CN105632187A - Low-power consumption side parking detection method based on geomagnetic sensor network - Google Patents

Low-power consumption side parking detection method based on geomagnetic sensor network Download PDF

Info

Publication number
CN105632187A
CN105632187A CN201511028573.6A CN201511028573A CN105632187A CN 105632187 A CN105632187 A CN 105632187A CN 201511028573 A CN201511028573 A CN 201511028573A CN 105632187 A CN105632187 A CN 105632187A
Authority
CN
China
Prior art keywords
signal
disturbance
threshold value
parking
correlation
Prior art date
Application number
CN201511028573.6A
Other languages
Chinese (zh)
Other versions
CN105632187B (en
Inventor
朱红梅
于峰崎
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Priority to CN201511028573.6A priority Critical patent/CN105632187B/en
Publication of CN105632187A publication Critical patent/CN105632187A/en
Application granted granted Critical
Publication of CN105632187B publication Critical patent/CN105632187B/en

Links

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/01Detecting movement of traffic to be counted or controlled
    • G08G1/042Detecting movement of traffic to be counted or controlled using inductive or magnetic detectors
    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/14Traffic control systems for road vehicles indicating individual free spaces in parking areas
    • G08G1/145Traffic control systems for road vehicles indicating individual free spaces in parking areas where the indication depends on the parking areas

Abstract

The invention provides a low-power consumption side parking detection method based on a geomagnetic sensor network. The detection method comprises following steps: a geomagnetic sensor node acquires geomagnetic disturbance signals by vehicles in a parking lot; preprocessing on the acquired signals is carried out and parking detection algorithms are selected according to signal intensity after the processing; if the signal intensity is lower than a choice threshold, a state machine detection algorithm with low complexity is selected; judgment for parking place states is obtained according to skipping results of a state machine, or else, a relatively complex cross-correlation detection method is selected; and states of a parking place are judged according to the correlation between a reference model obtained via K-means clustering and signals. According to the invention, by combining the state machine detection with the quite low complexity and the cross-correlation detection with the quite high precision, while detection precision is ensured, power consumption is considered; and service lifetime of the node is extended, thereby laying a foundation for intelligent parking management.

Description

Low-power consumption curb parking detection method based on geomagnetic sensor network

[technical field]

The present invention relates to vehicle testing techniques, particularly relate to a kind of low-power consumption curb parking detection method based on geomagnetic sensor network.

[background technology]

Curb parking detection is the pith realizing intelligent transportation, and detection method currently mainly has: the means such as ultrasonic, ground induction coil, video and geomagnetic sensor. Wherein, although ultrasound detection accuracy is higher, but be not suitable for parking lay-by install and safeguard; Ground induction coil detection stable performance, but installation and maintenance need cutting road surface, hinder normal traffic; Video Detection informative and directly perceived, however it is necessary that bigger storage capacity and computing capability, its detection performance is highly prone to the impact of illumination condition and sleety weather. Comparatively speaking, geomagnetic sensor has low-power consumption, low cost, high sensitivity, small size, is easily integrated and installs and the advantage such as easy to maintenance. Region geomagnetic sensor being positioned over vehicle process just can collect the vehicle disturbing signal to earth magnetism, and disturbing signal can be used to after processing detect the transport information such as the existence of vehicle, vehicle, speed.

Ground magnetic field intensity is thought uniform and stable in several kilometer range, earth's magnetic field will produce bigger disturbance when containing vehicle ferromagnetic in a large number through out-of-date, and the change of external magnetic field causes that geomagnetic sensor output valve changes. The signal of output can be used to detect the existence of vehicle after treatment, lay two geomagnetic sensor nodes and can estimate Vehicle Speed, the waveform of output signal has much relations with vehicle structure itself and the iron content quantity of magnetism, therefore can realize vehicle cab recognition according to this feature. Current existing algorithm is broadly divided into two kinds: based on the state machine detection of threshold value and the cross-correlation test method based on similarity.

Based on the state machine detection algorithm of threshold value, main thought is that vehicle detection process is divided into different states, and after meeting corresponding condition, state redirects, thus drawing parking space state. Wherein, state machine redirects condition and relies primarily on the comparison of its monitor signal and setting threshold value. Its advantage is that method is simple, and low in energy consumption, amount of calculation is little, it is simple to realize on sensor node, but its shortcoming is threshold value is empirical value, it is difficult to ensure its objectivity, when signal disturbance intensity is less, it is easy to erroneous judgement occurs or fails to judge. The background signal of curb parking is complicated, there is the interference signal of road travel, and signal to noise ratio is relatively low, and this simple state machine approach accuracy of detection in curb parking detects is difficult to ensure that.

