signal data set and the data volume to obtain a first ratio set; subtract the square of each first ratio in the first ratio set from a specific value to obtain the original uniformity index.

[0215] Optionally, the partition data set includes: a plurality of first sub-data sets; the hierarchical feature processor is specifically configured to determine the uniformity of each first sub-data set; calculate the ratio between the number of sub-data of each first sub-data set and the data volume of the signal data set to obtain the quantity ratio of each first sub-data set; multiply the quantity ratio of each first sub-data set by the uniformity to obtain an intermediate value of each first sub-data set; fuse the intermediate values of each first sub-data set to obtain the partition uniformity index.

[0216] Optionally, the signal data set includes: a plurality of second sub-data sets; the hierarchical feature processor is specifically configured to accumulate, based on each of the signal features, the influence values of the signal features on the uniform distribution degree of each of the second sub-data sets, to obtain the total sum of the uniformity improvement amounts of each of the signal features; for each of the signal features, the ratio between the total sum of the uniformity improvement amounts of the signal feature and the number corresponding to the plurality of second sub-data sets is the importance degree of the corresponding signal feature in interference detection.

[0217] Optionally, the convolutional neural network includes: an independent hierarchical feature convolutional neural network, a multi-level feature convolutional neural network; the number of levels of the multi-level feature convolutional neural network is determined by the number of levels in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network; the interference detection classification module 402 is further configured to input the multi-level signal features into the multi-level feature convolutional neural network in the order of levels correspondingly, to obtain the preprocessed feature data output by each level feature convolutional neural network in the multi-level feature convolutional neural network; and input the preprocessed feature data output by each level feature convolutional neural network into the independent hierarchical feature convolutional neural network for fusion, to obtain the interference detection classification result.

[0218] Optionally, the independent hierarchical feature convolutional neural network sequentially includes: an input layer, a first convolutional layer, a second convolutional layer, a flattening layer, a fully connected layer, a merging layer introducing a self-attention mechanism, and a final fully connected layer.

[0219] Optionally, the system further includes: a construction module of the convolutional neural network, configured to obtain a plurality of historical signal features corresponding to a historical signal data set; wherein, the historical signal data set is the signal received by the unmanned aerial vehicle in a historical time period; perform dynamic grading processing on the plurality of historical signal features based on the importance degree of the plurality of historical signal features in interference detection, to obtain multi-level historical signal features; and use the multi-level historical signal features to perform iterative training and performance evaluation on an initial neural network, to obtain the convolutional neural network.

[0220] Optionally, the hierarchical feature processor includes: a preprocessing unit, configured to perform normalization processing on the signal data set to obtain a normalized signal set; and perform signal parsing on the normalized signal set to obtain the plurality of signal features.

[0221] It should be noted that the description of the UAV interference detection and classification system 400 is similar to the description of the above-mentioned embodiments of the UAV interference detection and classification method, and has beneficial effects similar to those of the method embodiments. For the technical details not disclosed in the system embodiments of the present invention, please refer to the description of the method embodiments of the present invention for understanding.

[0222] Embodiment 3:

[0223] Based on the same inventive concept, as Figure 5 shown, the present invention also provides an electronic device, which may be a computer device, a single-chip microcomputer device, a smart mobile device, etc. The electronic device in this embodiment may include a processor 510, a memory 520, a transceiver component 530, etc. The processor 510, the memory 520, and the transceiver component 530 are connected through a bus 540; the memory 520 can be used to store an execution program, and an exemplary execution program may include instructions; the processor 510 is used to execute the instructions stored in the memory. The memory 520 can also be used to store data, and the data can be called and / or modified when the instructions are executed.

[0224] The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to implement the UAV interference detection and classification method in the above embodiments.

[0225] Embodiment 4:

[0226] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device-readable storage medium (Memory). The electronic device-readable storage medium is a memory device in the electronic device and is used to store programs and data. It can be understood that the storage medium here can include both the built-in storage medium in the electronic device and, of course, the extended storage medium supported by the electronic device. The storage medium provides a storage space, and this storage space stores the operating system of the terminal. Moreover, in this storage space, one or more instructions suitable for being loaded and executed by the processor are also stored. These instructions can be one or more executable programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium to implement the drone interference detection and classification method in the above embodiments.

[0227] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0228] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0229] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0230] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the functions specified in one process or a plurality of processes and / or boxes Figure 1 one process or a plurality of processes and / or boxes Figure 1 steps in one box or a plurality of boxes.

