CN103499350A - High-precision vehicle positioning method for fusing multi-source information under GPS (global positioning system) blind area and device - Google Patents

Abstract

The invention discloses a high-precision vehicle positioning method for fusing multi-source information under a GPS (global positioning system) blind area and a device. By a strap-down matrix algorithm of the method, INS (inertial navigation system) vehicle position information and position information obtained by a pavement matching technology are calculated according to angular rate information output by a gyroscope and acceleration information output by an accelerometer and are fused by a kalman filtering algorithm to output final fused positioning information. As an INS positioning algorithm has an accumulated error, vehicle position information is recalibrated by arranging an anchor node beside a road. According to the method, stable and reliable high-precision vehicle position information can be obtained; the method is suitable for non-GPS-signal environments such as urban roads with dense buildings, mountain areas and tunnels.

Description

Merge vehicle high-precision locating method and the device of multi-source information under the GPS blind area

Technical field

The invention belongs to traffic detection and vehicle positive location technical field, relate to the vehicle high-precision locating method under the GPS blind area, particularly the vehicle high-precision locating method based on Multi-source Information Fusion under a kind of GPS blind area.

Background technology

For the detection of vehicle high precision position information, be an important parameter in car networking field, particularly, under the complex environments such as mountain area, tunnel, vehicle high precision position information becomes particularly important.So the high precision position information of obtaining vehicle under the GPS blind area is significant.

At present, vehicle high precision acquisition methods more commonly used mainly contains: GPS location, inertial navigation technology, road surface matching technique etc.Wherein, GPS is the location technology be most widely used at present, and existing intelligent transportation system generally adopts the GPS method to carry out vehicle location, but the method is difficult to meet the application demand of future car networking, and mainly have following defect: (1) positioning precision is low.Normal domestic GPS positioning precision is 5～25 meters, and the positioning error of differential GPS is also 1 meter left and right.(2) there is the gps signal blind area.While travelling in the vehicle urban district intensive at mountainous area highway, tunnel, buildings, gps signal easily is blocked, and now vehicle will enter the gps signal blind area, can't use GPS to position.(3) output frequency is low.

INS(Inertial Navigation System). inertial navigation system, be called for short inertial navigation) be the navigation positioning system grown up 20 beginnings of the century, its ultimate principle is the mechanics law according to inertial space, utilize the inertance elements such as gyro and accelerometer to experience angular velocity of rotation and the acceleration of carrier in motion process, ground by servo-drive system hangs down and follows the tracks of or the coordinate system rotation conversion, integral and calculating in certain coordinate system, finally obtain the navigational parameters such as relative position, speed and attitude of carrier.But INS, due to the cumulative errors problem, is difficult to maintain for a long time hi-Fix.

The advantages such as the road surface Matching Navigation System possesses independence, high precision, round-the-clock, antijamming capability is strong, along with the continuous enhancing of development and the digital map bank preparative capacibility of image sensor technologies, road surface coupling airmanship development prospect is wide.But its technical disadvantages is also fairly obvious, is mainly manifested in: the fabrication cycle of (1) reference map is oversize; (2), when in facing on a large scale, road surface characteristic changes unconspicuous zone, the road surface Matching Navigation System will can't be used or make navigation accuracy greatly to reduce because of locating; (3) be difficult to the adaline moving-target.

Due to defects such as the existing adaptability of road surface matching algorithm, efficiency, it is the fuse information source that the vehicle high-precision locating method of most based on merging thought all adopts GPS, INS, but can't reach the ideal effect of expection in the situation of GPS blind area due to the cumulative errors problem.

Summary of the invention

The defect or the deficiency that while using separately for above-mentioned various technology, exist, the object of the invention is to propose to merge under a kind of GPS blind area the vehicle high-precision locating method of multi-source information, can avoid the drawback of prior art, thereby realize effectively obtaining under the GPS blind area function of vehicle high precision position information.

