JPS62249287A - Position measuring system - Google Patents

Abstract

PURPOSE:To attain an accurate position measurement by detecting only a part (blocked area) which is photographed in the image on one side of the correspondence between points of a pair of images. CONSTITUTION:The information to display a pair of images photographed by TV cameras 10 and 11 is stored into memory parts 12 and 13. Into a similarity table 14, the similarity between respective points obtained by using the image data of the memory parts 12 and 13. Corresponding determining parts 15 and 16 read the data of similarity from the table 14 and counts the corresponding point with one side image and other side image concerning respective images. These counting results are sent to a closed area extracting part 17, it is investigated whether or not the pair of the corresponding points is coincident, and the dissident point is extracted as the closed area. A position measuring part 18 counts the position of respective points excluding the closed area by using the dislocation quantity of the corresponding point. Thus, to execute a position measurement excluding the closed area, the reliability of the position measuring result can be increased.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a single-position measuring device, and in particular, uses a pair of image information obtained from the same object at different positions or times to determine the position or position of the object. The present invention relates to a position measurement method suitable for measuring distance. This type of device includes various devices that control equipment using image information (e.g. robots with vision,
It is indispensable for automatic vehicle control devices that respond to the external environment, process control devices that work with ITV, etc.).

[Conventional technology]

In position measuring devices of the type described above, it is necessary to determine by automatic processing which point in one image corresponds to which point in the other image. One of the conventional methods for this purpose is a method for forming depth information signals for stereoscopic television.
(Patent Publication (B2) Sho 60-10564). This method is.

This method compares two image signals and obtains depth information of an object from the amount of pixel shift when the signal waveforms almost match, and is suitable for high-speed processing. However, when comparing signal waveforms, one image is shifted and compared with the other image, so the position of a pixel for an object that is only visible in one image is determined. 1 for parts that are only visible in the image (occluded areas)
Originally, distance cannot be determined, so in order to obtain accurate position information, it is necessary to correctly detect and remove such occluded areas.

[Problem that the invention seeks to solve]

The above-mentioned conventional technology only finds corresponding points between one image and the other image, so it does not take into account occluded areas that appear only in one image, and incorrect position measurements may occur for this area. There was a problem with leaving.

An object of the present invention is to remove erroneous correspondence and perform precise position measurement by detecting occluded areas in correspondence between points in a pair of images.

[Means for solving problems]

In order to achieve the above purpose, for each point in one image (denoted as A), the corresponding point in the other image (denoted as B) is determined, and conversely, for each point in one image B, the corresponding point is calculated. Find the corresponding point of A. Thereafter, portions where the pairs of corresponding points obtained in each process do not match are detected. This part is a closed area, and by excluding this part and measuring the position, it is possible to perform position measurement with fewer erroneous correspondences.

[Effect]

To obtain location information from a pair of image information.

The object whose location information is to be obtained must be visible in both images. Corresponding points appearing in both images are the same pair no matter which image they are found from. However, points that appear only in one image do not have similar points in the other image, so when finding corresponding points from the other image,
There is no pair corresponding to this point. In this way, by finding corresponding points from both of a pair of images, erroneous correspondences to occluded areas can be removed.

〔Example〕

FIG. 1 shows an example of a position measuring device to which the present invention is applied. Information representing a pair of images taken by TV cameras 10 and 11 is temporarily stored in image storage units 12 and 13. In the case of stereoscopic viewing, two images taken at the same time from different positions are stored, but when the TV camera moves, only one TV camera is used and two images taken at different times are stored. , image storage units 12 and 1
In the similarity table 14, which may be stored in each of the image storage units 12 and 13, the similarity between each point determined using the image data in the image storage units 12 and 13 is stored. Corresponding point determination unit 15
In step 16, similarity data is read from the similarity table 14, and corresponding points between one image and the other image are calculated. In calculating the corresponding points, the corresponding point determination unit 15 calculates the corresponding points for each point of the image in the image storage unit 12, and the corresponding point determination unit 16 calculates the corresponding points for each point of the image in the image storage unit 13. Calculate. These calculation results are sent to the closed region extraction unit 17, which checks whether the pairs of corresponding points match and extracts points that do not match as closed regions.
The position measurement unit 18 calculates the position of each point excluding the occluded area using the shift amount of the corresponding points.

The details of the processing in each section will be explained below, taking the case of binocular stereopsis as an example.In the case of binocular stereopsis, the patent application
9-237600 (Method for determining corresponding points in a position measuring device using correspondence between r images), the search for corresponding points can be performed within the corresponding scanning lines of both images. It is assumed that the images stored in the storage units 12 and 13 are images captured by the TV cameras 10 and 11, converted in this manner, and stored.

The following discussion describes processing for a pair of scan lines of such an image.

The similarity table 14 stores the similarity obtained using the image data of the right eye image and the image data of the left eye image.

That is, as shown in FIG. 2, point A (ipj) has a 0 position where the similarity between the i-th pixel of the right eye image and the j-th pixel of the left eye image is stored. The width of this table can be determined appropriately. For example, if the deviation between the right eye image and the left eye image is within pixels, the table value can be calculated within the range of i-≦j≦i+. .

As shown in the following formula, the similarity is calculated as A(i,
j)=I R(i) −L(j)l...(
1) Or.

