CN113567974B - Multi-living-body intelligent detection device and method based on CPPWM radar - Google Patents
Multi-living-body intelligent detection device and method based on CPPWM radar Download PDFInfo
- Publication number
- CN113567974B CN113567974B CN202110694306.1A CN202110694306A CN113567974B CN 113567974 B CN113567974 B CN 113567974B CN 202110694306 A CN202110694306 A CN 202110694306A CN 113567974 B CN113567974 B CN 113567974B
- Authority
- CN
- China
- Prior art keywords
- cppwm
- signal
- life
- data
- representing
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000001514 detection method Methods 0.000 title claims abstract description 72
- 238000012545 processing Methods 0.000 claims abstract description 50
- 239000011159 matrix material Substances 0.000 claims description 58
- 238000004364 calculation method Methods 0.000 claims description 28
- 238000013507 mapping Methods 0.000 claims description 24
- 230000003068 static effect Effects 0.000 claims description 24
- 238000012549 training Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 16
- 238000010586 diagram Methods 0.000 claims description 14
- 238000001914 filtration Methods 0.000 claims description 13
- 238000003384 imaging method Methods 0.000 claims description 11
- 230000002093 peripheral effect Effects 0.000 claims description 7
- 230000001629 suppression Effects 0.000 claims description 7
- 241001465754 Metazoa Species 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 5
- 230000000295 complement effect Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 230000003190 augmentative effect Effects 0.000 claims description 3
- 230000000739 chaotic effect Effects 0.000 claims description 3
- 238000007477 logistic regression Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000008685 targeting Effects 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 230000003139 buffering effect Effects 0.000 claims description 2
- 230000007613 environmental effect Effects 0.000 abstract description 3
- 230000000241 respiratory effect Effects 0.000 description 4
- 230000036391 respiratory frequency Effects 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 3
- 230000036387 respiratory rate Effects 0.000 description 3
- 230000005855 radiation Effects 0.000 description 2
- 230000029058 respiratory gaseous exchange Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011068 loading method Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012634 optical imaging Methods 0.000 description 1
- 230000010363 phase shift Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
- G01S13/06—Systems determining position data of a target
- G01S13/08—Systems for measuring distance only
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V3/00—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation
- G01V3/12—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation operating with electromagnetic waves
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Remote Sensing (AREA)
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention belongs to the disaster detection field and discloses a multi-life intelligent detection device and method based on CPPWM-MIMO radar, wherein the device comprises a CPPWM signal generator, a plurality of power amplifiers, an array antenna, a plurality of low-noise amplifiers, a multichannel high-speed mixed sampler and a data processing system; the array antenna comprises a plurality of transmitting antennas and a plurality of receiving antennas, and the CPPWM signal generator simultaneously generates orthogonal multipath CPPWM detection signals and multipath CPPWM reference signals corresponding to the orthogonal multipath CPPWM detection signals; each CPPWM detection signal is connected with a transmitting antenna through a power amplifier, and the output end of each receiving antenna is connected with a multichannel high-speed mixed sampler through a low-noise amplifier; each CPPWM reference signal is respectively connected with the multi-channel high-speed mixed sampler; the multichannel high-speed mixed sampler is connected with the data processing module, and the data processing system is used for calculating and analyzing to obtain a life detection result. The invention has the advantages of being not influenced by environmental temperature, noise, site visibility, external electromagnetic interference and the like.
Description
Technical Field
The invention belongs to the field of disaster detection, and particularly relates to a CPPWM-MIMO radar-based intelligent multi-life-body detection device and method.
Background
The life detection radar actively transmits electromagnetic waves to penetrate through the obstacle, and receives and analyzes echoes reflected by trapped people to identify life signals such as respiration and the like, so that whether survivors are buried under ruins or not is judged. Compared with the traditional life detector (sensor and microsystem, vol.30, p.8-10,2011) based on infrared, audio, sound wave and optical imaging, the life detection radar has obvious advantages of being not influenced by environmental temperature, noise, site visibility and the like.
Existing life detection radars mainly include continuous wave Doppler radars (IEEE Trans. Microw. Thery techn., vol.61, p.2046-2060,2013), linear frequency modulated continuous wave radars (IEEE Trans. Microw. Thery techn., vol.62, p.1387-1399,2014), step frequency continuous wave radars (IEEE J. Sel. Topics appl. Earth observ., vol.7, p.775-782,2014), pulse ultra wideband radars (IEEE Trans. Geosci. Remote SEns., vol.52, p.7195-7204,2014), pseudo-random code ultra wideband radars (IEICE electron. Expr., vol.11, p.1-7,2014). The continuous wave Doppler radar emits single-frequency continuous waves as detection signals, and life signals such as respiration, heartbeat and the like are obtained by demodulating echo phase shift/frequency shift caused by a human body target, but the life signals cannot accurately position the human body target. The other radars are used for further obtaining respiratory signals by measuring the change rule of the distance along with the observation time on the basis of obtaining the one-dimensional distance of the human body target. The existing life detection radar mostly adopts a single-shot structure, and has the following defects: 1. only point-by-point searching is performed, the time consumption is long, the detection efficiency is low, and the search and rescue time is delayed; 2. only the respiratory signal and one-dimensional distance information of the human body target can be obtained, and the azimuth information of the human body target cannot be obtained, so that if a plurality of human bodies with similar respiratory frequency and depth are buried under ruins, the life detection radar cannot accurately distinguish and identify the human bodies.
In addition, the existing life detection algorithm adopts a combination of various clutter suppression and noise separation algorithms to suppress or eliminate various clutter and noise layer by layer so as to expect to obtain a clean life signal. For example: empirical Mode Decomposition (EMD) (technical guide, vol.32, p.36-42,2014), integrated empirical mode decomposition (EEMD) (Sensors, vol.16, 2016) and multiple higher order cumulants (IEEE Trans. Geosci. Remote sens., vol.50, p.1254-1265,2011) are used to suppress noise interference. Linear trend removal (LTS) (IEEE trans. Geoci. Remote sens., vol.50, p.3132-3142,2012) based on linear least squares fitting (IEEE trans. Geoci. Remote sens., vol.48, p.2005-2014,2009) is used to remove static clutter and linear trends. Singular Value Decomposition (SVD) (digit.signal process., vol.74, p.72-93,2018) is used to remove static clutter. However, the more clutter suppression/denoising algorithms are included in the life detection algorithm, the more control parameters are in the algorithm, so that the difficulty of self-adaptive adjustment is increased, the processing time is increased, and the robustness and the instantaneity of the algorithm are reduced.
Disclosure of Invention
The invention provides a multi-life-body intelligent detection device and method based on a CPPWM-MIMO radar, which are used for solving the problems of high false alarm rate and high false alarm rate when searching for multi-life bodies in a low signal-to-noise ratio environment after disaster of the existing life detection radar.
