CN215575652U - Dynamic monitoring device for pedestrian flow based on millimeter wave radar - Google Patents

Abstract

The utility model discloses a dynamic monitoring device for pedestrian flow based on a millimeter wave radar, which comprises a tripod, a computer, a data line, a millimeter wave radar and a door, wherein the computer and the millimeter wave radar are connected through the data line, a vertical rod at the top of the tripod is connected with a horizontal rod, the millimeter wave radar is arranged at the tail end of the horizontal rod, the millimeter wave radar is positioned at the same height as the door, the millimeter wave radar is positioned obliquely above a detection area, and a sector scanning area range of the radar is arranged below the millimeter wave radar. The millimeter wave radar is small in size, low in cost and high in resolution, people in public places can be effectively controlled through real-time people flow statistics of the equipment, and therefore the occurrence of safety accidents caused by too dense people is avoided.

Description

Dynamic monitoring device for pedestrian flow based on millimeter wave radar

Technical Field

The utility model relates to a dynamic monitoring device for man-flow of a millimeter wave radar in the field of radar target detection, in particular to a dynamic monitoring device for man-flow of a millimeter wave radar based on data screening and double-time-point detection.

Background

With the social development, the trip rate of people is continuously improved, and especially the traffic of people in public places such as markets, stations and scenic spots is increased, so that the safety problem is increasingly prominent. Through real-time people flow statistics, timely and effective shunting, dredging and control can be carried out on people, possible safety accidents can be avoided, and important reference values are configured for optimizing resources in public places.

The early people flow rate statistics of public places mainly adopts a manual calculation mode, the method cannot accurately count the number of people, the labor cost is high, the management is complex, and the method is not favorable for large data analysis and application of the method which are rapidly developed at present. In order to effectively count the human flow, save the human capital and guard the public area safety, domestic and foreign scholars study the automatic human flow counting method on the basis. The people flow monitoring method based on infrared sensor detection utilizes a light emitting diode emitting infrared rays and a receiving diode sensing the infrared rays to realize infrared ray induction and automatically count targets with certain temperature passing through an induction area, but the method has limited application scenes, cannot distinguish the motion direction of target individuals and is difficult to distinguish the condition that a plurality of people walk side by side. The people flow monitoring method based on ultrasonic detection realizes the position measurement of a human body target by processing a transmitting sound wave and an echo signal thereof, thereby realizing the statistics of the people flow, but the method has limited action distance, is easy to be interfered by the external environment, has a measurement blind area, and has certain limitation in the people flow monitoring application. With the rise of artificial intelligence algorithms, video monitoring becomes one of the main methods for monitoring the flow of people, and the current detection and tracking algorithms for counting the flow of people based on a video monitoring system can be generally divided into three categories: the method is based on the bottom layer characteristics of the image, the method is based on the motion trail of the characteristic points, and the detection tracking method is based on the moving target. The people flow monitoring technology based on video monitoring utilizes optical sensors such as a camera to limit application environment, has high sensitivity to environmental factors such as illumination, smoke and the like, is difficult to work in all weather, and simultaneously has the risk of personnel privacy disclosure in the collected monitoring data. Therefore, it is very necessary to develop a people flow rate monitoring method that can count people flow rate with high accuracy and overcome the disadvantages of the existing people flow rate counting method.

SUMMERY OF THE UTILITY MODEL

The real-time people flow statistics can effectively manage and control the personnel in public places so as to avoid safety accidents caused by excessively dense people. Aiming at the defects of the existing people flow monitoring method, the utility model provides a people flow dynamic monitoring device based on data screening and double-time-point detection by utilizing the characteristics of small size, low cost and high resolution of a millimeter wave radar.

The utility model provides the following technical scheme:

the utility model provides a flow of people developments monitoring devices based on millimeter wave radar, includes tripod support, computer, data line, millimeter wave radar and door, through data line connection computer and millimeter wave radar, the montant and the horizontal pole at tripod support top are connected, the terminal installation millimeter wave radar of horizontal pole, millimeter wave radar position is in the same eminence of door, the millimeter wave radar is located the oblique top in detection area, the millimeter wave radar below is the fan-shaped regional scope of strafing of radar.

Further, the inclination angle α of the millimeter wave radar is 40 to 50 °. The inclination angle α of the millimeter-wave radar is preferably 45 °.

Further, the working frequency of the millimeter wave radar is 60GHz, the bandwidth is 4GHz, the number of coded pulses in a unit frame is 128, and the frame period is 40 ms.