[0231] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific implementation manners of the present invention, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. A method for detecting and classifying UAV interference, characterized in that, The method includes: Obtaining a signal data set received by the UAV; Analyzing multiple signal features corresponding to the signal data set to obtain the importance of the multiple signal features in interference detection; Based on the importance, performing dynamic hierarchical processing on the multiple signal features to obtain multi-level signal features; Using a convolutional neural network to analyze and fuse the multi-level signal features to obtain the interference detection classification result of the UAV.

2. The method according to claim 1, characterized in that, The importance of each signal feature in interference detection is determined based on the influence value of each signal feature on the uniform distribution degree of the signal data set.

3. The method according to claim 2, wherein The obtaining method of the influence value of each signal feature on the uniform distribution degree of the signal data set includes: Obtaining the original uniformity index of the signal data set; Obtaining the partition uniformity index of the partition data set corresponding to each signal feature; wherein, the partition data set corresponding to each signal feature is obtained by partitioning the signal data set using each signal feature; Based on each signal feature, subtracting the original uniformity index from the partition uniformity index of the partition data set corresponding to the signal feature to obtain the influence value of each signal feature on the uniformity of the signal data set.

4. The method according to claim 3, characterized in that, The obtaining of the original uniformity index of the signal data set includes: Obtaining the data volume of the signal data set and the number of sub-signal data of different categories in the signal data set; Calculating the ratio between the number of sub-signal data of each category in the signal data set and the data volume to obtain a first ratio set; Subtracting a specific value from the square of each first ratio in the first ratio set to obtain the original uniformity index.

5. The method according to claim 3 or 4, characterized in that, The partition data set includes: a plurality of first sub-data sets; the obtaining of the partition uniformity index of the partition data set corresponding to each signal feature includes: Determining the uniformity of each first sub-data set; Calculating the ratio between the sub-data volume of each first sub-data set and the data volume of the signal data set to obtain the quantity ratio of each first sub-data set; Multiplying the quantity ratio of each first sub-data set by the uniformity to obtain the intermediate value of each first sub-data set; Fusing the intermediate values of each first sub-data set to obtain the partition uniformity index.

6. The method according to claim 1 or 2, characterized in that, The signal data set includes: a plurality of second sub-data sets; the analyzing of multiple signal features corresponding to the signal data set to obtain the importance of the multiple signal features in interference detection includes: Based on each signal feature, accumulating the influence values of the signal feature on the uniform distribution degree of each second sub-data set to obtain the total sum of the uniformity improvement amount of each signal feature; For each signal feature, the ratio between the total sum of the uniformity improvement amount of the signal feature and the number corresponding to the plurality of second sub-data sets is the importance of the corresponding signal feature in interference detection.

7. The method according to claim 1, characterized in that, The convolutional neural network includes: an independent hierarchical feature convolutional neural network, a multi-level feature convolutional neural network; the number of levels of the multi-level feature convolutional neural network is determined by the number of levels in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network; Using the convolutional neural network to analyze and fuse the multi-level signal features to obtain the interference detection classification result of the drone, including: Inputting the multi-level signal features into the multi-level feature convolutional neural network correspondingly in the order of levels to obtain the preprocessed feature data output by each level feature convolutional neural network in the multi-level feature convolutional neural network; Inputting the preprocessed feature data output by each level feature convolutional neural network into the independent hierarchical feature convolutional neural network for fusion to obtain the interference detection classification result.

8. The method according to claim 7, wherein The independent hierarchical feature convolutional neural network sequentially includes: an input layer, a first convolutional layer, a second convolutional layer, a flattening layer, a fully connected layer, a merging layer introducing a self-attention mechanism, and a final fully connected layer.

9. The method according to claim 1, wherein The construction process of the convolutional neural network includes: Obtaining a plurality of historical signal features corresponding to a historical signal dataset; wherein, the historical signal dataset is the signal received by the drone in a historical time period; Performing dynamic grading processing on the plurality of historical signal features based on the importance of the plurality of historical signal features in interference detection to obtain multi-level historical signal features; Using the multi-level historical signal features to perform iterative training and performance evaluation on an initial neural network to obtain the convolutional neural network.

10. The method according to claim 1, characterized in that, Before analyzing the plurality of signal features corresponding to the signal dataset to obtain the importance of the plurality of signal features in interference detection, the method further includes: Performing normalization processing on the signal dataset to obtain a normalized signal set; Performing signal parsing on the normalized signal set to obtain the plurality of signal features.

