CN100580371C - Building single image three-dimensional measurement method based on space right-angle relationship - Google Patents

Abstract

The invention discloses a building single image three-dimensional measuring method based on a space right angle relation. The steps are following: shooting the digital photos of the building; shooting a building corner point for the building while shooting; and three integral wall corner lines which are mutually vertical from the building corner point; two lines are parallel in the three integral wall corner lines, one line is the vertical line; except the building corner point, the three integral wall corner lines have a end point; the three integral wall corner lines are called image control point; measuring the image coordinate of the building corner point and the three integral wall corner lines; converting the coordinate of the building corner point and the three integral wall corner lines to the planar coordinate; building the object coordinate; confirming the edge of the measured wall corner line corresponding to the object coordinate; estimating the object coordinate of the wall corner edge and the corresponding image control point; calculating the known object coordinate and the shooting position and the shooting focus distance according to the object coordinate of the image control point, estimated object coordinate and the planar coordinate.

Description

Building single image three-dimensional measurement method based on space right-angle relationship

Technical field

The present invention proposes the space right-angle relationship that utilizes buildings and carry out the method that single image three-dimensional is measured, be widely used in computer vision, image measurement and digital urban construction field.

Background technology

From the nineties in 20th century so far, a lot of scholars have done research to the three-dimensional measurement that utilizes non-metric camera single width image to carry out buildings, and have obtained suitable progress.But this technology does not also obtain real practicality (Liu Yawen so far in the foundation in digital city, Zhang Zuxun, Zhang Jianqing, Wu Jun. utilize map and single image to carry out the new way [J] of buildings three-dimensional reconstruction. Wuhan University's journal information science version, in February, 2005,146～149).

" utilizing digital camera to carry out the meticulous three-dimensional reconstruction of house property measurement and buildings " ([doctorate paper] Wuhan: Wuhan University, 2005) and " acquisition methods of image direction element in the building scene three-dimensional reconstruction " (Zhang Zuxun, Wu Jun, Zhang Jianqing. Wuhan University's journal information science version, 2003.6,28 (3)) proposed to utilize non-metric camera list image to carry out the three-dimensional reconstruction of buildings, be divided into following step in theory: 1. the inside and outside element of orientation of image determines (f, x0, y0, XS, YS, ZS, φ, ω, κ); 2. how much CSG of application build entity (constructive solid geometry) theory is rebuild the three-dimensional model of buildings.Corresponding " vanishing point " determined the inside and outside element of orientation of image on the intersection point (∞ point) mostly its basic theories is to utilize three groups of mutually perpendicular parallel lines in space of buildings and the image, and generally one " angle point " of selected buildings is the initial point of object coordinates system, and three groups of orthogonal straight lines on this buildings are respectively X, Y, the Z axle of object coordinates system.At many uneven buildingss, utilize direct linear transformation, single image space resection to determine the initial value of image direction element; Utilize the straight line (the particularly pedal line of buildings) of whole buildingss to determine the exact value of image direction element then, to reach purpose to the modelings simultaneously of many solitary buildings.

As seen, present research mainly utilizes the parallel lines of buildings on three directions to calculate the element of orientation of image, realizes the single image three-dimensional modeling of buildings.Yet a lot of buildingss can't provide the parallel lines of complete three directions, make this method that significant limitation be arranged.

The primary content of city dimensional Modeling Technology is a large amount of basis, city 3-D data collections, and image data comprises planimetric position, ground elevation, height and the atural object texture of ground object target.Present 3-D data collection method is that the digital terrain model, the digital line that utilize photogrammetric and remote sensing technology to obtain the city are drawn map (Digital Line Graphic, abbreviation DLG), texture, the depth of building data of utilizing digital camera to obtain atural object are on the spot set up digital city model.

The depth of building data obtained a lot of methods, as: three-dimensional laser scanner collection, the photogrammetric collection of digital three-dimensional, total powerstation or hand-held stadimeter collection in worksite etc., these methods have the defective that efficient is low, cost is high, the cycle is long.Commercial digital city 3 d modeling software all is to calculate according to the buildings number of plies at present, and its defective is clearly, because altitude information is similar to, does not accomplish city three-dimensional model truly.

Summary of the invention

Technical matters to be solved by this invention is, overcomes the defective that prior art exists, and a kind of building single image three-dimensional measurement method based on space right-angle relationship is provided.

