JPH10332334A - Position measuring method by image processing and its device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a device which can detect a position accurately at high speed using a simple geometrical model satisfied with two reference points on an image produced with a camera installed obliquely upward. SOLUTION: A camera 2 is set at a position for photographing an oblique image to a measuring area where reference points P1, P2 are clearly expressed. A position measuring main body 1 is provided with an image memory 13 for storing an inputted image from the camera 2; an image processing means 12 for recognizing an article in the inputted image; a model constructing section 14 which assumes a middle coordinate face 7 to two-dimensionally connect between a world coordinate system for prescribing reference points P1, P2 and an image coordinate system corresponding to the picked up face of the camera 2 and prepares a geometrical model to define positional relation between the world coordinate system and the image coordinate system through the middle coordinate face 7; and a position measuring section 15 for obtaining a ground position from a position on the image of an article using the parameter of the geometrical model.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to position measurement of a moving object using a camera image, and more particularly to a position measurement method obtained from a diagonal image through a geometric model.

[0002]

2. Description of the Related Art A camera is installed in a measurement area such as a road or a water surface, and the image of the camera is processed to measure the position of the moving body and monitor the flow and speed of the moving body. Such position measurement by image processing is performed based on a two-dimensional position measurement principle by installing a camera directly above or beside an area where a target object moves.

[0003] However, depending on the object, there is a case where the camera cannot be physically installed directly above or beside the measurement area. In many cases, it is advantageous to install the camera obliquely above the surveillance area, such as a surveillance camera, and to extend the surveillance area by one camera to track the moving object.

[0004] Since an oblique camera image obtained by photographing an object from an oblique direction includes distortion due to a three-dimensional position, the object in the camera image appears to be large for a distant object and small for a nearby object. For this reason, the error from the position directly below the moving object increases as the distance from the camera increases, and an accurate position cannot be obtained by two-dimensional measurement in which the entire image is treated at the same scale.

As a method of measuring a three-dimensional position using an oblique camera image, a spot light, a slit light projection method and the like are known. In recent years, research on computer vision for recognizing the three-dimensional shape and space arrangement of an object from an image has been actively conducted. In this field, an object whose position is known in world coordinates in a three-dimensional space is taken into an image, and the image plane is acquired. By applying a projection model using the coordinates described above, it is general to obtain the relative relationship between the world coordinate system and the camera coordinate system, that is, the position and orientation of the camera and the optical system parameters.

[0006] Further, the literature "Geometry for computer vision (Koichirou Exit; Information Processing, Vol.37 No.6 '96 /
6 p549-p556) ”describes a perspective projection matrix composed of 12 components, which directly connects the world coordinates in space and the coordinates on the image with one matrix.

[0007]

The conventional three-dimensional position measurement by projection of spot light and slit light can provide good results in an environment where illumination can be appropriately controlled, such as in a small indoor space. However, measuring the position of a vehicle or the like traveling on a road requires a large light source capacity because of the outdoor environment that directly receives sunlight. In addition, two cameras are required for one scene, and there is a problem that equipment becomes large-scale.

The method using a perspective projection matrix introduced in the above-mentioned literature requires a minimum of six images which must not all be on one plane in determining the matrix. Requires a large number of images. Therefore, creation of a conversion model cannot be easily realized. In addition, since the calculation is performed using a three-dimensional model, real-time position measurement becomes difficult for a high-speed moving object.

An object of the present invention is to provide a method and an apparatus which overcome the problems of the prior art and which can accurately measure the position of a moving object on the ground surface with a simple conversion model based on an oblique camera image. It is in.

[0010]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides a method for measuring an object position recognized by image processing by capturing an image with an inclined camera line of sight to a measurement area on the ground. Assuming an intermediate coordinate plane that two-dimensionally connects the world coordinate system (ground coordinate system) based on the installation position and the image coordinate system corresponding to the imaging plane of the camera, the world coordinate system and the world coordinate system for the intermediate coordinate plane are assumed. A geometric model that defines a positional relationship of each image coordinate system is created, and a ground position is obtained from the position of the object on the image using the model.

The intermediate coordinate plane is set at a position that is perpendicular to the camera line of sight and that shares a ground viewpoint at which the camera line of sight intersects the ground with the world coordinate system.

