CN103345630A - Traffic sign positioning method based on spherical panoramic video - Google Patents

Abstract

The invention discloses a traffic sign positioning method based on spherical panoramic video. Firstly, an Adaboost classifier of traffic signs is trained by Haar characteristics, and the traffic signs are detected on the basis; on the basis of detection results, centroid positions of detection regions are calculated, and the centroid positions are used as the positions of the traffic signs; two spherical live-action images containing the same traffic sign are chosen, and on the basis of the position mapping relation of the two spherical live-action images, the position of the traffic sign in three-dimensional space is determined; the coordinate of the spatial position is converted into a geodetic spatial rectangular coordinate, then the geodetic spatial rectangular coordinate is converted into a geodetic coordinate, longitude and latitude information in the geodetic coordinate is extracted, and positioning of the traffic signs in a two-dimensional map is achieved. By means of the traffic sign positioning method, users only need to drive on a road once, all surrounding scenes can be shot, therefore, positioning of all the traffic signs in the scenes can be performed, longitude and latitude coordinates corresponding to the signs on the live-action images can be calculated, and positioning is performed. Time and labor are saved, the positioning algorithm is high in efficiency, and real-time performance is good.

Description

A kind of traffic sign localization method based on the spherical panorama video

Technical field

The present invention relates to the field of locating technology of traffic sign, be specifically related to a kind of traffic sign localization method based on the spherical panorama video.

Background technology

Traffic sign collection in the map vector has very important meaning to means of transportation management, Vehicular intelligent navigation etc.Social construction now is fast, traffic element is many, renewal is fast, and studying traffic sign collection fast and accurately is an important topic.The key of traffic sign information acquisition is the location of traffic sign, and on the basis of carrying out the traffic sign location accurately and rapidly, the positional information according to traffic sign is plotted in it on map vector.

Traditional artificial mapping method efficient is low, cost is high, can not satisfy the demand; Be subjected to the influence of pattern accuracy based on the localization method of aviation, satellite image, identification difficulty automatically; Based on the localization method of ordinary video then owing to be subjected to single camera to take the restriction at visual angle, can't in a gatherer process, carry out the location of all signs of road both sides, yet need carry out the judgement of video angular field of view during repeated acquisition, can cause repeated acquisition or leak the generation of gathering phenomenon.

The spherical panorama video is to be to constitute element with the panorama sketch, panorama sketch is mapped on the sphere model, formation be to observe viewpoint with the centre of sphere, be viewing area with the sphere, can reach the video of 360 degree observation visual angles.Each frame in the spherical panorama video is called a sphere outdoor scene.On the one hand, along with the development of panorama collecting device, can obtain spherical panorama video (the panorama collecting device can carry out taking rapidly, continuously, thereby obtains road scene information on the way) easily among the high speed of vehicle is advanced; On the other hand, the spherical panorama video has comprised viewpoint position 360 ° scene information on every side, only need once gather the location that just can carry out all signs of road both sides.

Therefore, the present invention is data source with the spherical panorama video that obtains based on the panorama collecting device, has proposed a kind of traffic sign localization method based on the spherical panorama video.

Summary of the invention

The present invention is the orientation problem for the transport solution sign, utilize the spherical panorama video to have and gather easy, the wide characteristics of angular field of view, a kind of traffic sign localization method based on the spherical panorama video has been proposed, the location fast and accurately that can realize traffic sign by this method.

Technical solution problem of the present invention adopts following technical scheme:

A kind of traffic sign localization method based on the spherical panorama video, described localization method may further comprise the steps:

(1) carries out the detection of traffic sign based on the spherical panorama video

1) extraction is no less than 3000 images that contain traffic sign as positive samples from the spherical panorama video, extracts to be no less than 9000 images that do not contain traffic sign as negative sample;

2) positive and negative samples is handled, utilized the Adaboost sorter of Haar features training traffic sign then, utilize the sorter that generates to carry out the detection of traffic sign;

3) if detect traffic sign, surveyed area is mapped on the spherical panorama figure;

(2) latitude and longitude coordinates of calculating traffic sign is to position

1) looking traffic sign is the isodensity object, and with the coordinate at its centre of form place coordinate as traffic sign, the position of calculating the surveyed area centre of form among the spherical panorama figure is mapped to it in sphere outdoor scene, calculates the spherical polar coordinates of its mapping point on the sphere outdoor scene;

2) choose the sphere outdoor scene that 2 frames comprise this traffic sign, its centre of sphere is respectively O ₁, O ₂, the gps data that obtains when extracting the sphere shoot on location is as the coordinate of centre of sphere position, and the elevation information of supposing traffic sign centre of form place is H, and then the terrestrial coordinate at virtual reality centre of sphere place is known; According to the transformational relation between earth coordinates and the earth rectangular coordinate system in space, be the earth rectangular space coordinate with this coordinate conversion;