A people flow dynamic monitoring method based on millimeter wave radar is characterized in that a human body target echo signal model is established, and distance and Doppler frequency information of scattering points of a human body are obtained by performing two-dimensional fast Fourier transform processing on the human body target echo signal; then, filtering clutter scattering points through constant false alarm processing, carrying out azimuth estimation on reserved scattering points, obtaining angle information of human body target scattering points, obtaining scattering point positions through two-dimensional coordinate transformation, and forming point cloud data together with corresponding Doppler frequency information; then, the motion direction of the human body is judged according to the positive and negative Doppler frequencies, and meanwhile, point cloud data are screened according to the difference of Doppler characteristics of different parts when the human body moves so as to reduce the number of interference points and avoid the false alarm problem of a density-based clustering algorithm; and finally, counting the number of people in a specific area at double time points, and further correcting the pedestrian flow data by using a clustering result obtained between the double time points, so that the problem of misjudgment caused by different walking speeds of the human body is solved.

Compared with the prior art, the utility model has the beneficial effects that:

(1) the utility model utilizes the advantages of high range resolution, low transmitting power and strong penetrating power of the millimeter wave radar to monitor the flow of people, is less influenced by environmental factors such as light, smoke and the like, and has no risk of personnel privacy disclosure.

(2) The utility model filters out the point cloud data with smaller frequency by setting a proper Doppler threshold so as to reduce the influence of interference scattering points of respiration, heartbeat and small arm amplitude swing of people on a DBSCAN clustering result, and simultaneously divides the point cloud data set into two data sets of people entering and exiting according to the positive and negative of the Doppler frequency, thereby not only improving the accuracy of human body target condensation, but also correctly distinguishing two situations of people entering and exiting.

(3) And counting the number of personnel in a specific area every delta t time, and correcting the counted data by traversing the personnel data on the same path in the detection area within the delta t time, so that the problems of missed detection and false detection caused by different walking speeds of the human body are solved.

Drawings

FIG. 1 is a flow chart of an implementation of the present invention.

Fig. 2 is a schematic diagram of a people flow statistics scenario.

Fig. 3a is a millimeter wave radar transmit array structure.

Fig. 3b is a millimeter wave radar receiving array structure.

Fig. 4 is a two-dimensional FFT processing flow.

Fig. 5a is a traditional DBSCAN clustering result when a single person enters and exits.

Fig. 5b is a DBSCAN cluster diagram based on doppler frequency filtering when a single person enters and exits.

Fig. 6 is a schematic view of a detection region.

FIG. 7 is a schematic view showing a case where the walking speed of a person is normal (t)₁Time point of person position t₂The location of the person at the point in time).

FIG. 8 is a diagram illustrating a case where the walking speed of a person is high (t)₁Time point of person position t₂The location of the person at the point in time).

FIG. 9 is a diagram illustrating a case where the walking speed of the person is slow (t)₁Time point of person position t₂The location of the person at the point in time).

Fig. 10 is an experimental scenario of the inventive apparatus.

1. The device comprises a tripod support, 11, vertical rods, 12, a cross rod, 2, a computer, 3, a data line, 4, a millimeter wave radar, 5, a door, 6 and a detection area; 7. the radar sector scans the area range.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

A people flow dynamic monitoring method based on a millimeter wave radar comprises the following steps:

step 1: the echo signal of the human target is processed by frequency mixing to obtain an intermediate frequency signal, then two-dimensional FFT processing, constant false alarm rate detection and phase method angle measurement are carried out,thereby obtaining a point cloud data set P of human body scattering points₀Can be represented as P₀:{(x_i,y_i,f_di) 1, 2.., I }, wherein x is_iAnd y_iRespectively the coordinate position of the i-th scattering point in the two-dimensional plane, f_diIs the Doppler frequency of the ith scattering point, and I is the number of scattering points of the human body target.