The present invention is based on the building single image three-dimensional measurement method of space right-angle relationship, its step is as follows:

The first step: the digital photograph of taking buildings; Require to absorb a room angle point of buildings during photography, and three mutually perpendicular completed wall linea angulatas that set out by this room angle point; Article three, two in the wall corner line are that horizontal line, one are perpendicular line; Except that the room angle point, respectively there is an end points in three wall corner lines, and three end points are called photo control point;

Second step: the image coordinate of measurement room angle point and three photo control points on digital photograph (u, v); Image coordinate system C ₀-uv is that the upper left corner with digital photograph is that initial point, level are that X-axis positive dirction, vertical downward direction are the Y-axis positive dirction to right;

The 3rd step: (u, v) converting to the digital photograph center is the right-handed system photo coordinate system C of initial point with the image coordinate of room angle point and photo control point ₁-xy coordinate (x, y), transformational relation is as follows:

$\left\{\begin{array}{c}x=u-\frac{\mathrm{width}}{2}\\ y=\frac{\mathrm{heigth}}{2}-v\end{array}\right.$

In the formula:

(u, v)---image coordinate system C ₀-uv coordinate, unit are pixel.

(x, y)---photo coordinate system C ₁-xy coordinate, unit are pixel.

The width of width---digital photograph, unit are pixel.

The height of heigth---digital photograph, unit are pixel.

The 4th step: set up the object coordinates system of buildings with space right-angle relationship, the initial point of object coordinates system is the room angle point, and three coordinate axis correspond respectively to by room angle point and three three wall corner lines that photo control point is determined, as Fig. 2;

The object coordinates system that the present invention relates to is meant that the space right-angle relationship of buildings is the rectangular space coordinate of the system of axis.In Fig. 2, be initial point, be X-axis, be Y-axis, be the Z axle with building object height direction OC with buildings Width OB with buildings length direction OA with room angle point O.

The 5th step: the object coordinates of determining its corresponding photo control point according to the length of side that predicts length of side wall corner line of buildings; Estimate all the other wall corner line length of sides, determine the estimation object coordinates of its corresponding photo control point;

The 6th step:, calculate the photography focal length of unknown point object coordinates value and camera positions and digital photograph according to object coordinates, estimation object coordinates and the picture planimetric coordinates of photo control point.Wherein camera positions are meant the coordinate of photography moment object lens center in object coordinates system;

According to known conditions difference in three photo control points, be divided into two kinds of computation schemas: " two push away one " pattern and " one pushes away two " pattern;

" two push away one " pattern is meant that the object coordinates that two points are arranged in three photo control points is a given value, calculates the computation schema of the object coordinates of another point;

" one pushes away two " pattern is meant that the coordinate that a point is arranged in three photo control points is a given value, calculates the computation schema of the object coordinates of two points in addition;

As Fig. 2, if two wall corner line length of OA and OB are known, the length of calculating OC is called " two push away one " pattern; If wall corner line OA is known, the length of calculating OB and OC is called " one pushes away two " pattern.

Concrete computing method are as follows:

S is the photo centre position, and any two i, j points are corresponding to photo control point I, the J point of extraterrestrial target on the digital photograph.The cosine value of the angle that i, j point and S point constitute is:

$\cos \left(\mathrm{iSj}\right)=\frac{x_i{x}_j+y_i{y}_j+{\mathrm{<figure-callout\; id="2"\; label="f"\; filenames="C2008100989110014H1.png"\; state="\{\{state\}\}">f</figure-callout>}}^2}{\mathrm{Si}\&CenterDot;\mathrm{Sj}}=\frac{x_i{x}_j+y_i{y}_j+f^2}{\sqrt{x_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="C2008100989110014H1.png"\; state="\{\{state\}\}">i</figure-callout>}}^2+y_i^2+f^2}\sqrt{x_{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="C2008100989110014H1.png"\; state="\{\{state\}\}">j</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="y\; j"\; filenames="C2008100989110014H1.png"\; state="\{\{state\}\}">y</figure-callout>}}_j^{}+{\mathrm{<figure-callout\; id="2"\; label="f"\; filenames="C2008100989110014H1.png"\; state="\{\{state\}\}">f</figure-callout>}}^2}}$

Wherein:

(x _i, y _i)---the picture planimetric coordinates of picture point i;

(x _j, y _j)---the picture planimetric coordinates of picture point j;

F---focal length

Equally, the included angle cosine value that is made of photo control point I, J and S point is:

$\cos \left(\mathrm{ISJ}\right)=\frac{(X_S-X_I)(X_S-X_J)+(Y_S-Y_I)(Y_S-Y_J)+(Z_S-Z_I)(Z_S-Z_J)}{\mathrm{SI}\&CenterDot;\mathrm{SJ}}$