In order to realize above-mentioned task, the present invention takes following technical solution:

Merge the vehicle high-precision locating method of multi-source information under a kind of GPS blind area, comprise the following steps:

Step 1, the INS localization process:

Step S10, obtain vehicle current acceleration a and angular velocity omega, and degree of will speed up and angular velocity, as the input data, connect according to victory the INS elements of a fix (x that matrix algorithms obtains vehicle ₁, y ₁);

Step 2, road surface coupling localization process:

Step S20, take after system initialization the first rectangular image edge, width road surface of obtaining to set up the frame of reference as x axle and y axle, records the now initial coordinate of vehicle;

Step S21, obtain continuous two road surface rectangular image: f _n(x, y) and f _n+1(x, y), and two images have the zone of coincidence on pixel;

Step S22, with the pavement image f obtained _n(x, y) and f _n+1(x, y), as the input data, obtains image f according to the SIFT matching algorithm _n(x, y) some P in overlapping zone ₁with a Q ₁at image f _n+1match point P on (x, y) ₂and Q ₂;

Step S23, at image f _n+12 P of (x, y) upper searching ₀and Q ₀, make P ₀and Q ₀at image f _n+1position on (x, y) and some P ₁and Q ₁at image f _nposition on (x, y) is identical;

Step S24, with image f _n+1the edge of (x, y) is that x axle and y axle are set up coordinate system XOY, in the XOY coordinate system with formula 1 computed image f _n+1(x, y) is with respect to f _nthe linear deflection amount of (x, y) (Δ x, Δ y):

$\left\{\begin{array}{c}\mathrm{\Δx}=P_0\left(x\right)-P_2\left(x\right)\\ \mathrm{\Δy}=P_0\left(y\right)-P_2\left(y\right)\end{array}\right.$ (formula 1)

In formula, P ₀and P (x) ₀(y) be a P ₀horizontal ordinate in coordinate system XOY and ordinate, P ₂and P (x) ₂(y) be a P ₂horizontal ordinate in coordinate system XOY and ordinate;

Step S25, in the XOY coordinate system, utilize formula 2 computed image f _n+1(x, y) and image f _n+1the anglec of rotation θ of (x, y):

$\mathrm{\θ}=\arctan \frac{Q_2\left(x\right)-P_2\left(x\right)}{Q_2\left(y\right)-P_2\left(y\right)}-\arctan \frac{Q_0\left(x\right)-P_0\left(x\right)}{Q_0\left(y\right)-P_0\left(y\right)}$ (formula 2)

In formula, Q ₀and Q (x) ₀(y) be a Q ₀horizontal ordinate in coordinate system XOY and ordinate, Q ₂and Q (x) ₂(y) be a Q ₂horizontal ordinate in coordinate system XOY and ordinate;

Step S26, with image f _nthe edge of (x, y) is that x axle and y axle are set up coordinate system xoy, in the xoy coordinate system with formula 3 computed image f _n+1(x, y) is with respect to f _nthe real offset of (x, y) (Δ x ', Δ y '):

(formula 3)

Step S27, perform step S21 to step S26 successively by two width road surface rectangular images of the arbitrary continuation that obtains in the Vehicle Driving Cycle process, can obtain vehicle now with respect to the track of the frame of reference; According to initial coordinate and the track of vehicle, obtain the road surface coupling elements of a fix (x ₂, y ₂),

Step 3, the elements of a fix merge:

Step S30, by the INS elements of a fix (x ₁, y ₁) mate the elements of a fix (x with road surface ₂, y ₂) utilize formula 4 to be merged, obtain merging coordinate (x ₀, y ₀):

(x ₀, y ₀)=a (x ₁, y ₁)+b (x ₂, y ₂) (formula 4)

In formula, a and b are fusion coefficients, by Orthogonal Rotational Regressive Tests, obtain, and wherein a gets 0.382, b and gets 0.618;

Step S31, utilize Kalman filtering algorithm to merging coordinate (x ₀, y ₀) carry out state estimation, obtain the final elements of a fix (x, y) of vehicle;

Step 4, the anchor node check and correction:

Step S41, utilize and be arranged on the camera on road at interval of 5km, at vehicle, through out-of-date video camera, catches vehicle image f (x, y);

Step S42, carry out gaussian filtering and dodging to vehicle image, the image after being processed;

Step S43, the image obtained after step S42 is processed, it is poor that the background image of taking at same position through out-of-date camera with vehicles failed is done, and then to doing after poor, the image obtained carries out histogram modification and erosion operation is processed;

Step S44, the image obtained after step S43 is processed carries out connected region scanning, obtains the connected region of vehicle image, the position in the middle of image using the minimum boundary rectangle center of connected region as vehicle;

Step S45, the position by vehicle in the middle of image is converted to the actual coordinate of vehicle in the middle of reality, and the elements of a fix (x, y) that this actual coordinate is obtained step 3 are replaced, to eliminate the cumulative errors of the elements of a fix (x, y).

