CN104217245A - People stream trajectory tracking and area dwell time statistics method and system based on heterogeneous network - Google Patents

Abstract

The invention discloses a people stream trajectory tracking and area dwell time statistics method based on a heterogeneous network. The method comprises the following steps: deploying wireless access points and iBeaocn base stations, collecting and recording AP (Access Point) RSSI (Received Signal Strength Indicator) values collected by a plurality of sampling points and the coordinate values of the corresponding sampling points by a handheld mobile terminal, taking collected data as training data to determine a BP (Back Propagation) neural network model, integrating training results as an output value which is used as a function of a current mobile terminal coordinate value, obtaining a current position coordinate through an iBeacon and RSSI-distance relational expression, and obtaining a current position according to the distance; connecting the position coordinates to form a people stream trajectory tracking route graph; and meanwhile, recording the dwell time of a people stream at each point. The invention also provides a corresponding system. The invention combines the advantages of high calculation speed of wifi technology and high precision of iBeacon technology, and realizes people stream trajectory tracking and area dwell time statistics in a complex large building by combing with the multiuser support of a smartphone.

Description

Based on stream of people's trajectory track of heterogeneous network and the method and system of region residence time statistics

Technical field

The present invention relates to wireless communication technology field, particularly the method and system of a kind of stream of people's trajectory track based on heterogeneous network and region residence time statistics.

Background technology

Along with the development of economy, various indoor large building is more and more, especially shopping mall, how better Resources allocation in heavy construction, for businessman's marketing activity obtains more real user data support more efficiently, comprise user's real time position, intensity of passenger flow, residence time, enter shop to analyze, resident area, visiting number of times etc., become the focus instantly studied.

So far, stream of people's trajectory track method mainly adopts following several scheme:

The first, adopted video people stream statistical technique, equipped by specific video monitoring, and by moving object detection, identify humanbody moving object, the steps such as pursuit movement target realize stream of people's trajectory track.

The second, the stream of people's track based on virtual door is added up, and by monitoring moving target, adopts optical flow method calculating kinematical vector, finally arranges virtual door, added up by virtual door.

Above method has shortcoming, first all needs to dispose specific equipment support, and maximum shortcoming is that above equipment can not do initiatively mutual with the stream of people, can only do one-side stream of people's track statistics.

Summary of the invention

In order to overcome the shortcoming of above-mentioned prior art, the object of the present invention is to provide the method and system of a kind of stream of people's trajectory track based on heterogeneous network and region residence time statistics, utilize main flow smart mobile phone on market directly can receive the feature of wifi signal and iBeacon signal, realize a kind of stream of people's trajectory track method that precision is higher, terminal adaptation is wider.

To achieve these goals, the technical solution used in the present invention is:

Based on stream of people's trajectory track of heterogeneous network and a method for region residence time statistics, comprise the following steps:

S1. dispose wireless access points (AP) and iBeaocn base station, and record deployed position;

S2. hand-held mobile terminal gathers and records the RSSI value of the AP that several sampled points receive and the coordinate figure of corresponding sampled point in deployment region;

S3. utilize the data gathered in S2 to determine BP neural network model as training data, input vector is the RSSI value of the diverse location AP that sampled point receives, and output vector is the coordinate figure of sampled point; Dispose N number of AP, then receive the RSSI value of N number of AP, input vector dimension is N, and output vector dimension is 2;

S4. final for S3 training result is integrated into a function, the parameter of this function input is the AP signal intensity that current mobile terminal receives, and output valve is current mobile terminal coordinate figure;

S5. the tuning on-line stage, user's hand-held mobile terminal, just starts the AP signal and the iBeacon signal that scan region after hand-held mobile terminal remains static and reaches certain hour;

S6. first by the RSSI-distance relation formula of iBeacon

L＝0.89976*{(rssi/txPower)^7.7095}+0.111

Obtain the distance L of the nearest iBeacon base station of current location, when this distance is less than predeterminable range, get the position that current location result is this iBeacon base station; Utilize the RSSI signal value of the wifi be currently received as parameter when this distance is greater than predeterminable range, the function trained in input S4 obtains current location; Wherein rssi is the iBeacon signal intensity that mobile phone receives, and txPower is the signal correction factor of iBeacon Base Transmitter;

S7. the position coordinates that S6 obtains is connected to form stream of people's track following road wiring diagram, is recorded in the time of each stop simultaneously.

