CN103914703B - Classification and identification method for pedestrian and vehicle micro-motion targets - Google Patents
Classification and identification method for pedestrian and vehicle micro-motion targets Download PDFInfo
- Publication number
- CN103914703B CN103914703B CN201410067084.0A CN201410067084A CN103914703B CN 103914703 B CN103914703 B CN 103914703B CN 201410067084 A CN201410067084 A CN 201410067084A CN 103914703 B CN103914703 B CN 103914703B
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- vehicle
- dimensional spectrogram
- spectrogram
- preprocessed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012706 support-vector machine Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims description 28
- 238000012360 testing method Methods 0.000 claims description 17
- 230000001629 suppression Effects 0.000 claims description 15
- 238000005070 sampling Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 230000035485 pulse pressure Effects 0.000 claims description 8
- 238000009825 accumulation Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 230000006835 compression Effects 0.000 claims description 4
- 238000007906 compression Methods 0.000 claims description 4
- 238000001228 spectrum Methods 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 description 5
- 230000007547 defect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
Landscapes
- Traffic Control Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention belongs to the technical field of micro-motion target classification and identification, and discloses a classification and identification method for pedestrian and vehicle micro-motion targets. The classification and identification method for the pedestrian and vehicle micro-motion targets includes the following steps that firstly, clutter rejection is carried out on pedestrian and vehicle echo signals by means of a CLEAN algorithm; secondly, time-frequency analysis is carried out on the echo signals after clutter rejection, proper spectrum energy threshold values are selected, windowing preprocessing is conducted on the proper spectrum energy threshold values, redundant information is removed, and image features are enhanced; finally, textural features of a spectrogram are extracted to serve as effective characteristic quantities, the effective characteristic quantities are combined with characteristic quantities extracted from a spectrogram obtained through pedestrian and vehicle real-time measured data by means of a support vector machine classification method, and the pedestrian and vehicle targets can be precisely classified.
Description
Technical Field
The invention belongs to the technical field of micro-motion target classification and identification, and particularly relates to a classification and identification method for micro-motion targets of pedestrians and vehicles.
Background
In the moving process of the radar target, the movement of the mass center exists, and the micro-movement of rotation, swing, oscillation and the like relative to the target is included, for example, the pedestrian and the vehicle not only have the movement of the mass center, but also the arm swing and the leg movement of the pedestrian and the wheel type wheel, the crawler type wheel and the crawler rotation of the vehicle respectively generate the micro-movement relative to the mass center of the pedestrian and the mass center of the vehicle. Due to the micro-motion of the target, the radar echo signal is subjected to micro-motion modulation on the basis of Doppler modulation, and the characteristics are extracted according to the micro-motion information of the target, so that the classification and identification of pedestrians and vehicles are realized.
For ground monitoring radars, the important targets for monitoring are pedestrian and vehicle targets, and the radars have important application values in the aspects of square monitoring, airport surface monitoring, battlefield monitoring and situation perception. The method is an important development direction for identifying ground micro-motion targets and ground surveillance radar targets by using micro-motion characteristics extracted by the micro-motion effect of pedestrians and vehicles. The characteristics of strong scattering number extracted by a scanning beam radar based on short residence time by using a pedestrian and vehicle time frequency distribution spectrogram are analyzed in a document of 'Bayesian classification of humans and vehicles using Micro-Doppler signals from a scanning-beam radar' (IEEEMicrowave and Wireless Compounds Letters, Vol. 19, No. 5, 2009) by Nanzer J A and Rogers R L and the like, and a Bayesian classifier is designed for classification and identification, although a better classification result is obtained under certain conditions. However, because the dwell time is short and no fine preprocessing is performed on the fine motion target features of the spectrogram, the doppler resolution of the target and the accuracy of extracting the fine motion information are generally limited.
Disclosure of Invention
The invention aims to provide a method for classifying and identifying micro-motion targets of pedestrians and vehicles. The invention overcomes the defects that the non-stationary characteristic of a micro-motion target signal cannot be fully described only by using Doppler one-dimensional information and the resolution ratio in a time-frequency spectrogram of a scanning beam radar based on short residence time is low in the prior art, obtains the time-frequency spectrogram by using a continuous wave radar to perform short-time Fourier transform on echoes of a pedestrian and a vehicle target, performs preprocessing, extracts effective textural features, and realizes classification and identification of the pedestrian and the vehicle target by using a support vector machine classifier.
In order to achieve the technical purpose, the invention is realized by adopting the following technical scheme.