3) with O ₁Be true origin, direction is set up rectangular coordinate system in space O for the X-axis positive dirction ₁XYZ supposes A ₁, A ₂Be respectively traffic sign with O ₁, O ₂Be the position in the sphere outdoor scene in the center of circle, O ₁A ₁With O ₂A ₂Meet at 1 A, then the A point is the position of traffic sign in the three dimensions; Based on the above-mentioned position mapping relations between 2 frame virtual realities, can calculate the A point at rectangular coordinate system O ₁The volume coordinate of XYZ;

4) volume coordinate that A is ordered is converted into the earth rectangular space coordinate, and then is converted into terrestrial coordinate, keeps the latitude and longitude information in the terrestrial coordinate, to position at two-dimensional map.

Compared with the prior art, beneficial effect of the present invention is embodied in:

(1) aspect Data Source, as data source, carries out the location of traffic sign with the spherical panorama video based on this data source, time saving and energy saving, only need to travel once at road, just can photograph all scenes on every side, thereby carry out the location of all traffic signs in the scene.

(2) on localization method, the specific position mapping relations between the spherical panorama video consecutive frame have been utilized, and the mutual transformational relation between earth coordinates and the earth rectangular coordinate system in space, calculate the corresponding latitude and longitude coordinates of sign on the sphere outdoor scene, thereby position, location algorithm efficient height, real-time are good.

Description of drawings

Fig. 1 is for carrying out the technology path synoptic diagram of traffic sign location based on the spherical panorama video.

Fig. 2 is the mapping relations synoptic diagram between spherical panorama figure and the sphere outdoor scene.

Fig. 3 is the position mapping relations synoptic diagram between the two frame sphere outdoor scenes.

Fig. 4 is O, O ₁, concern synoptic diagram between 3 of the P.

Embodiment

The present invention adopts the panorama collecting device to carry out obtaining of spherical panorama video, simultaneously, obtains the gps data of each position, and is associated with corresponding sphere outdoor scene.

As shown in Figure 1, this method is mainly carried out according to following two steps:

Step 1: carry out the detection of traffic sign in the spherical panorama video.

1) extraction is no less than 3000 images that contain traffic sign as positive samples from the spherical panorama video, extracts to be no less than 9000 images that do not contain traffic sign as negative sample, and positive and negative samples is carried out normalization respectively.

2) based on this sample, utilize the AdaBoost sorter of Haar features training traffic sign, the sorter that generates is used for the detection of traffic sign.

3) if detect traffic sign, surveyed area is mapped on the spherical panorama figure.

Step 2: calculate the latitude and longitude coordinates of traffic sign, in two-dimensional map, to position.

1) looking traffic sign is the isodensity object, and with the coordinate at its centre of form place coordinate as traffic sign, therefore, the first step need be calculated the surveyed area centre of form and be mapped to spherical polar coordinates in the sphere outdoor scene.

On spherical panorama figure, the coordinate Calculation formula of surveyed area centre of form S is as follows:

$\left\{\begin{array}{c}\stackrel{\‾}{X}=\frac{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{X}_i\mathrm{Gray}(X_i,Y_i)}{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{Gray}(X_i,Y_i)}\\ \stackrel{\‾}{Y}=\frac{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{Y}_i\mathrm{Gray}(X_i,Y_i)}{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{Gray}(X_i,Y_i)}\end{array}\right.---\left(1\right)$

(X wherein _i, Y _i) be the pixel coordinate of each pixel of surveyed area, N is the summation of surveyed area pixel, Gray (X _i, Y _i) be gray values of pixel points.

Based on formula (1), calculate the coordinate of the surveyed area centre of form.Suppose that the spherical panorama figure that adopts is of a size of panoWidth * panoHeight, calculates the spherical polar coordinates that centre of form S is mapped to the mapping point P on the sphere outdoor scene

(be true origin with the centre of sphere, suppose that r is 1, as shown in Figure 2), computing formula is:

2) choose the sphere outdoor scene that 2 frames comprise this traffic sign, its centre of sphere is O ₁And O ₂, the gps data that obtains when extracting the sphere shoot on location is designated as (B respectively as the coordinate of centre of sphere position ₁, L ₁) and (B ₂, L ₂).The elevation information of supposing traffic sign centre of form place is H, then O ₁, O ₂Terrestrial coordinate be (B ₁, L ₁, H) with (B ₂, L ₂, H).According to the transformational relation between earth coordinates and the earth rectangular coordinate system in space:

$\left(\begin{array}{c}X\\ Y\\ Z\end{array}\right)=\left(\begin{array}{c}N\cos B\cos L\\ N\cos B\sin L\\ N(1-e^2)\sin B\end{array}\right),---\left(3\right)$

Wherein, $N=\frac{a}{\sqrt{1-e^2{\sin }^{2}B}},$ $e^2=\frac{a^2-b^2}{a^2}$

Can be with O ₁, O ₂The terrestrial coordinate at place is converted into the earth rectangular space coordinate (X ₁, Y ₁, Z ₁) and (X ₂, Y ₂, Z ₂).