Cross-correlation test method based on similarity can effectively filter noise signal incoherent with reference signal so that useful signal is enhanced, thus improving Detection accuracy. The accuracy of detection of cross-correlation test is affected bigger by selected reference signal, the reference signal that existing cross-correlation test adopts has geomagnetic field variation model and Gauss curve fitting curve etc., these researchs are primarily directed to moving vehicle, it is not suitable for parking lot vehicle detection, therefore the present invention is found and the actual maximally related reference signal of vehicle disturbing signal by K-means clustering algorithm, simultaneously balanced power consumption and performance, combine state machine detection algorithm with cross-correlation test algorithm and obtain the curb parking vehicle checking method of a kind of low-power consumption.

[summary of the invention]

In view of this, it is necessary to the parking lay-by vehicle checking method of a kind of low-power consumption is provided. Described geomagnetic sensor node gathers the parking lot vehicle disturbing signal to earth magnetism, and the signal gathered is carried out pretreatment, parking detection method is selected to according to the signal intensity after processing, if signal intensity is lower than choice threshold value (TH_OP), select the state machine detection algorithm that complexity is relatively low, the result that redirects according to state machine obtains the judgement of parking space state, otherwise selecting the cross-correlation test method of relative complex, the dependency of the reference model obtained according to signal and K-means cluster judges the state of parking stall.

In a preferred embodiment, the pre-treatment step gathering signal is comprised the steps: that described collection signal carries out gaussian filtering process; Described filtered signal carries out modulus calculating. Wherein gaussian filtering adopts following formula:

F (x)=exp (-xs 2/(2*��2))

Wherein, �� determines the width of Gaussian function, xsBeing the signal after described x-axis smothing filtering, y-axis adopts the method identical with x-axis to carry out gaussian filtering with z-axis.

Calculate modulus value and adopt following formula:

S D = ( B X - B X 0 ) 2 + ( B Y - B Y 0 ) 2 + ( B Z - B Z 0 ) 2

Wherein, BX��BYAnd BZRepresent the signal after the gaussian filtering of x, y and z axes, B respectivelyX0��BY0And BZ0Represent x, y and z-axis baseline value respectively.

In a preferred embodiment, described curb parking vehicle detecting algorithm of choosing comprises the steps: to compare the modulus value of signal after pretreatment with choice threshold value (TH_OP), if it exceeds choice threshold value (TH_OP) then selects the state machine detection method that complexity is relatively low, otherwise select the cross-correlation test algorithm that degree of accuracy is higher.

In a preferred embodiment, described low complex degree state machine detection method comprises the following steps:

(1) initializing all parameters and baseline, arranging sign of flag _ Initial after having initialized is 1, and jumps to step (2);

(2) parking space state is idle (PS=0), and carries out baseline renewal in this step, and baseline more new formula is as follows:

B i 0 ( t ) = { B i 0 ( t - 1 ) · ( 1 - α ) + B i ( t ) · α , P S ( t - 1 ) = 0 B i 0 ( t - 1 ) , P S ( t - 1 ) = 1 , i ∈ { x , y , z }

In formula, Bi0(t-1) it is the i axle baseline value of previous moment, i �� { x, y, z}; BiT () is the signal after the gaussian filtering of current i axle; �� is forgetting factor, and its scope is [0-1]; The parking space state that PS (t-1) is previous moment;

When detection signal has a disturbance, and disturbing signal exceedes setting and sails threshold value (TH_AR) into and then arrange and sail disturbance mark (Fluctuation_Arrival=1) into, jumps to step (3);

(3) record exceed sail threshold value (TH_AR) into sail disturbance number of times (Cnt_Arrival) into, then arrange in the event of disturbance and sail disturbance into and be masked as 0 (Fluctuation_Arrival=0) and jump to step (4) lower than sailing threshold value (TH_AR) into, otherwise exceed when frequency threshold value (N_Arrival) is sailed in setting into jump to step (5) when vehicle sails disturbance number of times into;

(4) disturbance number of times Cnt_Arrival is sailed in clearing into, and record and sail disturbance number of times (Cnt_noArrival) into without vehicle, if disturbance is again above sailing threshold value (TH_AR) into, it is once again set up sailing disturbance into and is masked as 1 (Fluctuation_Arrival=1), jump to step (3); Otherwise, exceed setting without sailing frequency threshold value (N_noArrival) into when sailing disturbance number of times into without vehicle, then jump to step (2);

(5) parking stall occupied (mark PS=1 is set), if disturbance exceedes vehicle and rolls threshold value (TH_DP) away from, then jump procedure (6);

(6) vehicle is set and rolls disturbance mark (Fluctuation_departure=1) away from, and record rolls disturbance number of times (Cnt_Departure) away from, if disturbance is lower than rolling threshold value (TH_DP) away from, setting is rolled disturbance away from and is masked as 0 (Fluctuation_departure=0), and jumps to step (5); Otherwise, exceed setting and roll frequency threshold value (N_Departure) then jump procedure (2) away from when rolling disturbance number of times (Cnt_Departure) away from;

