CN104331882B - Method for measuring speed of aircraft - Google Patents

Abstract

The invention provides a method for measuring the speed of an aircraft. The speed of the aircraft is estimated by the aid of a sequential image shot by a camera. The method includes the steps: firstly, respectively correcting images in front time and rear time from current positions to positions with zero-degree attitude angles according to attitude angle information; secondly, automatically registering the two images based on an SIFT (scale invariant feature transform) algorithm and determining an overlap area of the images in the front time and the rear time to obtain a group of optimal matching points; thirdly, calculating the speed of the aircraft according to remote sensor height information and shooting time. The method for measuring the speed of the aircraft has the advantages of independent working performance, high interference resistance, simplicity and convenience in use and the like, and is suitable for future aerial development trend.

Description

A kind of method of measurement airborne vehicle speed

Technical field

The present invention relates to air remote sensing field, in particular to a kind of method of measurement airborne vehicle speed.

Background technology

Camera imaging model：As video camera is fixed on aircraft, therefore, the pose of video camera can be directly used for table Levy the pose of airborne vehicle.Camera imaging model can be described as follows：

Wherein, on the right of equation Section 1 be camera internal reference matrix, f_x=f/dx, f_y=f/dy, is referred to as in x-axis and y-axis Normalization focal length, f is camera focus, and dx and dy represents the size of unit pixel in sensor x-axis and y-axis respectively.c_x And c_yWhat is then represented is optical center, i.e. camera optical axis and the intersection point of the plane of delineation, is usually located at picture centre, therefore its value The half of resolution is taken often.

The pose of video camera represents that by attitude angle that is, Eulerian angles are used for describing rigid body in three-dimensional Euclidean space Orientation.For a referential in three dimensions, the orientation of any coordinate system, can be showed with three Eulerian angles. As shown in figure 1, including xyz- axles and XYZ- coordinate axess, intersecting for xy- planes and XY-plane is called the line of nodes (N), three Eulerian angles (α, β, γ) are respectively defined as：α is the angle of x- axles and the line of nodes；β is the angle of z-axis and Z- axles；γ is the line of nodes With the angle of X- axles.If α=0 °, β=30 °, γ=0 °, i.e. x- axles and X- overlapping of axles, and overlaps with the line of nodes, z-axis and Z- axles Angle be 30 °.

For any referential, the orientation of a rigid body is, according to order, to do the rotation of three Eulerian angles from this referential Then setting.So, the orientation of rigid body can be determined with three basic spin matrixs, i.e., any rotation with regard to rigid body rotation Torque battle array R is composited by three basic spin matrixs：

The content of the invention

Present invention aim at providing a kind of method of measurement airborne vehicle speed, the sequence image shot using video camera is estimated Calculate airborne vehicle speed.

The above-mentioned purpose of the present invention realized by the technical characteristic of independent claims, and dependent claims are selecting else or have The mode of profit develops the technical characteristic of independent claims.

To reach above-mentioned purpose, technical scheme of the present invention is as follows：

A kind of method of measurement airborne vehicle speed, the sequence image estimation airborne vehicle speed shot using video camera, in fact Now comprise the following steps：

Step 1, that the image at two moment of in front and back is corrected to attitude angle from current location respectively according to attitude angle information is equal For 0 position；

Step 2, autoregistration is carried out to aforementioned two width image based on SIFT algorithms, it is determined that the picture registration at two moment in front and back Region, obtains one group of optimal match point；And

Step 3, the speed that airborne vehicle is calculated with reference to remote sensor elevation information and shooting time.

From the above technical solution of the present invention shows that, the beneficial effects of the present invention is：

The present invention proposes the portable aviation speed-measuring method that a kind of service behaviour is independent, precision is higher, survey main at present Fast method has global positioning system (GPS) to test the speed, Doppler radar tests the speed and Pitot-static tube speed-measuring method.Wherein GPS tests the speed use On have larger restriction, and signal is easily disturbed；Doppler radar tests the speed, and volume and weight is big, and use cost is high；Pitot-static tube tests the speed It is easily affected by environment.The present invention program is based on sequence image, and with low cost, independence is strong, interference resistance is strong and uses simple The features such as facilitating, is adapted to following aviation development trend.

Description of the drawings

Schematic diagrams of the Fig. 1 for Eulerian angles.

Fig. 2 is the schematic flow sheet of the method that an embodiment of the present invention measures airborne vehicle speed.

Fig. 3 illustrates for the principle for calculating airborne vehicle speed according to remote sensor elevation information and shooting time in Fig. 2 embodiments Figure.