Further, according to the present invention, the first reference point is specified (marked) in the measurement area, and the first reference point is set at or near the ground viewpoint of the camera line of sight, and the second reference point is set at an arbitrary position. And a first vector having a point corresponding to the first reference point on the image coordinate system as a start point, a point corresponding to the second reference point as an end point, and the first vector on the intermediate coordinate plane. The world coordinate system and the image coordinate system with respect to the positional relationship between the first and second reference points, based on the ratio of the reference point to the second vector having the start point and the point mapped to the second reference point as the end point. Determine the distance conversion coefficient and the deviation angle between, set in advance as a parameter of the geometric model, on the other hand, when measuring the position of the object, determine the specific position of the object present in the measurement area on the image coordinate system, And a point corresponding to the first reference point as a starting point, Obtain an object position vector having a point corresponding to the fixed position as an end point, obtain a position on the intermediate coordinate plane with respect to the specific position by performing coordinate conversion on the object position vector using the distance conversion coefficient and the deviation angle, and perform the conversion. The specified position is converted into the world coordinate system to obtain a specific position on the ground.

[0013] The parameters of the geometric model include direction cosine information (lx, ly, lz) on a world coordinate system whose origin is the ground point immediately below the camera, and the specific position is a height higher than the ground. When information is provided, a ground position immediately below the position converted into the world coordinate system is obtained by a correction operation based on the height information and the direction cosine information.

[0014] Further, the present invention provides a method wherein, when the first reference point is set at a position within a range allowed as a temporary ground viewpoint of the camera's line of sight, the camera and the first reference point are set. A temporary intermediate coordinate plane is set at a position perpendicular to the connecting straight line and including the first reference point, and based on the ratio between the first vector and the second vector, the first and second reference points are determined. A distance conversion coefficient and a deviation angle between the world coordinate system and the image coordinate system with respect to the positional relationship are obtained, set as parameters of a temporary geometric model, and then the camera line of sight (light A third vector starting from an image viewpoint given as the position of the axis) and a starting point starting from a point corresponding to the first reference point is obtained, and the third vector is calculated using the parameters of the temporary geometric model. Mapping to the intermediate coordinate plane, Obtain the corresponding position of the viewpoint on the reference surface, further map this corresponding position to the ground coordinate system to obtain a real ground viewpoint, based on this real ground viewpoint, perpendicular to the camera line of sight, and Setting a true intermediate coordinate plane including the ground viewpoint, replacing the first reference point with the ground viewpoint, recalculating the distance conversion coefficient and the deviation angle, and constructing a true geometric model. It is characterized by.

The operation of the present invention will be described. A surveillance camera is installed so as to shoot from a diagonally above the moving object, and two reference points are clearly prepared on the ground surface of the surveillance area so that images can be captured. The first reference point is located near the center of the camera image. .
The second reference point may be at any position in the scene. Assuming that a straight line connecting the center of the camera and the first reference point is a camera line of sight, a geometric model for position measurement is constructed.

In the geometric model, an intermediate coordinate plane which is a plane parallel to the imaging plane of the camera and includes the first reference point on the ground is assumed, and this is defined as a relation between the world coordinate system and the image coordinate system. This is the third plane to be attached. The positional relationship between the first and second reference points is mapped from the world coordinate system and the image coordinate system to an intermediate coordinate plane by two-dimensional coordinate conversion, and the geometrical parameters between the two coordinate systems are mediated by this. , A geometric model that converts the position on the image to the position on the ground can be constructed.

In the above case, if the camera angle after installation can be freely adjusted, the first reference point is generally set to be at the center of the image, so that the ground viewpoint of the camera line of sight and the first viewpoint are set. Since the reference points coincide with each other, the position directly below the object can be measured with high accuracy from the tilt image. In addition, by using the intermediate coordinate plane, the ground position can be calculated by two-dimensional coordinate calculation using the distance conversion coefficient and the deviation angle, so that the arithmetic processing can be speeded up.