In a preferred embodiment, described high accuracy cross-correlation test method comprises the following steps:

(1) reference signal that after described pretreatment, signal and K-means cluster obtain carries out cross-correlation calculation;

C C ( τ ) = Σ t = 0 L - t - 1 ( S D ( t ) - S D ‾ ) ( R ( t + τ ) - R ‾ ) Σ t = 0 L - t - 1 ( S D ( t ) - S D ‾ ) 2 · Σ t = τ t - 1 ( R ( t ) - R ‾ ) 2

Wherein, SD (t) is modulus value after pretreatment,For modulus value average, R (t) is reference signal,For reference signal average, CC (��) represents the cross-correlation coefficient of two signals, and in [-1,1] scope, when two signal cross-correlation degree are higher, more levels off to 1, when two signals complete uncorrelated time, more level off to 0.

(2) cross-correlation calculation result CC (��) compares with setting cross-correlation threshold value (TH_CO), and parking space state more new formula is as follows

E v e n t ( t ) = 1 C C ( t ) &GreaterEqual; T H _ C O 0 C C ( t ) < T H _ C O

Wherein, Event (t) represents whether state changes, and represents change, represent that state is constant when being 0 when being 1; TH_CO is for setting cross-correlation threshold value; CC (��) is cross-correlation calculation result.

In a preferred embodiment, described judgement parking stall last state adopts following formula:

P S ( t ) = 0 , P S ( t - 1 ) = 0 , E v e n t ( t ) = 0 ; 1 , P S ( t - 1 ) = 0 , E v e n t ( t ) = 1 ; 1 , P S ( t - 1 ) = 1 , E v e n t ( t ) = 0 ; 0 , P S ( t - 1 ) = 1 , E v e n t ( t ) = 1 ;

Wherein, PS (t)=0 represents that parking stall is idle, and PS (t)=1 represents that parking stall is occupied.

In a preferred embodiment, described reference signal acquisition methods comprises the following steps: described preprocessed signal carries out intercepting process; Described intercept signal is interpolated process; After described interpolation, signal is normalized; After described normalized, signal carries out K-means cluster, and the cluster centre obtained is reference signal.

After described pretreatment, signal carries out characteristic curve intercepting, removes static part, adopts following formula;

SD_A=SD (t), { (t1-td)��t��(t2+td)}

SD_D=SD (t), { (t3-td)��t��(t4+td)}

Wherein, SD (t) is signal modulus value, t after pretreatment1It is the time of vehicle arrival, t2It it is the time point that after stopping, signal is changed into steady statue from labile state; Equally, t3It is time of leaving of vehicle, t4It is that car leaves rear signal and is changed into the time point of steady statue, t from labile statedIt it is time delay parameter.

The described characteristic curve to intercepting is interpolated process, makes characteristic curve length identical, and interpolation method comprises the following steps:

(1) characteristic curve length M is calculated, and matched curve y=f (x);

(2) on the interval of length M, N number of point (x it is evenly distributed1,x2,��,xN, N is the regular length after interpolation);

(3) calculate, according to matched curve, the functional value (y that N point is corresponding1,y2,��,yN);

(4) Y=[y1,y2,��,yN] it is characteristic curve after interpolation;

Described characteristic curve is normalized, and it is 1 that signal is normalized to maximum, and minima is in the scope of-1, adopts following formula:

Y = ( Y m a x - Y m i n ) * ( x - x m i n ) x max - x min + Y m i n

Wherein, YmaxAnd YminRepresenting maximum 1 and minima-1 of normalization scope respectively, X is primary signal, xmaxAnd xminRepresent the maximum in primary signal and minima respectively.

In a preferred embodiment, described to extract characteristic signal carry out k-means cluster comprise the steps:

(1) k cluster center of mass point is chosen, for ��1,��2,��,��k��Rn��

(2) calculate each sample class and adopt equation below:

C ( i ) : = &alpha; r g min j | | x ( i ) - &mu; j | | 2

(3) barycenter recalculating each class j adopts equation below:

&mu; j : = &Sigma; i = 1 m 1 { C ( i ) = j } x ( i ) &Sigma; i = 1 m 1 { C ( i ) = j }

Repeat (2) and (3) until restraining;

Wherein, K is known cluster numbers, C(i)That class that representative sample I and k apoplexy due to endogenous wind is closest, C(i)Value be in 1 to k, barycenter ��jRepresent the central point of same class sample.

In a preferred embodiment, described reference signal adopts cluster barycenter.

[accompanying drawing explanation]

Fig. 1 is the curb parking vehicle checking method flow chart of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 2 is the state machine overhaul flow chart of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 3 is the cross-correlation test flow chart of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 4 is the reference signal detecting flow chart of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 5 is the magnetic signal schematic diagram primitively of the collection of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 6 is the characteristic signal sectional drawing schematic diagram of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 7 is the AMR continuous collecting Geomagnetic signal schematic diagram of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

Fig. 8 is the testing result exemplary plot of the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention;

The system that Fig. 9 is the parking lay-by vehicle checking method of a kind of low-power consumption of the present invention lays schematic diagram.