Merge the device of the vehicle high-precision locating method of multi-source information under a kind of GPS blind area, comprise the accelerometer and the gyroscope that are arranged on vehicle, this device also comprises central processing module and is arranged on vehicle and camera lens is parallel to first camera on road surface, be connected with on central processing module and merge locating module and Zigbee wireless receiving module, described accelerometer is connected with the fusion locating module by INS localization process module with gyroscope, and the first camera mates locating module by road surface and is connected with the fusion locating module; Be provided with portal frame every 5km on the road of Vehicle Driving Cycle, second camera and RFID receiver are installed on portal frame, be connected with the anchor node correcting module on second camera and RFID receiver, be connected with the Zigbee wireless sending module with Zigbee wireless receiving module radio communication on the anchor node correcting module; Each parts are achieved as follows respectively function:

Accelerometer adopts YC-A150S-M type acceleration transducer Real-time Obtaining vehicle acceleration information;

Gyroscope adopts CMR3100-D01 type angular-rate sensor Real-time Obtaining vehicle angular velocity information;

Vehicle acceleration information and angular velocity information that INS localization process module is obtained according to accelerometer and gyroscope connect by victory the INS elements of a fix that matrix algorithms obtains vehicle;

The first camera adopts the image on road surface in XC-103c Sony CCD type video frequency pick-up head Real-time Collection Vehicle Driving Cycle process;

Road surface coupling localization process module is processed by the pavement image to the first camera collection, obtains the road surface coupling elements of a fix of vehicle with respect to initial coordinate;

Merge locating module and calculate the fusion coordinate according to the INS elements of a fix and the road surface coupling elements of a fix, and utilize Kalman filtering algorithm to carry out state estimation to merging coordinate, obtain the final elements of a fix of vehicle;

Second camera adopts XC-103c Sony CCD camera, when the vehicle process is installed the portal frame below of second camera, and the image of collection vehicle;

The RFID receiver coordinates with the IC-card on vehicle, and during vehicle process portal frame, the RFID receiver triggers second camera collection vehicle image;

The anchor node correcting module is processed the image of second camera collection, calculates the position of vehicle in the middle of image, and is converted to the coordinate in reality;

The final elements of a fix of car two that central processing module utilizes the coordinate in car two reality that obtain in the anchor node correcting module to calculate the fusion locating module are revised, to eliminate cumulative errors.

The present invention is merged the Multiple Source Sensors such as gyroscope, accelerometer, ccd video camera, RFID as information source, adopt Kalman filtering algorithm to carry out the positional information fusion treatment, use the anchor node Proofreading Algorithms to realize the positional information check and correction, finally realize stable, efficient, the high-precision fixed bit function of vehicle.Through the field experiment application, show, vehicle high-precision locating method based on Multi-source Information Fusion under GPS of the present invention blind area can effectively have been avoided the not enough problem of single location algorithm precision, road surface matching algorithm adaptability, efficiency defect have been improved simultaneously, solved the deficiency of inertial navigation algorithm cumulative errors, the method locating effect is better, is applicable to urban road that building are intensive, mountain area, tunnel etc. without the gps signal environment.

The accompanying drawing explanation

The program flow diagram that Fig. 1 is the inventive method;

The one-piece construction connection layout of Fig. 2 apparatus of the present invention;

Fig. 3 is is the installation site schematic diagram of second camera;

The workflow diagram that Fig. 4 is road surface coupling localization process module;

The schematic diagram that Fig. 5 is match point in the SIFT matching algorithm;

The track of vehicle that Fig. 6 is the inventive method acquisition and the comparison diagram of vehicle actual path;

Below in conjunction with accompanying drawing, to of the present invention, be described in further detail.

Embodiment

Merge the vehicle high-precision locating method of multi-source information under a kind of GPS blind area, as shown in Figure 1, comprise the following steps:

Thereby at first manually set gps coordinate or local relative coordinate initial point realize system INS the initialization of road surface coupling etc., if adopt the gps coordinate initialization, final Output rusults is gps coordinate, if adopt local relative coordinate, final Output rusults is local relative coordinate.