In order to solve the technical problems, the invention adopts the following technical scheme: a multi-living body intelligent detection device based on a CPPWM-MIMO radar comprises a CPPWM signal generator, a plurality of power amplifiers, a Vivaldi linear sparse array antenna, a plurality of low-noise amplifiers, a multi-channel high-speed mixed sampler and a data processing system;
the CPPWM signal generator simultaneously generates orthogonal multipath CPPWM detection signals and multipath CPPWM reference signals corresponding to the orthogonal multipath CPPWM detection signals by setting initial values of different Logistic mapping units and Tent mapping units; each CPPWM detection signal is respectively connected with a signal input end of a power amplifier, and each CPPWM reference signal is respectively connected with a signal input end of the multichannel high-speed mixed sampler;
the Vivaldi linear sparse array antenna comprises a plurality of transmitting antennas positioned in the middle and a plurality of receiving antennas positioned at two sides, wherein each transmitting antenna is respectively connected with a signal output end of a power amplifier, and the output end of each receiving antenna is respectively connected with a signal input end of a low-noise amplifier; the signal output end of each low-noise amplifier is connected with the signal input end of the multichannel high-speed mixed sampler;
The signal output end of the multichannel high-speed mixed sampler is connected with the data processing system, and the data processing system is used for calculating and analyzing to obtain a life detection result.
The number of the power amplifiers and the low-noise amplifiers is one, and the Vivaldi linear sparse array antenna comprises six transmitting antennas positioned in the middle and three receiving antennas positioned at two sides;
the data processing system comprises a multi-core central processing unit, a data processing module and a result display module;
the data processing module is used for calculating reference and echo data;
the result display module is used for displaying the multi-life body detection result;
the multi-core central processing unit is used for controlling the normal operation of the data processing module and the result display module.
The transmitting antenna and the receiving antenna are improved Vivaldi antennas, wherein the improved Vivaldi antennas are formed by arranging complementary opening resonance rings on radiation arms corresponding to the tail ends of feeder lines of the Vivaldi antennas, and a plurality of Y-shaped gaps with different lengths are loaded on two sides of a radiation patch.
The CPPWM signal generator includes:
the system comprises a logic mapper, a first counter, a second counter, a Tent mapper, a first comparator, a second comparator, a pulse generator and a signal storage output module;
The output of the Logistic mapper is X n+1 ＝4(1-X n )X n Wherein X represents the initial value of the Logistic mapping function, X n Representing the output value of the Logistic mapping function after the nth cycle calculation; the output ends of the first counter and the Logistic mapper are connected with the first comparator, and the output end of the Tent mapper is Y n+1 ＝1.99(0.5-|Y n -0.5|), wherein Y represents the initial value of the Tent mapping function, Y n Representing the output value of the Tent mapping function after the nth cycle calculation; the output ends of the second counter and the Tent mapper are connected with a second comparator, the output ends of the first comparator and the second comparator are connected with the input end of the pulse generator, and the output ends of the first comparator and the second comparator are divided into two partsThe high-level signal output by the pulse generator is used for enabling the control of the Tent mapper and the second counter to start working, and is also used for controlling the logic mapper and the first counter to stop working and resetting the first counter; the low-level signal output by the pulse generator is used for enabling the control Tent mapper and the second counter to stop working, resetting the second counter and enabling the control logic mapping and the first counter to start working and update to enter the next state; the level signal output by the pulse generator is stored and output by the signal storage output module.
The signal storage output module comprises: the device comprises a serializer, a register, a logic comparator, a write address counter, a static random access memory controller, a read address counter, a serial transceiving reset controller, a serial transceiver and a differential output module;
the serializer is used for converting the low-speed serial signal output by the pulse generator into parallel data, transmitting the parallel data to the register for caching, the logic comparator is used for extracting the data in the register according to the writing sequence and logically selecting the data and then transmitting the data to the static random access memory controller, and the serializer is also used for generating a pulse clock signal and transmitting the data to the static random access memory controller and the writing address counter; the write address data generated by the write address counter is matched with the pulse clock signal and the data sent by the logic comparator, so that the static random access memory controller writes the data into the static random access memory;
the serial transceiver reset controller is used for releasing the serial transceiver reset state, the serial transceiver () is used for generating a feedback clock and transmitting the feedback clock to the read address counter and the static random access memory controller, the read address data generated by the read address counter is matched with the feedback clock transmitted by the serial transceiver, the static random access memory controller reads out the data in the static random access memory and writes the data into the serial transceiver, and finally the serial transceiver deserializes the data and outputs CPPWM signals through the differential output module.
The multichannel high-speed hybrid sampler comprises a field programmable gate array and a plurality of data channels, wherein each data channel comprises a serial peripheral interface, a programmable delay chip and an analog-to-digital converter; the field programmable gate array is used for controlling the programmable delay chip through the serial peripheral interface and delaying the working clock of the analog-to-digital converter.
The invention also provides a CPPWM-MIMO radar-based intelligent multi-life body detection method, which is realized based on the device and comprises the following steps:
s1, chaos correlation filtering and ranging: performing pairwise cross correlation calculation on each CPPWM echo signal and each CPPWM reference signal to obtain each cross correlation curve, and performing matched filtering and ranging calculation;
s2, processing the cross-correlation curve to obtain life body information, and calculating and judging whether the life body is human or not according to chaos correlation filtering and ranging calculation results;
the specific steps for obtaining the life body information are as follows:
s21, imaging a cross-correlation curve based on a multi-scale weighted rapid BP algorithm;
s22, rapidly robust principal component analysis based on factor group sparse regularization suppresses artifacts in the image;
s23, performing two-dimensional space positioning and quantity detection on a plurality of living bodies in the image;
The specific steps of calculating and judging whether the living body is human include:
s31, accumulating chaotic correlation filtering and ranging results in a slow time domain to obtain an original slow time domain-distance matrix R, and preprocessing the R;
s32, intelligently extracting a multi-life body feature matrix based on a CenterNet life detection network;
s33, performing fast Fourier transform on the multi-living body feature matrix, performing wavelet entropy analysis on the Fourier transformed living body feature matrix, and judging whether the living body is a human or an animal based on the wavelet entropy.
The specific method of the step S21 is as follows: firstly, carrying out large-size meshing on the whole detection area, processing by a BP algorithm to obtain a low-resolution image, searching an energy maximum value by using a sliding window energy detection method, extracting a potential life area from the low-resolution image, carrying out small-size meshing on the area, processing by the BP algorithm to obtain a high-resolution image, and remaining areas are still low-resolution images; iterating for 2-3 times until the potential life region image reaches the resolution requirement, finally obtaining a weighted and overlapped image matrix W of the high-resolution image and the low-resolution image, and determining a weighting factor by calculating the mean value and standard deviation value of scattered data of each focus on the time delay curve;
The specific method of the step S22 is as follows: the constraint condition is removed by using an augmented Lagrangian multiplier method, and then an equation is solved by using an alternate direction multiplier method, so that low-rank artifacts L and sparse life bodies S are obtained, artifact suppression is realized, and the equation is as follows:
min||A|| 2,1 +α||B T || 2,1 +λ||S|| 1 s.t.W＝AB+S；
in the method, in the process of the invention,||·|| F represents f norm, alpha represents weight parameter, lambda represents regularization parameter, A and B represent decomposed clutter matrix, S represents target matrix, a j Representing the j-th column vector in matrix a;
in the step S23, the number of pixels in the imaging image after artifact suppression represents the number of living bodies, and the coordinates of the pixels in the imaging result diagram represent the two-dimensional spatial positions of the living bodies.