A classification and identification method for pedestrian and vehicle micro-motion targets comprises the following steps:
s1: respectively receiving pedestrian echo data and vehicle echo data by using a radar to obtain a pedestrian echo sample set and a vehicle echo sample set;
s2: selecting a pedestrian echo sample from the pedestrian echo sample set, and performing clutter suppression on the selected pedestrian echo sample to obtain clutter suppressed pedestrian echo data; selecting a vehicle echo sample from a vehicle echo sample set, and performing clutter suppression on the selected vehicle echo sample to obtain clutter suppressed vehicle echo data;
s3: carrying out short-time Fourier transform on the pedestrian echo data after clutter suppression to obtain a two-dimensional spectrogram of the pedestrian; preprocessing a two-dimensional spectrogram of the pedestrian to obtain a preprocessed two-dimensional spectrogram of the pedestrian; performing short-time Fourier transform on the clutter suppressed vehicle echo data to obtain a vehicle two-dimensional spectrogram; preprocessing a vehicle two-dimensional spectrogram to obtain a preprocessed vehicle two-dimensional spectrogram;
s4: respectively extracting texture features of the preprocessed pedestrian two-dimensional spectrogram and the preprocessed vehicle two-dimensional spectrogram;
s5: repeating the steps S2 to S4 until all pedestrian echo samples in the pedestrian echo sample set and all vehicle echo samples in the vehicle echo sample set are traversed to obtain a pedestrian two-dimensional spectrogram texture feature sample set and a vehicle two-dimensional spectrogram texture feature sample set;
s6: and training and testing through a support vector machine according to the pedestrian two-dimensional spectrogram texture feature sample set and the vehicle two-dimensional spectrogram texture feature sample set to obtain a corresponding classification recognition result.
The invention has the beneficial effects that:
firstly, the invention adopts a time-frequency domain joint analysis method of short-time Fourier transform to describe the target echo non-stationary signal, overcomes the defect that the correct recognition rate is low under the condition of low signal-to-noise ratio by only using a Doppler one-dimensional signal in the prior art, and has the advantages of comprehensiveness and stability.
Secondly, the two-dimensional spectrogram is subjected to windowing processing, redundant information is eliminated, and the defects that the extraction features are not obvious and the calculation amount is large due to the fact that the whole image is used in the prior art are overcome, so that the two-dimensional spectrogram has the advantages of being obvious in features and reducing the calculation amount.
Thirdly, the method extracts the spectrum entropy from the preprocessed two-dimensional spectrogram, counts texture features of the third moment and the directivity of the histogram and fully utilizes the spatial information of the image, so that the extracted texture features have the advantages of higher accuracy and accordance with the identification rule of biological vision.
Drawings
FIG. 1 is a flow chart of a method for classifying and identifying micro-motion targets of pedestrians and vehicles according to the present invention;
FIG. 2a is a two-dimensional spectrogram of a pedestrian of measured data;
FIG. 2b is a two-dimensional spectrogram of the measured data of the vehicle;
FIG. 2c is a diagram of a two-dimensional pedestrian spectrogram of measured data after being preprocessed;
fig. 2d is a schematic diagram of the classification result of the measured data.
Detailed Description
The invention will be further described with reference to the accompanying drawings in which:
referring to fig. 1, a flow chart of a method for classifying and identifying a micro-motion target of a pedestrian and a vehicle according to the present invention is shown. The classification and identification method of the pedestrian and vehicle micro-motion target comprises the following steps:
s1: and respectively receiving pedestrian echo data and vehicle echo data by utilizing a radar to obtain a target echo sample set, wherein the target echo sample set comprises a pedestrian echo sample set and a vehicle echo sample set. Specifically, narrow-band linear frequency modulation continuous wave signals are respectively transmitted to pedestrians and vehicles by utilizing a radar, and pedestrian echo data and vehicle echo data are respectively received by adopting a Dechirp receiving system.
S2: selecting a pedestrian echo sample from the pedestrian echo sample set, and performing clutter suppression on the selected pedestrian echo sample to obtain clutter suppressed pedestrian echo data; and selecting a vehicle echo sample from the vehicle echo sample set, and performing clutter suppression on the selected vehicle echo sample to obtain clutter suppressed vehicle echo data. The concrete description is as follows:
selecting a pedestrian echo sample from the pedestrian echo sample set, performing range-wise pulse compression processing on the selected pedestrian echo sample to obtain pedestrian echo data after pulse pressure processing, and then performing filtering processing on the pedestrian echo data after pulse pressure processing by using a Moving Target Indication (MTI) method to obtain the pedestrian echo data after filtering processing. Selecting a vehicle echo sample from the vehicle echo sample set, performing range-wise pulse compression processing on the selected vehicle echo sample to obtain vehicle echo data after pulse pressure processing, and then performing filtering processing on the vehicle echo data after pulse pressure processing by using a moving target display method to obtain the vehicle echo data after filtering processing.
In the pedestrian echo data after filtering processing, performing azimuth Fourier transform on data of a distance unit with the maximum energy to obtain pedestrian azimuth frequency domain data; and in the filtered vehicle echo data, performing azimuth Fourier transform on the data of the distance unit with the maximum energy to obtain vehicle azimuth frequency domain data.
Clutter is removed from the pedestrian azimuth frequency domain data by adopting a CLEAN algorithm, and then inverse Fourier transform is carried out on the pedestrian azimuth frequency domain data after the clutter removal to obtain pedestrian echo data after clutter suppression; and removing clutter from the vehicle azimuth frequency domain data by adopting a CLEAN algorithm, and performing inverse Fourier transform on the vehicle azimuth frequency domain data after the clutter removal to obtain vehicle echo data after clutter suppression.
