JPS59148976A - Estimating method of position and attitude of airframe - Google Patents

Abstract

PURPOSE:To estimate precisely the position and attitude parameters of an airframe which is necessary for correcting the shape distortion of a photographed picture precisely by using a sensor mounted on the airframe. CONSTITUTION:A picture display 30 reads out a part of a picture to be observed from a picture memory 33 and displays the picture. While observing the display, the operator sends the position of a ground stardard point in the picture to a parameter estimating device 31. Simultaneously, the positional vectors of the ground standard point and its difference convariance matrix are inputted. Said operation is repeated for all applied M ground standard points from an external terminal 28. The operation is repeated for all M pieces of the ground standard point. On the basis of these input information, the parameter estimating device 31 estimates the position and attitude parameters of the attitude. A distortion correcting model calculating device 32 forms a distortion correcting model by using the estimation parameter values. Subsequently, a resampling device 34 reads out picture element data of the picture to be observed successively from a picture memory 33. Then, resampling processing is executed in accordance with the distortion correcting model.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a method for correcting shape distortion included in images of the earth's surface taken by sensors mounted on artificial satellites, aircraft, etc. The present invention relates to an estimation method suitable for accurately estimating the position and attitude of an artificial satellite or an aircraft.

[Prior art]

Correction of the above-mentioned shape distortion has conventionally been performed as follows (for details, see, for example, Reference R, Bernstein).

”Digital Image Processin
g of Earth Ql) service 56
w5or [)ata”, IBM Journalo
f Research, Jan, 1976,
checking).

The outline of the conventional method is as follows: In Fig. 1, 1 is a flying object such as an artificial satellite or an aircraft, 2 is a senna mounted on the flying object 1, and 3 is the ground surface. A point 6 on the earth's surface forms an image at a point 5 on the sensor surface 4. At this time, the positional relationship between point 6 and image 5 is the position of the flying object in space.

Geometrical calculations are required using the attitude, model of the Senna optical system, and shape model of the ground surface. The image obtained by sampling the light reception signal on the sensor surface 4 is called an observed image, and each sample thereof is called a pixel.

In FIG. 2 (1), each pixel of the observed image is arranged in equidistant grid points (black dots in the figure), and this is called the observed image space.

Any point in this space is specified by coordinates (t,p). On the other hand, it is used for the purpose of image usage analysis.

As shown in FIG. 2(2), the desired image is one in which pixels are arranged at equidistant grid points (white points in the figure) on a coordinate system X-Y according to a predetermined map projection method. This is called a corrected image. To obtain a corrected image from an observed image, it is necessary to rearrange the pixel positions, that is, to perform resampling processing. This is done as follows. A distortion correction model that maps pixel positions (x, yl) in the corrected image space to pixel positions (t, p) in the observed image space is created using the geometric calculation that associates the above-mentioned ground points with pixel positions in the observed image. Next, this model is used to calculate the position of each pixel on the grid 8 of the observed image that corresponds to each pixel on the grid 9 of the corrected image.This pixel position is generally an integer grid point, that is, a black point of the grid 7. Therefore, the image intensity of each point of the grid 8 is interpolated from the image intensity of each point of the grid 7. This becomes the intensity of each point of the desired grid 9 in the corrected image. If each point is assumed to be equally spaced, the nearest neighbor method, pi-linear method, cubic convolution method, etc. described in the above-mentioned literature can be applied to the interpolation calculation.

The above image distortion correction method is based on the position and attitude of the flying object.

A geometric model consisting of a sensor model and an earth shape model is used. Among these, the measurement data of the position and attitude of the flying object includes measurement errors, so errors occur in the geometric model using this data, which results in distortion (correction error) of the corrected image. Therefore, in general, the accuracy of image correction is improved by more accurately estimating the position and attitude of a flying object by using information from feature points in the image whose positions on the ground surface are accurately known, that is, ground control points. A method is being taken.

First, the geometric model is given by the following equation.

El-9 (ball r t1+ pl+ tl)+shi...(1
) Here, the subscript i means the first ground control point 11 or the pixel 12 of the observation image corresponding thereto.

Let (t, p) represent the coordinates of the pixel 12, and t represent the imaging time. Now, the coordinate system 13 is arbitrarily set. Usually, an earth-centered fixed coordinate system is used. The position vector of the ground reference point in this coordinate system is y=(xyz)”... (2
). G in equation (1) is a vector function that associates the pixel coordinates (t, p) with the k ground point. In addition, el is the measurement error of the position vector y1 of the ground control point and the vector function G
is the sum of the errors. Also, the vector X which is the argument of the function G
are the position and attitude parameters of the flying object given as follows. Normally the position vector of the flying object y = (xyz)”
Each coordinate component X.

It can be approximated by a polynomial function of time 1 of y and z.

Here, N is the degree of the polynomial and is approximately 3. On the other hand, the attitude angle is roll angle (θF) + pitch angle (ra), yaw angle (θy)
Both of these can be approximated by a polynomial function of t.

At this time, the parameter vector X is a vector created by the coefficients of equations (3) and (4).