Step 1, the INS localization process:

Step S10, obtain vehicle current acceleration a and angular velocity omega, and degree of will speed up and angular velocity, as the input data, obtain acceleration and adopt YC-A150S-M type acceleration transducer, obtain angular velocity and adopt CMR3100-D01 type angular-rate sensor; INS localization process module is utilized acceleration information and angular velocity information, by victory, connects the INS elements of a fix (x that matrix algorithms obtains vehicle ₁, y ₁); The elements of a fix that herein obtain can be that gps coordinate can be also relative coordinate;

Step 2, road surface coupling localization process:

In road surface coupling location algorithm, road surface coupling localization process module is taken information of road surface as information source with first camera of usining that is arranged on vehicle and camera lens is parallel to road surface, adopts the SIFT matching algorithm to obtain locating information according to information of road surface; After adopting the SIFT matching algorithm to complete images match, then calculate side-play amount and the angle between match point, then according to side-play amount and the angle of correct match point, obtain the vehicle operating track, thereby obtain the vehicle real-time position information.The first camera adopts XC-103c Sony CCD type video frequency pick-up head; As shown in Figure 4, concrete steps are as follows for flow process:

Step S20, take after system initialization the first rectangular image edge, width road surface of obtaining to set up the frame of reference as x axle and y axle, and take first pixel position of this image is true origin; Due in processing procedure, be take before and after the two width images offset deviation of carrying out two width image relative positions obtain vehicle track, therefore after system initialization, the first width pavement image that while being vehicle launch, the first video camera is taken is benchmark image, set up the frame of reference, to obtain the running orbit of vehicle with respect to the frame of reference;

Step S21, obtain continuous two road surface rectangular image: f _n(x, y) and f _n+1(x, y), and two images have the zone of coincidence on pixel; In the present invention, adopt per second to take the speed of 5 pavement images, to guarantee in the situation that the two width images of the general travelling speed of vehicle by the first camera collection have coincidence on pixel, because shooting interval is short, therefore the latter half of front piece image there will be the first half at rear piece image, make two width images that coincidence be arranged on pixel, to meet the requirement of SIFT matching algorithm;

Step S22, with the pavement image f obtained _n(x, y) and f _n+1(x, y), as the input data, obtains image f according to the SIFT matching algorithm _n(x, y) some P in overlapping zone ₁with a Q ₁at image f _n+1match point P on (x, y) ₂and Q ₂, as shown in Figure 5, due to image f _n(x, y) and f _n+1(x, y) has the zone of coincidence, i.e. hatched example areas shown in figure, so match point P ₂with a P ₁can regard a point as on locus, but match point P ₂at image f _n+1(x, y) is upper, and some P ₁at image f _non (x, y);

Step S23, at image f _n+12 P of (x, y) upper searching ₀and Q ₀, make P ₀and Q ₀at image f _n+1position on (x, y) and some P ₁and Q ₁at image f _nposition on (x, y) is identical;

Step S24, the image gathered in the present invention is rectangular image, with image f _n+1the edge of (x, y) is that x axle and y axle are set up coordinate system XOY, in the XOY coordinate system with formula 1 computed image f _n+1(x, y) is with respect to f _nthe linear deflection amount of (x, y) (Δ x, Δ y):

$\left\{\begin{array}{c}\mathrm{\Δx}=P_0\left(x\right)-P_2\left(x\right)\\ \mathrm{\Δy}=P_0\left(y\right)-P_2\left(y\right)\end{array}\right.$ (formula 1)

In formula, P ₀and P (x) ₀(y) be a P ₀horizontal ordinate in coordinate system XOY and ordinate, P ₂and P (x) ₂(y) be a P ₂horizontal ordinate in coordinate system XOY and ordinate;

Step S25, in the XOY coordinate system, utilize formula 2 computed image f _n+1(x, y) and image f _n+1the anglec of rotation θ of (x, y):

$\mathrm{\θ}=\arctan \frac{Q_2\left(x\right)-P_2\left(x\right)}{Q_2\left(y\right)-P_2\left(y\right)}-\arctan \frac{Q_0\left(x\right)-P_0\left(x\right)}{Q_0\left(y\right)-P_0\left(y\right)}$ (formula 2)

In formula, Q ₀and Q (x) ₀(y) be a Q ₀horizontal ordinate in coordinate system XOY and ordinate, Q ₂and Q (x) ₂(y) be a Q ₂horizontal ordinate in coordinate system XOY and ordinate;