On the other hand, when the camera angle cannot be adjusted freely, a deviation occurs between the ground viewpoint and the first reference point, so that the model needs to be corrected. In the present invention, after constructing a temporary geometric model using the first reference point as a temporary ground viewpoint,
In general, based on the position (image viewpoint) of the camera's line of sight (optical axis) at the center of the image on the image coordinate system (image viewpoint), this image viewpoint is mapped to the ground surface using a temporary geometric model to provide a true ground viewpoint. Is obtained, and the true geometric model is reconstructed using the ground viewpoint, so that the accuracy of the model can be improved and the system configuration is easy and easy to use.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments 1 and 2 of the present invention will be described below in detail with reference to the drawings.

[Embodiment 1] FIG. 1 shows the configuration of a position measuring system for a moving object according to an embodiment. This system includes a position measuring device main body 1, a camera 2, a moving field 3, a moving body 4, a measuring area 5, and the like.

The main body 1 of the moving object measuring apparatus includes a control unit 11 for controlling each means, an image processing unit 12 for taking in and processing a camera image, an image memory 13 for storing a camera image, a geometric model construction unit 14, which will be described later, And a model table 16 for storing data of a geometric model using an intermediate coordinate plane as an intermediary.

The camera 2 is installed diagonally above the side of the moving field 3 (height H0). In this embodiment, the ground position immediately below the camera 2 is used as the origin of the world coordinate system (XYZ). However, the Z axis is in the height direction. Moving field 3
Is a space on the ground where the moving object 4 moves, and can be assumed to be a road, a water surface, a stadium, and the like. The measurement area 5 is set on the moving field 3 and corresponds to the field of view of the camera in this embodiment. The moving field 3 and the measurement area 5 are a two-dimensional space corresponding to the ground surface in the world coordinate system.

In this measurement area 5, a first reference point P1 and a second reference point P2 for constructing a geometric model are clearly shown so as to be photographable. The first reference point P1 is set at or near the intersection with the camera line of sight (optical axis) 6 of the camera 2, and the second reference point P2 is set at an arbitrary position. The positional relationship between the camera 5, the first reference point P1, the second reference point P2, and the like is as follows.
It is defined by the world coordinate system (XYZ).

The intermediate coordinate plane 7 is a concept on a geometric model peculiar to the present invention. The ground coordinate plane 7 is perpendicular to the camera line of sight 6 and intersects the camera line of sight 6 with the ground surface. It is a two-dimensional plane that passes. As described later, the intermediate coordinate plane 7 is parallel to an image plane coordinate system (hereinafter, referred to as an image coordinate system) of the image memory 13. Also, a two-dimensional coordinate system on the ground based on the world coordinate system (hereinafter referred to as a ground coordinate system)
Share the first reference point P1.

Therefore, the intermediate coordinate system that defines the intermediate coordinate plane 7 can perform two-dimensional coordinate conversion between the image coordinate system and the ground coordinate system, respectively. Position measurement can be easily realized.
Hereinafter, regarding the construction of the geometric model according to the present embodiment,
This will be described with reference to FIGS.

FIG. 2 is a conceptual diagram of a geometric model, showing a geometric relationship between an image coordinate system and an intermediate coordinate system with respect to an image on the ground. The reference point P1 is set in the measurement area 5 of the ground coordinate system.
Take P2 and look at camera 6 connecting P1 and the center of camera 2
, A two-dimensional intermediate coordinate plane 7 is set vertically. The point at which the line 8 connecting P2 and the camera center intersects the plane 7 is the mapping point P2 'on the intermediate coordinate plane 7 of P2. Vector P1P2 '
(R *) can be represented by an absolute value R and a declination Θ with respect to the Xm axis of the intermediate coordinate system. If the directions of the axes of the intermediate coordinate system and the image plane coordinate system match, the Xm axis becomes x
The axis and the deviation angle become zero.

The image of the lens 21 of the camera 2 is formed on the image pickup plate 22 at the position of the focal length f. P on the imaging plate 22
The images 1 and p2 are inverted images of the real images of P1 and P2 on the ground, and will be described using images on the image memory that look in the same direction as the ground. The image on the image memory can be understood well if, for example, an image coordinate system 23 parallel to the imaging plate 22 is set at a position of the focal length f in front of the lens 21. As shown in an enlarged manner in FIG.
The vector p1p2 (r *) based on the images p1 and p2 corresponding to the reference points P1 and P2 can be expressed by the absolute value = r and the argument θ of the image coordinate system 23 with respect to the x-axis.