[detailed description of the invention]

Illustrate below in conjunction with concrete embodiment and accompanying drawing.

As it is shown in figure 1, a kind of parking lay-by method for detecting parking stalls, comprise the following steps:

S110, gathers signal and carries out gaussian filtering, and gaussian filtering adopts equation below:

F (x)=exp (-xs 2/(2*��2))

Wherein, �� determines the width of Gaussian function, xsBeing the x-axis component of described locality magnetic signal, y-axis adopts the method identical with x-axis to carry out gaussian filtering with z-axis.

S120, after gaussian filtering, signal carries out modulus calculating, adopts equation below:

S D = ( B X - B X 0 ) 2 + ( B Y - B Y 0 ) 2 + ( B Z - B Z 0 ) 2

Wherein, BX��BYAnd BZRepresent the signal after the gaussian filtering of x, y and z axes, B respectivelyX0��BY0And BZ0Represent x, y and z-axis baseline value respectively.

S130, modulus value compares with choice threshold value (TH_OP), if it exceeds choice threshold value (TH_OP) then selects the state machine detection method that complexity is relatively low, otherwise selects the cross-correlation test algorithm that degree of accuracy is higher.

S140, low complex degree state machine detects. This step comprises the steps, as shown in Figure 2:

S141, initializes all parameters and baseline, and arranging sign of flag _ Initial after having initialized is 1, and jumps to step S142;

S142, parking space state is idle (PS=0), and carries out baseline renewal in this step, and baseline more new formula is as follows:

B i 0 ( t ) = { B i 0 ( t - 1 ) &CenterDot; ( 1 - &alpha; ) + B i ( t ) &CenterDot; &alpha; , P S ( t - 1 ) = 0 B i 0 ( t - 1 ) , P S ( t - 1 ) = 1 , i &Element; { x , y , z }

In formula, Bi0(t-1) it is the i axle baseline value of previous moment, i �� { x, y, z}; BiT () is the signal after the gaussian filtering of current i axle; �� is forgetting factor, and its scope is [0-1]; The parking space state that PS (t-1) is previous moment;

When detection signal has a disturbance, and disturbing signal exceedes setting and sails threshold value (TH_AR) into and then arrange disturbance mark (Fluctuation_Arrival=1), jumps to step S143;

S143: record exceed sail threshold value (TH_AR) into sail disturbance number of times (Cnt_Arrival) into, in the event of disturbance lower than sailing threshold value (TH_AR) into, then it is once again set up sailing disturbance into be masked as 0 (Fluctuation_Arrival=0) and jump to step S144, otherwise exceedes when frequency threshold value (N_Arrival) is sailed in setting into jump to step S145 when vehicle sails disturbance number of times (Cnt_Arrival) into;

S144: reset and sail disturbance number of times Cnt_Arrival into, and record and sail disturbance number of times (Cnt_noArrival) into without vehicle, if disturbance is again above sailing threshold value (TH_AR) into, setting is sailed disturbance into and is masked as 1 (Fluctuation_Arrival=1), jumps to step S143; Otherwise, exceed setting without sailing frequency threshold value (N_noArrival) into when sailing disturbance number of times (Cnt_noArrival) into without vehicle, then jump to step S142;

S145: parking stall occupied (arranging mark PS=1), if disturbance exceedes vehicle and rolls threshold value (TH_DP), then jump procedure S146 away from;

S146: vehicle is set and rolls disturbance mark (Fluctuation_departure=1) away from, and record rolls disturbance number of times (Cnt_Departure) away from, if disturbance rolls threshold value (TH_DP) away from lower than setting, then setting is rolled disturbance away from and is masked as 0 (Fluctuation_departure=0), jumps to step S145; Otherwise, exceed setting and roll frequency threshold value (N_Departure) then jump procedure S142 away from when rolling disturbance number of times away from;

S150: high accuracy cross-correlation test. This step comprises the following steps. As shown in Figure 3.

S151: obtain reference signal, comprise the following steps. As shown in Figure 4.

S1511: carry out characteristic curve intercepting, the magnetic signal primitively that node gathers, as it is shown in figure 5, its characteristic of correspondence curve intercepts as shown in Figure 6, is removed static part, is adopted following formula;

SD_A=SD (t), { (t1-td)��t��(t2+td)}

SD_D=SD (t), { (t3-td)��t��(t4+td)}

Wherein, SD (t) is signal modulus value, t after pretreatment1It is the time of vehicle arrival, t2It it is the time point that after stopping, signal is changed into steady statue from labile state; Equally, t3It is time of leaving of vehicle, t4It is that car leaves rear signal and is changed into the time point of steady statue, t from labile statedIt it is time delay parameter.