Step S26, with image f _nthe edge of (x, y) is that x axle and y axle are set up coordinate system xoy, in the xoy coordinate system with formula 3 computed image f _n+1(x, y) is with respect to f _nthe real offset of (x, y) (Δ x ', Δ y '):

$\left\{\begin{array}{c}\mathrm{\Δ}{x}^{\′}=\mathrm{\Δ}x\cos \mathrm{\θ}-\mathrm{\Δ}y\sin \mathrm{\θ}\\ \mathrm{\Δ}{y}^{\′}=\mathrm{\Δ}x\sin \mathrm{\θ}+\mathrm{\Δ}y\cos \mathrm{\θ}\end{array}\right.$ (formula 3)

Step S27, perform step S21 to step S26 successively by two width road surface rectangular images of the arbitrary continuation that obtains after self-initialize in the Vehicle Driving Cycle process, can obtain vehicle now with respect to the track of the frame of reference; The initial coordinate of frame of reference during initialization is known, therefore can be according to initial coordinate and the track of vehicle, according to the transformational relation of gps coordinate and distance, calculate this algorithm Output rusults if be initialized as gps coordinate, if the elements of a fix (x, directly as this algorithm Output rusults, is mated thereby obtain road surface in the relative coordinate position ₂, y ₂),

Step 3, the elements of a fix merge:

Step S30, merge locating module by the INS elements of a fix (x ₁, y ₁) mate the elements of a fix (x with road surface ₂, y ₂) utilize formula 4 to be merged, obtain merging coordinate (x ₀, y ₀):

(x ₀, y ₀)=a (x ₁, y ₁)+b (x ₂, y ₂) (formula 4)

In formula, a and b are fusion coefficients, by Orthogonal Rotational Regressive Tests, obtain, and wherein a gets 0.382, b and gets 0.618 o'clock, the two fitting effect the best, and the matching test is as follows:

At certain proving ground, adopt INS location algorithm as above and pavement image location algorithm respectively vehicle to be positioned 30 times, each Vehicle Driving Cycle 1KM, can obtain 30 groups of INS elements of a fix (x ' ₁, y ' ₁); With 30 groups of road surfaces coupling elements of a fix (x ' ₂, y ' ₂), the actual coordinate of vehicle (x ' ₀, y ' ₀) known, this actual coordinate is the fusion coordinate (x that will try to achieve ₀, y ₀); There is the 4-1 relation between the actual coordinate of vehicle and two kinds of location algorithms:

(x ' ₀, y ' ₀)=a (x ' ₁, y ' ₁)+b (x ' ₂, y ' ₂) formula (4-1)

In above formula, it is unknown only having coefficient a and b, by 30 groups of data substitution above formulas respectively, the line retrace analytical calculation of going forward side by side, find out make all groups with actual value (x ' ₀, y ' ₀) coefficient a and the b of error minimum, finally calculate to such an extent that return fusion coefficients a and get 0.382, b and get 0.618.

Step S31, utilize Kalman filtering algorithm to merging coordinate (x ₀, y ₀) carrying out state estimation, the fusion coordinate that Kalman filtering algorithm can obtain constantly according to front, estimate the coordinate at lower a moment, the elements of a fix (x, y) that the coordinate that Kalman filtering algorithm is obtained is final as vehicle;

Step 4, the anchor node check and correction:

Step S41, second camera is set on road at interval of 5km, this camera and RFID receiver coordinate, IC-card is housed on vehicle, when vehicle through out-of-date, the RFID receiver obtains IC-card information and learns that vehicle arrives, and triggers second camera and catches vehicle image f (x, y) at vehicle through out-of-date video camera; Second camera adopts XC-103c Sony CCD type video frequency pick-up head, on road, at interval of 5km, portal frame is set, and second camera is arranged on portal frame, gathers image;

Step S42, the anchor node correcting module carries out gaussian filtering and dodging to vehicle image, eliminates as much as possible external interference, the image after being processed;

Step S43, the image obtained after step S42 is processed, it is poor that the background image of taking at same position through out-of-date camera with vehicles failed is done, and then to doing after poor, the image obtained carries out histogram modification and erosion operation is processed; Can improve the contrast of image, the dynamic range of expansion pixel; Erosion operation can be eliminated little bright spot noise.

