Background
Along with the wide use of civilian unmanned aerial vehicle of present society, when bringing a great deal of facility for the human society, also can bring some unexpected circumstances, unmanned aerial vehicle out of control can lead to the unexpected injury of pedestrian or building. Therefore, there is a need for a mechanism to detect and identify an unmanned aerial vehicle, thereby distinguishing legitimate from illegitimate unmanned aerial vehicles and improving management and control efficiency.
The main existing unmanned aerial vehicle detection technologies include radio frequency spectrum monitoring, radar detection, sound wave identification, visible light/infrared detection and the like. The radio frequency spectrum detection technology mainly utilizes radio monitoring equipment to detect wireless signals of a target unmanned aerial vehicle, and finds and identifies the target through comprehensive analysis and processing. The radar detection technology of the unmanned aerial vehicle utilizes a radar scanning technology to realize the detection of the unmanned aerial vehicle according to the reflection wave phenomenon generated when electromagnetic waves pass through different transmission media. Unmanned aerial vehicle is when flying, and its motor work all can produce the noise of certain degree with rotor vibrations to the noise that every unmanned aerial vehicle produced has the uniqueness, can regard as unmanned aerial vehicle's "audio frequency fingerprint". The sound wave identification technology of the unmanned aerial vehicle just utilizes 'audio fingerprints' to discover and detect the unmanned aerial vehicle. The visible light/infrared detection is to detect the unmanned aerial vehicle by using visible light or thermal infrared reflection of a target and adopting a visible light camera and an infrared thermal imager sensor combination with beyond visual range and high definition. The advantages and disadvantages of the detection and identification of the unmanned aerial vehicle by using the technology are shown in fig. 1, and as can be seen from fig. 1, the equipment of the traditional unmanned aerial vehicle detection technology is complex and the identification precision is not high.
The unmanned aerial vehicle detection and identification technology based on the radio frequency fingerprint provided by the invention not only can solve the problem of complex equipment in the conventional unmanned aerial vehicle detection system, but also has high identification precision. The radio frequency fingerprint refers to a characteristic that a radio frequency circuit still generates tiny randomness on the premise of ensuring the product to be qualified in the process of manufacturing the wireless communication equipment. These features are unique, universal, robust, and short-time invariant, and a biometric-like fingerprint that uniquely identifies an individual is referred to as a radio frequency fingerprint. Because unmanned aerial vehicle just adopts wireless communication to carry out the information interaction, consequently unmanned aerial vehicle based on radio frequency fingerprint technique surveys and has very big feasibility with the discernment, and can discern unmanned aerial vehicle identity effectively, can improve unmanned aerial vehicle's management and control efficiency.
Disclosure of Invention
The invention aims to provide an unmanned aerial vehicle detection and identification method based on radio frequency fingerprints, which effectively solves the problems that the existing unmanned aerial vehicle detection and identification equipment is complex in system and low in identification efficiency.
The invention provides the following technical scheme:
an unmanned aerial vehicle detection and identification method based on radio frequency fingerprints comprises the following steps:
and the signal receiving module is used for acquiring wireless signals through the radio frequency front end, and performing a series of preprocessing on the received signals and storing the preprocessed signals.
And the signal processing module is used for carrying out comprehensive analysis processing on the stored signals and realizing the detection of the unmanned aerial vehicle.
The unmanned aerial vehicle identification module is used for selecting the characteristics of the detected unmanned aerial vehicle signals and extracting the signal characteristics, the extracted signal characteristics are used as the fingerprint of the unmanned aerial vehicle, and the unmanned aerial vehicle identification algorithm is used for realizing the classification and identification of the unmanned aerial vehicle.
An unmanned aerial vehicle detection and identification method based on radio frequency fingerprints comprises the following steps:
step 1, receiving a wireless signal through a radio frequency front end. The method specifically comprises the following steps: the 2.4GHz omnidirectional antenna is used for receiving wireless signals, the wireless signals received by the antenna are amplified by the 2.4GHz low-noise high-frequency amplifier and then sent to the filter for filtering, the filtered signals are subjected to down-conversion processing by the mixer, and relatively stable intermediate-frequency signal frequency spectrum information is output.