A(i,j)=(R(i)−L(j))”−(
2). Furthermore, using the amount of change in brightness DR(i) and DR(j) at each point of the right eye and the right eye image, A
(i, j)"IDR(i)-DL(j)I...
・(3) Or A(i,j)=lDR(i)−DL(j)l”・
...(4) etc. may also be used.

The corresponding point calculation units 15 and 16 perform the following processing to obtain corresponding points. First, the corresponding point calculation unit 15 calculates a small area of an appropriate width centered on one point (pixel) of the left eye image into a third
Set as shown in the shaded area in the figure and find. In equation (5), if da is the value of d where ARta(d) has the maximum value, then da represents the amount of deviation between the ia-th pixel of the right eye image and the left eye image, that is, the right eye image The ia-th Il element of the left-eye image and the ia+da-th pixel of the left eye image are a pair of corresponding points.

This process is performed on all pixels of the right eye image to find corresponding points between the right eye image and the left eye image. Next, the corresponding point calculation unit 16
Now, set a small area of an appropriate width centered on one point (pixel ') of the left eye image, as shown in the shaded area in Figure 4.

seek. In Equation (6), if da is the value of d where ALja(d) has the maximum value, then da indicates the amount of deviation between the j&th pixel of the left eye image and the right eye image, that is, the left eye The j&th pixel of the image and the ja+d&th pixel of the right eye image are a pair of corresponding points.

The closed region extraction section 17 extracts a closed region using the results of the corresponding point determination sections 15 and 16. This procedure will be explained with reference to FIG. FIGS. 5(a) and 5(b) are a right eye image and a left eye image, respectively, and AA' is a pair of scanning lines for finding corresponding points. Now, the corresponding point determination unit 15 calculates the points of the left eye image to which each point of the right eye image corresponds, and the result is determined by the fifth
FIG. 5(c) shows the results obtained by the corresponding point determination unit 16 of the points of the right eye image to which each point of the left eye image corresponds, and FIG. 5(d) shows the results.

The closed area extraction unit 17 extracts the images shown in FIG. 5(c) and FIG. 5(
The common part of the corresponding points in d) is determined and removed to correspond to the closed area, thereby obtaining the corresponding points shown in FIG. 5(e). Specifically, the corresponding points of each 1iWs of the right eye image are sequentially extracted, and for example, if the corresponding point of the ia-th pixel of the right-eye image is the ja-th pixel of the left-eye image, the ja-th pixel of the left-eye image is It is only necessary to check whether the corresponding point of element a matches the pixel number La of the right eye image. If they do not match, it is a corresponding point to the occluded area.

Remove from pairs of corresponding points.

tJRm region extraction region extraction part 7. The position of the object is calculated in the 1-position measurement unit 18 based on the pair of corresponding points from which the 8 corresponding points in the closed sM region have been removed. The position of the object is calculated from the coordinates of each pair of corresponding points in the right eye image and left eye image, the center coordinates of the optical system of the TV camera, and the focal length of the lens (see Japanese Patent Application No. 59-2376000). ).

The present invention, which has been described above with respect to embodiments in binocular images, is applicable not only to cases where corresponding points need to be found only on corresponding scanning lines, such as binocular images, but also to objects that are relatively displaced in arbitrary directions. This method can also be applied to cases where it is necessary to find corresponding points in a two-dimensional range, such as images taken at different times.

〔Effect of the invention〕

According to the present invention, it is possible to easily extract the part of the outside world that appears only in one of the paired images (y!4 occluded area), and to perform position measurement by removing this part, the position measurement results can be This has the effect of increasing reliability.

[Brief explanation of drawings]

Fig. 1 is a block diagram of an apparatus for implementing the present invention, Fig. 2 is an explanatory diagram of a similarity table for each pixel of the right eye image and left eye image, and Fig. 3 is a diagram of the right eye in the case of binocular vision. FIG. 4 is a diagram illustrating a method for finding corresponding points in images; FIG. 4 is a diagram explaining a method for finding corresponding points in left-eye images in the case of binocular vision;
Figure 5 shows the main engine Figure 1 Figure 2 Item 3 Figure 4 Left side (I41 number bar %5 z

Claims (1)

[Scope of Claims] 1. Means for generating image information representing an image of the external world, processing this image information to determine pairs of mutually corresponding points in a pair of images, and determining points of two points forming each pair. In a position measuring device that includes a calculation unit that calculates the position of a point in the outside world that corresponds to both points based on the positional relationship, for a given point on each image, the corresponding point on the paired image is Find from other images,
A position measurement method characterized by extracting only corresponding points whose degree of coincidence is within a predetermined range from a pair of corresponding points obtained from both images. 2. When determining the above-mentioned corresponding point pairs, for a given point on one image, the corresponding point on the other image to be paired is determined, and there are multiple corresponding points for each point on the other image. If there are two corresponding points, the corresponding points are determined from the other image, and only the corresponding points whose degree of coincidence with the corresponding points determined from the one image is within a predetermined range are considered as true corresponding points. The position measurement method of the first term. 3. The position measurement method according to item 1 or 2, wherein the predetermined point is a feature point in each image. 4. The position measurement method according to item 3, wherein the feature points in each image are selected according to the amount of change in brightness of the image in the vicinity of each point. 5. Items 1 to 4, characterized in that the corresponding points between the images are determined based on the degree of similarity between a small area set on one image and a small area set on the other image. Position measurement method for any one of the terms.