In described S1, iBeacon base station is at least deployed in the position of turnover and multi-obstacle avoidance.

In described S3, in the BP neural network model set up, by node in hidden layer location 30, form the BP neural network of a N-30-2 structure, wherein the neuron of hidden layer uses tansig tangent S type activation function, and output layer uses the linear activation function of purelin, frequency of training 1000 times, traingdm function is adopted to carry out function training, training error performance index 1e-5.

In described S4, use the input and output structure of the determined BP neural network of S3, interstitial content, training time and error are trained, the weights and bias obtaining each node of hidden layer and output layer is the most at last preserved as BP neural network parameter record, obtains the BP neural network model trained.

In described S5, certain hour is set as 10 seconds.

Described hand-held mobile terminal is smart mobile phone, and operating system can be Android4.3 system or IOS system.

Present invention also offers the system of the method for the stream of people's trajectory track based on heterogeneous network described in realization and region residence time statistics, comprising:

Location server, storage AP and iBeacon base station position information, stores route track and the residence time; Complete BP neural metwork training, accept mobile terminal locations request;

Hand-held mobile terminal, completes sampled point signal scanning and is uploaded to location server neural network training; Complete live signal scanning, up-delivering signal is to location server, and waiting for server position data returns; Real-time update stream of people track completes interface display and the display of the region residence time;

IBeacon base station, provides signal for hand-held mobile terminal collection;

AP, provides signal for hand-held mobile terminal collection.

Compared with prior art, the method of a kind of stream of people's trajectory track based on heterogeneous network of the present invention and region residence time statistics, fully merge the advantage that wifi technique computes speed is fast and iBeacon technology acuracy is high, the multi-user of combined with intelligent mobile phone supports, fast achieves the stream of people's trajectory track in complex large-sized building and region residence time statistics exactly.

Accompanying drawing explanation

Fig. 1 is overall flow figure of the present invention.

Fig. 2 is wireless access points of the present invention and iBeaocn base station indoor deployment planimetric map.

Fig. 3 is the network number of plies selection figure of the BP neural network model that the present invention sets up.

Fig. 4 is the network structure of the BP neural network model that the present invention sets up.

Fig. 5 is BP neural network model of the present invention training schematic diagram.

Fig. 6 is the locator data table of the embodiment of the present invention.

Fig. 7 is result of calculation and the actual result comparison diagram of the embodiment of the present invention.

Embodiment

Embodiments of the present invention are described in detail below in conjunction with drawings and Examples.

As shown in Figure 1, based on stream of people's trajectory track of heterogeneous network and the method for region residence time statistics, comprise the following steps:

Data acquisition phase:

S1 selects the planimetric map of test zone as shown in Figure 2, and the area of room area to be measured is 10m*10m, disposes the router that 3 bench-types number are TP-Link respectively use as AP in room.Use based on the millet 2s of Android4.3 as mobile terminal.Dispose iBeacon base station on doorway, room and corridor place, AP and iBeacon base station deployment position coordinates is stored in location server.

S2 hand-held mobile terminal is slow mobile collection data equably in central area, and often mobile 1 step gathers a secondary data after pausing 10 seconds, gather the signal intensity of 3 wireless apss received for (RSSI1, RSSI2, RSSI3), the position coordinates (X, Y) of region is gathered.