FIG. 3 shows a procedure for creating a geometric model, and FIG. 4 shows a geometric relationship corresponding to the procedure. This creation processing is executed by the geometric model construction means 14.

First, the reference point P set on the measurement area 5
The camera images of P1 and P2 are captured (s101), and the camera's line of sight (optical axis) is adjusted to the reference point P1 (s102). Next, the direction cosine of the camera line of sight (lx ₀ , ly ₀ , lz ₀ )
Then, data in the world coordinate system (XYZ) display of the intermediate coordinate plane 7 passing through the reference point P1 perpendicular to the camera line of sight is acquired (s103).

The direction cosine of the line of sight of the camera connecting the center C of the camera 2 and the reference point P1 is represented by the world coordinates (Cx =
0, Cy = 0, Cz = H0), the world coordinates (P1xx, P1y, P1z = 0) of the reference point P1 and the distance Pc1 from the camera center C to P1 are calculated by Equation 1.

[0031]

Lx ₀ = P1x / Pc1 ly ₀ = P1y / Pc1 lz ₀ = P1z / Pc1 Here, Pc1 = √ (P1x ² + P1y ² + H0 ² ).

If an arbitrary point on the intermediate coordinate plane 7 is represented by (XYZ) in the world coordinate system, it can be expressed by Equation 2 from the direction cosine of the camera's line of sight perpendicular to the intermediate coordinate plane 7 and the distance Pc1.

[0033]

Lx ₀ * X + ly ₀ * Y + lz ₀ * (Z−H0) = Pc1 The intermediate coordinate plane 7 is defined on the world coordinate system by Expression 2. The definition formula of the direction cosine of the camera's line of sight and the intermediate coordinate plane is
It is stored in the model table 16 as a part of the geometric model.

Next, as shown in FIG.
3, the points p1 and p2 corresponding to the ground reference points P1 and P2 are acquired by image processing (s10).
4). Then, a vector p1p2 (r *) = (r, θ) having a start point p1 and an end point p2 is obtained (s105).

Further, a position P2 'where the reference point P2 on the ground is mapped on the intermediate coordinate plane 7 is obtained (s106). FIG. 4 (b)
, The reference point P2 (P2x, P2y, P2z = 0)
Of intersection P between the straight line 8 connecting the camera center C and the intermediate coordinate plane 7
2 ′ is the result of conversion of P2 on the ground to the intermediate coordinate plane 7. Here, the straight line 8 can be expressed as in Equation 3.

[0036]

Y = (P2y / P2x) * X Z = ((P2z−H0) / P2x) * x + H0 Therefore, the intersection P2 ′ is obtained from the unknowns X and
Y and Z are determined and obtained as coordinates (XYZ).

Further, as shown in FIG.
1. A vector P1P2 ′ (R, Θ) having an end point P2 ′ is obtained (s107). Then, from the ratio between the vector r * and the vector R *, the distance conversion coefficient ε and the deviation angle △ θ are obtained by Expression 4 (s108).

[0038]

Ε = R / r Δθ = Θ−θ Here, when the direction of the intermediate coordinate plane 7 is matched with the coordinate axes of the image memory, Δθ = 0. The distance conversion coefficient ε and the deviation angle △ θ are stored in the model table 16 as parameters of the geometric model (s109). Thus, the construction of the geometric model is completed.

Next, using the constructed geometric model,
The position measurement of the moving object will be described. This processing is executed by the position measurement processing unit 5. FIG. 5 is a flowchart of the position measurement process, and FIG. 6 is an explanatory diagram showing a geometric relationship corresponding to the measurement procedure.

First, as shown in FIG. 6A, a camera image including a moving object (Px) whose position is unknown is input from the measurement area (s201). A moving object is recognized from the input image by image processing (s202), and an object position px on the image coordinate system is acquired as shown in FIG. 6B (s203).

Next, referring to the model table 16, the first
A vector p1px (rx, θx) having a point p1 corresponding to the reference point P1 as a start point and a point px as an end point is obtained (s20).
4). Further, as shown in FIG. 6C, the point px is mapped onto the intermediate coordinate plane 7, and the point Px 'on the intermediate coordinate system corresponding to the point px is obtained.
, And a vector P1Px ′ (Rx, Θx) having the first reference point P1 as a start point and the point Px ′ as an end point is obtained (s20).
5). Vector P1Px ′ (Rx, Θ
x) can be obtained from Equation 5 from rx and θx obtained in step s204, the distance conversion coefficient ε, and the deviation angle Δθ stored in the model table 16.