S1512: the described characteristic curve to intercepting is interpolated process, makes characteristic curve length identical, and interpolation method comprises the following steps.

(1) characteristic curve length M is calculated, and matched curve y=f (x);

(2) on the interval of length M, N number of point (x it is evenly distributed1,x2,��,xN, N is the regular length after interpolation);

(3) calculate, according to matched curve, the functional value (y that N point is corresponding1,y2,��,yN);

(4) Y=[y1,y2,��,yN] it is characteristic curve after interpolation;

S1513: described characteristic curve is normalized, it is 1 that signal is normalized to maximum, and minima is in the scope of-1, adopts following formula:

Y = ( Y m a x - Y m i n ) * ( x - x m i n ) x max - x min + Y m i n

Wherein, YmaxAnd YminRepresenting maximum 1 and minima-1 of normalization scope respectively, X is primary signal, xmaxAnd xminRepresent the maximum in primary signal and minima respectively.

S1514: cluster obtains reference signal and comprises the following steps:

(1) k cluster center of mass point is chosen, for ��1,��2,��,��k��Rn��

(2) calculate each sample class and adopt equation below:

C ( i ) : = &alpha; r g min j | | x ( i ) - &mu; j | | 2

(3) barycenter recalculating each class j adopts equation below:

&mu; j : = &Sigma; i = 1 m 1 { C ( i ) = j } x ( i ) &Sigma; i = 1 m 1 { C ( i ) = j }

Repeat (2) and (3) until restraining;

Wherein, K is known cluster numbers, C(i)That class that representative sample I and k apoplexy due to endogenous wind is closest, C(i)Value be in 1 to k, barycenter ��jRepresent the central point of same class sample.

S152: the reference signal that after pretreatment, signal and K-means cluster obtain carries out cross-correlation calculation, adopts formula calculated as below:

C C ( &tau; ) = &Sigma; t = 0 L - t - 1 ( S D ( t ) - S D &OverBar; ) ( R ( t + &tau; ) - R &OverBar; ) &Sigma; t = 0 L - t - 1 ( S D ( t ) - S D &OverBar; ) 2 &CenterDot; &Sigma; t = &tau; t - 1 ( R ( t ) - R &OverBar; ) 2

Wherein, SD (t) is modulus value after pretreatment,For modulus value average, R (t) is reference signal,For reference signal average, CC (��) represents the cross-correlation coefficient of two signals, and in [-1,1] scope, when two signal cross-correlation degree are higher, more levels off to 1, when two signals complete uncorrelated time, more level off to 0.

S153: cross-correlation calculation result CC (��) compares with setting cross-correlation threshold value (TH_CO), and parking space state more new formula is as follows

E v e n t ( t ) = 1 C C ( t ) &GreaterEqual; T H _ C O 0 C C ( t ) < T H _ C O

Wherein, Event (t) represents whether state changes, and represents change, represent that state is constant when being 0 when being 1; TH_CO is for setting cross-correlation threshold value; CC (��) is cross-correlation calculation result. Magnetic signal primitively that parking sensor node detects as it is shown in fig. 7, its testing result as shown in Figure 8. This method interior joint is laid as shown in Figure 9.

S154: judge that parking stall last state adopts following formula:

P S ( t ) = 0 , P S ( t - 1 ) = 0 , E v e n t ( t ) = 0 ; 1 , P S ( t - 1 ) = 0 , E v e n t ( t ) = 1 ; 1 , P S ( t - 1 ) = 1 , E v e n t ( t ) = 0 ; 0 , P S ( t - 1 ) = 1 , E v e n t ( t ) = 1 ;

Wherein, PS (t)=0 represents that parking stall is idle, and PS (t)=1 represents that parking stall is occupied.

The above is the preferred embodiment of the present invention; it should be pointed out that, for those skilled in the art, under the premise without departing from principle of the present invention; can also making some improvements and modifications, these improvements and modifications also should be regarded as protection scope of the present invention.

Claims (12)