Neural network structure is determined:

1. S3 sets up neural network model

Kolmogorov theorem is verified, and any one continuous function can be realized by three layers of BP network.Because receive the RSSI value of 3 AP, so the dimension of input vector is 3, again because export as coordinate, so the dimension of output vector is 2.Through repeatedly testing, node in hidden layer is decided to be 30, experimental data as shown in Figure 3.Result in formation of the BP nerve net of 3-30-2 structure, as shown in Figure 4.

2. the selection of learning algorithm

If containing the network model being total to L layer and n node, every layer unit accepts the output of front one deck and exports to each unit of lower floor, and each node adopts continuously differentiable Sigmoid type function.If given N number of sample (x _k, y _k) (k=1,2 ..., N), wherein the output information of any node i is o _i, be x for wherein certain node inputs information _k, then the output information of place network is y _k, the output of node i is o _ik, now using the jth of a l layer unit as research object, when inputting kth sample, the input information of node j is

${\mathrm{net}}_{\mathrm{ij}}^l=\underset{j}{\mathrm{\Σ}}{w}_{\mathrm{ij}}^l{o}_{\mathrm{jk}}^{l-1}---\left(1\right)$

$o_{\mathrm{jk}}^l=f\left({\mathrm{net}}_{\mathrm{jk}}^l\right)---\left(2\right)$

representing l-1 layer, representing when inputting kth sample, the output information of a jth cell node.

The error function wherein adopted is

$E_k=\frac{1}{2}\underset{l}{\mathrm{\Σ}}{(y_{\mathrm{lk}}-{\stackrel{\‾}{y}}_{\mathrm{lk}})}^2---\left(3\right)$

for the actual output information of unit j.So wherein total error is

$E=\frac{1}{2N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{E}_k---\left(4\right)$

Definition ${\mathrm{\δ}}_{\mathrm{jk}}^l=\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}$

So $\frac{{\∂E}_k}{{\∂w}_{\mathrm{ij}}^l}=\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}\frac{{\∂\mathrm{net}}_{\mathrm{jk}}^l}{{\∂w}_{\mathrm{ij}}^l}=\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}{o}_{\mathrm{jk}}^{l-1}={\mathrm{\δ}}_{\mathrm{jk}}^l{o}_{\mathrm{jk}}^{l-1}---\left(5\right)$

There are two kinds of situations:

(1) if node j is output unit, then

${\mathrm{\δ}}_{\mathrm{jk}}^l=\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}=\frac{{\∂E}_k}{\∂{\stackrel{\‾}{y}}_{\mathrm{jk}}}\frac{\∂{\stackrel{\‾}{y}}_{\mathrm{jk}}}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}=-(y_k-{\stackrel{\‾}{y}}_k)f^{\′}\left({\mathrm{net}}_{\mathrm{jk}}^l\right)---\left(6\right)$

(2) if node j is not output unit, then

${\mathrm{\δ}}_{\mathrm{jk}}^l=\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}=\frac{{\∂E}_k}{\∂{\stackrel{\‾}{y}}_{\mathrm{jk}}}\frac{\∂o_{\mathrm{jk}}^l}{{\∂\mathrm{net}}_{\mathrm{jk}}^l}=\frac{{\∂E}_k}{{\∂o}_{\mathrm{jk}}^l}{f}^{\′}\left({\mathrm{net}}_{\mathrm{jk}}^l\right)---\left(7\right)$

Wherein the input information delivering to lower one deck (l+1) layer, so calculate to return from (l+1) layer.