S3: carrying out short-time Fourier transform on the pedestrian echo data after clutter suppression to obtain a two-dimensional spectrogram of the pedestrian; preprocessing a two-dimensional spectrogram of the pedestrian to obtain a preprocessed two-dimensional spectrogram of the pedestrian; performing short-time Fourier transform on the clutter suppressed vehicle echo data to obtain a vehicle two-dimensional spectrogram; and preprocessing the vehicle two-dimensional spectrogram to obtain a preprocessed vehicle two-dimensional spectrogram. The concrete description is as follows:
carrying out short-time Fourier transform on the pedestrian echo data after clutter suppression to obtain a two-dimensional spectrogram of the pedestrian; and performing short-time Fourier transform on the clutter suppressed vehicle echo data to obtain a vehicle two-dimensional spectrogram.
Accumulating the energy of the two-dimensional spectrogram of the pedestrian in the time dimension according to the following formula to obtain the frequency dimension energy distribution E of the two-dimensional spectrogram of the pedestrianx(nx):
Wherein S isx(mx,nx) Representing the energy distribution of a two-dimensional spectrogram of a pedestrian, mxNumber of sampling points in time dimension corresponding to two-dimensional spectrogram of pedestrian, nxThe number of sampling points of the frequency dimension corresponding to the two-dimensional spectrogram of the pedestrian is obtained.
According to the following formula, energy accumulation is carried out on the two-dimensional spectrogram of the vehicle in the time dimension to obtain the frequency dimension energy distribution E of the two-dimensional spectrogram of the vehiclec(nc):
Wherein S isc(mc,nc) Representing the energy distribution of a two-dimensional map of the vehicle, mcThe number of sampling points in the time dimension corresponding to the two-dimensional spectrogram of the vehicle, ncIs a vehicleAnd sampling point numbers of frequency dimensions corresponding to the vehicle two-dimensional spectrogram.
Frequency dimension energy distribution E of pedestrian two-dimensional spectrogramx(nx) Peak position of (F)xRespectively selecting frequency units symmetrically towards two sides for a symmetry axis, and performing energy accumulation until the energy sum exceeds 60% of the total energy of the frequency dimension of the pedestrian two-dimensional spectrogram, at the moment, recording the minimum frequency and the maximum frequency in all the selected frequency units, and constructing a rectangular window function corresponding to the pedestrian two-dimensional spectrogram on the basis of the recorded minimum frequency and maximum frequency; and windowing the two-dimensional spectrogram of the pedestrian according to the rectangular window function corresponding to the two-dimensional spectrogram of the pedestrian to obtain the preprocessed two-dimensional spectrogram of the pedestrian.
Frequency dimension energy distribution E of vehicle two-dimension spectrogramc(nc) Peak position of (F)cRespectively selecting frequency units symmetrically towards two sides for a symmetry axis, and performing energy accumulation until the energy sum exceeds 60% of the total energy of the frequency dimension of the two-dimensional spectrogram of the vehicle, at the moment, recording the minimum frequency and the maximum frequency in all the selected frequency units, and constructing a rectangular window function corresponding to the two-dimensional spectrogram of the vehicle on the basis of the recorded minimum frequency and maximum frequency; and windowing the vehicle two-dimensional spectrogram according to the rectangular window function corresponding to the vehicle two-dimensional spectrogram to obtain the preprocessed vehicle two-dimensional spectrogram.
S4: and respectively extracting texture features of the preprocessed pedestrian two-dimensional spectrogram and the preprocessed vehicle two-dimensional spectrogram. The method specifically comprises the following steps of extracting the textural features of the preprocessed two-dimensional spectrogram of the pedestrian:
accumulating the energy of the preprocessed pedestrian two-dimensional spectrogram in a time dimension to obtain frequency dimension energy distribution of the preprocessed pedestrian two-dimensional spectrogram, normalizing the frequency dimension energy distribution of the preprocessed pedestrian two-dimensional spectrogram to obtain normalized data p of the frequency dimension energy distribution of the preprocessed pedestrian two-dimensional spectrogramx(nx) Then, the entropy E of the two-dimensional spectrogram of the pedestrian is calculated according to the following formulax:
According to the energy distribution of the preprocessed two-dimensional spectrogram of the pedestrian, a plurality of equally divided energy intervals are obtained through energy equally dividing processing (such as ten equally dividing processing); obtaining a pedestrian two-dimensional spectrogram energy statistical histogram based on the plurality of equally divided energy intervals, and then calculating a third moment M of the pedestrian two-dimensional spectrogram energy statistical histogramx。
Defining two mutually perpendicular directions HxAnd VxCalculating the gradient vector magnitude of the preprocessed two-dimensional spectrogram of the pedestrian according to the following formulaAnd gradient vector angle thetax:
Wherein,the energy of the pixel points of the preprocessed two-dimensional spectrogram of the pedestrian is HxThe gradient in the direction of the magnetic field,the energy of pixel points of a two-dimensional spectrogram of the preprocessed pedestrian is VxA gradient in direction.