Wherein:

(X _S, Y _S, Z _S)---the object coordinates of S;

(X _I, Y _I, Z _I)---the object coordinates of photo control point I;

(X _J, Y _J, Z _J)---the object coordinates of photo control point J;

SI ²＝(X _S-X _I) ²+(Y _S-Y _I) ²+(Z _S-Z _I) ²

SJ ²＝(X _S-X _J) ²+(Y _S-Y _J) ²+(Z _S-Z _J) ²

Set up equation:

F＝cos(ISJ)-cos(iSj)＝0

Between any two photo control points, all set up such equation between room angle point and the photo control point, can set up six equations between four points;

" two push away one " pattern

Known photo control point A, B object coordinates, and photo control point C object coordinates (0,0, Z _C) the unknown, then there is X _S, Y _S, Z _S, f, Z _CFive unknown numbers, the error equation general formula is:

$v=\frac{\&PartialD;F}{{\&PartialD;X}_S}\&CenterDot;{\mathrm{\&Delta;X}}_S+\frac{\&PartialD;F}{{\&PartialD;Y}_S}\&CenterDot;{\mathrm{\&Delta;Y}}_S+\frac{\&PartialD;F}{{\&PartialD;Z}_S}\&CenterDot;{\mathrm{\&Delta;Z}}_S+\frac{\&PartialD;F}{\&PartialD;f}\&CenterDot;\mathrm{\&Delta;f}+\frac{\&PartialD;F}{{\&PartialD;Z}_C}\&CenterDot;{\mathrm{\&Delta;Z}}_C-l$

Be expressed in matrix as:

$\underset{6\&times;1}{V}=\underset{6\&times;5}{B}\underset{5\&times;1}{X}-\underset{6\&times;1}{L}$

" one pushes away two " pattern

Known photo control point A object coordinates, and photo control point B, C object coordinates (0, Y _B, 0), (0,0, Z _C) the unknown, then there is X _S, Y _S, Z _S, f, Y _B, Z _CSix unknown numbers, the error equation general formula is:

$v=\frac{\&PartialD;F}{{\&PartialD;X}_S}\&CenterDot;{\mathrm{\&Delta;X}}_S+\frac{\&PartialD;F}{{\&PartialD;Y}_S}\&CenterDot;{\mathrm{\&Delta;Y}}_S+\frac{\&PartialD;F}{{\&PartialD;Z}_S}\&CenterDot;{\mathrm{\&Delta;Z}}_S+\frac{\&PartialD;F}{\&PartialD;f}\&CenterDot;\mathrm{\&Delta;f}+\frac{\&PartialD;F}{{\&PartialD;Y}_B}\&CenterDot;{\mathrm{\&Delta;Y}}_B+\frac{\&PartialD;F}{{\&PartialD;Z}_C}\&CenterDot;\mathrm{\&Delta;}{Z}_C-l$

Be expressed in matrix as:

$\underset{6\&times;1}{V}=\underset{6\&times;6}{B}\underset{6\&times;1}{X}-\underset{6\&times;1}{L}$

Constant term is:

l＝cos(iSj)-cos(ISJ)

Wherein: the desirable o of i, j, a, b, c, i ≠ j

I, J are and the corresponding O of i, j, A, B, C, I ≠ J

Solve an equation with the method for indirect adjustment and to calculate the object coordinates of camera positions, objective focal length and unknown point.

The inventive method is utilized the space right-angle relationship of buildings, passes through building single image, just can obtain the buildings three-dimensional data, can reduce outdoor surveying work amount,, can try to achieve with the inventive method for the data that can't survey or measure difficulty owing to the outdoor condition reason of buildings.Simultaneously can also learn camera positions and objective focal length, can be applicable to photogrammetric and field such as computer vision.

Description of drawings

Fig. 1 is image coordinate system and conversion synoptic diagram thereof;

Fig. 2 is a synoptic diagram for the object coordinates of space right-angle relationship;

Fig. 3 is photo centre, photo control point and photo control point picture point spatial relationship synoptic diagram;

Embodiment

Below in conjunction with specific embodiments and the drawings, the inventive method is described in further detail.

Embodiment: as shown in Figure 2,

(1) given data: the buildings among Fig. 2 is carried out field survey, and the result of measurement is: the length OA=54.73 rice of buildings, width OB=17.67 rice, four layers height value OC=13.13 rice.In object coordinates system, comprise four points (room angle point O, photo control point A, B, C) altogether, its three-dimensional coordinate data such as table 1.