The utility model adopts a Frequency Modulated Continuous Wave (FMCW) radar, has the characteristics of larger bandwidth, almost no range blind area, low transmitting power, high resolution and the like, and is suitable for the application scene of monitoring the pedestrian volume in public places. Fig. 2 is a schematic view of a flow monitoring scene of the inventor, and a millimeter wave radar is arranged obliquely above a detection area, so that the radar can better detect a plurality of human body targets. The millimeter wave radar module selects an IWR6843 evaluation board of Texas Instruments (TI), an onboard 3-transmitting 4-receiving antenna can transmit 60GHz frequency modulation continuous wave signals, a radio frequency front end, a low noise amplifier, an ARM processor, a DSP, a memory and the like are integrated inside the millimeter wave radar module, and the processing of intermediate frequency sampling cache, multi-dimensional FFT, constant false alarm detection and the like of echo signals can be met. A millimeter wave radar transmitting antenna (figure 3a), a millimeter wave radar receiving antenna (figure 3b),

the transmitting antenna and the receiving antenna have a distance d_tAnd d_r. Assuming that a plurality of human objects are in the detection area, I scattering points are provided, wherein the azimuth angle of the scattering point I (I ═ 1, 2.. said., I) is θ_i. It is assumed that the FMCW radar emits a sawtooth chirp continuous wave with a transmission signal of

In the formula, A_TRepresents a transmit power; f. of_cRepresenting the starting frequency of Chirp; b represents the bandwidth of Chirp; t is_cRepresenting the duration of Chirp;

representing phase noise. The echo signal of FMCW radar is

In the formula (I), the compound is shown in the specification,

for indicating electromagnetic waves in radar and range radar R_iThe round trip time between the ith target scatter point of (a), where v_diIs the radial velocity of the ith target relative to the radar, positive in the direction of approach to the radar, c is the speed of light, f_diRepresents the Doppler frequency of the ith scattering point; alpha is alpha_iIt is related to the return loss of the ith scattering point. Will transmit signal x_T(t) and a received signal x_R(t) analysis by mixing and then combining the I/Q signals, wherein the frequency signal can be approximately represented as

In the formula, A_RiRepresenting the received power of the ith scattering point, f_bi＝2B(R_i+v_dit)/(cT_c) Representing the difference frequency, phase of the ith scattering point

Is the phase of the ith scattering point,

is the residual noise of the ith target. The residual noise is caused by the close distance between the human target and the radar

And phase

In (1)

The values of the terms are small and negligible.

Sampling the mixed intermediate frequency signal, and assuming that the number of sampling points in each period is N and the number of sampling periods is M, the intermediate frequency signal can be represented in a sampling matrix form

Where N is 1,2, …, N denotes the corresponding index on the fast-time sampling axis, M is 1,2, …, M denotes the corresponding index on the slow-time sampling axis, T_f,T_sThe sampling intervals on the fast and slow time axes, respectively.

For the radar sampling data matrix Y shown in formula (4), the intermediate frequency signal frequency f_biAnd Doppler frequency f_diThe distance information and the doppler information of each scattering point are included, so that the distance information of the target can be obtained by performing N-point FFT on each row of the radar sampling data matrix Y, the doppler information of the target can be obtained by performing M-point FFT on each column, and the processing flow is shown in fig. 4.

Since the use scenario of the people flow monitoring method is mostly in a crowded environment, target information, background noise and clutter interference are often included in the echo signal. In order to realize effective detection of the human target, clutter scattering points need to be filtered. The CFAR algorithm is a target detection algorithm based on a threshold, and judges whether a target exists or not by estimating background noise and clutter power of a reference unit adjacent to a test unit and effective signals and noise received by a receiver. The most common of the CFAR algorithms is the cell average constant false alarm detection algorithm (CA-CFAR), which estimates the background clutter power by averaging neighboring cells outside the protected cell. The utility model adopts a two-dimensional CA-CFAR detector to respectively carry out constant false alarm detection on the distance dimension and the Doppler frequency dimension of a radar sampling data matrix Y.

For the same scattering point I (I ═ 1, 2.., I), the angle can be estimated from the echo phase difference received by the two receiving antennas. Suppose the receiving echo path difference of two receiving antennas is DeltaR_iThen the phase difference of the received signal at the ith scattering point at the time t is

As can be seen from FIG. 3b, the receive echo path difference between two adjacent receive antennas can be approximated to

ΔR_i＝d_t sinθ_i (6)

As can be seen from the equations (5) and (6), the azimuth angle θ of the i-th scattering point_iCan be expressed as

Thus, the two-dimensional coordinates of the ith scattering point can be expressed as

In the formula, x_iAnd y_iRespectively the coordinate position of the ith point target on the two-dimensional plane. To obtain point cloud data containing scattering point position and Doppler information, a data set P is used₀Is shown as

P₀:{(x_i,y_i,f_di),i＝1,2,...,I} (9)

Step 2: in order to reduce the influence of scattering points of human respiration, heartbeat and arm small-amplitude swing on human body target condensation, the scattering points with small Doppler frequency are filtered by setting a proper Doppler threshold, and only the scattering points generated by the translation of the whole human trunk are reserved. Setting the Doppler threshold to f_tIn point cloud data set P₀Data set P obtained after filtering out points of small Doppler frequency₁:{(x_j,y_j,f_dj)||f_dj|＞f_tJ ═ 1,2,.. J }, where x is_jAnd y_jRespectively representing the coordinate position of the jth scattering point with the Doppler frequency larger than the threshold value on the two-dimensional plane，f_djDenotes the Doppler frequency of the jth scattering point, J denotes that | f is satisfied_dj|＞f_tThe number of all scattering points, | · | is an absolute value.