[0042]

Rx = rx * εΘx = θx−Δθ The coordinates of the point Px ′ on the intermediate coordinate plane 7 obtained from this are expressed as (P
3x ′, P3y ′, P3z ′).

Next, Px 'is converted into a position on the ground coordinate system, and the position Px of the moving object on the ground is obtained (s20).
6). As shown in FIG. 6D, Px on the ground coordinate system is a straight line 8 ′ connecting the camera center C and a point Px ′ on the intermediate coordinate plane 7.
And the point of intersection with the ground surface. Straight line 8 '
Can be represented by Equation 6.

[0044]

Y = (P3y '/ P3x') * X Z = ((P3z'-H0) / P3x ') * X Therefore, the intersection Px is obtained from the unknowns X, Y,
Z is determined and obtained as coordinates (XYZ).

Here, a description will be given of a height correction process for finding a ground point immediately below a feature point when the feature point of the measurement object is located above the ground. FIG. 7 shows an explanatory diagram of the height correction. When the characteristic point of the moving body 4 (for example, the mast of the hull) is at the height of the ground h, the value of the point Px acquired in step s206 is corrected as follows.

There is a feature point at S1 at the altitude h above the ground, and as viewed from the camera 2, it looks as if it is at a point S2 on the ground. What is desired to be measured is the ground position S0 immediately below the feature point S1 of the object 4. As described above, since S2 is obtained as Px, the height is corrected by obtaining the correction amount ΔPx by Expression 7 with reference to the direction cosine (lx, ly, lz) of the model data table 16 (s207).

[0047]

S0 = Px + ΔPx X-axis component of ΔPx = h * lx / lz Y-axis component of ΔPx = h * ly / lz The height information h is obtained by searching a database for a known object. May be directly input by the operator.

As described above, according to the first embodiment, the intermediate coordinate plane which moves in parallel to the image coordinate system in a direction perpendicular to the camera's line of sight and shares one of the two reference points set on the ground with the ground coordinate system is used. Suppose. Through the intermediate coordinate system, the relative relationship between the image coordinate system and the ground coordinate system can be linked by a simple geometric model to the positional relationship between the two reference points that form the oblique image. According to this, the three-dimensional positional relationship in the oblique image can be accurately modeled only by two reference point images and two-dimensional conversion.

Further, the position of the moving object in the measurement area is
Using the distance conversion coefficient ε and the deviation angle Δθ, which are the main parameters of this model, can be easily calculated from the vector value on the image coordinate system, so that the real-time position measurement of the moving object becomes possible. Further, since it is not necessary to install the camera directly above or beside the device as in the related art, the configuration of the monitoring system and the like is simplified. In addition, since the area monitored by one camera can be widened, the system cost can be reduced.

[Embodiment 2] Next, Embodiment 2 of the present invention.
Will be described. In the first embodiment, the first reference point P1
The camera direction is adjusted with respect to
Is the center of the image plane, and the line of sight of the camera (optical axis)
Can be set to the reference point P1 or very close. However, there are cases where the camera direction after installation cannot be freely adjusted, or monitoring is performed while switching the camera direction. In such a case, since the reference point P1 is set according to the camera line of sight, an error occurs between the camera line of sight and the reference point P1. In a second embodiment of the present invention, an example will be described in which this error is corrected by a line-of-sight point and an accurate geometric model is constructed. Note that the system configuration is the same as that of FIG. 1, and this geometric model is also created by the model construction unit 14.

FIG. 8 is a schematic diagram showing a construction process of a geometric model according to the second embodiment. First, a temporary geometric model is constructed using two reference points on the ground (box A).
If the first reference point P1 is the center of the scene and is on the camera line of sight 6, no correction is necessary. However, this is not the case in practice, so a temporary geometric model is constructed using the first reference point P1 as a temporary ground line-of-sight point. This processing is performed in the same manner as the procedure shown in FIG. 3, and the distance conversion coefficient ε and the deviation angle △ θ are obtained and stored in the model data table 16.