Gradient vector angle theta of preprocessed pedestrian two-dimensional spectrogramxDiscretizing and equally dividing the vector data to obtain a plurality of equally divided gradient vector angle sections, wherein each equally divided gradient vector angle section isi is taken from 2 to L,is thetaxA discretized maximum; obtaining a pedestrian two-dimensional spectrogram directivity statistical histogram based on the gradient vector angle intervals of the plurality of equal divisions
Finding out the gradient vector angle α corresponding to the peak point of the directional statistical histogram of the two-dimensional spectrogram of the pedestrianxCalculating the directional characteristic F of the preprocessed two-dimensional spectrogram of the pedestrian according to the following formulax:
Extraction of Ex、Mx、And FxAnd the obtained data is used as the texture feature of the preprocessed two-dimensional spectrogram of the pedestrian.
In step S4, the step of extracting the texture features of the preprocessed vehicle spectrogram specifically includes the following steps:
accumulating the energy of the preprocessed vehicle two-dimensional spectrogram in a time dimension to obtain frequency dimension energy distribution of the preprocessed vehicle two-dimensional spectrogram, normalizing the frequency dimension energy distribution of the preprocessed vehicle two-dimensional spectrogram to obtain normalized data p of the frequency dimension energy distribution of the preprocessed vehicle two-dimensional spectrogramc(nc) Then, the entropy value E of the two-dimensional spectrogram of the vehicle is calculated according to the following formulax:
According to the energy distribution of the preprocessed vehicle two-dimensional spectrogram, a plurality of equally divided energy intervals are obtained through energy equally dividing processing; based on the plurality of equally divided energy intervals, obtaining a vehicle two-dimensional spectrogram energy statistical histogram, and then calculating a third moment M of the vehicle two-dimensional spectrogram energy statistical histogramc。
Defining two mutually perpendicular directions HcAnd VcCalculating the gradient vector magnitude of the preprocessed vehicle two-dimensional spectrogram according to the following formulaAnd gradient vector angle thetac:
Wherein,the energy of the pixel points of the two-dimensional spectrogram of the preprocessed vehicle is HcThe gradient in the direction of the magnetic field,the energy of pixel points of a two-dimensional spectrogram of the preprocessed vehicle is VcA gradient in direction.
Gradient vector angle theta of preprocessed vehicle two-dimensional spectrogramcDiscretizing and equally dividing the vector data to obtain a plurality of equally divided gradient vector angle sections, wherein each equally divided gradient vector angle section isi is taken from 2 to L,is thetacA discretized maximum; obtaining a two-dimensional spectrogram directivity statistical histogram of the vehicle based on the gradient vector angle intervals of the plurality of equal divisions
Finding out the corresponding gradient vector angle α of the peak point of the directional statistical histogram of the two-dimensional spectrogram of the vehiclecCalculating the directional characteristic F of the preprocessed vehicle two-dimensional spectrogram according to the following formulac:
Extraction of Ec、Mc、And FcAnd the texture features are used as the texture features of the preprocessed vehicle two-dimensional spectrogram.
S5: and repeating the steps S2 to S4 until all pedestrian echo samples in the pedestrian echo sample set and all vehicle echo samples in the vehicle echo sample set are traversed to obtain a pedestrian two-dimensional spectrogram texture feature sample set and a vehicle two-dimensional spectrogram texture feature sample set.
In step S5, the sample set of the two-dimensional spectrogram texture features of the pedestrian is: a set of texture features of the preprocessed pedestrian two-dimensional spectrogram corresponding to all the pedestrian echo samples in the pedestrian echo sample set; the texture feature sample set of the two-dimensional spectrogram of the vehicle is as follows: and the texture feature set of the preprocessed vehicle two-dimensional spectrogram corresponding to all the vehicle echo samples in the vehicle echo sample set.
S6: and training and testing through a support vector machine according to the pedestrian two-dimensional spectrogram texture feature sample set and the vehicle two-dimensional spectrogram texture feature sample set to obtain a corresponding classification recognition result. The concrete description is as follows:
selecting 30% of samples from the pedestrian two-dimensional spectrogram texture feature sample set and the vehicle two-dimensional spectrogram texture feature sample set to form a target two-dimensional spectrogram training sample set; the target two-dimensional spectrogram test sample set comprises: the rest samples of the pedestrian two-dimensional spectrogram texture feature sample set and the rest samples of the vehicle two-dimensional spectrogram texture feature sample set;
sending the target two-dimensional spectrogram training sample set into a Support Vector Machine (SVM), and carrying out training and learning; and then, sequentially sending each test sample of the target two-dimensional spectrogram test sample set into a support vector machine for testing (sending the test sample into an SVM classifier for testing) to obtain a corresponding test result, wherein each test result is a corresponding classification recognition result.
After the corresponding classification recognition results are obtained, the correct recognition rates of the pedestrian target and the vehicle target in the classification recognition results are respectively calculated, and weighted average is carried out to obtain the average recognition rate.
The effects of the present invention can be further illustrated by the following experiments:
in the experiment, the radar transmitting signal is a narrow-band linear frequency modulation continuous wave signal, the radar receiving system adopts a Dechirp receiving system, and the sampling frequency of the Dechirp receiving system is 3.125 MHz. Target sample parameters referring to table 1, targets include both pedestrians and vehicles.