””(XoX+ =・Xw'f1'!2-ywZIZ2
- ZNθrQθ1. ...θrNθ, 1θ, 2...
Conventionally, the position and attitude of a flying object have been accurately estimated using the following method according to a geometric model of θ, -7°θa,...θyN)T(F5) and using information on ground control points. (For details, see the literature R1H1Coron and K.
W, Simon @t AttitL1de'pim
e-8eries Estimator for 1 (
certificationof 5pacel)orn
e Imagery”, Journal of Spa
cecraft, VOl, 12, AI,
1975. checking). That is, M GCs
When P data (i=1, . . . , M) is given, x=i is estimated by applying the Kalman filter method. At this time, equation (1) becomes the observation equation for the Kalman filter, but since it is nonlinear with respect to X, it is linearized and approximated before the Kalman filter is applied.

FIG. 3 shows the processing flow of the conventional method described above. Processing step 150 by Kalman filter Input data 14
As, the position vector 3'l of each ground standard point + error e1
covariance matrix W+.

Its error covariance 1) Give o. A distortion correction model is created in step 16 from the parameter estimation results.

Using the model, a corrected image 19 is obtained by applying resampling processing to the observed image 17 previously stored on a magnetic tape or the like in step 18.

This conventional method had the following drawbacks. That is, as described above, the result of estimating X obtained by linear approximation of equation (1) is not optimal. For example, for the weight matrix Wi ('=t+..., M) of the error e1, the estimated value that minimizes the evaluation function, that is, the least squares estimated value is not obtained. However, from equation (1), eI is eI −yt G(X, t+, [)++'+)
(7).

[Purpose of the invention]

An object of the present invention is to provide a method suitable for accurately estimating the position and attitude parameters of a flying object, which are necessary for precisely correcting the shape distortion of an image captured by a sensor mounted on a flying object. There is a particular thing.

[Summary of the invention]

Here, an overview of the parameter estimation method used in the present invention will be explained. First, the estimated value of the unknown parameter X is (
6) Define the evaluation function J of the expression as the one that minimizes it, and if
If the conditions are satisfied that the a priori information Xo is a normally distributed random variable and the error e1 is an independent normally distributed random variable with respect to i, then the obtained estimated value is This is the maximum likelihood estimate. Normally, it can be considered that the above conditions are approximately satisfied. Now, since equation (6) is nonlinear with respect to X, X that gives the minimum value of J cannot be found analytically. Therefore, the present invention is characterized in that the Newton-Roughnon method is used as the sequential solution method.

The processing flow is shown in FIG. First step 20~
Initial value setting is performed in step 22. ΔX is the amount of correction of the estimated value X in the sequential calculation process.

Steps 23 to 27 are calculations in each iteration. In step 24, the estimated value X is corrected. The content of calculation of the correction amount ΔX in step 25 is as follows.

(8) where M is the number of ground feature points and n is the number of unknown parameters. J (X) in step 26 is calculated using equation (6). The determination step 27 is for determining the convergence of J, and ε is a sufficiently small positive number.

[Examples of practical application of the invention] The present invention will be described in detail below based on examples. Fifth
The figure is an overall configuration diagram of an image distortion correction device according to the present invention. A priori information Xo of an unknown parameter X and its error covariance matrix pO are supplied from an external terminal 28 to a parameter estimating device 31 based on the least squares method. It is assumed that the one-sided image memory 33 stores observed images. Image display (9) The play 30 reads a part of the observed image from the image memory 33 and displays it. The displayed contents are as shown in FIG. 6, for example. Observation image 37 near the ground control point on the screen 36
is displayed. The operator looks at this display and inputs the position of the ground reference point in the image from the coordinate input device 29. This is sent to device 31. At the same time, the position vector y+ of the ground reference point and its error covariance matrix WI are inputted from the external terminal 28. This operation is repeated for all M ground control points given. Thereafter, the parameter estimation device 31 determines the position of the flying object according to the processing flow shown in FIG. 4 based on this input information.

Estimate the pose parameters. Distortion correction model calculation device 32
creates a distortion correction model using these estimated parameter values. Thereafter, the resampling device 34 sequentially reads out the pixel data of the observed image from the image memory 33, performs resampling processing according to the distortion correction model, and sequentially writes the pixel data of the corrected image onto the magnetic tape 35. The resampling device 34 repeats the above operations until a corrected image (10) having the number of pixels corresponding to a predetermined size is created.

〔Effect of the invention〕

As described above, according to the present invention, the position of a flying object is determined from information about several ground control points selected in an observation image by using a method of directly obtaining a numerical solution without linear approximation of a nonlinear estimation problem. Since the pose parameters can be estimated accurately, it is possible to accurately correct the shape distortion of the observed image.

[Brief explanation of drawings]

Figure 1 is a diagram showing the state in which the ground surface is observed by a sensor mounted on a flying object, Figure 2 is a diagram showing the relationship between observed images and corrected images, and Figure 3 is a diagram showing the flow of distortion correction processing using the conventional method. , FIG. 4 is a diagram showing a solution flow for a nonlinear estimation problem using the method of the present invention, FIG. 5 is an overall configuration diagram of an image distortion correction device according to the present invention, and FIG. It is a figure which shows the example of a display of a standard point. 1... Flying object, 31... Parameter estimation device. Agent Patent Attorney Akio Takahashi (11) 1st Figure 1 η 2 (1n (2,
) Ka 3 Figure 4 Figure z f Figure 1 z th b

Claims (1)

[Claims] Estimating the position and attitude parameters of the flying object from multiple ground standard data, creating an image distortion correction model for the Shizukushu based on the parameters, and resampling the observed image using the model to create a corrected image. An image distortion correction method for estimating the position and attitude parameters of a flying object, characterized in that the position and attitude parameters are estimated based on the nonlinear least squares method from the ground standard data.