In the above provisional geometric model, the first reference point is assumed to be on the line of sight of the camera, so that the measured value contains an error. In the present embodiment, the true gaze position is obtained and included in the geometric model, and the error is removed using this gaze position during position measurement. The position of the true camera line of sight is given as one point (= p0) in the coordinate system of the image memory if the camera is specified. Thus, using the temporary geometric model, the intersection of the camera line of sight from the image coordinate system p0 and the ground coordinate plane is determined as described later, and the true ground start point P0 is determined (box B). Further, the geometric model is reconstructed using this P0 (box C). The details will be described below.

FIG. 9 shows a conceptual diagram of a geometric model including the reference point P1 and the true ground viewpoint P0. The point P0 where the optical axis 6 passing through the center of the camera lens 21 intersects the ground surface is a true ground viewpoint. Here, the line 6 'connecting the lens center and the first reference point is a temporary line of sight.

FIG. 10 shows a flowchart of the ground viewpoint calculation process. First, the line-of-sight position p0 on the image coordinate system 23 is determined (box B-10). The line-of-sight position of the image coordinate system 23 is a point where the images passing through the optical axis 6 are connected, and generally coincides with the center of the image. However, the camera's line of sight and the center of the image on the memory may not coincide with each other due to the camera, lens, image processing, or the like. , This coordinate information (p0
).

Next, a vector p1p0 (r0, θ0) between the point p1 corresponding to the first reference point position P1 on the image memory coordinate system 23 and the line-of-sight position p0 is created. As shown in FIG. 9 (b), this vector has a start point p1 and an end point p0.
, The absolute value r0 and the argument θ0 (B-20). Next, the vector p1p0 (r0, θ0) on the image memory coordinates is mapped on the intermediate coordinate plane 7 to create the vector P1P0 ′ (R0 ′, Θ0) (B-30). Further, a point P on the intermediate coordinate plane 7
0 ′ is mapped to a ground coordinate system to obtain a true ground start point P0. That is, a line connecting the center of the camera 2 and P0 'is extended to the ground surface, and the intersection of the intersection is defined as P0 (B-4).
0).

Next, the correction processing of the geometric model will be described. FIG. 11 shows a conceptual diagram of the modified geometric model. The line of sight 6 in the figure coincides with the optical axis of the lens 21 and intersects the ground surface at the true ground viewpoint P0. The line 8 connects the second reference point P2 to the lens center, and the intersection with the intermediate coordinate plane 7 is P2 '.

FIG. 12 shows a processing procedure for modifying the geometric model. First, pass the line-of-sight point P0 on the ground, and
A plane perpendicular to the straight line 6 connecting the camera and the camera center is newly defined as an intermediate coordinate plane 7 (C-10). All points in the camera scene on the ground can be mapped to this intermediate coordinate plane. That is, when the camera center and an arbitrary point in the scene on the ground coordinate system are connected by a straight line, and an intersection with the intermediate coordinate plane 7 is obtained, a mapping of the arbitrary point is obtained. Conversely, mapping from the intermediate coordinate plane 7 to the ground coordinate system is also possible. If a straight line connecting the camera center and an arbitrary point on the intermediate coordinate plane 7 is extended and an intersection with the ground surface is obtained, an arbitrary point is mapped. .

Next, the position vector p0p2 of the point p2 corresponding to the second reference point from the viewpoint p0 on the image coordinate system 23 is calculated as follows:
Let it be rf * as shown in FIG. Rewrite, rf * =
(rf, θf) (C-20). Further, the second reference point P2 on the ground is mapped on the intermediate coordinate plane 7, and is set as P2 '(C-30). Next, on the intermediate coordinate plane 7, a vector Rf * having the viewpoint P0 as a start point and the point P2 'as an end point is created. Rewritten and expressed as Rf * = (Rf, Θf) (C-4
0). Finally, the norms of the two vectors rf * and Rf * are compared, and the data of the geometric model (distance conversion constant ε, deviation angle Δθ) is recalculated according to Equation 8, and stored in the model data table 16 (C -50).