3) suppose A ₁, A ₂Be respectively traffic sign with O ₁, O ₂Be the position in the sphere outdoor scene in the center of circle, O ₁A ₁With O ₂A ₂Meet at 1 A, then the A point is the position (as shown in Figure 3) of traffic sign in the three dimensions.The above-mentioned relation that exists between two place's sphere outdoor scenes is called the position mapping relations.Be true origin with O1,

Direction is set up rectangular coordinate system in space O1XYZ for the X-axis positive dirction, calculates the coordinate (x that A is ordered under this rectangular coordinate system ₁, y ₁, z ₁), as follows:

$\left\{\begin{array}{c}{x}_1=-H_1\cos \frac{2\mathrm{\π}{X}_{\mathrm{pixel}}}{\mathrm{panoWidth}}\tan \frac{\mathrm{\π}{Y}_{\mathrm{pixel}}}{\mathrm{panoHeight}}\\ y_1=-H_1\sin \frac{2\mathrm{\π}{X}_{\mathrm{pixel}}}{\mathrm{panoWidth}}\tan \frac{\mathrm{\π}{Y}_{\mathrm{pixel}}}{\mathrm{panoHeight}}\\ z_1=-H_1\\ \end{array}\right.,$

Wherein, (X _Pixel, Y _Pixel) being the pixel coordinate of A1 point correspondence on spherical panorama figure, the height on collecting device centre distance ground when H1 is collection can be by measuring acquisition.

4) true origin of supposing the earth rectangular coordinate system in space is O, then

(as shown in Figure 4), namely can calculate the coordinate of A point under the earth rectangular coordinate system in space.This coordinate is converted into terrestrial coordinate again, does not consider elevation information, only keep longitude and the latitude information of terrestrial coordinate, on the map that at last latitude and longitude information is added to, the location that namely can realize traffic sign.

Claims (1)

1. the traffic sign localization method based on the spherical panorama video is characterized in that, may further comprise the steps:

(1) carries out the detection of traffic sign based on the spherical panorama video

1) extraction is no less than 3000 images that contain traffic sign as positive samples from the spherical panorama video, extracts to be no less than 9000 images that do not contain traffic sign as negative sample;

2) positive and negative samples is handled, utilized the Adaboost sorter of Haar features training traffic sign then, utilize the sorter that generates to carry out the detection of traffic sign;

3) if detect traffic sign, surveyed area is mapped on the spherical panorama figure;

(2) latitude and longitude coordinates of calculating traffic sign is to position

1) looking traffic sign is the isodensity object, and with the coordinate at its centre of form place coordinate as traffic sign, the position of calculating the surveyed area centre of form among the spherical panorama figure is mapped to it in sphere outdoor scene, calculates the spherical polar coordinates of its mapping point on the sphere outdoor scene;

2) choose the sphere outdoor scene that 2 frames comprise this traffic sign, its centre of sphere is respectively O ₁ , O ₂ , the gps data that obtains when extracting the sphere shoot on location supposes that as the coordinate of the centre of sphere position elevation information at traffic sign centre of form place is H, then the terrestrial coordinate at virtual reality centre of sphere place is known; According to the transformational relation between earth coordinates and the earth rectangular coordinate system in space, be the earth rectangular space coordinate with this coordinate conversion;

3) with O ₁ Be true origin, direction is XThe axle positive dirction is set up rectangular coordinate system in space O ₁ XYZ, suppose A ₁ , A ₂ Be respectively traffic sign with O ₁ , O ₂ Be the position in the sphere outdoor scene in the center of circle, O ₁ A ₁ With O ₂ A ₂ Meet at a bit A, then APoint is the position of traffic sign in the three dimensions; Based on the above-mentioned position mapping relations between 2 frame virtual realities, can calculate the A point in rectangular coordinate system O ₁ XYZVolume coordinate;

4) volume coordinate that A is ordered is converted into the earth rectangular space coordinate, and then is converted into terrestrial coordinate, keeps the latitude and longitude information in the terrestrial coordinate, to position at two-dimensional map.

Images