TABLE 1
Referring to fig. 2a, it is a two-dimensional spectrogram of the measured data; wherein, the abscissa is time, and the unit is second; the ordinate is frequency in Hz. Referring to fig. 2b, the two-dimensional vehicle spectrogram is measured data. In fig. 2b, the abscissa is time in seconds; the ordinate is frequency in Hz. Referring to fig. 2c, the two-dimensional spectrogram of the pedestrian is preprocessed from the measured data, wherein the abscissa is time and the unit is second; the ordinate is frequency in Hz. Referring to fig. 2d, a diagram of the classification result of the measured data is shown. In fig. 2d, the triangle represents a vehicle object and the circle represents a pedestrian object, it can be seen that the two types of objects have been substantially separated. In this experiment, the recognition rate of the pedestrian target was found to be 100%, while the recognition rate of the vehicle target was 98.67%, and the average recognition rate was 99.33%.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (7)
1. A classification and identification method for micro-motion targets of pedestrians and vehicles is characterized by comprising the following steps:
s1: respectively receiving pedestrian echo data and vehicle echo data by using a radar to obtain a pedestrian echo sample set and a vehicle echo sample set;
s2: selecting a pedestrian echo sample from the pedestrian echo sample set, and performing clutter suppression on the selected pedestrian echo sample to obtain clutter suppressed pedestrian echo data; selecting a vehicle echo sample from a vehicle echo sample set, and performing clutter suppression on the selected vehicle echo sample to obtain clutter suppressed vehicle echo data;
s3: carrying out short-time Fourier transform on the pedestrian echo data after clutter suppression to obtain a two-dimensional spectrogram of the pedestrian; preprocessing a two-dimensional spectrogram of the pedestrian to obtain a preprocessed two-dimensional spectrogram of the pedestrian; performing short-time Fourier transform on the clutter suppressed vehicle echo data to obtain a vehicle two-dimensional spectrogram; preprocessing a vehicle two-dimensional spectrogram to obtain a preprocessed vehicle two-dimensional spectrogram;
the step S3 specifically includes the following steps:
s31: carrying out short-time Fourier transform on the pedestrian echo data after clutter suppression to obtain a two-dimensional spectrogram of the pedestrian; performing short-time Fourier transform on the clutter suppressed vehicle echo data to obtain a vehicle two-dimensional spectrogram;
s32: accumulating the energy of the two-dimensional spectrogram of the pedestrian in the time dimension according to the following formula to obtain the frequency dimension energy distribution E of the two-dimensional spectrogram of the pedestrianx(nx):
Wherein S isx(mx,nx) Representing the energy distribution of a two-dimensional spectrogram of a pedestrian, mxNumber of sampling points in time dimension corresponding to two-dimensional spectrogram of pedestrian, nxSampling point numbers of frequency dimensions corresponding to the two-dimensional spectrogram of the pedestrian;
for a vehicle according to the following formulaAccumulating the energy of the two-dimensional spectrogram in the time dimension to obtain the frequency dimension energy distribution E of the two-dimensional spectrogram of the vehiclec(nc):
Wherein S isc(mc,nc) Representing the energy distribution of a two-dimensional map of the vehicle, mcThe number of sampling points in the time dimension corresponding to the two-dimensional spectrogram of the vehicle, ncThe sampling point number of the frequency dimension corresponding to the vehicle two-dimensional spectrogram is obtained; m is a time dimension coordinate of the pedestrian two-dimensional spectrogram and the vehicle two-dimensional spectrogram;
s33: frequency dimension energy distribution E of pedestrian two-dimensional spectrogramx(nx) Peak position of (F)xRespectively selecting frequency units symmetrically towards two sides for a symmetry axis, and performing energy accumulation until the energy sum exceeds 60% of the total energy of the frequency dimension of the pedestrian two-dimensional spectrogram, at the moment, recording the minimum frequency and the maximum frequency in all the selected frequency units, and constructing a rectangular window function corresponding to the pedestrian two-dimensional spectrogram on the basis of the recorded minimum frequency and maximum frequency; windowing the two-dimensional spectrogram of the pedestrian according to the rectangular window function corresponding to the two-dimensional spectrogram of the pedestrian to obtain a preprocessed two-dimensional spectrogram of the pedestrian;
frequency dimension energy distribution E of vehicle two-dimension spectrogramc(nc) Peak position of (F)cSelecting frequencies symmetrically to both sides for the symmetry axisThe unit carries out energy accumulation until the energy sum exceeds 60% of the total energy of the frequency dimension of the two-dimensional spectrogram of the vehicle, at the moment, the minimum frequency and the maximum frequency are recorded in all selected frequency units, and a rectangular window function corresponding to the two-dimensional spectrogram of the vehicle is constructed on the basis of the recorded minimum frequency and maximum frequency; windowing the vehicle two-dimensional spectrogram according to the rectangular window function corresponding to the vehicle two-dimensional spectrogram to obtain a preprocessed vehicle two-dimensional spectrogram;
s4: respectively extracting texture features of the preprocessed pedestrian two-dimensional spectrogram and the preprocessed vehicle two-dimensional spectrogram;
s5: repeating the steps S2 to S4 until all pedestrian echo samples in the pedestrian echo sample set and all vehicle echo samples in the vehicle echo sample set are traversed to obtain a pedestrian two-dimensional spectrogram texture feature sample set and a vehicle two-dimensional spectrogram texture feature sample set;
s6: and training and testing through a support vector machine according to the pedestrian two-dimensional spectrogram texture feature sample set and the vehicle two-dimensional spectrogram texture feature sample set to obtain a corresponding classification recognition result.