In the application scene of people flow monitoring, except the integral trunk translation of a human body, the breathing, heartbeat and small-amplitude arm swing of a person can generate a micro Doppler effect, and the micro Doppler effect corresponds to point cloud data with small Doppler frequency. A clustering method based on Doppler frequency screening is characterized in that the points with low frequency are removed by setting a proper Doppler threshold, so that the influence of interference points of human respiration, heartbeat and arm small-amplitude swing on a clustering result is reduced. Setting the Doppler threshold to f_tThen the data set P after the point with the smaller Doppler frequency is removed₁Is composed of

P₁:{(x_j,y_j,f_dj)||f_dj|＞f_t,j＝1,2,...,J} (10)

In the formula, x_jAnd y_jRespectively representing the target coordinate positions of the jth points with Doppler frequency larger than a threshold value; f. of_djIndicating the Doppler frequency of the j point target; j represents satisfying | f_dj|＞f_tThe number of all scattering points; | is an absolute value.

And step 3: the data set P is divided according to the positive and negative of the Doppler frequency₁Two data sets of person entering and exiting are obtained

In the formula, P_inAnd P_outPoint cloud datasets, J, representing person ingress and egress, respectively₁And J₂The number of scattering points of corresponding personnel. Using DBSCAN algorithm to respectively process two sets P_inAnd P_outClustering the midpoint cloud data, and aggregating scattering points of people entering and exiting into a position set of human body targets

Wherein the content of the first and second substances,

and

respectively representing the position sets of the clustered human body targets which enter and exit; x is the number of_pAnd y_pIs the coordinate position of the entering person; x is the number of_qAnd y_qThe coordinate position of the person who goes out;

and

the number of human targets after respective coagulation.

Clustering methods generally cluster closely spaced points, with one cluster corresponding to one target. Due to the human body target point cloud data set P₀The size is small, the shape is irregular, and the number of human body targets in the detection area is unknown, so that the DBSCAN algorithm is selected, the clustering algorithm takes the density of points as a clustering basis, the clustering shape is not biased, and the influence of noise is not easy to influence.

The radial velocity of the target relative to the radar is positive in the direction of approach to the radar and the direction of motion of the target determines the positive or negative of the doppler frequency, so that the data set P is divided according to the positive or negative of the doppler frequency₁Two data sets of person entering and exiting are obtained

In the formula, P_inAnd P_outPoint cloud datasets, J, representing person ingress and egress, respectively₁And J₂Respectively representing the number of scattering points of the person entering and exiting. Are respectively to P_inAnd P_outClustering two groups of point cloud data by using a DBSCAN algorithm, and condensing the number of scattering points of people entering and exiting into a position set of human body targets

In the formula (I), the compound is shown in the specification,

and

respectively representing the clustered position sets of the in-and-out human body targets,

and

the number of human targets after respective coagulation.

Fig. 5a and 5b are clustering results of the conventional DBSCAN method and the method when there is only one moving human target. Because the micro Doppler effect is generated by the human body micromotion, a false target appears after the traditional DBSCAN method is clustered, scattering points with smaller frequency brought by the human body micromotion are filtered, and the clustering result of the DBSCAN is correct.

And 4, step 4: clustering collection of personnel positions in detection area at intervals of delta t

And

counting the number of in and out human body targets, wherein t is t₂-t₁，t₁And t₂Two detection time points are respectively, and the width of the detection area on the y axis is y_r-y_l<v delta t and v are the normal walking speed of the human body, and the human body walks linearly along the y-axis direction in the detection area.

According to t₁And t₂Time-point location clustering set

It can be seen that there is only one time point t₁Or t₂If a human body target is detected on the same path in the detection area, counting the number of the detected people at the time point as the number of the entering people; if t₁And t₂If the human body target is not detected on the same path in the detection area at the time point, turning to step 41; if t₁And t₂And when the human body target is detected to exist on the same path in the detection area at the time point, the step 42 is switched to.