[0059]

８ = Rf / rf Δθ = Θf−θf As described above, according to the second embodiment, if there is an error between the ground viewpoint based on the camera's line of sight and the set reference point, the set reference point is temporarily determined. A temporary geometric model is constructed by defining an intermediate coordinate plane that passes through this reference point as the ground viewpoint, and then the true gaze position is determined based on the temporary model, and the true position is calculated as a new ground viewpoint. Recalculate the geometric model as

According to this, in building a model in which one of the two reference points set on the ground is the ground viewpoint, even if there is a deviation from the true ground viewpoint, the camera direction and the reference point position are changed. Since the correction can be performed by software processing without performing, simple and accurate position measurement can be realized.

[0061]

According to the present invention, a geometric model for processing an oblique image of an object on the ground and measuring the position is constructed via an intermediate coordinate plane that connects the image plane and the ground plane two-dimensionally. Therefore, it is easy to create a model, and only two reference points need to be set on the ground.

Further, since a true geometric viewpoint can be obtained and an accurate geometric model can be reconstructed by using a temporary model in which one of the reference points is the camera ground viewpoint, it is necessary to change the camera direction and the reference point position. Therefore, it is possible to realize an easy-to-use high-accuracy position measurement system.

Further, since the position can be calculated by a simple algorithm based on two-dimensional conversion, real-time measurement of a moving object is possible, which is suitable for monitoring traffic flow and the like.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a position measurement system of a moving object according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram of a geometric model.

FIG. 3 is a flowchart showing a procedure for creating a geometric model according to the first embodiment.

FIG. 4 is an explanatory diagram showing a geometric relationship in a process of creating a geometric model.

FIG. 5 is a flowchart showing a procedure of a position measurement process.

FIG. 6 is an explanatory diagram showing a geometric relationship corresponding to a measurement procedure.

FIG. 7 is an explanatory diagram showing a geometric relationship of height correction.

FIG. 8 is a flowchart showing an outline of a construction process of a geometric model according to the second embodiment.

FIG. 9 is a conceptual diagram of a geometric model according to the second embodiment.

FIG. 10 is a flowchart showing a ground viewpoint calculation process.

FIG. 11 is a conceptual diagram of a modified geometric model.

FIG. 12 is a flowchart showing a procedure for modifying a geometric model.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Position measuring device main body, 2 ... Camera, 3 ... Moving field, 4 ... Moving body, 5 ... Measurement area, 6 ... Camera line of sight (optical axis), 7 ... Intermediate coordinate plane, 11 ... Control part, 12 ... Image processing Unit, 13 ... image memory, 14 ... geometric model construction unit, 1
5 position measuring unit 16 model data table 21
Lens 22, 22 imaging plate, 23 image coordinate system.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Toshiharu Watanabe 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture Inside the Omika Plant, Hitachi, Ltd. (72) Inventor Katsuyasu Kato 5-5-2 Omika-cho, Hitachi City, Ibaraki No. 1 Hitachi, Ltd. Omika Plant (72) Inventor Kunizo Sakai 5-2-1, Omika-cho, Hitachi City, Ibaraki Prefecture Inside Hitachi, Ltd. Omika Plant

Claims (7)