2. The method for classifying and identifying the micro-moving objects of the pedestrians and the vehicles according to claim 1, wherein in step S1, the radar is used to transmit narrow-band chirp continuous wave signals to the pedestrians and the vehicles, respectively, and a Dechirp receiving system is used to receive the echo data of the pedestrians and the echo data of the vehicles, respectively.
3. The method for classifying and identifying the micro-motion object of the pedestrian and the vehicle as claimed in claim 1, wherein the step S2 specifically comprises the steps of:
s21: selecting a pedestrian echo sample from a pedestrian echo sample set, performing range-wise pulse compression processing on the selected pedestrian echo sample to obtain pedestrian echo data after pulse pressure processing, and then performing filtering processing on the pedestrian echo data after pulse pressure processing by using a moving target display method to obtain the pedestrian echo data after filtering processing;
selecting a vehicle echo sample from a vehicle echo sample set, performing range-wise pulse compression processing on the selected vehicle echo sample to obtain vehicle echo data after pulse pressure processing, and then performing filtering processing on the vehicle echo data after the pulse pressure processing by using a moving target display method to obtain the vehicle echo data after the filtering processing;
s22: in the pedestrian echo data after filtering processing, performing azimuth Fourier transform on data of a distance unit with the maximum energy to obtain pedestrian azimuth frequency domain data;
in the filtered vehicle echo data, performing azimuth Fourier transform on the data of the distance unit with the maximum energy to obtain vehicle azimuth frequency domain data;
s23: clutter is removed from the pedestrian azimuth frequency domain data by adopting a CLEAN algorithm, and then inverse Fourier transform is carried out on the pedestrian azimuth frequency domain data after the clutter removal to obtain pedestrian echo data after clutter suppression;
and removing clutter from the vehicle azimuth frequency domain data by adopting a CLEAN algorithm, and performing inverse Fourier transform on the vehicle azimuth frequency domain data after the clutter removal to obtain vehicle echo data after the clutter suppression.
4. The method for classifying and identifying the micro-motion target of the pedestrian and the vehicle as claimed in claim 1, wherein in the step S4, the step of extracting the texture features of the preprocessed two-dimensional spectrogram of the pedestrian specifically comprises the steps of:
s411: accumulating the energy of the preprocessed pedestrian two-dimensional spectrogram in a time dimension to obtain frequency dimension energy distribution of the preprocessed pedestrian two-dimensional spectrogram, normalizing the frequency dimension energy distribution of the preprocessed pedestrian two-dimensional spectrogram to obtain normalized data p of the frequency dimension energy distribution of the preprocessed pedestrian two-dimensional spectrogramx(nx) Then, the entropy E of the two-dimensional spectrogram of the pedestrian is calculated according to the following formulax:
Wherein n isxSampling point numbers of frequency dimensions corresponding to the two-dimensional spectrogram of the pedestrian;
s412: according to the energy distribution of the preprocessed two-dimensional spectrogram of the pedestrian, a plurality of equally divided energy intervals are obtained through energy equally dividing processing; obtaining a pedestrian two-dimensional spectrogram energy statistical histogram based on the plurality of equally divided energy intervals, and then calculating a third moment M of the pedestrian two-dimensional spectrogram energy statistical histogramx;
S413: defining two mutually perpendicular directions HxAnd VxCalculating the gradient vector magnitude of the preprocessed two-dimensional spectrogram of the pedestrian according to the following formulaAnd gradient vector angle thetax:
Wherein,the energy of the pixel points of the preprocessed two-dimensional spectrogram of the pedestrian is HxThe gradient in the direction of the magnetic field,the energy of pixel points of a two-dimensional spectrogram of the preprocessed pedestrian is VxA gradient in direction;
s414: gradient vector angle theta of preprocessed pedestrian two-dimensional spectrogramxDiscretizing and equally dividing the vector data to obtain a plurality of equally divided gradient vector angle sections, wherein each equally divided gradient vector angle section isi is taken from 2 to L,is thetaxA discretized maximum; obtaining a pedestrian two-dimensional spectrogram directivity statistical histogram based on the plurality of equally divided gradient vector angle intervalsL-1 is the number of the pedestrian two-dimensional spectrogram gradient vector angle interval and the vehicle two-dimensional spectrogram gradient vector angle interval;
s415, finding out a gradient vector angle α corresponding to the peak point of the directional statistical histogram of the two-dimensional spectrogram of the pedestrianxCalculating the directional characteristic F of the preprocessed two-dimensional spectrogram of the pedestrian according to the following formulax:
S416: extraction of Ex、Mx、And FxAnd the obtained data is used as the texture feature of the preprocessed two-dimensional spectrogram of the pedestrian.