And 2, processing based on the signals obtained in the step 1, and if the signal amplitude of a certain frequency band of the received signals is stably larger than a preset threshold value sigma, judging that signals enter and marking the signals as suspected unmanned aerial vehicle signals. The suspected signal is judged, the wireless signal of the unmanned aerial vehicle is a periodic signal, the autocorrelation function of the periodic signal still has periodicity, and the autocorrelation function of the interference signals such as noise does not have periodicity, so that the suspected signal can be judged by calculating whether the autocorrelation function of the suspected signal has periodicity, and the detection of the unmanned aerial vehicle is realized.
And 3, acquiring data based on the unmanned aerial vehicle signals obtained in the step 2, and dividing the data into an off-line stage and an on-line stage. An off-line stage: the signal feature extraction algorithm is used to extract the drone signal features as the "fingerprint" for the drone device. Finally, storing the extracted fingerprint data and the label representing the unmanned aerial vehicle category as training data; an online stage: and (4) extracting the signal characteristics of the unmanned aerial vehicle signals obtained in the step (2), and storing the signal characteristics as test data.
And 4, based on the unmanned aerial vehicle signal training data and the test data obtained in the step 3, utilizing an unmanned aerial vehicle classification recognition algorithm to realize classification recognition of the unmanned aerial vehicle.
Preferably, the unmanned aerial vehicle signal feature extraction algorithm includes the following steps:
s1: converting the signal x (t) to xI(t)+xQ(t) as input signal, modulating by I/Q quadrature modulator to obtain output signal s (t), wherein XI(t) is the in-phase component, XQ(t) is the quadrature component. Obtaining the complex number of S (t) and simplifying the complex number to SB(t)=αx(t)+βx*(t) where α ═ cos θ + jssin θ, β ═ cos θ + jssin θ, and θ are the gain mismatch and phase mismatch parameters of the quadrature modulator, respectively.
S2: for S in S1
B(t) analysis was carried out. Defining a received signal
Complex conjugation is taken from r (t) to obtain r
*(t)=β
*x(t)+α
*x
*(t) definition of
Calculating an autocorrelation matrix R of Y (t)
Y=E[Y(t)Y(t)
H]To R, to R
YIs simplified to obtain
Wherein sigma
x 2Is the energy of x (t) (. sigma.)
s 2Is S
B(t) energy.
S3: based on the autocorrelation matrix R obtained in S2
YDefinition of
Substituting alpha and beta for simplification
As can be seen from the result of the simplification,
dependent only on theta, the phase mismatch theta and gain mismatch due to I/Q mismatch are different for different drone devices, and thus for different drones
The values are different, then
The value of (c) is taken as the "fingerprint" of the drone device.
Preferably, the K nearest neighbor algorithm of the classification and identification algorithm of the unmanned aerial vehicle includes the following steps:
s1: and after the training data S and the data T to be tested of the unmanned aerial vehicle are obtained, loading the S and the T into an algorithm, and designating the value of K as K.
S2: based on the data S and T of S1, the data is normalized to obtain new training data S 'and new test data T', where the normalization formula is X '═ X-minX)/(maxX-minX), where X is the original data, X' is the new data after X normalization, and maxX and minX are the maximum and minimum values in X, respectively.
S3: selecting data T in the data T ' to be tested based on the new data S ' and T ' after normalization in S2, and calculating data S from T to SiOf Euclidean distance diWhere i ═ n (1, 2,. n), and n is the total number of data in S'.
S4: based on Euclidean distance d obtained in S3iTo d is pairediSorting in ascending order to obtain a distance set D ═ D1,d2,...,dn) And n is the total number of data in S'.
S5: based on the distance D in S4, the first k euclidean distances in D are selected to correspond to the data S "in the training data S', respectively.
S6: and counting the number of classes in S 'based on the data S' in S5, and taking the class label with the highest occurrence frequency as the class of the test data T, so that the label is the class of the unmanned aerial vehicle to be tested, namely the identification of the unmanned aerial vehicle is completed.