Step 41: by traversing t₁And t₂And correcting the personnel data of the rest time points in the same path in the detection area, counting the number of the entering personnel if more than half of the time points in the detection area can detect the existence of the human body target on the same walking path within the time interval delta t, and otherwise, no personnel enters.

Step 42: by traversing t₁And t₂And correcting the personnel data of the rest time points in the two time points on the same path in the detection area, counting that different personnel enter if no more than half of the time points in the detection area detect the existence of the human body target on the same walking path within the interval delta t time, and otherwise, counting that the same personnel enter.

Clustering collections at locations of personal departure

The number of persons counted out is combined with the above cluster

The steps for counting the number of the entering personnel are the same.

Fig. 6 is a schematic diagram of a detection area, a radar is located at an origin, a sector area surrounded by a black dotted line is a radar scanning coverage area, a shadow part is a human flow detection area, scattering points which are not located in the area are filtered, only the in-and-out conditions of human targets located in the area are counted to detect personnel in real time, and measurement errors caused by multipath propagation and other clutters can be effectively reduced. A two-dimensional coordinate system is established by taking a radar as a coordinate origin, and a pedestrian flow detection area is assumed as follows: x is the number of_i∈[x_l,x_r]，y_i∈[y_l,y_r]Wherein x is_l,x_r,y_l,y_rRespectively, the boundary conditions of the detection area. Clustering collection of personnel positions in detection area at intervals of delta t

And

counting the number of in and out human body targets, wherein t is t₂-t₁，t₁And t₂Two detection time points are provided. Assuming that the normal walking speed of the person is v, the person walks linearly in the detection area along the y-axis direction, and the normal walking speed is t₁And t₂At the time point the person can only be detected once in the detection area, so that the width of the detection area must satisfy y_r-y_l<v.DELTA.t. Taking the entering of a person as an example, FIG. 7 is a schematic diagram of a person with normal walking speed according to t₁And t₂Time-point location clustering set

It can be seen that there is only one time point t₁Or t₂And if the human body target is detected in the detection area, counting the number of the detected people at the time point as the number of the entering people.

In practical situations, the walking speeds of the persons are different, and the detection method can cause the problems of excessive detection and missed detection of the persons. The utility model is thus based on t₁And t₂Position set of the remaining time points in between

And

the human flow data is corrected, so that the misjudgment caused by different walking speeds of the human body is avoided. The following discussion is based on two possible scenarios, respectively.

(1) The walking speed of the person is higher than the normal speed

Taking the person entering as an example, when the walking speed of the person is higher than the normal speed, t may occur₁And t₂The time point does not detect the existence of the human body target in the detection area, and the method corresponds to two conditions: one is no person entering, the other is due to the fast walking speed, when the person passes through the detection area quickly in the time of delta t, which results in t₁And t₂At no time point, the presence of a human target is detected, as shown in fig. 8. To distinguish the two cases, by traversing t₁And t₂And correcting the personnel data in the detection area at the rest time points, counting the number of the entering personnel if more than half of the time points in the detection area can detect the existence of the human body target on the same walking path within the time interval delta t, and otherwise, no personnel enters.

(2) The walking speed of the person is lower than the normal speed

Taking the person entering as an example, when the walking speed of the person is lower than the normal speed, t may occur₁And t₂The time point detects that the human body target exists in the detection area, and the time point corresponds to two conditions: one is that different persons enter, and the other is that the same person still stays in the detection area after delta t time due to slower walking speed, thereby causing t₁And t₂The same human target was detected at all time points as shown in fig. 9. In order to distinguish the two cases, the utility model still traverses t₁And t₂And correcting the personnel data in the detection area at the rest time points in the two time points, counting that different personnel enter if no more than half of the time points in the detection area within the time interval delta t detect that the human body target on the same walking path exists, and otherwise, counting that the same personnel enter.

The technical effects of the present invention can be further illustrated by the following experimental test results. The utility model relates to a people flow monitoring method which is mainly applied to public places such as building buildings, scenic spot entrances and exits, and the like, the people flow monitoring scene in real life is simulated in a laboratory, the method is tested by building an experimental scene shown in figure 10, a testing device comprises a three-leg support 1, a computer 2, a data line 3, a millimeter wave radar 4 and a door 5, the computer 2 and the millimeter wave radar 4 are connected through the data line 3, a vertical rod 11 at the top of the three-leg support 1 is connected with a cross rod 12, the millimeter wave radar 4 is installed at the tail end of the cross rod 12, the position of the millimeter wave radar 4 is located at the same height as the door 5, the millimeter wave radar 4 is located obliquely above a detection area 6, and the lower part of the millimeter wave radar 4 is a radar sector scanning area range 7.