5. The method for classifying and identifying the micro-motion target of the pedestrian and the vehicle as claimed in claim 1, wherein in the step S4, the step of extracting the texture feature of the preprocessed vehicle spectrogram specifically comprises the steps of:
s421: accumulating the energy of the preprocessed vehicle two-dimensional spectrogram in a time dimension to obtain frequency dimension energy distribution of the preprocessed vehicle two-dimensional spectrogram, normalizing the frequency dimension energy distribution of the preprocessed vehicle two-dimensional spectrogram to obtain normalized data p of the frequency dimension energy distribution of the preprocessed vehicle two-dimensional spectrogramc(nc) Then, the entropy value E of the two-dimensional spectrogram of the vehicle is calculated according to the following formulax:
Wherein n iscThe sampling point number of the frequency dimension corresponding to the vehicle two-dimensional spectrogram is obtained;
s422: according to the energy distribution of the preprocessed vehicle two-dimensional spectrogram, a plurality of equally divided energy intervals are obtained through energy equally dividing processing; obtaining a vehicle two-dimensional spectrogram energy statistical histogram based on the plurality of equally divided energy intervals, and then calculating a third moment M of the vehicle two-dimensional spectrogram energy statistical histogramc;
S423: defining two mutually perpendicular directions HcAnd VcCalculating the gradient vector magnitude of the preprocessed vehicle two-dimensional spectrogram according to the following formulaAnd gradient vector angle thetac:
Wherein,the energy of the pixel points of the two-dimensional spectrogram of the preprocessed vehicle is HcThe gradient in the direction of the magnetic field,the energy of pixel points of a two-dimensional spectrogram of the preprocessed vehicle is VcA gradient in direction;
s424: gradient vector angle theta of preprocessed vehicle two-dimensional spectrogramcDiscretizing and equally dividing the vector data to obtain a plurality of equally divided gradient vector angle sections, wherein each equally divided gradient vector angle section isi is taken from 2 to L,is thetacA discretized maximum; obtaining a two-dimensional spectrogram directivity statistical histogram of the vehicle based on the plurality of equally divided gradient vector angle intervalsL-1 is the number of the pedestrian two-dimensional spectrogram gradient vector angle interval and the vehicle two-dimensional spectrogram gradient vector angle interval;
s425, finding out a gradient vector angle α corresponding to a peak point of a directional statistical histogram of a two-dimensional spectrogram of the vehiclecCalculating the directional characteristic F of the preprocessed vehicle two-dimensional spectrogram according to the following formulac:
S426: extraction of Ec、Mc、And FcAnd the texture features are used as the texture features of the preprocessed vehicle two-dimensional spectrogram.
6. The method for classifying and identifying the micro-motion object between the pedestrian and the vehicle as claimed in claim 1, wherein in step S5, the sample set of texture features of the two-dimensional spectrogram of the pedestrian is as follows: the pedestrian echo sample set comprises a set of texture features of a preprocessed pedestrian two-dimensional spectrogram corresponding to all pedestrian echo samples;
the texture feature sample set of the vehicle two-dimensional spectrogram is as follows: and the set of texture features of the preprocessed vehicle two-dimensional spectrogram corresponding to all vehicle echo samples in the vehicle echo sample set.
7. The method for classifying and identifying the micro-motion object of the pedestrian and the vehicle as claimed in claim 1, wherein the step S6 specifically comprises the steps of:
s61: selecting 30% of samples from the pedestrian two-dimensional spectrogram texture feature sample set and the vehicle two-dimensional spectrogram texture feature sample set to form a target two-dimensional spectrogram training sample set; the target two-dimensional spectrogram test sample set comprises: the rest samples of the pedestrian two-dimensional spectrogram texture feature sample set and the rest samples of the vehicle two-dimensional spectrogram texture feature sample set;
s62: sending the target two-dimensional spectrogram training sample set into a support vector machine for training and learning; and then, sequentially sending each test sample of the target two-dimensional spectrogram test sample set into a support vector machine for testing to obtain a corresponding test result, wherein each test result is a corresponding classification recognition result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410067084.0A CN103914703B (en) | 2014-05-12 | 2014-05-12 | Classification and identification method for pedestrian and vehicle micro-motion targets |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410067084.0A CN103914703B (en) | 2014-05-12 | 2014-05-12 | Classification and identification method for pedestrian and vehicle micro-motion targets |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103914703A CN103914703A (en) | 2014-07-09 |
CN103914703B true CN103914703B (en) | 2017-02-15 |
Family
ID=51040370
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410067084.0A Expired - Fee Related CN103914703B (en) | 2014-05-12 | 2014-05-12 | Classification and identification method for pedestrian and vehicle micro-motion targets |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103914703B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108459311A (en) * | 2018-03-22 | 2018-08-28 | 三明学院 | Micro-doppler objective classification method under convolutional neural networks based on Hough transform |