Step S44, the image obtained after step S43 is processed carries out connected region scanning, obtain the connected region of vehicle image, obtain the geometric properties of connected region simultaneously, comprise that connected region bears line segment frontier point, minimum boundary rectangle, area, girth and the centre of form etc.Target area area parameters S () can be used as a kind of scale of measurement, and for connected region R (x, y), S () is defined as number of pixels in this zone,

$S\left(R(x,y)\right)=\underset{(x,y)\∈R_t(x,y)}{\mathrm{\Σ}}f(x,y)$

Wherein, f (x, y) is the image when pre-treatment; The position in the middle of image using the minimum boundary rectangle center of connected region as vehicle;

Step S45, the position by vehicle in the middle of image is converted to the actual coordinate of vehicle in the middle of reality; Because the position of second camera is fixed, coordinate is known, and in the photo of its shooting, the coordinate of any point is also known, and therefore according to vehicle, the position in the middle of image can obtain the vehicle actual coordinate; Central processing module obtains this actual coordinate elements of a fix (x, y) to step 3 are replaced, to eliminate the cumulative errors of the elements of a fix (x, y); ; in the normal vehicle operation process; the final elements of a fix (x, y) that utilize step S31 to obtain are as the current coordinate of vehicle, but in order to eliminate cumulative errors; be provided with the anchor node correction module in the present invention; carry out the coordinate position correction every 5km, with the vehicle actual coordinate after proofreading and correct, replace the elements of a fix (x, y); can eliminate the cumulative errors of location, keep the high precision of location.

The present invention also provides a kind of device of realizing said method, and structure is shown in Fig. 2:

Merge the device of the vehicle high-precision locating method of multi-source information under a kind of GPS blind area, comprise the accelerometer and the gyroscope that are arranged on vehicle, this device also comprises central processing module and is arranged on vehicle and camera lens is parallel to first camera on road surface, be connected with on central processing module and merge locating module and Zigbee wireless receiving module, described accelerometer is connected with the fusion locating module by INS localization process module with gyroscope, and the first camera mates locating module by road surface and is connected with the fusion locating module; Be provided with portal frame every 5km on the road of Vehicle Driving Cycle, second camera and RFID receiver are installed on portal frame, be connected with the anchor node correcting module on second camera and RFID receiver, be connected with the Zigbee wireless sending module with Zigbee wireless receiving module radio communication on the anchor node correcting module; Each parts are achieved as follows respectively function:

Accelerometer adopts YC-A150S-M type acceleration transducer Real-time Obtaining vehicle acceleration information;

Gyroscope adopts CMR3100-D01 type angular-rate sensor Real-time Obtaining vehicle angular velocity information;

Vehicle acceleration information and angular velocity information that INS localization process module is obtained according to accelerometer and gyroscope connect by victory the INS elements of a fix that matrix algorithms obtains vehicle;

The first camera adopts the image on road surface in XC-103c Sony CCD type video frequency pick-up head Real-time Collection Vehicle Driving Cycle process;

Road surface coupling localization process module is processed by the pavement image to the first camera collection, obtains the road surface coupling elements of a fix of vehicle with respect to initial coordinate;

Merge locating module and calculate the fusion coordinate according to the INS elements of a fix and the road surface coupling elements of a fix, and utilize Kalman filtering algorithm to carry out state estimation to merging coordinate, obtain the final elements of a fix of vehicle;

Second camera adopts XC-103c Sony CCD camera, when the vehicle process is installed the portal frame below of second camera, and the image of collection vehicle;

The RFID receiver coordinates with the IC-card on vehicle, and during vehicle process portal frame, the RFID receiver triggers second camera collection vehicle image;

The anchor node correcting module is processed the image of second camera collection, calculates the position of vehicle in the middle of image, and is converted to the coordinate in reality;

The final elements of a fix of car two that central processing module utilizes the coordinate in car two reality that obtain in the anchor node correcting module to calculate the fusion locating module are revised, to eliminate cumulative errors.