[Claims] 1. A method of capturing an image with an inclined camera line of sight to a measurement area on the ground and measuring the position of an object recognized by image processing, comprising: a world coordinate system (ground coordinate system) based on a camera installation position; Assuming an intermediate coordinate plane that two-dimensionally connects the image coordinate systems corresponding to the imaging planes of the camera, and a geometrical shape that defines the positional relationship between the world coordinate system and the image coordinate system with respect to the intermediate coordinate plane A position measurement method based on image processing, wherein a position on the ground is determined from a position on the image of the object using the model. 2. The method according to claim 1, wherein the intermediate coordinate plane is set at a position that is perpendicular to the camera line of sight and that shares a ground viewpoint with the world coordinate system where the camera line of sight intersects the ground. Position measurement method using image processing. 3. A method for capturing an image in which a camera's line of sight is tilted with respect to a measurement area on the ground and measuring the position of an object recognized by image processing, wherein the measurement area on the world coordinate system (ground coordinate system) Specifically, a first reference point is set at or near the ground viewpoint of the camera line of sight, and a second reference point is set at an arbitrary position, parallel to the image coordinate system of the camera, perpendicular to the camera line of sight, and An intermediate coordinate plane is set at a position including the first reference point, and a point corresponding to the first reference point on the image coordinate system is set as a start point, and a point corresponding to the second reference point is set as an end point. From the ratio of a first vector and a second vector having the first reference point on the intermediate coordinate plane as a start point and a point on which the second reference point is mapped as an end point, the first and second vectors are obtained. The world coordinate system and the front for the positional relationship of the reference point A distance conversion coefficient and a deviation angle between the image coordinate system are obtained and set in advance as parameters of a geometric model. When the position of the object is measured, a specific position of the object present in the measurement area is determined on the image coordinate system. Sought and said first
An object position vector having a point corresponding to the reference point as a start point and a point corresponding to the specific position as an end point is obtained. A position measuring method by image processing, wherein a position on a coordinate plane is obtained, and the converted position is converted to the world coordinate system to obtain a specific position on the ground. 4. The method according to claim 3, wherein the parameters of the geometric model include direction cosine information (lx, ly, lz) on a world coordinate system whose origin is a ground point immediately below the camera. When the position has height information higher than the ground, a ground position immediately below the position converted into the world coordinate system is obtained by a correction operation using the height information and the direction cosine information. Position measurement method using image processing. 5. A method for capturing an image in which a camera's line of sight is inclined with respect to a measurement area on the ground and measuring the position of an object recognized by image processing, wherein the measurement area on the world coordinate system (ground coordinate system) is Specifically, the first reference point is set at a position within a range allowed as a temporary ground viewpoint of the camera line of sight, and the second reference point is set at an arbitrary position. The camera and the first reference point A temporary intermediate coordinate plane is set at a position that is perpendicular to a straight line connecting points and includes the first reference point, and a point corresponding to the first reference point on the image coordinate system is a start point, and the second A first vector whose end point is a point corresponding to the reference point, and a second vector whose start point is the first reference point on the temporary intermediate coordinate plane and whose end point is a point on which the second reference point is mapped.
The distance conversion coefficient and the deviation angle between the world coordinate system and the image coordinate system with respect to the positional relationship between the first and second reference points are determined from the ratio of the first and second reference points, and are set as parameters of a temporary geometric model. Then, a third vector having an image viewpoint given as a position of the camera line of sight (optical axis) on the image coordinate system as an end point and a point corresponding to the first reference point as a start point is obtained. The third vector is mapped to the intermediate coordinate plane using the parameters of the temporary geometric model, a corresponding position of the viewpoint on the intermediate coordinate plane is determined, and the corresponding position is converted to a ground coordinate system. Mapping to obtain a real ground viewpoint, based on the real ground viewpoint, perpendicular to the camera line of sight, and set a true intermediate coordinate plane including the ground viewpoint, the first reference point the Replace the distance conversion coefficient with the ground viewpoint Recalculate the deviation angle to construct a true geometric model, and when measuring the position of the object, determine the specific position of the object existing in the measurement area on the image coordinate system, and correspond to the ground viewpoint. To obtain an object position vector having a point corresponding to the specific position as a start point and an end point corresponding to the point corresponding to the specific position, mapping the object position vector on the intermediate coordinate plane by a recalculated distance conversion coefficient and a deviation angle, and further mapping the world coordinate. A position measuring method by image processing, wherein a specific position on the ground is obtained by mapping to a system. 6. An image memory in which a camera is installed at a position where a tilt image is photographed with respect to a measurement area on the ground, and an image memory for storing an input image from the camera, and image processing for recognizing an object in the input image by image processing An apparatus for measuring the position of an object by image processing, comprising: means for specifying first and second reference points in the measurement area, a world coordinate system defining the reference points, and imaging by the camera Assuming an intermediate coordinate plane that two-dimensionally connects the image coordinate systems corresponding to the planes, and creating a geometric model that defines the positional relationship between the world coordinate system and the image coordinate system via the intermediate coordinate plane Model construction means, model data storage means for storing parameters of the created model, and position measurement for obtaining a ground position from a position on the image of the object using the parameters of the geometric model. Position measuring device by image processing, characterized by comprising a stage, a. 7. The model construction unit according to claim 6, wherein the model construction means has a function of correcting an error of the geometric model due to an error between a ground viewpoint where a camera line of sight intersects and the first reference point. A position measuring device by image processing characterized by the following.