1., based on a low-power consumption curb parking detection method for geomagnetic sensor network, comprise the following steps:
Geomagnetic sensor node gathers institute's curb parking vehicle disturbing signal to earth magnetism;
The disturbing signal gathered is carried out pretreatment, obtains preprocessed signal;
By the amplitude of preprocessed signal with choice threshold ratio relatively, and select state machine detection method that complexity is low or the high cross-correlation test algorithm of degree of accuracy;
If preprocessed signal amplitude is higher than choice threshold value, then redirects scheme according to state machine and judge current parking space state; Otherwise calculate the dependency of the reference signal that preprocessed signal obtains with K-means cluster, compare to judge whether parking space state changes by cross-correlation calculation result and cross-correlation threshold value, exceed cross-correlation threshold value and then think that parking space state changes, judge parking stall last state according to parking stall previous moment state thinking after parking space state changes.
2. method according to claim 1, it is characterised in that to gather signal carry out pretreatment particularly as follows:
The disturbing signal of collection is carried out gaussian filtering process, finally calculates the modulus value of filtered signal, obtain preprocessed signal.
3. method according to claim 2, it is characterised in that described gaussian filtering processes and adopts following formula:
F (x)=exp (-xs 2/(2*��2))
Wherein, �� determines the width of Gaussian function, xsBeing the x-axis signal of described collection, y-axis adopts the method identical with x-axis to carry out gaussian filtering with z-axis.
4. method according to claim 2, it is characterised in that calculate the modulus value of signal after gaussian filtering processes and adopt following formula:
S D = ( B X - B X 0 ) 2 + ( B Y - B Y 0 ) 2 + ( B Z - B 20 ) 2
Wherein, BX��BYAnd BZRepresent the signal after the gaussian filtering of x, y and z axes, B respectivelyX0��BY0And BZ0Represent x, y and z-axis baseline value respectively.
5. method according to claim 1, it is characterised in that described in choose curb parking vehicle detecting algorithm method particularly as follows:
The modulus value of preprocessed signal is compared with choice threshold value, if it exceeds choice threshold value then selects the state machine detection method that complexity is low, otherwise selects the cross-correlation test algorithm that degree of accuracy is high.
6. method according to claim 5, it is characterised in that described low complex degree state machine detection method comprises the following steps:
(1) initialize all parameters and baseline, arrange after having initialized and be masked as 1, and jump to step (2);
(2) parking space state is idle, and carries out baseline renewal in this step, and baseline more new formula is as follows:
B i 0 ( t ) = B i 0 ( t - 1 ) &CenterDot; ( 1 - &alpha; ) + B i ( t ) &CenterDot; &alpha; , P S ( t - 1 ) = 0 B i 0 ( t - 1 ) , P S ( t - 1 ) = 0 , i &Element; { x , y , z }
In formula, Bi0(t-1) it is the i axle baseline value of previous moment, i �� { x, y, z}; BiT () is the signal after the gaussian filtering of current i axle; �� is forgetting factor, and its scope is [0-1]; The parking space state that PS (t-1) is previous moment;
When detection signal has a disturbance, and disturbing signal exceedes setting and sails threshold value into and then arrange and sail disturbance into and be masked as 1, jumps to step (3);
(3) record exceed sail threshold value into sail disturbance number of times into, in the event of disturbance lower than sailing threshold value into, then arrange and sail disturbance into and be masked as 0 and jump to step (4), otherwise exceed when frequency threshold value is sailed in setting into, jump to step (5) when vehicle sails disturbance number of times into;
And reset (4), and record without sailing disturbance number of times into, if disturbance is again above sailing threshold value into, is once again set up sailing disturbance into and is masked as 1, jump to step (3); Otherwise, sail disturbance number of times into when nothing and exceed setting without sailing frequency threshold value into, then jump to step (2);
(5) parking stall is occupied, arranges mark PS=1, if disturbance exceedes vehicle and rolls threshold value away from, then and jump procedure (6);
(6) arrange vehicle to roll disturbance away from and be masked as 1, and record and roll disturbance number of times away from, if disturbance rolls threshold value away from lower than setting, arrange and roll disturbance away from and be masked as 0, then jump to step (5); Otherwise, exceed setting and roll frequency threshold value then jump procedure (2) away from when rolling disturbance number of times away from;
Described parking stall measure result updates in step (2) and (5).
7. method according to claim 5, it is characterised in that described high accuracy cross-correlation test method comprises the following steps:
(1) reference signal that after described pretreatment, signal and K-means cluster obtain carries out cross-correlation calculation;
C C ( &tau; ) = &Sigma; t = 0 L - t - 1 ( S D ( t ) - S D &OverBar; ) ( R ( t + &tau; ) - R &OverBar; ) &Sigma; t = 0 L - t - 1 ( S D ( t ) - S D &OverBar; ) 2 &CenterDot; &Sigma; t = &tau; t - 1 ( R ( t ) - R &OverBar; ) 2
Wherein, SD (t) is modulus value after pretreatment,For modulus value average, R (t) is reference signal,For reference signal average, CC (��) represents the cross-correlation coefficient of two signals, and in [-1,1] scope, when two signal cross-correlation degree are higher, more levels off to 1, when two signals complete uncorrelated time, more level off to 0;
(2) cross-correlation calculation result CC (��) compares with setting cross-correlation threshold value, and parking space state more new formula is as follows
E v e n t ( t ) = 1 C C ( t ) &GreaterEqual; T H _ C O 0 C C ( t ) < T H _ C O
Wherein, Event (t) represents whether state changes, and represents change, represent that state is constant when being 0 when being 1; TH_CO is for setting cross-correlation threshold value; CC (��) is cross-correlation calculation result;
(3) judge that parking stall last state adopts following formula according to Event (t):
P S ( t ) = 0 , P S ( t - 1 ) = 0 , E v e n t ( t ) = 0 ; 1 , P S ( t - 1 ) = 0 , E v e n t ( t ) = 1 ; 1 , P S ( t - 1 ) = 1 , E v e n t ( t ) = 0 ; 0 , P S ( t - 1 ) = 1 , E v e n t ( t ) = 1 ;
Wherein, PS (t)=0 represents that parking stall is idle, and PS (t)=1 represents that parking stall is occupied.
8. method according to claim 7, it is characterised in that described reference signal acquisition methods comprises the following steps:
(1) preprocessed signal intercepting processes;
(2) signal is interpolated process after intercepting;
(3) after interpolation processing, signal is normalized;
(4) after normalized, signal carries out K-means cluster, obtains cluster centre and is adopted as reference signal.
9. method according to claim 8, it is characterised in that described signal intercepting processes and adopts formula as follows:
SD_A=SD (t), { (t1-td)��t��(t2+td)}
SD_D=SD (t), { (t3-td)��t��(t4+td)}
Wherein, SD (t) is signal modulus value, t after pretreatment1It is the time of vehicle arrival, t2It is the time point that after stopping, signal is changed into steady statue from labile state, t3It is time of leaving of vehicle, t4It is that car leaves rear signal and is changed into the time point of steady statue, t from labile statedIt it is time delay parameter.
10. method according to claim 8, it is characterised in that described interpolation processing comprises the following steps:
(1) signal calculated length M, and matched curve y=f (x);
(2) on the interval that length is M, N number of point, X it are evenly distributed1, X2..., XN(N is the regular length after interpolation);
(3) functional value (y corresponding to N number of point is calculated according to matched curve1, y2..., yN);
(4) Y=[y1, y2..., yN] it is characteristic curve after interpolation.
11. method according to claim 8, it is characterised in that described normalization adopts following formula:
Y = ( Y m a x - Y m i n ) * ( x - x m i n ) x max - x min + Y m i n
Wherein, YmaxAnd YminRepresenting maximum 1 and minima-1 of normalization scope respectively, X is primary signal, xmaxAnd xminRepresent the maximum in primary signal and minima respectively.
12. method according to claim 8, it is characterised in that described K-means sorting procedure is as follows:
(1): choose k cluster center of mass point, for ��1, ��2..., ��k��Rn;
(2): calculating the classification of each sample, computing formula is as follows:
C(i):=argminj||x(i)-��j||2;
(3): recalculating the barycenter of each class j, computing formula is as follows:
&mu; j : = &Sigma; i = 1 m 1 { C ( i ) = j } x ( i ) &Sigma; i = 1 m 1 { C ( i ) = j } ;
Repeat step (2) and (3) until restraining;
Wherein, K is known cluster numbers, C(i)That class that representative sample I and k apoplexy due to endogenous wind is closest, C(i)Value be in 1 to k, barycenter ��jRepresent the central point of same class sample.
CN201511028573.6A 2015-12-31 2015-12-31 Low-power consumption curb parking detection method based on geomagnetic sensor network CN105632187B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201511028573.6A CN105632187B (en) 2015-12-31 2015-12-31 Low-power consumption curb parking detection method based on geomagnetic sensor network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201511028573.6A CN105632187B (en) 2015-12-31 2015-12-31 Low-power consumption curb parking detection method based on geomagnetic sensor network