Embodiment:

The inventor tests with the inventive method and GPS localization method, and experimentation is as follows:

The inventor is tested at certain proving ground, adopt as above method, every 5km, portal frame and second camera are set on runway, on vehicle, inner installation gyroscope, accelerometer obtain angular speed, acceleration, outside vehicle, section part is installed the first camera and is taken pavement of road, adopts as above method to be tested final testing result as shown in Figure 6.Wherein indicate the track of leg-of-mutton lines for adopting the method to obtain, the line that indicates circle is the vehicle actual path.As can be seen from Figure 6, triangle line is almost completely identical with the circular lines trajectory diagram, proves that locating effect is quite desirable, and the actual effect of algorithm is good.

Comparative Examples:

The inventor adopts respectively journey ZT410GPS and Jia Ming 3790TGPS to be tested at certain proving ground, and the whole gps signal of this test site is good, on average can connect number of satellite and can reach 3 more than star.Each test 1km, test respectively 3 times, and track of vehicle and the vehicle actual travel track that adopts GPS to obtain carried out to error analysis, as shown in table 1 below.As can be seen from the table, the inventive method can reach the GPS positioning precision, proves that locating effect is quite desirable, and the actual effect of algorithm is good.

Table 1 error testing result

Claims (2)

1. merge the vehicle high-precision locating method of multi-source information under a GPS blind area, it is characterized in that, comprise the following steps:

Step 1, the INS localization process:

Step S10, obtain vehicle current acceleration a and angular velocity omega, and degree of will speed up and angular velocity, as the input data, connect according to victory the INS elements of a fix (x that matrix algorithms obtains vehicle ₁, y ₁);

Step 2, road surface coupling localization process:

Step S20, take after system initialization the first rectangular image edge, width road surface of obtaining to set up the frame of reference as x axle and y axle, records the now initial coordinate of vehicle;

Step S21, obtain continuous two road surface rectangular image: f _n(x, y) and f _n+1(x, y), and two images have the zone of coincidence on pixel;

Step S22, with the pavement image f obtained _n(x, y) and f _n+1(x, y), as the input data, obtains image f according to the SIFT matching algorithm _n(x, y) some P in overlapping zone ₁with a Q ₁at image f _n+1match point P on (x, y) ₂and Q ₂;

Step S23, at image f _n+12 P of (x, y) upper searching ₀and Q ₀, make P ₀and Q ₀at image f _n+1position on (x, y) and some P ₁and Q ₁at image f _nposition on (x, y) is identical;

Step S24, with image f _n+1the edge of (x, y) is that x axle and y axle are set up coordinate system XOY, in the XOY coordinate system with formula 1 computed image f _n+1(x, y) is with respect to f _nthe linear deflection amount of (x, y) (Δ x, Δ y):

$\left\{\begin{array}{c}\mathrm{\Δx}=P_0\left(x\right)-P_2\left(x\right)\\ \mathrm{\Δy}=P_0\left(y\right)-P_2\left(y\right)\end{array}\right.$ (formula 1)

In formula, P ₀and P (x) ₀(y) be a P ₀horizontal ordinate in coordinate system XOY and ordinate, P ₂and P (x) ₂(y) be a P ₂horizontal ordinate in coordinate system XOY and ordinate;

Step S25, in the XOY coordinate system, utilize formula 2 computed image f _n+1(x, y) and image f _n+1the anglec of rotation θ of (x, y):

$\mathrm{\θ}=\arctan \frac{Q_2\left(x\right)-P_2\left(x\right)}{Q_2\left(y\right)-P_2\left(y\right)}-\arctan \frac{Q_0\left(x\right)-P_0\left(x\right)}{Q_0\left(y\right)-P_0\left(y\right)}$ (formula 2)

In formula, Q ₀and Q (x) ₀(y) be a Q ₀horizontal ordinate in coordinate system XOY and ordinate, Q ₂and Q (x) ₂(y) be a Q ₂horizontal ordinate in coordinate system XOY and ordinate;

Step S26, with image f _nthe edge of (x, y) is that x axle and y axle are set up coordinate system xoy, in the xoy coordinate system with formula 3 computed image f _n+1(x, y) is with respect to f _nthe real offset of (x, y) (Δ x ', Δ y '):

$\left\{\begin{array}{c}\mathrm{\Δ}{x}^{\′}=\mathrm{\Δ}x\cos \mathrm{\θ}-\mathrm{\Δ}y\sin \mathrm{\θ}\\ \mathrm{\Δ}{y}^{\′}=\mathrm{\Δ}x\sin \mathrm{\θ}+\mathrm{\Δ}y\cos \mathrm{\θ}\end{array}\right.$ (formula 3)