Specific embodiment

In order to know more about the technology contents of the present invention, especially exemplified by specific embodiment and institute's accompanying drawings are coordinated to be described as follows.

As shown in Fig. 2 preferred embodiment of the invention, a kind of method of measurement airborne vehicle speed, using video camera The sequence image estimation airborne vehicle speed of shooting, its realization are comprised the following steps：

Step 1, that the image at two moment of in front and back is corrected to attitude angle from current location respectively according to attitude angle information is equal For 0 position；

Step 2, autoregistration is carried out to aforementioned two width image based on SIFT algorithms, it is determined that the picture registration at two moment in front and back Region, obtains one group of optimal match point；And

Step 3, the speed that airborne vehicle is calculated with reference to remote sensor elevation information and shooting time.

Below in conjunction with the accompanying drawings shown in 2-3, being embodied as above steps is described in detail.

Step 1, that the image at two moment of in front and back is corrected to attitude angle from current location respectively according to attitude angle information is equal For 0 position

Updating formula is：

In formula：λ is scale factor, and K is intrinsic parameters of the camera matrix, and R is the orthogonal spin matrix of unit.

Specifically, in the present embodiment, it is corrected as steps described below：

Converted images size is determined first；Then the coordinate figure inversion after correction is shifted to into original image coordinate points；Finally Sub-pixel positioning is carried out by quadratic interpolattion, the image after being corrected.

Step 2, autoregistration is carried out to aforementioned two width image based on SIFT algorithms, it is determined that the picture registration at two moment in front and back Region, obtains one group of optimal match point

In the present embodiment, following steps are specifically included：

(1) detection of key point：By gaussian kernel function continuous filtering and down-sampling of the input picture by different scale, shape Into gaussian pyramid image, then again two Gaussian images of adjacent yardstick are subtracted each other and obtains DoG pyramid multiscale space tables Show；

To DoG metric spaces, each point is compared one by one with the point of adjacent yardstick and adjacent position, the local pole for obtaining Value position is key point location and corresponding yardstick；

(2) SIFT description are constructed：Using the gradient direction distribution characteristic of key point neighborhood territory pixel, it is that each key point refers to Determine directioin parameter：Coordinate axess are rotated to into characteristic point direction, to ensure rotational invariance；To any one key point, in its institute Metric space, take the neighborhood of 16 × 16 pixel sizes centered on key point, then this neighborhood be evenly divided into into 4 × 4 Every sub-regions are calculated gradient orientation histogram by sub-regions, and the rectangular histogram is uniformly divided into 8 directions；Then, to 4 × 4 8 histograms of oriented gradients of subregion are sorted successively according to position, constitute the vector of 4 × 4 × 8=128 dimension, and the vector is For SIFT description；

(3) characteristic matching：Measured using similarity determination of the Euclidean distance as key point in two width images, when this distance During less than given threshold, then judge two key points as possible match point；

(4) remove error hiding：To the be possible to matching double points having determined, less tolerance is set, using RANSAC Robust method carries out Geometrical consistency inspection, obtains one group of optimal match point；

(5) affine transformation：Two width images are mapped under same coordinate system and determine overlapping region.

Step 3, the speed that airborne vehicle is calculated with reference to remote sensor elevation information and shooting time

Before and after corrected, the match point of two moment sequence images, was parallel to each other on differing heights and parallel to the moon In the plane in face, therefore the displacement in the x, y, z-directions of video camera photocentre can be obtained respectively with reference to elevation information.

To a change in coordinate axis direction and two moment t₁And t₂, as shown in figure 3, p₁And p₂It is matching double points, is same on lunar surface Point P is in picture point not in the same time, O_i(i=1,2) be camera position, therefore can be obtained according to similar triangle theory：

In formula：I=1,2, O_iC_iFor focal length of camera f, O_iT_iFor camera height H_i, point C_iFor t_iThe center of time chart picture Point, C_iP_iFor pixel P in image_iTo the distance of central pixel point.Therefore, in focal length of camera f, camera height O_iT_iIt is known In the case of, by match point in measurement image to picture centre pixel apart from C_iP_i, we can obtain：

Thus, between two moment, video camera photocentre relative to the displacement on ground is

With reference to shooting time, the displacement to all directions carries out the instantaneous velocity that difference is obtained all directions.

V=T₁T₂/(|t₂-t₁|)

Although the present invention is disclosed above with preferred embodiment, so which is not limited to the present invention.Skill belonging to of the invention Has usually intellectual in art field, without departing from the spirit and scope of the present invention, when can be used for a variety of modifications and variations.Cause This, protection scope of the present invention ought be defined depending on those as defined in claim.