Assuming that the width of the door is about 1.5m and the height is about 2.2m, the millimeter wave radar 4 is fixed at the same height with the door 5 by the triangular support 1, namely, the millimeter wave radar 4 is obliquely arranged above the detection area 6, the inclination angle alpha is approximately equal to 45 degrees, the millimeter wave radar 4 is connected with the computer 2 through the USB data line 3, and the human flow entering and exiting results are displayed on the software of the upper computer.

The bandwidth of the millimeter wave radar is 4GHz, the number of encoding pulses (Chirp) in a unit frame is 128, the frame period is 40ms, and the distance resolution can reach 3.75cm theoretically. In practical tests, a Doppler threshold f is set_tSelecting a detection area as 125 Hz: x is the number of_l＝-1m，x_r＝1m，y_l＝1.2m，y_r1.6 m. Assuming that the normal walking speed v of the person is 1.1m/s, and the time interval Δ t for counting the pedestrian flow in the detection area is 0.4s, the requirement (y) is satisfied_r-y_l)cosα<vΔt。

Five scenarios were tested multiple times: (a) a scene is continuously entered and exited by one person, and only one person enters and exits a preset door each time; (b) two persons enter and exit the scene continuously in the same direction, and two persons enter and exit a preset door at the same time each time; (c) a double-person reverse continuous access scene, wherein every time two persons simultaneously access a preset door in opposite directions; (d) the double persons enter and exit scenes continuously at a distance of 1m, only one person enters and exits a preset door each time, and other persons enter and exit at a distance of 1 m; (e) the distance between two persons is 0.6m, the entrance and exit scenes are continuously followed, only one person enters and exits the preset door each time, and the distance between the person and the person next to the person is smaller and is about 0.6 m.

The statistics of the above five sets of data from the five scene tests are shown in table 1, and it is assumed that the number of actual entering persons and the number of actual leaving persons are respectively M₁And M₂System ofThe number of entering and leaving persons is N respectively₁And N₂The monitoring accuracy is calculated by

As can be seen from table 1, it is,

table 1 personnel access test results.

Aiming at the test of five personnel access scenes, the method can count personnel access with higher accuracy. The scenes a, c and d all obtain monitoring accuracy rate not lower than 95%, and when a single person continuously enters and exits (scene a) and two persons continuously follow the entrance and exit (scene d) at a distance of 1m, the monitoring accuracy rate is higher due to less multipath interference among the persons. In addition, the point cloud data is divided into an incoming data set and an outgoing data set according to the positive and negative Doppler frequencies for statistics, so that the accuracy similar to that of the single person in the process of entering and exiting the double persons in a reverse continuous mode (scene c) can be achieved. Due to the limitation of the azimuth and distance resolution of the radar, the monitoring accuracy of the two-person same-direction continuous access (scene b) and the two-person interval 0.6m continuous following access (scene e) is reduced, but the monitoring accuracy is still not lower than 90%.

Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the utility model, the scope of which is defined in the appended claims and their equivalents.

Claims (4)

1. The utility model provides a people flow dynamic monitoring device based on millimeter wave radar which characterized in that: including tripod support (1), computer (2), data line (3), millimeter wave radar (4) and door (5), through data line (3) connection computer (2) and millimeter wave radar (4), montant (11) at tripod support (1) top are connected with horizontal pole (12), horizontal pole (12) end installation millimeter wave radar (4), millimeter wave radar (4) position is in door (5) with the eminence, millimeter wave radar (4) are located detection area (6) top to one side, millimeter wave radar (4) below is the fan-shaped regional scope of sweeping of radar (7).

2. The dynamic human flow monitoring device based on the millimeter wave radar as claimed in claim 1, wherein: the inclination angle alpha of the millimeter wave radar (4) is 40-50 degrees.

3. The dynamic human flow monitoring device based on the millimeter wave radar as claimed in claim 2, wherein: the preferable inclination angle α of the millimeter wave radar (4) is 45 °.

4. The dynamic human flow monitoring device based on the millimeter wave radar as claimed in claim 1, wherein: the working frequency of the millimeter wave radar (4) is 60GHz, the bandwidth is 4GHz, the number of coded pulses in a unit frame is 128, and the frame period is 40 ms.

Images