JPH01191013A - Automatic generation system for altitude data from stereoscopic image - Google Patents

Abstract

PURPOSE:To generate altitude data with high accuracy by correcting differences in sale and light-shade level between two stereoscopic images. CONSTITUTION:The two stereoscopic images are contained in a file 101 and read in a computer 3. The computer 3 has a scale correcting means 4, a light- shade level correcting means 5, a matching means 6, and an altitude calculating means 7 and writes the altitude data in a file 1-2 through a memory 2. The altitude data which is obtained as mentioned above is displayed in three dimensions by an image display device 8. The means 4 and 5 operate so as to correct the difference in scale and light-shade level between stereoscopic images which are photographed in different directions. Consequently, the matching means easily retrieves a corresponding point, so the probability of misrecognition is reduced.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to elevation measurement using stereo images, and in particular to highly accurate elevation data obtained from images of the same place viewed from a high altitude in different directions or on different days. The present invention relates to an image processing method suitable for obtaining .

[Conventional technology]

The conventional method extracts small areas of the same size or different sizes from two images of the same place viewed from different directions from a high altitude, as in Japanese Patent Application Laid-Open No. 62-134773.
The point where the correlation coefficient between two small area images is maximum is recognized as a corresponding point, and the elevation is calculated based on the geometric positional relationship of the correspondence.

there was.

[Problem to be solved by the invention]

The above-mentioned conventional technology automatically searches for stereo corresponding points based only on image shading or color information, so the shading level may vary due to differences in the scale of the images when viewed from different directions or differences in the observation dates of the two stereo images. No consideration was given to the fact that the correlation coefficient decreases due to differences, and there was a problem in that erroneous recognition of corresponding points occurred.

The purpose of the present invention is to provide an automatic elevation data creation method from stereo images that solves the above problems.

[Means to solve the problem]

The above object includes a scale correction means for correcting a difference in image scale between two stereo images, a gray level correction means for correcting a difference in gray level of pixels of the two stereo images,
A matching means for determining a position on the other stereo image that corresponds to a point for which the altitude is to be determined on the one stereo image, and an elevation calculating means for determining the altitude from the positional relationship between the corresponding points on the two stereo images are provided. This is achieved by

[Effect]

The vertical scale correction means and the gray level correction means were used on different days,
It operates to correct differences in scale and gray level of stereo images taken from different directions. This makes it easier for the matching means to search for corresponding points, thereby reducing the probability of misrecognition in the automatic search for corresponding points in stereo images.

Therefore, the elevation calculation means can obtain elevation data from these corresponding points with high accuracy.

That is, according to the present invention, highly accurate elevation data can be automatically created.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is an overall configuration diagram of an embodiment of the present invention.

The two stereo images are stored in advance in file 1-1 and read into computer 3. The calculator 3 includes a scale correction means 4. Shade level correction means 5, matching means 6
and an altitude calculating means 7, which writes altitude data to the file 1-2 via the memory 2. The altitude data obtained in this way can also be displayed three-dimensionally on the image display device 8.

FIG. 2 is a diagram showing two stereo images 21 and 22 stored in file 1-1. To automatically create elevation data, first, measurement points 23, 24, 25, 26, etc., at which elevation data will be created, are determined on one stereo image 21. The criteria for deciding is that the size of one pixel required to create a topographic map of approximately 1:500,000 (contour interval 50m) is approximately 30m, so for example, the size of one pixel of a stereo image 21.22. If the distance is 10 m, it is necessary to select measuring points 23, 24°, 25.26, etc. at intervals of at most 3 pixels. In this way, for example, stereo image 2
Once a measurement point 23 is determined on 1, consider a window W of the size of MXM centered on this measurement point 23. It is also assumed that a search area S of size LXL (L is larger than M) is set on the stereo image 22.

FIG. 3 is an explanatory diagram of the operation of this embodiment. First, the image of the window W and the image of the search area S are stored in a file 1-1.
The data is transferred from the memory 2 of the computer 3 to the memory 2 of the computer 3 (31). In this way, for one measurement point, the processing time can be shortened by accessing the memory 2 instead of accessing the file 1-1. Next, the scale of the image of the window W in the memory 2 is corrected by the scale correction means 4 (32). Furthermore, the shading level of the image of the window W is corrected by the shading label correction means 5 (33). Then, the image of the window W subjected to these two corrections and the image of the search area S are used by the matching means 6 to find a point 27 on the stereo image 22 corresponding to the measurement point 23 of the stereo image 21 (34). Furthermore, the position of the measurement point 23 on the stereo image, the position of the corresponding point 27 on the stereo image, and photographing conditions, for example, if the stereo image was photographed from a satellite, the position and speed of the satellite.

Based on the posture and sensor information, the altitude calculation means 7 calculates the altitude of the measurement point 23 and determines the geographical position of the measurement point 23 (
For example, the altitude data is written to the file 1-2 together with information on latitude φ and longitude λ (35). Finally, the second
It is checked whether elevation data has been obtained for measurement points 23, 24, 25, 26, etc. defined in the figure, and if there are any measurement points that have not yet been processed, the above process is repeated (36). In this way, elevation data is automatically created in file 1-2, and as mentioned above, it can be displayed in 3D on the image display device 8, or a contour map can be created by connecting measurement points of equal elevation, or It becomes possible to transfer the numerical data of Facil 1-2 to another medium (magnetic tape, optical disk device) and use it as a database.