The observation data table of table 1 space right-angle relationship

(2) image measurement data

Carry out image measurement with Visual C#.NET2003 platform, the O among Fig. 2, A, B, C have been carried out 10 times measured, with O, the A, B, the C image coordinate conversion imaging planimetric coordinates that measure, O, A, B, C see Table 2 as planimetric coordinates.

Table 2 picpointed coordinate measurement result (unit: pixel)

(3) " two push away one " result of calculation

Utilize this buildings of " two push away one " computation model to calculate, with the length OA=54.73 rice of buildings, width OB=17.67 rice is known number, with four layers height OC as unknown-value.

The object coordinates of known point A, B is (54.73,0,0), (0,17.67,0);

The estimation object coordinates of unknown point C is (0,0,12), and wherein 12 for estimating numerical value.

See Table 3 with " two push away one " computation model result calculated.

Table 3 " two push away one " computation model result of calculation

Calculate content	Xs (rice)	Ys (rice)	Zs (rice)	F (pixel)	Z C(rice)
						Result of calculation	-21.23	-19.92	-1.42	2891.9	13.147

The C point coordinate that calculates is compared with actual measured results, and its difference is 0.012 meter.As seen, " two push away one " computation model all has higher computational accuracy.

(4) " one pushes away two " result of calculation

Length OA=54.73 rice with buildings is known number, with the height OC of buildings width OB and four layers as unknown-value.

The object coordinates of known point A is (54.73,0,0);

The estimation object coordinates of unknown point B, C is (0,17,0), (0,0,12).Wherein 17,12 is estimated value.

Result of calculation sees Table 4.

Table 4 " one pushes away two " computation model result of calculation

Calculate content	Xs (rice)	Ys (rice)	Zs (rice)	F (pixel)	Y B(rice)	Z C(rice)
							Result of calculation	-20.52	-18.99	-1.37	2872.4	17.03	12.69

The B, the C point coordinate that calculate are compared with actual measured results, and its difference is respectively 0.64 meter and 0.445 meter.As seen, " one pushes away two " computation model can't be eliminated the systematic error in resolving because the calculating unknown number is more, causes computational accuracy lower.

Claims (1)

1, a kind of building single image three-dimensional measurement method based on space right-angle relationship, its step is as follows:

The first step: the digital photograph of taking buildings; Require to absorb a room angle point of buildings during photography, and three mutually perpendicular completed wall linea angulatas that set out by this room angle point; Article three, two in the wall corner line are that horizontal line, one are perpendicular line; Except that the room angle point, respectively there is an end points in three wall corner lines, and three end points are called photo control point;

Second step: the image coordinate of measurement room angle point and three photo control points on digital photograph (u, v); Image coordinate system C ₀-uv is that the upper left corner with digital photograph is that initial point, level are that X-axis positive dirction, vertical downward direction are the Y-axis positive dirction to right;

The 3rd step: (u, v) converting to the digital photograph center is the right-handed system photo coordinate system C of initial point with the image coordinate of room angle point and photo control point ₁-xy coordinate (x, y), transformational relation is as follows:

$\left\{\begin{array}{c}x=u-\frac{\mathrm{width}}{2}\\ y=\frac{\mathrm{heigth}}{2}-v\end{array}\right.$

In the formula:

(u, v)---image coordinate system C ₀-uv coordinate, unit are pixel;

(x, y)---photo coordinate system C ₁-xy coordinate, unit are pixel;

The width of width---digital photograph, unit are pixel;

The height of heigth---digital photograph, unit are pixel;

The 4th step: set up the object coordinates system of buildings with space right-angle relationship, the initial point of object coordinates system is the room angle point, and three coordinate axis correspond respectively to by room angle point and three three wall corner lines that photo control point is determined;

The 5th step: the object coordinates of determining its corresponding photo control point according to the length of side that predicts length of side wall corner line of buildings; Estimate all the other wall corner line length of sides, determine the estimation object coordinates of its corresponding photo control point;

The 6th step: according to object coordinates, estimation object coordinates and the picture planimetric coordinates of photo control point, calculate the photography focal length of unknown point object coordinates value and camera positions and digital photograph, wherein camera positions are meant the coordinate of photography moment object lens center in object coordinates system;

According to known conditions difference in three photo control points, the photography focal length that calculates unknown point object coordinates value and camera positions and digital photograph is divided into two kinds of computation schemas: " two push away one " pattern and " one pushes away two " pattern;