In the step S31, the specific method for preprocessing the time-distance matrix R is as follows:
firstly, eliminating static clutter and linear trend interference by using a linear trend removal method, wherein the processing procedure is as follows:
wherein x= [ M/M1 ] M ]，m＝[0,1,···,M-1] T ，1 M Is an mx 1 vector containing a unit value,representing a signal matrix obtained after the linear trend removal method;
and then the automatic gain control is utilized to enhance the weak life signal, and the signal enhancement process is as follows: for signal matrixBy calculating->Adjusting the gain by the signal power in a time window of given magnitude ω, where k and m represent matrices, respectively K=0, 1, … …, N k ，m＝0,1,……,M m ，N k And M is as follows m Representing fast and slow time domain sampling points and associated with a predetermined maximum gain g MAX Comparing, wherein the calculation formula is as follows:
wherein i=0, 1, … …, N- ω, g [ i, m]G is a gain based on signal power norm [i,m]Is a normalized representation of all gains with minimum gain, g mask [i,m]For gain mask, use gain mask pairPerforming weight adjustment; g min (m) represents the minimum gain, g, for each value of m, for all i norm [i,m]＝g[i,m]/g min (m) carrying out normalization representation on all gains, and finally carrying out gray scale processing and outputting a gray scale image;
the specific method of step S32 is as follows:
screening gray level images without life bodies to form a data set, and marking life body labels to obtain a life feature data set for network training and testing;
inputting a gray matrix of training set data into an improved DLA-34 feature extraction network, and respectively training a center point, a center point offset and a size of a target to obtain a target boundary frame representing a vital body feature matrix;
the method for training the target center point comprises the following steps: for the center point P of each vital sign frame, where P ε R 2 First, calculate a low resolution central point equivalent value R is the output stepping amplitude, and the Gaussian kernel is used for mapping the central point of each vital sign frame to thermodynamic diagram +.>Wherein W and H represent the width and height of the image, respectively,indicating that this point is the center point, < >>Representing the point as background, then performing pixel-level logistic regression using a Focal Loss function, the expression of which is:
wherein alpha and beta respectively represent super parameters of a loss function, and N is the number of center points in the image; y is Y xyz Representing a gaussian kernel, the expression of which is:
in sigma p Representing target size adaptationThe standard deviation should be set to be equal to the standard deviation,and->Respectively representing the abscissa of the center point P, and x and y respectively representing the abscissa of the random calculation point;
the method for training the offset of the target center point comprises the following steps: the local offset of each center point isTraining the offset of all object center points by using an L1 loss function; the expression of the L1 loss function is:
in the method, in the process of the invention,p represents the position coordinate of the central point for the local offset;
the method for training the target size comprises the following steps: order theBoundary box of object k, center position isx i And y i Respectively integer coordinates, and regressing the size of each object k to bePerforming target size training by using a Smooth L1 loss function; targeting with a single size predictor +. >In the target detection, let ∈ ->In order to detect n sets of center points, obtaining a boundary box by extracting 100 peak points on a thermodynamic diagram, and reserving if the peak point is larger than 8 neighborhood points of the peak point;
the target bounding box representing the vital sign matrix is obtained as follows:
in the method, in the process of the invention,for the predicted centre point position +.>For the predicted amount of center point offset,for the predicted target width and height, the width can be directly taken as the slow time domain observation length;
finally, extracting a multi-vital-body feature matrix from a plurality of slow time domain-distance maps obtained by single measurement, and generating a final vital-body feature matrix after comprehensive analysis and judgment;
the specific method of the step S33 is as follows:
firstly, performing fast Fourier transform on a vital body feature matrix; and then carrying out wavelet entropy calculation on the vital sign matrix after the fast Fourier transformation, wherein the calculation formula is as follows:
in the method, in the process of the invention,representing the relative energy of signal i at the j scale,/-, and>representing the temporal behavior of wavelet entropy, H WT Represents the average wavelet entropy, N T Representing wavelet coefficients at a resolution T contained in a time window i;
then, calculating a corresponding wavelet entropy standard deviation SWT, wherein the calculation formula is as follows:
and finally judging whether the human body is according to the wavelet entropy and the wavelet entropy standard deviation.
Compared with the prior art, the invention has the following beneficial effects:
1. according to the method, the broadband orthogonal CPPWM signal is transmitted to serve as a MIMO radar detection signal, the non-fuzzy positioning and quantity detection of multiple life bodies are realized based on high-resolution imaging in the distance direction and the azimuth direction, and the anti-electromagnetic interference measurement is realized based on good autocorrelation of the CPPWM signal;
2. the invention uses the CenterNet life detection network for intelligent extraction of the life feature matrix, thereby enhancing the robustness of the life detection method;
3. the invention can efficiently, accurately and intelligently detect the quantity, the position, the type and the respiratory frequency of multiple life bodies in a search and rescue environment with low signal-to-noise ratio after disaster, and solves the problem that the existing life detection radar has high false alarm rate and high false alarm rate in search and rescue after disaster.
In conclusion, the invention can realize the efficient, accurate and intelligent detection of the number, the position, the type and the respiratory frequency of multiple living bodies in the search and rescue environment with low signal-to-noise ratio after disaster, and has the advantages of being not influenced by the environmental temperature, noise, site visibility, external electromagnetic interference and the like.
Drawings
FIG. 1 is a schematic structural diagram of a multi-living body intelligent detection device based on CPPWM-MIMO radar provided by the embodiment of the invention;
FIG. 2 is a schematic diagram of a CPPWM signal generator in accordance with an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an improved Vivaldi antenna according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a multi-channel high-speed hybrid sampler according to an embodiment of the present invention;
fig. 5 is an algorithm flow chart of a multi-living body intelligent detection method based on a CPPWM-MIMO radar provided by the embodiment of the invention.