11. A UAV interference detection and classification system, characterized in that, The system includes: A hierarchical feature processor, configured to obtain a signal dataset received by the drone; analyze the plurality of signal features corresponding to the signal dataset to obtain the importance of the plurality of signal features in interference detection; perform dynamic grading processing on the plurality of signal features based on the importance to obtain multi-level signal features; An interference detection classification module, configured to use a convolutional neural network to analyze and fuse the multi-level signal features to obtain the interference detection classification result of the drone.

12. The system according to claim 11, wherein The importance of each signal feature in interference detection is determined based on the influence value of each signal feature on the uniform distribution degree of the signal dataset.

13. The system according to claim 12, wherein, The hierarchical feature processor is specifically configured to obtain the original uniformity index of the signal data set; obtain the partition uniformity index of the partition data set corresponding to each signal feature, where the partition data set corresponding to each signal feature is obtained by partitioning the signal data set using each signal feature; based on each signal feature, subtract the partition uniformity index of the partition data set corresponding to the signal feature from the original uniformity index to obtain the influence value of each signal feature on the uniformity of the signal data set.

14. The system according to claim 13, wherein The hierarchical feature processor is specifically configured to obtain the data volume of the signal data set and the number of sub-signal data of different categories in the signal data set; calculate the ratio between the number of sub-signal data of each category in the signal data set and the data volume to obtain a first ratio set; subtract a specific value from the square of each first ratio in the first ratio set to obtain the original uniformity index.

15. The system according to claim 13 or 14, characterized in that The partition data set includes: a plurality of first sub-data sets; The hierarchical feature processor is specifically configured to determine the uniformity of each first sub-data set; calculate the ratio between the sub-data volume of each first sub-data set and the data volume of the signal data set to obtain the quantity ratio of each first sub-data set; multiply the quantity ratio of each first sub-data set by the uniformity to obtain the intermediate value of each first sub-data set; fuse the intermediate values of each first sub-data set to obtain the partition uniformity index.

16. The system according to claim 11 or 12, characterized in that, The signal data set includes: a plurality of second sub-data sets; The hierarchical feature processor is specifically configured to, based on each signal feature, accumulate the influence values of the signal feature on the uniform distribution degree of each second sub-data set to obtain the total sum of the uniformity improvement amounts of each signal feature; for each signal feature, the ratio between the total sum of the uniformity improvement amounts of the signal feature and the quantity corresponding to the plurality of second sub-data sets is the importance of the corresponding signal feature in interference detection.

17. The system according to claim 11, wherein The convolutional neural network includes: an independent hierarchical feature convolutional neural network, a multi-level feature convolutional neural network; the number of layers of the multi-level feature convolutional neural network is determined by the number of layers in the multi-level signal features, and the network structure complexity of the high-level feature convolutional neural network in the multi-level feature convolutional neural network is higher than that of the low-level feature convolutional neural network; The interference detection classification module is specifically configured to input the multi-level signal features into the multi-level feature convolutional neural network in the order of layers to obtain the preprocessed feature data output by each level feature convolutional neural network in the multi-level feature convolutional neural network; input the preprocessed feature data output by each level feature convolutional neural network into the independent hierarchical feature convolutional neural network for fusion to obtain the interference detection classification result.

18. The system according to claim 17, wherein The independent hierarchical feature convolutional neural network sequentially includes: an input layer, a first convolutional layer, a second convolutional layer, a flattening layer, a fully connected layer, a merging layer introducing self-attention mechanism, and a final fully connected layer.

19. The system according to claim 11, wherein The system further includes: A building block of a convolutional neural network for obtaining multiple historical signal features corresponding to a historical signal dataset; wherein the historical signal dataset is the signal received by the drone in a historical time period; based on the importance of the multiple historical signal features in interference detection, dynamically classify the multiple historical signal features to obtain multi-level historical signal features; use the multi-level historical signal features to iteratively train and evaluate the performance of an initial neural network to obtain the convolutional neural network.

20. The system according to claim 11, characterized in that The hierarchical feature processor includes: A preprocessing unit for normalizing the signal dataset to obtain a normalized signal set; and parsing the normalized signal set to obtain the multiple signal features.

21. An electronic device, characterized in that, Including: At least one processor and a memory; The memory and the processor are connected by a bus; The memory is used for storing one or more programs; When the one or more programs are executed by the at least one processor, the drone interference detection and classification method according to any one of claims 1 to 10 is implemented.

22. A readable storage medium, characterized in that, A program is stored thereon, and when the executed program is executed, the drone interference detection and classification method according to any one of claims 1 to 10 is implemented.

Images