Specifically be calculated as follows:

S is the photo centre position, and any two some i, j are corresponding to photo control point I, the J point of extraterrestrial target on the digital photograph; The cosine value of the angle that i, j point and S point constitute is:

$\cos \left(\mathrm{iSj}\right)=\frac{x_i{x}_j+y_i{y}_j+f^2}{\mathrm{Si}\&CenterDot;\mathrm{Sj}}=\frac{x_i{x}_j+y_i{y}_j+f^2}{\sqrt{x_i^2+y_i^2+f^2}\sqrt{x_j^2+y_j^2+f^2}}$

Wherein:

(x _i, y _i)---the picture planimetric coordinates of picture point i;

(x _j, y _j)---the picture planimetric coordinates of picture point j;

F---focal length;

Equally, the included angle cosine value that is made of photo control point I, J and S point is:

$\cos \left(\mathrm{ISJ}\right)=\frac{(X_S-X_I)(X_S-X_J)+(Y_S-Y_I)(Y_S-Y_J)+(Z_S-Z_I)(Z_S-Z_J)}{\mathrm{SI}\&CenterDot;\mathrm{SJ}}$

Wherein:

(X _S, Y _S, Z _S)---the object coordinates of S;

(X _I, Y _I, Z _I)---the object coordinates of photo control point I;

(X _J, Y _J, Z _J)---the object coordinates of photo control point J;

SI ²＝(X _S-X _I) ²+(Y _S-Y _I) ²+(Z _S-Z _I) ²

SJ ²＝(X _S-X _J) ²+(Y _S-Y _J) ²+(Z _S-Z _J) ²

Set up equation:

F＝cos(ISJ)-cos(iSj)＝0

Between any two photo control points, all set up an above-mentioned equation between room angle point and the photo control point, can set up six equations between four points;

" two push away one " pattern: if known photo control point A, B object coordinates, and photo control point C object coordinates (0,0, Z _C) the unknown, then there is X _S, Y _S, Z _S, f, Z _CFive unknown numbers, the error equation general formula is:

$v=\frac{\&PartialD;F}{\&PartialD;X_S}\&CenterDot;\mathrm{\&Delta;}{X}_S+\frac{\&PartialD;F}{\&PartialD;Y_S}\&CenterDot;\mathrm{\&Delta;}{Y}_S+\frac{\&PartialD;F}{\&PartialD;Z_S}\&CenterDot;\mathrm{\&Delta;}{Z}_S+\frac{\&PartialD;F}{\&PartialD;f}\&CenterDot;\mathrm{\&Delta;f}+\frac{\&PartialD;F}{\&PartialD;Z_C}\&CenterDot;\mathrm{\&Delta;}{Z}_C-l$

Be expressed in matrix as:

$\underset{6\&times;1}{V}=\underset{6\&times;5}{B}\underset{5\&times;1}{X}-\underset{6\&times;1}{L}$

" one pushes away two " pattern: if known photo control point A object coordinates, and photo control point B, C object coordinates (0, Y _B, 0), (0,0, Z _C) the unknown, then there is X _S, Y _S, Z _S, f, Y _B, Z _CSix unknown numbers, error mode general formula is:

$v=\frac{\&PartialD;F}{\&PartialD;X_S}\&CenterDot;\mathrm{\&Delta;}{X}_S+\frac{\&PartialD;F}{\&PartialD;Y_S}\&CenterDot;\mathrm{\&Delta;}{Y}_S+\frac{\&PartialD;F}{\&PartialD;Z_S}\&CenterDot;\mathrm{\&Delta;}{Z}_S+\frac{\&PartialD;F}{\&PartialD;f}\&CenterDot;\mathrm{\&Delta;f}+\frac{\&PartialD;F}{\&PartialD;Y_B}\&CenterDot;\mathrm{\&Delta;}{Y}_B+\frac{\&PartialD;F}{\&PartialD;Z_C}\&CenterDot;\mathrm{\&Delta;}{Z}_C-l$

Be expressed in matrix as:

$\underset{6\&times;1}{V}=\underset{6\&times;6}{B}\underset{6\&times;1}{X}-\underset{6\&times;1}{L}$

Constant term is:

l＝cos(iSj)-cos(ISJ)

Wherein: i, j=o, a, b, c, i ≠ j;

I, J are and the corresponding O of i, j, A, B, C, I ≠ J;

Solve an equation with the method for indirect adjustment and to calculate the object coordinates of camera positions, objective focal length and unknown point.

Images