In the figure: 1: CPPWM signal generator; 2 a power amplifier; 3: vivaldi linear sparse array antenna; 4: a low noise amplifier; 5: a multi-channel high-speed hybrid sampler; 6: a multi-core central processing unit; 7: a data processing module; 8: a result display module; 9: a Logistic mapping unit; 10: a 1 st counter; 11: a 2 nd counter; 12: a Tent mapping unit; 13: a 1 st comparator; 14: a 2 nd comparator; 15: a pulse generator; 16: a serializer; 17: a register; 18: a logic comparator; 19: a write address counter; 20: a static random access memory; 21: a static random access memory controller; 22: a read address counter; 23: a serial transceiving reset controller; 24: a serial transceiver; 25: a differential output module; 26: a serial transceiving dynamic reconfiguration controller; 27: a field programmable gate array; 28: a serial peripheral interface; 29: a programmable delay chip; 30: an analog-to-digital converter; 31: a transmitting antenna; 32: a receiving antenna; 33: a radiating arm; 34: a complementary split-ring resonator; 35: a radiating patch; 36: y-shaped slits; 37: the operating clock of the analog-to-digital converter.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments; all other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
As shown in fig. 1, the first embodiment of the present invention provides a multi-life body intelligent detection device based on cpwm-MIMO radar, which is characterized by comprising a cpwm signal generator 1, a plurality of power amplifiers 2, vivaldi linear sparse array antenna 3, a plurality of low noise amplifiers 4, a multi-channel high-speed hybrid sampler 5 and a data processing system; the CPPWM signal generator 1 simultaneously generates orthogonal multipath CPPWM detection signals and multipath CPPWM reference signals corresponding to the orthogonal multipath CPPWM detection signals by setting initial values of different Logistic mapping units and Tent mapping units. Orthogonal multipath CPPWM signals do not interfere with each other and do not affect each other. Each CPPWM detection signal is respectively connected with a signal input end of one power amplifier 2, and each CPPWM reference signal is respectively connected with a signal input end of the multi-channel high-speed mixed sampler 5; the array antenna 3 comprises a plurality of transmitting antennas 31 positioned in the middle and a plurality of receiving antennas 32 positioned at two sides, each transmitting antenna 31 is respectively connected with the signal output end of one power amplifier 2, and the output end of each receiving antenna 32 is respectively connected with the signal input end of one low-noise amplifier 4; the signal output end of each low-noise amplifier 4 is connected with the signal input end of the multichannel high-speed mixing sampler 5; the signal output end of the multichannel high-speed mixed sampler 5 is connected with the data processing system, and the life detection result is obtained through calculation and analysis of the data processing system.
Specifically, as shown in fig. 1, in the present embodiment, the number of the power amplifiers 2 and the low-noise amplifiers 4 is six, and the Vivaldi linear sparse array antenna 3 includes six transmitting antennas 31 located in the middle and three receiving antennas 32 located on both sides.
Specifically, in the present embodiment, the data processing system includes a multi-core central processing unit 6, a data processing module 7, and a result display module 8; the data processing module 7 is used for calculating reference and echo data, and then returning the results to the multi-core central processing unit 6 and displaying the multi-life body detection results in the result display module 8. The result display module 8 is used for displaying the multi-life body detection result; the multi-core central processing unit 6 is used for controlling the normal operation of the data processing module 7 and the result display module 8.
Specifically, in this embodiment, the transmitting antenna 31 and the receiving antenna 32 are both improved Vivaldi antennas, as shown in fig. 3, where the improved Vivaldi antennas are formed by arranging complementary split resonant rings 34 on radiating arms 33 corresponding to feeder ends of the Vivaldi antennas, and loading a plurality of Y-shaped slots 36 with different lengths on two sides of a radiating patch 35.
Further, as shown in fig. 2, in this embodiment, the cpwm signal generator 1 includes: the logic mapper 9, the first counter 10, the second counter 11, the Tent mapper 12, the first comparator 13, the second comparator 14, the pulse generator 15 and the signal storage output module.
Wherein the output of the Logistic mapper 9 is X n+1 ＝4(1-X n )X n Wherein X is 0 Representing the initial value, X, of the Logistic mapping function n And the output value of the Logistic mapping function after the nth cycle calculation is represented. The output ends of the first counter 5 and the Logistic mapper 9 are connected with a first comparator 13, and the output of the Tent mapper 12 is Y n+1 ＝1.99(0.5-|Y n -0.5|), wherein Y 0 Representing the initial value of the Tent mapping function, Y n Representing the output value of the Tent mapping function after the nth cycle calculation; the output ends of the second counter 11 and the Tent mapper 12 are connected with the second comparator 14, the output ends of the first comparator 13 and the second comparator 14 are connected with the input end of the pulse generator 15, and are respectively used for triggering the pulse generator 15 to continuously output a high-level signal and a low-level signal, and the high-level signal output by the pulse generator 15 is used for enabling to control the Tent mapper 12 and the second counter 11 to start working, controlling the logic mapper 9 and the first counter 10 to stop working and resetting the first counter 10; the low level signal output by the pulse generator 15 is used for enabling the control Tent mapper 12 and the second counter 11 to stop working and resetting the second counter 11, and is also used for enabling the control logic map 9 and the first counter 10 to start working and updating to enter the next state; the level signal output from the pulse generator 15 is stored and output through the signal storage output module.
Further, as shown in fig. 2, the signal storage output module includes: serializer 16, register 17, logic comparator 18, write address counter 19, sram 20, sram controller 21, read address counter 22, serial transmit-receive reset controller 23, serial transceiver 24, differential output module 25.
The serializer 16 is configured to convert the low-speed serial signal output by the pulse generator 15 into parallel data, then transmit the parallel data to the register 17 for buffering, and the logic comparator 18 is configured to extract and logically select the data in the register 17 according to the writing sequence, and then send the data to the sram controller 21, and further generate a pulse clock signal to send the sram controller 21 and the write address counter 19; the write address data generated by the write address counter 19 cooperates with the pulse clock signal and the data sent from the logic comparator 18 to cause the sram controller 21 to write the data into the sram 20.
The serial transceiver reset controller 23 is configured to release the reset state of the serial transceiver 24, the serial transceiver 24 is configured to generate a feedback clock, send the feedback clock to the read address counter 22 and the sram controller 21, match the read address data generated by the read address counter 22 with the feedback clock sent by the serial transceiver 24, enable the sram controller 21 to read the data in the sram 20, write the data into the serial transceiver 24, and finally deserialize the data by the serial transceiver 24 and output the cpwm signal through the differential output module 25.
Further, as shown in fig. 4, the multi-channel high-speed hybrid sampler 5 includes a field programmable gate array 27 and a plurality of data channels, each of which includes a serial peripheral interface 28, a programmable delay chip 29 and an analog-to-digital converter 30; the field programmable gate array 27 is used for controlling the programmable delay chip 29 through the serial peripheral interface 28 to delay the working clock 37 of the analog-to-digital converter 30.
Example two
As shown in fig. 5, a second embodiment of the present invention provides a method for intelligently detecting a multiple living body based on a cpwm-MIMO radar, which is implemented based on a device shown in fig. 1, and includes the following steps:
s1, chaos correlation filtering and ranging: and carrying out pairwise cross-correlation calculation on each CPPWM echo signal and each CPPWM reference signal to obtain each cross-correlation curve, and carrying out matched filtering and ranging calculation.
Taking the Vivaldi linear sparse array antenna 3 including six transmitting antennas 31 in the middle and three receiving antennas 32 on two sides as an example, performing two-by-two cross-correlation calculation on the 6 paths of cpwm echo signals and the 6 paths of cpwm reference signals to obtain 36 (6×6) channel cross-correlation curves, so as to realize matched filtering and distance estimation.