Advantageous effects
The invention has the beneficial effects that: the invention firstly uses a radio frequency front-end platform to receive signals, and specifically comprises the following steps: receiving wireless signals of different frequency points from 2.412GHz to 2.472GHz by using a 2.4GHz omnidirectional antenna, and preprocessing the received signals; detecting a suspected unmanned aerial vehicle signal by comparing whether the amplitude of the received signal is stably larger than a preset threshold value sigma or not, and analyzing whether an autocorrelation function of the suspected signal has periodicity or not to realize the detection of the unmanned aerial vehicle; after the unmanned aerial vehicle is successfully detected, signal characteristics of the unmanned aerial vehicle are extracted by using a signal characteristic extraction algorithm to serve as the fingerprint of unmanned aerial vehicle equipment; and finally, using an unmanned aerial vehicle classification recognition algorithm to realize classification recognition of the unmanned aerial vehicle. The extracted signal features have uniqueness and short-time invariance, and different unmanned aerial vehicle devices can be well distinguished. The unmanned aerial vehicle classification recognition algorithm uses a K nearest neighbor algorithm, the Euclidean distance is selected in the algorithm to calculate the distance between an unknown point and a known point, the recognition rate of the algorithm is the highest by setting different K values, and the problems of complex system and low recognition efficiency in the existing unmanned aerial vehicle detection and recognition equipment are solved.
Detailed Description
The invention will be further described with reference to the accompanying drawings in which:
the invention provides an unmanned aerial vehicle detection and identification method based on radio frequency fingerprints, which is applied to an unmanned aerial vehicle detection and identification system, and specifically the unmanned aerial vehicle detection and identification system comprises: signal receiving module, signal processing module and unmanned aerial vehicle identification module
As shown in fig. 2, a method for detecting and identifying an unmanned aerial vehicle based on a radio frequency fingerprint specifically includes the following steps:
step 1, receiving a wireless signal through a radio frequency front end. The method specifically comprises the following steps: the 2.4GHz omnidirectional antenna is used for receiving wireless signals, the wireless signals received by the antenna are amplified by the 2.4GHz low-noise high-frequency amplifier and then sent to the filter for filtering, the filtered signals are subjected to down-conversion processing by the mixer, and relatively stable intermediate-frequency signal frequency spectrum information is output.
And 2, processing based on the signals obtained in the step 1, and if the signal amplitude of a certain frequency band of the received signals is stably larger than a preset threshold value sigma, judging that signals enter and marking the signals as suspected unmanned aerial vehicle signals. Whether the suspected signal is an unmanned aerial vehicle signal is judged by calculating whether the autocorrelation function of the suspected signal has periodicity, so that the unmanned aerial vehicle detection is realized.
And 3, performing data acquisition based on the unmanned aerial vehicle signals obtained in the step 1-2, wherein the data acquisition is divided into training data acquisition in an online stage and test data acquisition in an offline stage. The method comprises the following specific steps: the received signal is r (t) and order
r
*(t) calculating the autocorrelation function of Y (t) for the complex conjugate of r (t)
Wherein sigma
x 2Is the energy, σ, of the ideal signal x (t)
s 2Is the energy of r (t). Defining signal characteristics
Will be provided with
The value of (a) is used as the radio frequency 'fingerprint' of the unmanned aerial vehicle device, and the acquired training data and the test data are respectively stored.
And 4, based on the unmanned aerial vehicle signal training data and the test data obtained in the step 3, carrying out classification and identification on the unmanned aerial vehicle by using a K nearest neighbor algorithm. The flow chart of the K-nearest neighbor algorithm is shown in fig. 3, and the specific steps are as follows: (1) for dataAnd normalizing the set to obtain new training data S 'and new test data T', wherein the normalization formula is X '═ X-minX)/(maxX-minX), wherein X is original data, X' is new data after normalization, and maxX and minX are respectively the maximum value and the minimum value in X. (2) Selecting data T in the data T' to be tested, and calculating data S from T to SiOf Euclidean distance diWhere i is (1, 2, … n), and n is the total number of data in S'. (3) To diSorting in ascending order to obtain a distance set D ═ D1,d2,...,dn) And n is the total number of data in S'. (4) And selecting the first k Euclidean distances in the D to respectively correspond to the data points S 'in the training set S'. (5) And (4) counting the number of the classes in the s' and taking the class label with the highest frequency of occurrence as the class of the test data T, so that the label is the class of the unmanned aerial vehicle to be tested, and the identification of the unmanned aerial vehicle is completed.