Claims (3)

1. a kind of method of measurement airborne vehicle speed, estimates airborne vehicle speed, its feature using the sequence image that video camera shoots It is to comprise the following steps：

Step 1, the image at two moment of in front and back is corrected to from current location respectively by attitude angle according to attitude angle information it is 0 Position；

Step 2, autoregistration is carried out to aforementioned two width image based on SIFT algorithms, it is determined that the picture registration area at two moment in front and back Domain, obtains one group of optimal match point；And

Step 3, the speed that airborne vehicle is calculated with reference to remote sensor elevation information and shooting time；

Wherein, in abovementioned steps 3, the match point to corrected two moment sequence images in front and back, positioned at phase on differing heights In mutually parallel and plane parallel to lunar surface, therefore video camera photocentre can be obtained respectively in the x, y, z-directions with reference to elevation information Displacement, the then displacement to all directions carry out the instantaneous velocity that difference is obtained all directions, specifically include：

To a change in coordinate axis direction and two moment t₁And t₂, p₁And p₂Matching double points, be on lunar surface same point P not in the same time Picture point, thus it is available according to similar triangle theory：

$\frac{C_i{P}_i}{T_{i}P}=\frac{O_i{C}_i}{O_i{T}_i}$

In formula：I=1,2, O_iIt is the position of camera, O_iC_iFor focal length of camera f, O_iT_iFor camera height H_i, point C_iFor t_iMoment The central point of image, C_iP_iFor pixel P in image_iTo the distance of central pixel point；

Due to focal length of camera f, camera height O_iT_i, it is known that by measuring match point p in image₁And p₂To picture centre pixel Point apart from C_iP_i, obtain：

$T_{i}P=\frac{f}{H_i}{C}_i{P}_i$

Thus, try to achieve displacement T of the video camera photocentre relative to ground between two moment₁T₂：

$T_1{T}_2=T_{1}P+{\mathrm{PT}}_2=\frac{f}{H_1}{C}_1{P}_1+\frac{f}{H_2}{C}_2{P}_2$

With reference to shooting time, the displacement to all directions carries out the instantaneous velocity that difference is obtained all directions：

V=T₁T₂/(|t₂-t₁|)。

2. it is according to claim 1 measurement airborne vehicle speed method, it is characterised in that the realization of abovementioned steps 1 includes Following steps：

Converted images size is determined first, and the coordinate figure inversion after correction is shifted to into original image coordinate points then, finally by Quadratic interpolattion carries out sub-pixel positioning, and the image after being corrected, wherein updating formula are：

$\mathrm{\λ}\left[\begin{array}{c}{x}^{\′}\\ y^{\′}\\ 1\end{array}\right]={\mathrm{KR}}^{-1}{K}^{-1}\left[\begin{array}{c}x\\ y\\ 1\end{array}\right]$

In formula：λ is scale factor, and K is intrinsic parameters of the camera matrix, and R is the orthogonal spin matrix of unit.

3. the method for measurement airborne vehicle speed according to claim 1, it is characterised in that abovementioned steps 2 are implemented Comprise the following steps：

(1) detection of key point：By gaussian kernel function continuous filtering and down-sampling of the input picture by different scale, form high Then two Gaussian images of adjacent yardstick are subtracted each other and obtain DoG pyramid multiscale spaces and represent by this pyramid diagram picture again；

To DoG metric spaces, each point is compared one by one with the point of adjacent yardstick and adjacent position, the local extremum position for obtaining Put as key point location and corresponding yardstick；

(2) SIFT description are constructed：Using the gradient direction distribution characteristic of key point neighborhood territory pixel, it is each key point designated parties To parameter：Coordinate axess are rotated to into characteristic point direction, to ensure rotational invariance；To any one key point, it is located at which Metric space, takes the neighborhood of 16 × 16 pixel sizes centered on key point, then by this neighborhood be evenly divided into 4 × 4 it is sub Every sub-regions are calculated gradient orientation histogram by region, and the rectangular histogram is uniformly divided into 8 directions；Then, to 4 × 4 sub-districts 8 histograms of oriented gradients in domain are sorted successively according to position, constitute the vector of 4 × 4 × 8=128 dimension, and the vector is SIFT description；

(3) characteristic matching：Measured using similarity determination of the Euclidean distance as key point in two width images, when this distance is less than During given threshold, then judge two key points as possible match point；

(4) remove error hiding：To the be possible to matching double points having determined, less tolerance is set, using RANSAC robusts Method carries out Geometrical consistency inspection, obtains one group of optimal match point；

(5) affine transformation：Two width images are mapped under same coordinate system and determine overlapping region.