S2, processing the cross-correlation curve to obtain life body information, and simultaneously calculating and judging whether the life body is human or not according to chaos correlation filtering and ranging calculation results.
The specific steps for obtaining the life body information are as follows:
s21, imaging a cross-correlation curve based on a multi-scale weighted rapid BP algorithm.
The specific method of the step S21 is as follows: firstly, carrying out large-size meshing on the whole detection area, processing by a BP algorithm to obtain a low-resolution image, searching an energy maximum value by using a sliding window energy detection method, extracting a potential life area from the low-resolution image, carrying out small-size meshing on the area, processing by the BP algorithm to obtain a high-resolution image, and remaining areas are still low-resolution images; and iterating for 2-3 times until the potential life region image reaches the resolution requirement, finally obtaining an image matrix W of weighted superposition of the high-resolution image and the low-resolution image, and determining the weighting factor by calculating the mean value and standard deviation value of scattered data of each focus on the time delay curve. The iteration herein refers to an iteration of dividing the mesh size. The smaller the grid size, the higher the resolution and the iteration stops when the discrete grid size meets the imaging resolution requirement.
S22, suppressing artifacts in the image based on fast Robust Principal Component Analysis (RPCA) of factor group sparse regularization.
W is modeled as: w=l+s, where L represents low-rank artifacts and S represents sparse life; l is decomposed into A= [ a ] 1 ,a 2 ,···,a d ]∈R m×k ，B＝[b 1 ,b 2 ,···,b d ]∈R k×n And rank (L) k is more than or equal to min (m, n), a j And b j Is a column vector; is provided withAn sp norm that is the power of p of L, as shown in the following equation:
in sigma i Is the ith maximum singular value of L; when p=1, the number of the groups,when p=0, the number of the active groups,when p is more than 0 and less than 1, the smaller the p value is, the more approximate rank is, and 1/2 can be taken; according to factor set sparse regularization, equation (1) can be expressed as follows:
in the method, in the process of the invention,||·|| F representing f norms; solving the following formula can obtain L and S:
min||A|| 2,1 +α||B T || 2,1 +λ||S|| 1 s.t.W＝AB+S； (3)
and removing constraint conditions by using an augmented Lagrangian multiplier method, and solving the problems by using an alternate direction multiplier method to realize artifact suppression.
S23, performing two-dimensional space positioning and quantity detection on a plurality of living bodies in the image.
The imaging result diagram obtained through the processing of the steps S21 to S23 is provided with a plurality of pixel points, the number of the pixel points represents the number of the living bodies, and the coordinates of the pixel points in the imaging result diagram represent the two-dimensional space positions of the living bodies.
The specific steps of calculating and judging whether the living body is a human being include:
s31, accumulating chaotic correlation filtering and ranging results in a slow time domain to obtain an original slow time domain-distance matrix R, and preprocessing the R.
In the step S31, the specific method for preprocessing the slow time domain-distance matrix R is as follows:
firstly, eliminating static clutter and linear trend interference by using a linear trend removal method, wherein the processing procedure is as follows:
wherein x= [ M/M1 ] M ]，m＝[0,1,···,M-1] T ，1 M Is an mx 1 vector containing a unit value,representing a signal matrix obtained after the linear trend removal method;
and then the automatic gain control is utilized to enhance the weak life signal, and the signal enhancement process is as follows: for signal matrixBy calculating->Adjusting the gain by the signal power in a time window of given magnitude ω, where k and m represent matrices, respectivelyK=0, 1, … …, N k ，m＝0,1,……,M m ，N k And M is as follows m Representing fast and slow time domain sampling points and associated with a predetermined maximum gain g MAX Comparing, wherein the calculation formula is as follows:
wherein i represents a row, i=0, 1, … …, N- ω, g [ i, m]G is a gain based on signal power norm [i,m]Is a normalized representation of all gains with minimum gain, g mask [i,m]For gain mask, use gain mask pairPerforming weight adjustment; g min (m) represents the minimum gain, g, for all rows i for each m columns norm [i,m]＝g[i,m]/g min (m) normalized representation of all gains. And finally, gray scale processing is carried out, and a gray scale image is output.
S32, intelligently extracting a multi-life body feature matrix based on the CenterNet life detection network.
The specific method of step S32 is as follows:
screening gray level images without life bodies to form a data set, and marking life body labels to obtain a life feature data set for network training and testing;
inputting a gray matrix of training set data into an improved DLA-34 feature extraction network, and respectively training a center point, a center point offset and a size of a target to obtain a target boundary frame representing a vital body feature matrix;
the method for training the target center point comprises the following steps: for the center point P of each vital sign frame, its center point value P ε R at each resolution 2 First, calculate a low resolution central point equivalent valueR is the output stepping amplitude, and the Gaussian kernel is used for mapping the central point of each vital sign frame to thermodynamic diagram +.>Wherein, among them,wherein W and H represent the width and height of the image, respectively, < >>Indicating that this point is the center point, < >>Representing the point as background, then performing pixel-level logistic regression using a Focal Loss function, the expression of which is:
wherein alpha and beta respectively represent super parameters of the loss function, N is the number of center points in the image, and N is selected for normalization. The expression of the gaussian kernel is:
In sigma p Representing the target size adaptive standard deviation of the target,and->Respectively, the abscissa of the center point P, and x and y respectively, the abscissa of the randomly calculated point.
The method for training the offset of the target center point comprises the following steps: the local offset of each center point isTraining the offset of all object center points by using an L1 loss function; the expression of the L1 loss function is:
in the method, in the process of the invention,for the local offset, p represents the center point position coordinate. The method for training the target size comprises the following steps: order theFor the bounding box of object k, the center position is +.>x i And y i Respectively, integer coordinates. And regressing the size of each object k to +.>Performing target size training by using a Smooth L1 loss function; targeting with a single size predictor +.>In the target detection, let ∈ ->In order to detect n sets of center points, obtaining a boundary box by extracting 100 peak points on a thermodynamic diagram, and reserving if the peak point is larger than 8 neighborhood points of the peak point; />
The target bounding box representing the vital sign matrix is obtained as follows:
in the method, in the process of the invention,for the predicted centre point position +.>For the predicted amount of center point offset,for the predicted target width and height, the width may be taken directly as the slow time domain observation length. And extracting the multi-vital-body characteristic matrix from the 36 (6 multiplied by 6) Zhang Man time-distance graph obtained by single measurement, and comprehensively analyzing and judging to generate a final vital-body characteristic matrix. If a certain vital sign matrix exists in the half of the slow time domain-distance graphs, judging that the vital sign exists; otherwise, it is not present.
Finally, extracting a multi-vital-body feature matrix from a plurality of slow time domain-distance maps obtained by single measurement, and comprehensively analyzing and judging to generate a final vital-body feature matrix.