When (l+1) layer m unit

$\frac{{\∂E}_k}{{\∂o}_{\mathrm{jk}}^l}=\underset{m}{\mathrm{\Σ}}\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{mk}}^{l=1}}\frac{{\∂\mathrm{net}}_{\mathrm{mk}}^{l+1}}{{\∂o}_{\mathrm{jk}}^l}=\underset{m}{\mathrm{\Σ}}\frac{{\∂E}_k}{{\∂\mathrm{net}}_{\mathrm{mk}}^{l+1}}{w}_{\mathrm{mj}}^{l+1}=\underset{m}{\mathrm{\Σ}}{\mathrm{\δ}}_{\mathrm{mk}}^{l+1}{w}_{\mathrm{mj}}^{l+1}---\left(8\right)$

Formula (38) is substituted in formula (37), then

${\mathrm{\δ}}_{\mathrm{jk}}^l=\underset{m}{\mathrm{\Σ}}{\∂}_{\mathrm{mk}}^{l+1}{w}_{\mathrm{mj}}^{l+1}\left({\mathrm{net}}_{\mathrm{jk}}^l\right)---\left(9\right)$

In sum, have

$\left\{\begin{array}{c}{\mathrm{\δ}}_{\mathrm{jk}}^l=\underset{m}{\mathrm{\Σ}}{\mathrm{\δ}}_{\mathrm{mk}}^{l+1}{w}_{\mathrm{mj}}^{l+1}{f}^{\′}\left({\mathrm{net}}_{\mathrm{jk}}^l\right)\\ \frac{{\∂E}_k}{{\∂w}_{\mathrm{ij}}^l}={\mathrm{\δ}}_{\mathrm{jk}}^l{o}_{\mathrm{jk}}^{l-1}\end{array}\right.---\left(10\right)$

Therefore, the step setting up BP neural network can be summarized as follows:

(1) selected weight coefficient initial value;

(2) following process is repeated, until error criterion meets accuracy requirement, that is:

$E=\frac{1}{2N}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{E}_k<\mathrm{\&epsiv;},$ ε: precision

1. to k=1 to N

Forward process computation: calculate every layer unit with k=2 ..., N.

Reverse procedure calculates: to each layer (l=L-1 is to 2), to every layer of each unit, calculates

2. weights are revised

$w_{\mathrm{ij}}=w_{\mathrm{ij}}-\mathrm{\&mu;}\frac{\&PartialD;E}{\&PartialD;w_{\mathrm{ij}}}\mathrm{\&mu;}>0---\left(11\right)$

(3) terminate.

S4. utilize the BP neural algorithm that Matlab Neural Network Toolbox provides, realize above-mentioned steps, performing step as shown in Figure 5.Wherein the neuron of hidden layer uses tansig tangent S type activation function, and output layer uses the linear activation function of purelin, frequency of training 1000 times, adopts traingdm function to carry out function training, training error performance index 1e-5.The weights and bias obtaining each node of hidden layer and output layer is the most at last preserved as BP neural network parameter record, obtains the BP neural network model trained.

S5. the tuning on-line stage, user's hand-held mobile terminal (smart mobile phone), obtain mobile phone motion state by smart mobile phone API, after 10 seconds being reached when smart mobile phone remains static, start the AP signal (i.e. WIFI signal) and the iBeacon signal that scan region.

S6. first by testing the RSSI-distance relation formula of the iBeacon of gained in advance

L＝0.89976*{(rssi/txPower)^7.7095}+0.111

Wherein rssi is the iBeacon signal intensity that mobile phone receives, txPower is the signal correction factor that iBeacon launches, obtain current location distance to the distance of nearest iBeacon base station, when this distance is less than predeterminable range, get the position that current positioning result is this iBeacon base station.Utilize the RSSI signal value of the wifi be currently received as parameter when this distance is greater than predeterminable range, the BP neural network model trained in input S4 obtains current position.Utilize the network model after training to carry out 20 tests, result as shown in Figure 6.Wherein X1, Y1 represent the predicted value of model, and X2, Y2 represent actual value.Wherein coordinate with the interior architecture lower left corner for true origin, unit centimetre.Training result and Measured Coordinates error contrast as shown in Figure 7.

S7. the last position coordinates obtained by S6 is connected to form stream of people's track following road wiring diagram, is recorded in the time of each stop simultaneously.