Publications (2)

Publication Number Publication Date
CN105632187A true CN105632187A (en) 2016-06-01
CN105632187B CN105632187B (en) 2019-05-31

Family

ID=56047060

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201511028573.6A CN105632187B (en) 2015-12-31 2015-12-31 Low-power consumption curb parking detection method based on geomagnetic sensor network

Country Status (1)

Country Link
CN (1) CN105632187B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106327913A (en) * 2016-08-31 2017-01-11 广州地理研究所 Parking lot management method and device based on automobile electronic identification
CN106530817A (en) * 2016-12-21 2017-03-22 迈锐数据(北京)有限公司 Parking space detection method and device
CN106652479A (en) * 2016-11-22 2017-05-10 广西大学 Geomagnetic sensing-based wireless low-power consumption parking space detection system
CN106781531A (en) * 2016-12-21 2017-05-31 迈锐数据(北京)有限公司 A kind of method for detecting parking stalls and device
CN107610481A (en) * 2017-08-31 2018-01-19 深圳先进技术研究院 Stop detection method, device and terminal device and computer-readable recording medium
CN109308814A (en) * 2017-07-28 2019-02-05 南宁富桂精密工业有限公司 Method for detecting parking stalls, parking sensor and computer readable storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101034496A (en) * 2007-04-17 2007-09-12 中国科学院计算技术研究所 Detecting method and device for communication flow
KR20090106201A (en) * 2008-04-04 2009-10-08 (주)에스엔알 System and method for managing the admittance to parking place based on geomagnetic sensor
CN103117003A (en) * 2013-01-17 2013-05-22 杭州快泊信息技术有限公司 Self-calibration threshold self-adjusting parking space detection method
CN103794058A (en) * 2014-03-05 2014-05-14 武汉慧联无限科技有限公司 Method and device for detecting vehicle based on state machine
CN203882450U (en) * 2014-04-24 2014-10-15 纳米新能源(唐山)有限责任公司 Parking-stall state detector
CN104361765A (en) * 2014-11-24 2015-02-18 天津理工大学 Parking space detecting system based on magnetoresistive sensor and ZigBee