S33, performing Fast Fourier Transform (FFT) on the multi-vital-body feature matrix, wherein the peak position represents the respiratory frequency, performing wavelet entropy analysis on the vital-body feature matrix after the fast Fourier transform, and judging whether the vital body is a human or an animal based on the wavelet entropy.
After fast Fourier transform FFT is performed on the multi-vital-feature matrix, a two-dimensional graph with the abscissa representing the respiratory rate and the ordinate representing the respiratory amplitude can be obtained, and the abscissa corresponding to the peak position of the peak in the two-dimensional graph represents the respiratory rate. The target can be judged as the living body by acquiring the respiratory rate, and whether the living body is a human or an animal is further distinguished by calculating the wavelet entropy of the living body feature matrix.
The specific method of the step S33 is as follows:
firstly, performing fast Fourier transform on a vital body feature matrix; then wavelet entropy analysis is carried out on the vital sign matrix after the fast Fourier transformation, and the energy of the respiratory signal after the wavelet transformation in the j scale in the time window i is calculated Obtaining the sum of signal energy of each scale as +.>The relative energy of each scale signal is +.>The wavelet entropy of the vital signal in the time window is:
in the method, in the process of the invention,representing the relative energy of signal i at the j scale,/-, and>representing the temporal behavior of wavelet entropy, H WT Represents the average wavelet entropy, N T Representing the wavelet coefficients at the resolution T contained in the time window i. The corresponding wavelet entropy standard deviation is as follows:
typically, wavelet entropy and wavelet entropy standard deviation greater than 1 are both expressed as human and near 0.5 as animal, thereby distinguishing human from animal. And finally, the result display module displays the multi-life body detection result.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (8)
1. The multi-life intelligent detection device based on the CPPWM radar is characterized by comprising a CPPWM signal generator (1), a plurality of power amplifiers (2), a Vivaldi linear sparse array antenna (3), a plurality of low-noise amplifiers (4), a multi-channel high-speed mixed sampler (5) and a data processing system;
The CPPWM signal generator (1) simultaneously generates orthogonal multipath CPPWM detection signals and multipath CPPWM reference signals corresponding to the orthogonal multipath CPPWM detection signals by setting initial values of different Logistic mapping units and Tent mapping units; each CPPWM detection signal is respectively connected with a signal input end of a power amplifier (2), and each CPPWM reference signal is respectively connected with a signal input end of a multi-channel high-speed mixed sampler (5);
the Vivaldi linear sparse array antenna (3) comprises a plurality of transmitting antennas (31) positioned in the middle and a plurality of receiving antennas (32) positioned at two sides, each transmitting antenna (31) is respectively connected with a signal output end of one power amplifier (2), and the output end of each receiving antenna (32) is respectively connected with a signal input end of one low-noise amplifier (4); the signal output end of each low-noise amplifier (4) is connected with the signal input end of the multichannel high-speed mixed sampler (5);
the signal output end of the multichannel high-speed mixed sampler (5) is connected with the data processing system, and the data processing system is used for calculating and analyzing to obtain a life detection result;
the CPPWM signal generator (1) comprises:
the system comprises a Logistic mapper (9), a first counter (10), a second counter (11), a Tent mapper (12), a first comparator (13), a second comparator (14), a pulse generator (15) and a signal storage output module;
The output of the Logistic mapper (9) is X n+1 ＝4(1-X n )X n Wherein X is 0 Representing the initial value, X, of the Logistic mapping function n Representing the output value of the Logistic mapping function after the nth cycle calculation; the output ends of the first counter (10) and the Logistic mapper (9) are connected with the first comparator (13), and the output end of the Tent mapper (12) is Y n+1 ＝1.99(0.5-|Y n -0.5|), wherein Y 0 Representation TenInitial value of t mapping function, Y n Representing the output value of the Tent mapping function after the nth cycle calculation; the output ends of the second counter (11) and the Tent mapper (12) are connected with the second comparator (14), the output ends of the first comparator (13) and the second comparator (14) are connected with the input end of the pulse generator (15), the output ends are respectively used for triggering the pulse generator (15) to continuously output a high-level signal and a low-level signal, the high-level signal output by the pulse generator (15) is used for enabling and controlling the Tent mapper (12) and the second counter (11) to start working, and also used for controlling the Logistic mapper (9) and the first counter (10) to stop working and resetting the first counter (10); the low-level signal output by the pulse generator (15) is used for enabling the control Tent mapper (12) and the second counter (11) to stop working and resetting the second counter (11), and is also used for enabling the control logic mapper (9) and the first counter (10) to start working and updating to enter the next state; the level signal output by the pulse generator (15) is stored and output by the signal storage output module.
2. The cpwm radar-based multi-living-body intelligent detection apparatus according to claim 1, wherein the number of the power amplifier (2) and the low-noise amplifier (4) is 6, and the Vivaldi linear sparse array antenna (3) includes six transmitting antennas (31) located in the middle and three receiving antennas (32) located at both sides;
the data processing system comprises a multi-core central processing unit (6), a data processing module (7) and a result display module (8);
the data processing module (7) is used for calculating reference and echo data;
the result display module (8) is used for displaying the multi-life body detection result;
the multi-core central processing unit (6) is used for controlling the normal operation of the data processing module (7) and the result display module (8).
3. The multi-life intelligent detection device based on the CPPWM radar according to claim 1, wherein the transmitting antenna (31) and the receiving antenna (32) are improved Vivaldi antennas, the improved Vivaldi antennas are formed by arranging complementary split resonant rings (34) on radiating arms (33) corresponding to the tail ends of feeder lines of the Vivaldi antennas, and a plurality of Y-shaped slits (36) with different lengths are loaded on two sides of a radiating patch (35).
4. The cpwm radar-based intelligent multi-living body detection apparatus according to claim 3, wherein said signal storage output module comprises: a serializer (16), a register (17), a logic comparator (18), a write address counter (19), a static random access memory (20), a static random access memory controller (21), a read address counter (22), a serial transceiving reset controller (23), a serial transceiver (24) and a differential output module (25);
The serializer (16) is used for converting the low-speed serial signal output by the pulse generator (15) into parallel data, transmitting the parallel data to the register (17) for buffering, and the logic comparator (18) is used for extracting and logically selecting the data in the register (17) according to the writing sequence and then transmitting the data to the static random access memory controller (21) and generating a pulse clock signal to transmit the static random access memory controller (21) and the writing address counter (19); the write address data generated by the write address counter (19) is matched with the pulse clock signal and the data sent by the logic comparator (18), so that the static random access memory controller (21) writes the data into the static random access memory (20);
the serial transceiver reset controller (23) is used for releasing the reset state of the serial transceiver (24), the serial transceiver (24) is used for generating a feedback clock and sending the feedback clock to the read address counter (22) and the static random access memory controller (21), the read address data generated by the read address counter (22) is matched with the feedback clock sent by the serial transceiver (24), the static random access memory controller (21) reads out the data in the static random access memory (20) and writes the data into the serial transceiver (24), and finally the serial transceiver (24) deserializes the data and outputs CPPWM signals through the differential output module (25).