Above embodiment is only for illustration of the present invention; and be not limitation of the present invention; the those of ordinary skill of relevant technical field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all equivalent technical schemes also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (8)

1., based on stream of people's trajectory track of heterogeneous network and a method for region residence time statistics, it is characterized in that, comprise the following steps:

S1. dispose wireless access points (AP) and iBeaocn base station, and record deployed position;

S2. hand-held mobile terminal gathers and records the RSSI value of the AP that several sampled points receive and the coordinate figure of corresponding sampled point in deployment region;

S3. utilize the data gathered in S2 to determine BP neural network model as training data, input vector is the RSSI value of the diverse location AP that sampled point receives, and output vector is the coordinate figure of sampled point; Dispose N number of AP, then receive the RSSI value of N number of AP, input vector dimension is N, and output vector dimension is 2;

S4. final for S3 training result is integrated into a function, the parameter of this function input is the AP signal intensity that current mobile terminal receives, and output valve is current mobile terminal coordinate figure;

S5. the tuning on-line stage, user's hand-held mobile terminal, just starts the AP signal and the iBeacon signal that scan region after hand-held mobile terminal remains static and reaches certain hour;

S6. first by the RSSI-distance relation formula of iBeacon

L＝0.89976*{(rssi/txPower)^7.7095}+0.111

Obtain the distance L of the nearest iBeacon base station of current location, when this distance is less than predeterminable range, get the position that current location result is this iBeacon base station; Utilize the RSSI signal value of the wifi be currently received as parameter when this distance is greater than predeterminable range, the function trained in input S4 obtains current location; Wherein rssi is the iBeacon signal intensity that mobile phone receives, and txPower is the signal correction factor of iBeacon Base Transmitter;

S7. the position coordinates that S6 obtains is connected to form stream of people's track following road wiring diagram, is recorded in the time of each stop simultaneously.

2. the method for the stream of people's trajectory track based on heterogeneous network according to claim 1 and region residence time statistics, is characterized in that, in described S1, iBeacon base station is at least deployed in the position of turnover and multi-obstacle avoidance.

3. the method for the stream of people's trajectory track based on heterogeneous network according to claim 1 and region residence time statistics, it is characterized in that, in described S3, in the BP neural network model set up, by node in hidden layer location 30, form the BP neural network of a N-30-2 structure, wherein the neuron of hidden layer uses tansig tangent S type activation function, output layer uses the linear activation function of purelin, frequency of training 1000 times, traingdm function is adopted to carry out function training, training error performance index 1e-5.

4. the method for the stream of people's trajectory track based on heterogeneous network according to claim 1 and region residence time statistics, it is characterized in that, in described S4, use the input and output structure of the determined BP neural network of S3, interstitial content, training time and error are trained, and the weights and bias obtaining each node of hidden layer and output layer is the most at last preserved as BP neural network parameter record, obtains the BP neural network model trained.

5. the method for the stream of people's trajectory track based on heterogeneous network according to claim 1 and region residence time statistics, it is characterized in that, in described S5, certain hour is set as 10 seconds.

6. the method for the stream of people's trajectory track based on heterogeneous network according to claim 1 and region residence time statistics, it is characterized in that, described hand-held mobile terminal is smart mobile phone.

7. the method for the stream of people's trajectory track based on heterogeneous network according to claim 6 and region residence time statistics, it is characterized in that, described operation system of smart phone is Android4.3 system or IOS system.

8. realize the system of the method for the stream of people's trajectory track based on heterogeneous network according to claim 1 and region residence time statistics, it is characterized in that, comprising:

Location server, storage AP and iBeacon base station position information, stores route track and the residence time; Complete BP neural metwork training, accept mobile terminal locations request;

Hand-held mobile terminal, completes sampled point signal scanning and is uploaded to location server neural network training; Complete live signal scanning, up-delivering signal is to location server, and waiting for server position data returns; Real-time update stream of people track completes interface display and the display of the region residence time;

IBeacon base station, provides signal for hand-held mobile terminal collection;

AP, provides signal for hand-held mobile terminal collection.