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101034496A (en) * 2007-04-17 2007-09-12 中国科学院计算技术研究所 Detecting method and device for communication flow
KR20090106201A (en) * 2008-04-04 2009-10-08 (주)에스엔알 System and method for managing the admittance to parking place based on geomagnetic sensor
CN103117003A (en) * 2013-01-17 2013-05-22 杭州快泊信息技术有限公司 Self-calibration threshold self-adjusting parking space detection method
CN103794058A (en) * 2014-03-05 2014-05-14 武汉慧联无限科技有限公司 Method and device for detecting vehicle based on state machine
CN203882450U (en) * 2014-04-24 2014-10-15 纳米新能源(唐山)有限责任公司 Parking-stall state detector
CN104361765A (en) * 2014-11-24 2015-02-18 天津理工大学 Parking space detecting system based on magnetoresistive sensor and ZigBee

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106327913A (en) * 2016-08-31 2017-01-11 广州地理研究所 Parking lot management method and device based on automobile electronic identification
CN106652479A (en) * 2016-11-22 2017-05-10 广西大学 Geomagnetic sensing-based wireless low-power consumption parking space detection system
CN106530817A (en) * 2016-12-21 2017-03-22 迈锐数据(北京)有限公司 Parking space detection method and device
CN106781531A (en) * 2016-12-21 2017-05-31 迈锐数据(北京)有限公司 A kind of method for detecting parking stalls and device
CN109308814A (en) * 2017-07-28 2019-02-05 南宁富桂精密工业有限公司 Method for detecting parking stalls, parking sensor and computer readable storage medium
CN107610481A (en) * 2017-08-31 2018-01-19 深圳先进技术研究院 Stop detection method, device and terminal device and computer-readable recording medium
CN107610481B (en) * 2017-08-31 2019-09-17 深圳先进技术研究院 Stop detection method, device and terminal device and computer readable storage medium

Also Published As

Publication number Publication date
CN105632187B (en) 2019-05-31

Similar Documents

Publication Publication Date Title
Balid et al. Intelligent vehicle counting and classification sensor for real-time traffic surveillance
CN102717765B (en) Fatigue driving detection method and anti-fatigue driving auxiliary device
Lan et al. Landslide hazard spatial analysis and prediction using GIS in the Xiaojiang watershed, Yunnan, China
CN103903452B (en) Forecasting Approach for Short-term Traffic Flow
Taghvaeeyan et al. Portable roadside sensors for vehicle counting, classification, and speed measurement
CN103971520B (en) A kind of traffic flow data restoration methods based on temporal correlation
CN102156282B (en) Method for detecting radar target based on micro-Doppler effect
CN101739815B (en) On-line recognition method of road traffic congestion state
CN101807352B (en) Method for detecting parking stalls on basis of fuzzy pattern recognition
CN103413441B (en) Road weather condition method for supervising
CN102855638B (en) Detection method for abnormal behavior of vehicle based on spectrum clustering
CN102967854B (en) Multi-fractal detection method of targets in FRFT (Fractional Fourier Transformation) domain sea clutter
CN103196465B (en) Phase sensitive optical time-domain reflectometer (phi-OTDR) sensing signal noise separation and signal extraction method
CN101633358A (en) Adaptive vehicle control system with integrated driving style recognition
CN103323749A (en) Multi-classifier information fusion partial discharge diagnostic method
CN102360526B (en) Real-time monitoring method for road section state of high road
CN102997926B (en) A kind of method for obtaining navigation data
Yang et al. Experimental study of a hand-drawn cart for measuring the bridge frequencies
CN102253377B (en) Target detection method for polarimetric interferometry synthetic aperture radar on basis of eigenvalue analysis
Becker et al. Spatiotemporal analysis of precipitation trends in the Yangtze River catchment
Song et al. Distribution characteristics of vehicle-specific power on urban restricted-access roadways
CN106314438B (en) The detection method and system of abnormal track in a kind of driver driving track
CN105489008B (en) Urban road congestion computational methods and system based on Floating Car satellite location data
CN103236063B (en) Based on the SAR image oil spilling detection method of multiple dimensioned spectral clustering and decision level fusion
CN102750703B (en) Remote sensing image airport automatic detecting method based on linear cutting

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
GR01 Patent grant