Next, the scale correction means 32 will be explained using FIGS. 4 and 5. FIG. 4 is a flowchart of the scale correction means 32, and FIG. 5 is a conceptual diagram of the scale correction means 32.

Since the pixel size of the window W and the search area S differ only in the horizontal direction of the image, only the horizontal direction is corrected. First, the distance corresponding to the ground of one pixel of the image of the window W and the search area S is calculated (41), and the scale ratio (the distance corresponding to the ground of one pixel of the window) is determined (42). The line-of-sight angle and the altitude of the flying object when the stereo images 21 and 22 were taken are respectively set to θ1.
If θ2; H1, H2, then α can be calculated using the equation (1): Hs CO52θ2 H2CO5”θworking.Next, based on this α, horizontal direction sampling of the image of the window W is performed (43).

For example, resampling can be performed using the convolution operation. Here, I (i) the density level 51 of the position owner before beam sampling, Io'(
k) Shading level 53 at the position after beam sampling. [
αkl is the largest integer not exceeding αk. 5W(ki) is an appropriate weighting function, and for example, in the case of resampling by the cubic convolution method, a=-1 may be used.

Next, the gray level correction means 33 will be explained based on FIGS. 6 and 7. Figure 6 shows the shading level correction means 3.
3, the flowchart in FIG. 7 shows the gray level correction means 33.
It is a conceptual diagram. First, calculate the histogram 72.71 of the image of the window W and the image of the search area S after the pixel scale has been corrected by the scale correction means 32 (61).
do.

Then calculate the mean and variance of these two histograms and calculate them as μm, μS; σw2. Let it be σsz. The gray level of the image of the window W is corrected so that the average and variance of the histogram of the window S after the gray level correction are equal to the average and variance of the histogram of the search area S (63). To do this, if the density level before density level correction is calculated, and the density level after density level correction is good.

FIG. 8 is a flowchart of the matching means 34.

Here, we first find coarse corresponding points using the 5SDA method that minimizes the sum of the absolute values of differences in gray levels (81).
Then, a method (82) is used in which precise corresponding points are calculated by maximizing the correlation coefficient in the vicinity of the rough corresponding points. As a result, 5SDA has a small amount of calculations but low accuracy, while the correlation method has good accuracy but requires redundant calculations.By combining these characteristics as described above, high speed can be achieved. Furthermore, it is possible to perform matching with high precision.

FIG. 9 is a diagram explaining the principle of the altitude calculation means. The matching means 34 finds the point 27 on the stereo image 22 corresponding to the measurement point 23 of the stereo image 21, and the image shows the gradation level corresponding to the elevation point 93 on the ground surface 97 as the measurement point 23 and its corresponding point. There are 27 gray levels.

Also, if the position 23°27 on the stereo image 21.22 is given, the altitude Hs, Hz and line-of-sight angle θ, θ2 of the flying object at that time are determined, and the height on the earth's spheroid (plane of altitude O) is determined. Positions 95 and 96 are determined. If the difference between the positions 95 and 96 is ΔX, the altitude of the point 93 on the ground surface 94 can be calculated by tanθ1+tanθ2.

〔Effect of the invention〕

According to the present invention, errors occur when recognizing corresponding points due to a difference in the scale of two stereo images caused by viewing the same place from different directions, or a difference in gray level due to a difference in the shooting date of the two stereo images. Since the recognition problem can be solved by the scale correction means and the gray level correction means, it is possible to recognize corresponding points with high precision, and therefore, there is an effect that altitude data can be automatically created with high precision.

[Brief explanation of the drawing]

FIG. 1 is an overall configuration diagram of an embodiment of the present invention, FIG. 2 is a diagram showing two stereo images, FIG. 3 is an explanatory diagram of the operation of this embodiment, and FIG. 4 is a flowchart of the scale correction means. FIG. 5 is a conceptual diagram of the scale correction means, FIG. 6 is a flowchart of the gray level correction means, FIG. 7 is a conceptual diagram of the gray level correction means, FIG. 8 is a flowchart of the matching means, and FIG. 9 is an elevation calculation means. FIG. 1-1...File, 2...Memory, 3...Computer, 8...Image display device. Agent: Patent Attorney Katsutaka Ogawa (,1, 1st prisoner, 3rd judge, 4th figure, 6th day, 9th session)

Claims (1)

[Claims] 1. In elevation measurement using two stereo images taken of the same place from different directions, a point on one image for which the elevation is to be determined is defined, and a point on the other image corresponding to this point is defined. A scale correction means for correcting the difference in pixel scale between the two stereo images; gray level correction means for correcting differences in gray level of pixels; matching means for determining a position on the other stereo image that corresponds to a point on the one stereo image for which the elevation is to be determined; An automatic elevation data creation method from stereo images characterized by providing an elevation calculation means for calculating elevation from the positional relationship of corresponding points.