5. The cpwm radar-based multi-living-body intelligent detection apparatus of claim 1, said multi-channel high-speed hybrid sampler (5) comprising a field programmable gate array (27) and a plurality of data channels, each data channel comprising a serial peripheral interface (28), a programmable delay chip (29) and an analog-to-digital converter (30); the field programmable gate array (27) is used for controlling the programmable delay chip (29) through the serial peripheral interface (28) and delaying the working clock (37) of the analog-to-digital converter (30).
6. A multi-living body intelligent detection method based on cpwm radar, characterized by being realized based on the device of claim 1, comprising the steps of:
s1, chaos correlation filtering and ranging: performing pairwise cross correlation calculation on each CPPWM echo signal and each CPPWM reference signal to obtain each cross correlation curve, and performing matched filtering and ranging calculation;
s2, processing the cross-correlation curve to obtain life body information, and calculating and judging whether the life body is human or not according to chaos correlation filtering and ranging calculation results;
the specific steps for obtaining the life body information are as follows:
s21, imaging a cross-correlation curve based on a multi-scale weighted rapid BP algorithm;
S22, rapidly robust principal component analysis based on factor group sparse regularization suppresses artifacts in the image;
s23, performing two-dimensional space positioning and quantity detection on a plurality of living bodies in the image;
the specific steps of calculating and judging whether the living body is human include:
s31, accumulating chaotic correlation filtering and ranging results in a slow time domain to obtain an original slow time domain-distance matrix R, and preprocessing the R;
s32, intelligently extracting a multi-life body feature matrix based on a CenterNet life detection network;
s33, performing fast Fourier transform on the multi-living body feature matrix, performing wavelet entropy analysis on the Fourier transformed living body feature matrix, and judging whether the living body is a human or an animal based on the wavelet entropy.
7. The intelligent multi-life detection method based on CPPWM radar of claim 6, wherein,
the specific method of the step S21 is as follows: firstly, carrying out large-size meshing on the whole detection area, processing by a BP algorithm to obtain a low-resolution image, searching an energy maximum value by using a sliding window energy detection method, extracting a potential life area from the low-resolution image, carrying out small-size meshing on the area, processing by the BP algorithm to obtain a high-resolution image, and remaining areas are still low-resolution images; iterating for 2-3 times until the potential life region image reaches the resolution requirement, finally obtaining a weighted and overlapped image matrix W of the high-resolution image and the low-resolution image, and determining a weighting factor by calculating the mean value and standard deviation value of scattered data of each focus on the time delay curve;
The specific method of the step S22 is as follows: the constraint condition is removed by using an augmented Lagrangian multiplier method, and then an equation is solved by using an alternate direction multiplier method, so that low-rank artifacts L and sparse life bodies S are obtained, artifact suppression is realized, and the equation is as follows:
min||A|| 2,1 +α||B T || 2,1 +λ||S|| 1 s.t.W＝AB+S；
in the method, in the process of the invention,represents f norm, alpha represents weight parameter, lambda represents regularization parameter, A and B represent decomposed clutter matrix, S represents target matrix, a j Representing the j-th column vector in matrix a;
in the step S23, the number of pixels in the imaging image after artifact suppression represents the number of living bodies, and the coordinates of the pixels in the imaging result diagram represent the two-dimensional spatial positions of the living bodies.
8. The intelligent multi-life detection method based on CPPWM radar of claim 6, wherein,
in the step S31, the specific method for preprocessing the time-distance matrix R is as follows:
firstly, eliminating static clutter and linear trend interference by using a linear trend removal method, wherein the processing procedure is as follows:
wherein x= [ M/M1 ] M ]，m＝[0,1,···,M-1] T ，1 M Is an mx 1 vector containing a unit value,representing a signal matrix obtained after the linear trend removal method;
and then the automatic gain control is utilized to enhance the weak life signal, and the signal enhancement process is as follows: for signal matrix By calculating->The gain is adjusted by the signal power in a time window of given size ω, where k and m represent the matrix +.>K=0, 1, … …, N k ，m＝0,1,……,M m ，N k And M is as follows m Representing fast and slow time domain sampling points and associated with a predetermined maximum gain g MAX Comparing, wherein the calculation formula is as follows:
wherein i=0, 1, … …, N- ω, g [ i, m]G is a gain based on signal power norm [i,m]Is to make all gains according to the minimum gainNormalized representation, g mask [i,m]For gain mask, use gain mask pairPerforming weight adjustment; g min (m) represents the minimum gain, g, for each value of m, for all i norm [i,m]＝g[i,m]/g min (m) carrying out normalization representation on all gains, and finally carrying out gray scale processing and outputting a gray scale image;
the specific method of step S32 is as follows:
screening gray level images without life bodies to form a data set, and marking life body labels to obtain a life feature data set for network training and testing;
inputting a gray matrix of training set data into an improved DLA-34 feature extraction network, and respectively training a center point, a center point offset and a size of a target to obtain a target boundary frame representing a vital body feature matrix;
the method for training the target center point comprises the following steps: for the center point P of each vital sign frame, where P ε R 2 First, calculate a low resolution central point equivalent valueR is the output stepping amplitude, and the Gaussian kernel is used for mapping the central point of each vital sign frame to thermodynamic diagram +.>Wherein W and H represent the width and height of the image, respectively, < >>Indicating that this point is the center point, < >>Representing the point as background, then performing pixel-level logistic regression using a FocalLoss loss function, whose expression is:
wherein alpha and beta respectively represent super parameters of a loss function, and N is the number of center points in the image; y is Y xyz Representing a gaussian kernel, the expression of which is:
in sigma p Representing the target size adaptive standard deviation of the target,and->Respectively representing the abscissa of the center point P, and x and y respectively representing the abscissa of the random calculation point;
the method for training the offset of the target center point comprises the following steps: the local offset of each center point isTraining the offset of all object center points by using an L1 loss function; the expression of the L1 loss function is:
in the method, in the process of the invention,p represents the position coordinate of the central point for the local offset;
the method for training the target size comprises the following steps: order theBoundary box of object k, center position isx i And y i Respectively integer coordinates, and regressing the size of each object k to be Performing target size training by using a Smooth L1 loss function; targeting with a single size predictor +.>In the target detection, let ∈ ->In order to detect n sets of center points, obtaining a boundary box by extracting 100 peak points on a thermodynamic diagram, and reserving if the peak point is larger than 8 neighborhood points of the peak point;