Step S27, perform step S21 to step S26 successively by the two continuous width road surface rectangular images that constantly obtain in the Vehicle Driving Cycle process, can obtain vehicle now with respect to the track of the frame of reference; According to initial coordinate and the track of vehicle, obtain the road surface coupling elements of a fix (x ₂, y ₂),

Step 3, the elements of a fix merge:

Step S30, by the INS elements of a fix (x ₁, y ₁) mate the elements of a fix (x with road surface ₂, y ₂) utilize formula 4 to be merged, obtain merging coordinate (x ₀, y ₀):

(x ₀, y ₀)=a (x ₁, y ₁)+b (x ₂, y ₂) (formula 4)

In formula, a and b are fusion coefficients, by Orthogonal Rotational Regressive Tests, obtain, and wherein a gets 0.382, b and gets 0.618;

Step S31, utilize Kalman filtering algorithm to merging coordinate (x ₀, y ₀) carry out state estimation, obtain the final elements of a fix (x, y) of vehicle;

Step 4, the anchor node check and correction:

Step S41, utilize and be arranged on the camera on road at interval of 5km, at vehicle, through out-of-date video camera, catches vehicle image f (x, y);

Step S42, carry out gaussian filtering and dodging to vehicle image, the image after being processed;

Step S43, the image obtained after step S42 is processed, it is poor that the background image of taking at same position through out-of-date camera with vehicles failed is done, and then to doing after poor, the image obtained carries out histogram modification and erosion operation is processed;

Step S44, the image obtained after step S43 is processed carries out connected region scanning, obtains the connected region of vehicle image, the position in the middle of image using the minimum boundary rectangle center of connected region as vehicle;

Step S45, the position by vehicle in the middle of image is converted to the actual coordinate of vehicle in the middle of reality, and the elements of a fix (x, y) that this actual coordinate is obtained step 3 are replaced, to eliminate the cumulative errors of the elements of a fix (x, y).

2. the device for the vehicle high-precision locating method of realizing under GPS as claimed in claim 1 blind area merging multi-source information, comprise the accelerometer and the gyroscope that are arranged on vehicle, it is characterized in that, this device also comprises central processing module and is arranged on vehicle and camera lens is parallel to first camera on road surface, be connected with on central processing module and merge locating module and Zigbee wireless receiving module, described accelerometer is connected with the fusion locating module by INS localization process module with gyroscope, the first camera mates locating module by road surface and is connected with the fusion locating module, be provided with portal frame every 5km on the road of Vehicle Driving Cycle, second camera and RFID receiver are installed on portal frame, be connected with the anchor node correcting module on second camera and RFID receiver, be connected with the Zigbee wireless sending module with Zigbee wireless receiving module radio communication on the anchor node correcting module, each parts are achieved as follows respectively function:

Accelerometer adopts YC-A150S-M type acceleration transducer Real-time Obtaining vehicle acceleration information;

Gyroscope adopts CMR3100-D01 type angular-rate sensor Real-time Obtaining vehicle angular velocity information;

Vehicle acceleration information and angular velocity information that INS localization process module is obtained according to accelerometer and gyroscope connect by victory the INS elements of a fix that matrix algorithms obtains vehicle;

The first camera adopts the image on road surface in XC-103c Sony CCD type video frequency pick-up head Real-time Collection Vehicle Driving Cycle process;

Road surface coupling localization process module is processed by the pavement image to the first camera collection, obtains the road surface coupling elements of a fix of vehicle with respect to initial coordinate;

Merge locating module and calculate the fusion coordinate according to the INS elements of a fix and the road surface coupling elements of a fix, and utilize Kalman filtering algorithm to carry out state estimation to merging coordinate, obtain the final elements of a fix of vehicle;

Second camera adopts XC-103c Sony CCD camera, when the vehicle process is installed the portal frame below of second camera, and the image of collection vehicle;

The RFID receiver coordinates with the IC-card on vehicle, and during vehicle process portal frame, the RFID receiver triggers second camera collection vehicle image;

The anchor node correcting module is processed the image of second camera collection, calculates the position of vehicle in the middle of image, and is converted to the coordinate in reality;

The final elements of a fix of car two that central processing module utilizes the coordinate in car two reality that obtain in the anchor node correcting module to calculate the fusion locating module are revised, to eliminate cumulative errors.

Images