CN110275147A (en) * | 2018-03-13 | 2019-09-24 | 中国人民解放军国防科技大学 | Human behavior micro-Doppler classification and identification method based on migration depth neural network |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104215952B (en) * | 2014-08-27 | 2017-06-06 | 苏州闻捷传感技术有限公司 | Vehicle-mounted target identification system based on micro-motion characteristics and identification method thereof |
CN105137426A (en) * | 2015-08-07 | 2015-12-09 | 河海大学 | High-resolution distance image obtaining method |
CN105718856B (en) * | 2016-01-13 | 2019-01-18 | 杭州电子科技大学 | A kind of extraction of vehicle micro-tremor signal and classification method based on EEMD |
CN109190673B (en) * | 2018-08-03 | 2021-09-03 | 西安电子科技大学 | Ground target classification method based on random forest and data rejection |
CN110907906B (en) * | 2018-09-14 | 2023-01-10 | 深圳市道通智能航空技术股份有限公司 | Object classification method and related device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102788969A (en) * | 2012-07-04 | 2012-11-21 | 中国人民解放军海军航空工程学院 | Sea surface micromotion target detection and feature extraction method based on short-time fractional Fourier transform |
CN103245937A (en) * | 2013-05-27 | 2013-08-14 | 四川大学 | Micro moving target feature extracting method based on micro Doppler effect |
-
2014
- 2014-05-12 CN CN201410067084.0A patent/CN103914703B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102788969A (en) * | 2012-07-04 | 2012-11-21 | 中国人民解放军海军航空工程学院 | Sea surface micromotion target detection and feature extraction method based on short-time fractional Fourier transform |
CN103245937A (en) * | 2013-05-27 | 2013-08-14 | 四川大学 | Micro moving target feature extracting method based on micro Doppler effect |
Non-Patent Citations (3)
Title |
---|
Bayesian Classification of Humans and Vehicles Using Micro-Doppler Signals From a Scanning-Beam Radar;Jeffrey A.Nanzer et al;《IEEE MICROWAVE AND WIRELESS COMPONENTS LETTERS》;20090531;第19卷(第5期);第338-339页 * |
一种抑制维格纳分布交叉项的方法及在故障诊断中应用;程发斌 等;《中国机械工程》;20080731;第19卷(第14期);第1727页 * |
基于微多普勒的窄带飞机目标分类方法研究;杨自豪;《中国优秀硕士学位论文全文数据库信息科技辑》;20131215(第S2期);第I136-1228页 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110275147A (en) * | 2018-03-13 | 2019-09-24 | 中国人民解放军国防科技大学 | Human behavior micro-Doppler classification and identification method based on migration depth neural network |
CN110275147B (en) * | 2018-03-13 | 2022-01-04 | 中国人民解放军国防科技大学 | Human behavior micro-Doppler classification and identification method based on migration depth neural network |
CN108459311A (en) * | 2018-03-22 | 2018-08-28 | 三明学院 | Micro-doppler objective classification method under convolutional neural networks based on Hough transform |
Also Published As
Publication number | Publication date |
---|---|
CN103914703A (en) | 2014-07-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103914703B (en) | Classification and identification method for pedestrian and vehicle micro-motion targets | |
US11346942B2 (en) | Target positioning device and method based on plecotus auritus double-pinna bionic sonar | |
US20130082858A1 (en) | Classification of subsurface objects using singular values derived from signal frames | |
CN104851097B (en) | The multichannel SAR GMTI methods aided in based on target shape and shade | |
CN103293521B (en) | Method for detecting water depth of offshore sea by X-band radar | |
CN102914773B (en) | Multi-pass circumference SAR three-dimensional imaging method | |
Lv et al. | Ground moving target indication in SAR images by symmetric defocusing | |
CN102043148A (en) | Ground moving target classification method based on general matched filtering | |
CN102721952A (en) | Method for steadily classifying ground moving target based on super-resolution Doppler spectrum | |
DE102017204496A1 (en) | Method and radar device for determining radial relative acceleration of at least one target | |
CN112137620A (en) | Ultra-wideband radar-based human body weak respiration signal detection method | |
CN106556833A (en) | Based on the ISAR imaging simulation methods that time domain Shooting and bouncing rays fast near-field is calculated | |
CN105738888A (en) | Double-characteristic sea surface floating small-target detection method based on sea clutter suppression | |
CN107942323B (en) | Precession target time-frequency curve extraction method based on frequency domain entropy | |
Cammenga et al. | Combined high range resolution and micro-Doppler analysis of human gait | |
JP2017211348A (en) | Locus estimation device, locus estimation method, and program | |
CN108693530B (en) | Orientation entropy extraction method based on circular synthetic aperture radar data | |
CN106707278A (en) | Sparse representation-based doppler beam sharpened imaging method and device | |
CN112305541A (en) | SAR imaging method based on sampling sequence length constraint condition | |
CN101251595B (en) | Method for estimation of SAR image goal position angle based on non-sample Contourlet transformation | |
CN114217284B (en) | Radar moving target detection and interference suppression method and system based on characteristics | |
CN114355329A (en) | Method for detecting vital sign distance of frequency modulated continuous wave radar | |
Mason et al. | Passive synthetic aperture radar imaging based on low-rank matrix recovery | |
Li et al. | Deep learning for interference mitigation in time-frequency maps of fmcw radars | |
Wang et al. | Feature analysis for forward-looking landmine detection using GPR |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170215 Termination date: 20170512 |
|
CF01 | Termination of patent right due to non-payment of annual fee |