the target bounding box representing the vital sign matrix is obtained as follows:
in the method, in the process of the invention,for the predicted centre point position +.>For the predicted amount of center point offset,taking the slow time domain observation length for the predicted target width and height;
finally, extracting a multi-vital-body feature matrix from a plurality of slow time domain-distance maps obtained by single measurement, and generating a final vital-body feature matrix after comprehensive analysis and judgment;
the specific method of the step S33 is as follows:
firstly, performing fast Fourier transform on a vital body feature matrix; and then carrying out wavelet entropy calculation on the vital sign matrix after the fast Fourier transformation, wherein the calculation formula is as follows:
in the method, in the process of the invention,representing the relative energy of signal i at the j scale,/-, and>representing the temporal behavior of wavelet entropy, H WT Represents the average wavelet entropy, N T Representing wavelet coefficients at a resolution T contained in a time window i;
then, calculating a corresponding wavelet entropy standard deviation SWT, wherein the calculation formula is as follows:
And finally judging whether the human body is according to the wavelet entropy and the wavelet entropy standard deviation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110694306.1A CN113567974B (en) | 2021-06-22 | 2021-06-22 | Multi-living-body intelligent detection device and method based on CPPWM radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110694306.1A CN113567974B (en) | 2021-06-22 | 2021-06-22 | Multi-living-body intelligent detection device and method based on CPPWM radar |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113567974A CN113567974A (en) | 2021-10-29 |
| CN113567974B true CN113567974B (en) | 2023-11-03 |
Family
ID=78162569
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110694306.1A Active CN113567974B (en) | 2021-06-22 | 2021-06-22 | Multi-living-body intelligent detection device and method based on CPPWM radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113567974B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5321409A (en) * | 1993-06-28 | 1994-06-14 | Hughes Missile Systems Company | Radar system utilizing chaotic coding |
| CN101089655A (en) * | 2006-06-14 | 2007-12-19 | 中国科学院电子学研究所 | Synthetic aperture radar system using chaos signal |
| DE60139793D1 (en) * | 2001-12-21 | 2009-10-15 | St Microelectronics Srl | Device for distance measurement by means of chaotic signals |
| CN104765031A (en) * | 2015-03-02 | 2015-07-08 | 太原理工大学 | Ultra-wide bandwidth microwave chaos life detection radar device |
| CN107121677A (en) * | 2017-06-02 | 2017-09-01 | 太原理工大学 | Avoidance radar method and device based on ultra wide band cognition CPPM signals |
| CN107293077A (en) * | 2017-06-02 | 2017-10-24 | 太原理工大学 | Perimeter intrusion detecting device and method based on orthogonal CPPM signals |
| WO2019170878A1 (en) * | 2018-03-08 | 2019-09-12 | Iee International Electronics & Engineering S.A. | Method and system for target detection using mimo radar |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9442189B2 (en) * | 2010-10-27 | 2016-09-13 | The Fourth Military Medical University | Multichannel UWB-based radar life detector and positioning method thereof |
-
2021
- 2021-06-22 CN CN202110694306.1A patent/CN113567974B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5321409A (en) * | 1993-06-28 | 1994-06-14 | Hughes Missile Systems Company | Radar system utilizing chaotic coding |
| DE60139793D1 (en) * | 2001-12-21 | 2009-10-15 | St Microelectronics Srl | Device for distance measurement by means of chaotic signals |
| CN101089655A (en) * | 2006-06-14 | 2007-12-19 | 中国科学院电子学研究所 | Synthetic aperture radar system using chaos signal |
| CN104765031A (en) * | 2015-03-02 | 2015-07-08 | 太原理工大学 | Ultra-wide bandwidth microwave chaos life detection radar device |
| CN107121677A (en) * | 2017-06-02 | 2017-09-01 | 太原理工大学 | Avoidance radar method and device based on ultra wide band cognition CPPM signals |
| CN107293077A (en) * | 2017-06-02 | 2017-10-24 | 太原理工大学 | Perimeter intrusion detecting device and method based on orthogonal CPPM signals |
| WO2019170878A1 (en) * | 2018-03-08 | 2019-09-12 | Iee International Electronics & Engineering S.A. | Method and system for target detection using mimo radar |
Non-Patent Citations (4)
| Title |
|---|
| Pattern synthesis of MIMO radar based on chaotic differential evolution algorithm;Jiang Yi 等;《Optik》;第140卷;第794-801页 * |
| 一种便携式伪随机编码超宽带人体感知雷达设计;夏正欢 等;《雷达学报》;第4卷(第5期);第527-537页 * |
| 一种基于超宽带雷达的多观测点人体呼吸信号检测方法;姚思奇;吴世有;张经纬;叶盛波;方广有;;电子测量技术(10);第194-201页 * |
| 基于调频网络的混沌雷达信号产生;邹云;《中国优秀硕士学位论文全文数据库信息科技辑》(第02期);第I136-924页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113567974A (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107132535B (en) | ISAR sparse band imaging method based on variational Bayesian learning algorithm | |
| US9075129B2 (en) | Method and system for forming images by comparing subsets of image data | |
| Dong et al. | A novel compressive sensing algorithm for SAR imaging | |
| CN111796272B (en) | Real-time gesture recognition method and computer equipment for through-wall radar human body image sequence | |
| Cheng et al. | Compressive sensing radar imaging with convolutional neural networks | |
| CN110333489B (en) | Processing method for SAR echo data sidelobe suppression by adopting CNN and RSVA combination | |
| CN111062321B (en) | SAR detection method and system based on deep convolutional network | |
| Kılıç et al. | Through-wall radar classification of human posture using convolutional neural networks | |
| Domínguez et al. | A multisquint framework for change detection in high-resolution multitemporal SAR images | |
| Wang et al. | Multi-classification of UWB signal propagation channels based on one-dimensional wavelet packet analysis and CNN | |
| CN110554384A (en) | imaging method based on microwave signal | |
| WO2021250943A1 (en) | Graph-based array signal denoising for perturbed synthetic aperture radar | |
| Agarwal et al. | A novel neural network based image reconstruction model with scale and rotation invariance for target identification and classification for Active millimetre wave imaging | |
| CN112198506A (en) | Method, device and system for learning and imaging ultra-wideband through-wall radar and readable storage medium | |
| CN117148353B (en) | Close-range multi-target detection method and security inspection system based on millimeter wave SAR | |
| Cheong et al. | Phased-array design for biological clutter rejection: Simulation and experimental validation | |
| CN107656271A (en) | Terahertz radar imagery algorithm based on compressed sensing reconstruct | |
| CN108562901B (en) | ISAR high-resolution imaging method based on maximum signal-to-noise-and-noise ratio criterion | |
| CN108919263B (en) | ISAR high-resolution imaging method based on maximum mutual information criterion | |
| CN113567974B (en) | Multi-living-body intelligent detection device and method based on CPPWM radar | |
| CN102722878A (en) | SAR (synthetic aperture radar) image despeckle method based on target extraction and PPB (probabilistic patch-based filter) algorithm | |
| CN111931570B (en) | Through-wall imaging radar human body target detection method based on full convolution network | |
| Shao et al. | Signal classification for ground penetrating Radar using sparse kernel feature selection | |
| CN114114246A (en) | Through-wall radar imaging method and system, terminal device and readable storage medium | |
| Xu et al. | Backward projection imaging of through-wall radar based on airspace nonuniform sampling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |