WO2012144848A2

WO2012144848A2 - Methods of three-dimensional recognition via the acquisition of a two-dimensional image consisting of multiple beams of transmitted light having a preset magnification angle

Info

Publication number: WO2012144848A2
Application number: PCT/KR2012/003052
Authority: WO
Inventors: 이태경; 정제교; 박철우; 김영준
Original assignee: Yi Tae Kyoung; Jung Je Kyo; Park Chul Woo; Kim Young Jun
Priority date: 2011-04-21
Filing date: 2012-04-20
Publication date: 2012-10-26
Also published as: WO2012144848A3; KR101188357B1

Abstract

The present invention relates to a method which involves projecting multiple point sources of light having a preset projection angle to a space, acquiring, as two-dimensional information, images of projected points created through the reflection of each point source of light from an ambient environment or from a surface of an object via one light-intercepting unit having a preset geometric relationship to a unit for projecting point sources of light, and recognizing three-dimensional characteristics of the ambient environment or of the object based on said two-dimensional information.

Description

3D Recognition Method through 2D Image Acquisition of Multiple Projection Light with Preset Magnification Angle

The present invention projects a plurality of point light sources having a preset projection angle in a space, and displays an image of the projection points made by reflecting each projected point light source from a surface of an environment or an object (hereinafter referred to as an object). The present invention relates to a method of recognizing three-dimensional characteristics of an object based on the two-dimensional information obtained through one light receiving unit having a predetermined geometric relationship with respect to the information.

In general, three-dimensional recognition is regarded as an essential element in the operation of automated devices and devices and the voluntary functioning. There are various fields that require 3D recognition, from not only toys, cars and robots used in everyday life, but also to drones and missiles.

The development of the Global Positioning System (GPS) has resulted in the commercialization of 3D coordinates at any point in the open area, but the spatial recognition in GPS jamming or indoor environments where GPS is relatively unacceptable. The technology does not implement standardized recognition methods. At present, a method of recognizing indoor space using stereo vision and a laser range finder using two cameras is introduced.

However, although stereo vision can distinguish three-dimensional recognition and specific objects, the extraction of feature points necessary for matching images acquired by two cameras is very sensitive to the change in the illumination of the surroundings. Because the pattern must be recognized, the computational amount is very large compared to the algorithm of simply finding a color or detecting an edge. In addition, the use of two cameras mechanically increases the system size, the alignment of the cameras must be very precise, and additional difficulties arise in that the calibration of the two images must be performed again by a program.

In addition, since the laser range finder can only detect the relative distance on the two-dimensional plane, there are inherent limitations as a method for obtaining three-dimensional spatial information.

Therefore, there is a need for an efficient method that enables three-dimensional recognition through a simple hardware configuration based on a simple algorithm. If such an efficient method is developed, a three-dimensional space recognition field (for example, a cleaning robot) is developed. It is expected to be widely used in many fields that require 3D space recognition, from military applications such as small drones that can be used for the game recognition interface, etc.) as well as independent missions without being affected by GPS jamming. do.

An object of the present invention to overcome the above limitations of the prior art is to provide a method capable of three-dimensional spatial recognition using only one two-dimensional image acquisition means.

In particular, the present invention provides a three-dimensional recognition method through the two-dimensional image acquisition of the multi-projection light that can speed up the image processing and downsizing the image processing processor by eliminating the burden of the excessive amount of the conventional three-dimensional recognition technology stereo vision There is that purpose.

A three-dimensional recognition method by acquiring two-dimensional images of multi-projection light according to the present invention includes a first step of projecting a plurality of projection light beams having a predetermined projection angle with respect to the center projection line to a three-dimensional object; and centering on the center projection line A second step of acquiring a pattern of projection points formed on the three-dimensional object as a plurality of acquisition points having respective coordinate values through the two-dimensional image detection surface of the light receiving sensor while the acquisition lines are located on the same plane; and A third step of determining a one-to-one correspondence between the plurality of projection lights and an acquisition point; And a fourth step of calculating each acquisition distance from a geometric relationship between the projection angle and an acquisition angle and a focal length of the acquisition point.

Wherein the fourth step is to calculate the acquisition distance through the following equation (1) and (2).

Expression (1)

..... Equation (2)

Where L is the acquisition distance, f is the focal length, α is the projection angle and β is the acquisition angle.

The present invention further includes a step 4-1 of converting the acquisition distance into an origin distance through Equation (3) below.

Expression (3)

Where D is the origin distance.

When the center projection line and the center acquisition line do not coincide in the second step, an angle formed between the center projection line and the center acquisition line based on the center projection point may be 20 ° or less.

The present invention may further include a fifth step of calculating the surface shape of the three-dimensional object by curvedly fitting the origin distance information of the plurality of acquisition points.

Meanwhile, the third step may be performed by measuring the gradient characteristic of the measurable property values applied to the projection lights at the acquisition point and correlating each other.

In addition, the first step may allow the plurality of projection lights to form a projection pattern of equally spaced grid arrays or conformal concentric arrays based on the center projection line, and the third step may extend radially about the center projection line. The acquisition points may correspond one-to-one in correspondence with the arrangement order of a series of projection lights existing on a line.

In this case, the projection pattern may be made by projecting a single light source onto a diffraction grating having a preset dot pattern.

Meanwhile, the present invention may further include a step 2-1 of comparing the number of projection light beams with the number of acquisition points to determine whether a missing acquisition point exists.

And if it is determined that the acquisition point lost in step 2-1 exists, the method further includes step 2-2 of adjusting the direction of the center projection line such that the number of the acquisition points is equal to the number of the projection light beams. have.

Alternatively, when it is determined that the acquisition point lost in step 2-1 exists, a part of the plurality of projection lights is blocked by the light blocking unit, and it is determined whether the acquisition point that does not disappear among the acquisition points exists in the lost acquisition point. The method may further include performing a 2-2 ′ step of repeatedly determining the corresponding projection light for all the projection light.

Here, the step 2-2 'simultaneously blocks a series of projection light beams having the same radial distance with respect to the center projection line among the projection light beams of the projection pattern to the light blocking unit, and radially arranges the blocked projection light beams. By determining whether there is an acquisition point that does not disappear among the acquisition points corresponding to the order, the process of determining the projection light corresponding to the lost acquisition point can be repeated for all the projection lights.

In particular, the light blocking unit may sequentially block the series of projection lights having the largest radial distance or the series of projection lights having the smallest radial distance.

The light blocking unit may be a physical shutter or an optical shutter which is a liquid crystal glass in which some regions are converted to non-transmissive light by application of a current.

In this case, when the projection pattern is a concentric concentric array, the physical shutter may have a structure of a camera shutter.

Alternatively, when the projection pattern is an equally spaced grid array, the physical shutters may be a pair of shutters having "b" and "a" shapes and moving in a diagonal direction toward the center projection line.

On the other hand, the sequential blocking of the light blocking unit is performed in a preset pattern, and when the light blocking unit is positioned in a preset positional relationship with respect to the projection focus and sequential blocking is performed for each projection light, the projection at a certain point of time It is also possible to designate or confirm the projection angle of the projection light just before blocking according to the blocking phase with respect to the focus.

In addition, the projection light is preferably a laser light having a wavelength that is distinct from the wavelength of light existing around the three-dimensional object.

The present invention having the configuration as described above has the effect that can be three-dimensional recognition by only one two-dimensional image acquisition means.

In addition, since the projection point is identified and three-dimensional recognition is performed through simple filtering without complex image processing in the two-dimensional image, the information processing amount is very small, thereby speeding up the image processing and miniaturizing the image processing processor. There is an advantage.

In addition, by giving the projection light source a specific wavelength that is distinguished from the wavelength emitted from the object periphery, the projection point can be stably recognized regardless of the weakness of the peripheral illumination, which is one of the inherent problems of the image acquisition means. .

1 is a diagram illustrating the basic principle of triangulation used in the present invention.

2 is a diagram schematically showing a three-dimensional geometric relationship between a projection point and an acquisition point.

3 is an orthogonal projection of the three-dimensional projection line and the acquisition line shown in FIG. 2 in the XZ plane with respect to the reference origin;

4 is a view schematically showing that a plurality of projection light is detected as an acquisition point on the two-dimensional image detection surface of the light receiving sensor after forming the projection point on the three-dimensional object.

FIG. 5 is a diagram of a first projection pattern in which a plurality of projection lines form an equally spaced grid array based on a center projection line. FIG.

FIG. 6 is a view of a second projection pattern in which a plurality of projection lines form an isoconcentric array based on a center projection line. FIG.

FIG. 7 schematically illustrates a configuration of forming a constant projection pattern using a diffraction grating. FIG.

8 is a diagram illustrating a phenomenon in which a projection point is formed in an undercut portion of a three-dimensional object to lose an acquisition point.

9 is a view showing a configuration of a light blocking portion for a projection pattern of a conformal concentric array.

10 is a view for explaining the principle of determining the lost acquisition point through the sequential blocking using the light shield.

FIG. 11 is a diagram showing a configuration of a light blocking unit for a projection pattern of an equally spaced grid array.

12 is a flowchart illustrating a three-dimensional recognition method through two-dimensional image acquisition of multi-projection light according to the present invention.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

In the description of one embodiment of the present invention, descriptions of well-known configurations that will be apparent to those skilled in the art will be omitted so as not to obscure the subject matter of the present invention.

In addition, when referring to the drawings it should be considered that the thickness of the lines shown in the drawings or the relative size between the components may be exaggerated for clarity of explanation and convenience of understanding.

According to the present invention, after a projection image having multiple projection lights having a predetermined magnification angle formed on a surface of a three-dimensional object 10 is obtained as a two-dimensional image, an object 10 is obtained through pattern analysis of the multiple projection light shown in the two-dimensional image. It relates to a method of recognizing a three-dimensional surface shape of).

In more detail, when a plurality of point light sources are projected to reach the surface of the three-dimensional object 10 to form an image, the two-dimensional orthogonal image of the images is the surface of the three-dimensional object 10. Depending on the shape, it will be different from the original projection pattern (hereinafter, a plurality of projected light sources are projected light P, and the image formed by the point light source reaching the surface of the three-dimensional object 10 is' Projection point (PP) ').

Therefore, since the pattern of projection points formed on the surface of the three-dimensional object 10 contains the surface information of the three-dimensional object 10, analyzing the pattern information inversely infers the three-dimensional surface shape of the object 10. It can be intuitively understood that it will be possible.

The method of inferring the three-dimensional surface shape of the object 10 from the pattern of the projection points formed on the surface of the three-dimensional object 10 calculates the distance from one point for observing (acquiring) the two-dimensional pattern of the projection points to each projection point. It is. If the distance to each projection point is known, finite number of discontinuous coordinate information about the surface shape of the object 10 seen from the observation point is obtained, and the coordinate information of these projection points is obtained by using a known algorithm. 3D Surface Fitting can infer the approximate surface shape.

Although the most basic outline of the present invention is described above, there are some technical difficulties to be solved in actual implementation.

The first relates to a method of calculating the distance from one point of observation of a two-dimensional pattern of projection points to each projection point PP. This can be overcome relatively simply among the various difficulties according to the principle of triangulation to be described later.

Second, when a plurality of projection light beams P projected from a light source reach the surface of the object 10 and acquire a projection point PP formed as a two-dimensional image from an observation point, a plurality of acquisition points exist in the two-dimensional image. (CP) Determine which projection light P corresponds to each. This is important because the principle of triangulation itself cannot be applied unless this one-to-one correspondence is determined.

Third, depending on the surface shape of the object 10, a certain projection point may not be obtained at the observation point. If the loss of the acquisition point CP is not found, the acquisition point CP and the projection light P Errors can occur in a one-to-one correspondence between them. The loss of the acquisition point CP may occur when a projection point is formed in the undercut portion of the object 10.

In addition, it is necessary to consider a configuration for forming a plurality of point light sources (projected light) projected onto the object 10, a configuration for determining or determining the angle at which each point light source is projected.

Hereinafter, the configuration of the present invention to solve the above technical difficulties will be described in detail.

The basic principle of calculating the distance from one point of observation to the two-dimensional pattern of projection points is the principle of triangulation.

The principle of triangulation is a classical geometry-based survey method, if the observer occupying the position of A knows the distance between AB and BC for the other two points B and C, then the distance between the unknown AC It is based on the principle of knowing.

In practice, measuring the distance between A-B and B-C to find the distance between point A and point C is not only inefficient, indirect, but also difficult. Rather, it would be straightforward and easy to measure the distance between A and C immediately. It may also be impossible to measure the distance between A-B or B-C directly due to the influence of the terrain.

This can be solved by a blending method that measures the angle between the imaginary lines connecting the three points to the distance measurement. As shown in FIG. 1, another point that can be easily measured by the measurer at point A, for example, For example, if you measure only one distance to point B, and then measure the angles from the point A and point B to the point that you actually need to survey, you can find the distance between AC and BC.

The present invention also calculates the distance from one point of observing the two-dimensional pattern to each projection point (PP) through the combination of the above-described distance and angle measurement. In more detail, the light reception for the light receiving sensor 200 in which an image of the projection focal point Pf, which is the focal point where the plurality of projection light beams P converge, and the projection point PP that is formed on the three-dimensional object 10 is obtained. Acquisition angle CA calculated through the distance between the focal points Cf, the projection angle PA of each of the plurality of projection light beams P, and the coordinates of the respective acquisition points CP detected on the light receiving sensor 200. If we know, we can finally calculate the distance to the projection point PP.

Here, the projection angle PA is easily determined because it is determined by the optical design of the light transmitting part 100 that generates a plurality of projection light beams P, but the acquisition angle CA is the surface shape of the three-dimensional object 10. You have to find it by calculation because it is changed from time to time. As will be described in detail later, the acquisition angle CA uses a trigonometric function to determine the two-dimensional coordinates of the acquisition point CP formed on the light receiving sensor 200 and the geometric relationship between the acquisition point CP and the known light reception focal point Cf. Can be calculated by solving.

A method of calculating a distance between the acquisition point CP and the projection point PP conceptually described above will be described in detail with reference to FIG. 2.

FIG. 2 schematically illustrates the projection point PP for the two projection lights P and the acquisition point CP formed on the two-dimensional image detection surface 210 of the light receiving sensor 200. In addition, prior to the detailed description, each point and virtual line, length, angle, etc. shown in FIG. 2 will be described as follows, and this definition will be maintained throughout the detailed description as well as in the claims unless otherwise specified.

The projection line PL refers to a traveling path through which the projection light P whose projection angle PA is finally adjusted (determined) is projected toward the object 10. In this case, the final determination of the projection angle PA is performed after the intermediate process of dividing the light source from a single light source into a plurality of light sources to form a projection pattern and adjusting the projection magnification. This is because the traveling path of the projection light P toward (10) is determined.

The center projection line PLc refers to one projection line PL which is a reference among the plurality of projection lines PL. Although it can be arbitrarily determined, there is no change in straightness even after a single light source is divided into multiple light sources and the projection magnification is adjusted. It is convenient to use the projection line PL as the center projection line PLc. Accordingly, the angle formed by each projection line (except for the center projection line) with the center projection line PLc is defined as the projection angle PA. The projection point PP corresponding to the center projection line PLc is referred to as the center projection point PPc.

The projection focus Pf means a point at which all the projection lines PL converge, and do not necessarily coincide with a single light source. However, although the projection focus Pf may be set for each projection line PL according to the projection pattern and the projection magnification, it is obvious that matching the projection focus Pf to one is simple and convenient.

The acquisition line CL is an imaginary extension line connecting the projection point PP formed on the three-dimensional object 10 and the light reception focus Cf, and the acquisition line CL is a two-dimensional image detection surface of the light reception sensor 200 ( An intersection formed through 210 is called an acquisition point CP. Therefore, the acquisition point CP has a coordinate value on the two-dimensional image detection surface 210. As will be described later, the coordinate value of this acquisition point CP is used to calculate the distance to each projection point PP.

The light receiving focus Cf means a point at which all the acquisition lines CL converge. In an embodiment of the present invention, the projection focus Pf and the light receiving focus Cf are located on a straight line along the center projection line PLc. Is arranged to. Accordingly, since all the acquisition lines CL converge to the light reception focal point Cf, the center acquisition line CLc that returns to the light reception focal point Cf after the center projection line PLc forms the projection point PP is the center projection line. (PLc).

In addition, the acquisition point CP formed by the central acquisition line CLc is defined as the central acquisition point CPc, and the acquisition angle CA is the central acquisition line CLc and each acquisition line similarly to the projection angle PA. It is defined as the angle formed by (CL) (excluding the center line). In addition, the straight line distance between the straight line projection focus Pf and the light reception focus Cf is defined as the focal length f. Of course, the projection focus Pf and the light reception focal point Cf can be produced by a conventional optical apparatus using a lens.

The reference origin O is a three-dimensional coordinate origin existing on the central acquisition line CLc or on an imaginary line extending from the central acquisition point CPc, and refers to a reference point recognized by a device in which the present invention is implemented. Therefore, unless otherwise mentioned below, the distance to the three-dimensional object 10 should be understood as referring to the origin distance OD which is the distance between the reference origin O and the projection point PP. Of course, the light receiving focus Cf may be set as the reference origin O, and the coordinate axis definition of the reference origin O when describing an embodiment of the present invention is directed from the reference origin O toward the center acquisition point CPc. The coordinate axis will be the Z axis.

However, since the distance from the coordinate of the acquisition point CP, which is the intersection point of the two-dimensional image detection surface 210 of the acquisition line CL and the light receiving sensor 200, to the projection point PP must be calculated, the initial calculated value Denotes an acquisition distance CD, which is a distance between the projection point PP and the acquisition point CP, before the origin distance OD, which is the distance between the reference origin O and the projection point PP. However, since the geometric relationship between the two-dimensional image detection surface 210 and the reference origin O of the light receiving sensor 200 is set in advance, the conversion of the origin distance OD from the acquisition distance CD is a simple calculation. It can be done easily.

Based on the above definition, the two-dimensional image detection surface 210 of the projection focus Pf, the light reception focus Cf, the projection line PL, the acquisition line CL, and the light receiving sensor 200 shown in FIG. A relational formula for calculating the origin distance OD is derived from the geometric relations of the back. 3 shows the three-dimensional projection line PL and the acquisition line CL shown in FIG. 2 in the XZ plane with respect to the reference origin O, and all distances and angles are represented by orthogonal projections to the XZ plane. . Accordingly, when referring to each distance and angle, in principle, a description indicating "XZ orthogonal projection" should be attached, but the description of "XZ orthogonal projection" will be omitted for convenience.

First, using the first law of sine, the following equation (1) holds.

Expression (1)

In Equation (1), L denotes an acquisition distance, f denotes a focal length, α denotes a projection angle PA, and β denotes an acquisition angle CA. The focal length (f) and the projection angle (PA) are known values, and the acquisition angle (CA) is the "acquisition point (CP) and center" for the "distance between the light receiving focus (Cf) and the center acquisition point (CPc)". The tangent function of "distance between acquisition points CPc" is calculated by the following equation (2). Of course, the distance between the acquisition point CP and the central acquisition point CPc may be calculated by the coordinate value of each acquisition point CP.

..... Equation (2)

Therefore, according to equations (1) and (2), the acquisition distance CD is obtained from the calculation of a simple trigonometric function.

Next, the origin distance OD can be obtained from Equation (3) applying the second cosine law.

Expression (3)

Applying these equations (1) to (3) to each projection line PL and acquisition line CL, the origin distance OD with respect to the projection point PP formed on the surface of the three-dimensional object 10 can be obtained. By performing the surface fitting using the plurality of origin distances OD, an approximate surface shape of the 3D object 10 can be recognized. Of course, it is obvious that the center projection line PLc and the center acquisition line CLc having the projection angle PA and the acquisition angle CA of 0 ° serve as a reference but cannot obtain distance information therefrom.

4 illustrates that a plurality of projection lights P are detected as acquisition points CP on the two-dimensional image detection surface 210 of the light receiving sensor 200 after forming the projection points PP on the three-dimensional object 10. It is shown schematically, by performing the above series of calculation process for each acquisition point (CP) to calculate the respective acquisition distance (CD) (convertable to the origin distance) to determine the surface shape of the three-dimensional object (10) It is clearly shown that it is possible to calculate roughly. Here, the two-dimensional image detection surface 210 of the light receiving sensor 200 is disposed perpendicular to the plane where the center projection line PLc and the center acquisition line CLc are located. In addition, as can be seen from the above equation, since the origin distance (OD) calculation for calculating the surface shape of the three-dimensional object 10 is performed using a simple trigonometric function, the information throughput is extremely small, which is considerably compared to the conventional three-dimensional recognition method. It can be seen that high speed processing can be enabled.

On the other hand, the embodiment of the present invention has been described that the projection focus (Pf), the light receiving focus (Cf) and the reference origin (O) are all arranged in a straight line, in practice, the center projection line (PLc) and the center acquisition line (CLc) It can also be expected that this design does not match completely and is designed to open slightly around the projection point (PP). This is because, as the projection focus Pf and the light reception focus Cf are arranged on a straight line, the installation of the optical system to separate and process the projection line PL and the acquisition line CL, which overlap or intersect, may be complicated. Therefore, an embodiment of the present invention includes a configuration in which the center projection line PLc and the center acquisition line CLc do not coincide with each other but are arranged on at least one plane. 3, the light receiving focus Cf has a height difference with respect to the projection focus Pf along the Y axis perpendicular to the ground.

When the projection focal point Pf and the light receiving focal point Cf are spaced apart from each other, an error occurs as much as the center acquisition line CLc is converted into orthogonal projection to the XZ plane. However, the error due to the orthogonal transformation of the center acquisition line CLc to the XZ plane is proportional to the cosine of the angle between the center projection line PLc and the center acquisition line CLc with respect to the projection point PP. If is small, the error due to the orthogonal transformation is negligible from the practical point of view.

For example, if the angle θ between the center projection line PLc and the center acquisition line CLc is 20 °, the cos20 ° ≒ 0.940, so that the center projection line PLc and the center acquisition line CLc coincide with each other. As a result, the error level is about 94%, that is, about 6%, so that a fairly accurate value can be obtained. When the light receiving focus Cf to the acquisition distance CD is converted into the ratio of the distances spaced apart from the projection focus Pf along the Y axis, this is a value corresponding to cotan20 ° (752.75). If the distance to) is approximately three times the distance of the light receiving focal point Cf, it means that an error level of about 6% can be maintained geometrically. That is, even if the light receiving focus Cf is spaced about 10 cm apart, the surface contour of the three-dimensional object 10 separated by 28 cm or more can be calculated with considerable precision, which is a practically meaningful result.

On the other hand, the wavelength of the projection light (P) has a specific wavelength that is distinct from the wavelength of the light around the three-dimensional object 10 to solve the weakness that is sensitive to the peripheral illumination, which is one of the inherent problems of the image acquisition means It is preferable. For example, the projection light P having the wavelength of the infrared band which is distinguished from the wavelength of the visible light band can be applied.

As described above, it was confirmed that the calculation of the distance from one point of observing the two-dimensional pattern of the projection points PP, in particular, the reference origin O to each projection point PP can be solved using the principle of triangulation. The next problem to be solved is to determine which projection light P corresponds to each of the plurality of acquisition points CP present in the 2D image.

In the simplest case, a gradient characteristic is given to physically measurable physical properties such as wavelength (frequency) or output included in a plurality of projection lights P. That is, each projection light P imparts a physical property distinguished from other projection light P, measures the physical property at each acquisition point CP, and corresponds it to each projection light P, thereby projecting light To determine the one-to-one correspondence between (P) and the acquisition point (CP).

However, this method has a disadvantage in that it is structurally complicated because it is necessary to provide a structure having a meaningful gradient characteristic to each of the plurality of projection lights P and detecting them again. In some cases, physical properties of the projection light P may be altered in the process of being scattered or passed through heterogeneous media (eg, water droplets or dust in the air). Therefore, there is a need for a simpler and more reliable alternative.

5 and 6 show projection patterns of the multiple projection light beams P in the embodiment of the present invention. FIG. 5 is a first projection pattern 112 in which a plurality of projection lines PL form an equally spaced grid array (the same distance along the X and Y axes) based on the center projection line PLc, and FIG. 6 is a center projection line PLc. ) Shows a second projection pattern 114 in which a plurality of projection lines PL form a concentric concentric array.

When the projection light beams P are arranged to be symmetrical with respect to the center projection line PLc as described above, in the embodiment of the present invention in which the projection focus Pf and the light reception focus Cf are in a straight line, each acquisition point ( CP is moved only along the line connecting the corresponding projection line PL and the center projection line PLc in the radial direction. That is, as shown in FIG. 5 and FIG. 6, the acquisition point CP corresponding to each projection line PL indicated by a large circle only moves its position on a line extending radially about the center projection point PPc. . This occurs because the center projection line PLc and the center acquisition line CLc coincide.

Using this phenomenon, the one-to-one correspondence between the plurality of projection lights P and the acquisition point CP is very simple. The series of projection lights P existing on the line extending radially with respect to the center projection line PLc only changes its position little by little, and the arrangement order is maintained. Therefore, the projection light P and the All you have to do is match the acquisition points one-to-one. However, the maintenance of the arrangement order is effective only along a line extending radially with respect to the center projection line PLc, so that the second projection pattern 114 constituting the conformal concentric arrangement of FIG. 6 is more convenient.

In this case, the first and

second projection patterns

112 and 114 may be formed using a diffraction grating 110 as shown in FIG. 7. When a dot pattern is formed on the diffraction grating 110 by etching or laser processing, a plurality of projection light beams P having a constant pattern, such as the first /

second projection patterns

112 and 114 shown by the dot pattern, are formed. The projection angle PA of each projection light P has a predetermined value according to the design of the dot pattern. In addition, the light source adopts a laser light source having almost no light spread and excellent straightness.

Next, to solve the problem of the loss of the acquisition point (CP) that can occur according to the surface shape of the object (10). As described above, the loss of the acquisition point CP may occur when the projection point PP is formed in the undercut portion of the object 10, an example of which is illustrated in FIG. 8. This is because, as the projection focus Pf and the light reception focus Cf are arranged in a straight line, the projection focus Pf must be disposed earlier in the projection direction than the light receiving sensor 200, so that the projection angle PA is This is because it is larger than the acquisition angle CA.

If the acquisition point CP is lost, an error occurs in the one-to-one correspondence between the projection light P and the acquisition point CP. If the loss of the acquisition point CP has occurred, the acquisition of the projection light P It is necessary to determine whether the point CP is lost.

Whether the acquisition point CP is lost can be determined by determining whether the number of the acquisition points CP is short compared to the number of the projection lights P. FIG. In this case, by adjusting the directions of the center projection line PLc and the center acquisition line CLc, the number of acquisition points CP may be the same as the number of projection light beams P, thereby avoiding the loss of the acquisition points CP. .

However, the above avoidance measures are not always valid, and there is a fear that the adjustment of the direction of the center projection line PLc and the center acquisition line CLc may interfere with the shape of the three-dimensional object 10.

Therefore, in the present invention, when the loss of the acquisition point CP occurs, the lost acquisition point (loss point) is calculated when calculating the origin distance OD by determining which projection light P corresponding to the projection point P is lost. ). This utilizes the property that a series of projection lights P existing on a line extending radially about the center projection line PLc only changes its position little by little but maintains the arrangement order.

As shown in FIG. 9, the outermost projection light P of the grid array of the light transmission pattern (in FIG. 9 using the conformal concentric array of FIG. 6 as an example), that is, the series of projection light furthest from the center projection line PLc ( When the light blocking unit 120 sequentially moves inwardly from P), the acquisition point CP corresponding to the projection light beams P obscured by the light blocking unit 120 is determined by the light receiving sensor 200. It is disappeared, and by determining the change of the acquisition point CP before and after being covered by the light blocking unit 120, it is possible to know which projection light P corresponds to the lost acquisition point CP.

The second projection pattern 114 and its acquisition point CP will be described in more detail with reference to FIG. 10. The two-digit subscripts shown in the projection light (P) and the acquisition point (CP) are given to distinguish each projection light (P) and the acquisition point (CP). The first digit is centered on the center projection line (PLc). The radially extending line is displayed in order along the counterclockwise direction, and the second digit is numbered along the direction away from the center projection line PLc (outer direction).

Except for the center projection point (PPc) and the center acquisition point c), each projection light (P) and the acquisition point (CP) are arranged up to 01 ... 04, 11 ... 14, ..., 41 ... 44 The light blocking unit 120 sequentially covers the projection light beams P in the circumferential direction of which the second digit is the same.

If the outermost acquisition point CP ₃₃ in radial line 3 of the acquisition points CP is still present in the light receiving sensor 200 when the projection light P to subscripts 04 ... 44 is covered, PP It can be determined that the projection point PP formed by the projection light P denoted as ₃₄ is lost (the acquisition points CP indicated by gray circles in FIG. 10 are the outermost acquisition disappeared by the blocking of the projection light P). Points), the origin distance OD calculation for this PP ₃₄ projection light P is not performed.

The vanishing point determination is sequentially performed on the same circumferential projection light beams P, and it is determined whether or not the projection light beams P which have been vanishing point determination remain blocked at the next vanishing point determination or are unblocked. It doesn't matter. Also, it does not matter whether the projection light P is blocked from the outermost or starts from the center projection line PLc. In other words, this is only a difference between sequential opening or sequential blocking, and there is no difference in effect.

In addition, the sequential light blocking or light opening as described above may be used as an auxiliary means for designating the projection angle PA of the projection line PL. That is, when the light blocking unit 120 is positioned in a preset positional relationship with respect to the projection focus Pf and the blocking or opening is sequentially performed with respect to the projection light P in a preset pattern, the projection focus at any temporary point If the blocking phase with respect to (Pf) can be known, the projection angle PA of the immediately before the open or blocked projection line PL can be known. This is possible because the projection pattern of the projection light P is already designed, and an auxiliary means (such as specifying or confirming the projection angle) in the disappearance of the projection point PP causing a problem of one-to-one correspondence of the preset projection angle PA. It can be used as.

In the case of the second projection pattern 114 having the conformal concentric array, the light blocking unit 120 may preferably adopt the same structure as a general camera shutter, and the first projection pattern 112 having the equal interval grid array. As shown in FIG. 11, a pair of shutters having "b" and "a" shapes and moving along the diagonal direction toward the center projection line PLc may be used.

In addition, the light shielding for vanishing point determination may be applied not only to this physical shutter method but also to an optical shutter such as the liquid crystal glass 124. The liquid crystal glass 124 is installed as a light transmission window disposed on the optical path of the projection light P, and controls a current applied to the liquid crystal glass 124 to sequentially convert some regions into non-transmissive light. The function of the light blocking unit 120 may be implemented. Such an optical shutter has advantages in that it is structurally simple, does not generate mechanical vibration, and relatively freely designs a part that does not transmit light, compared to a physical shutter.

Of course, it is also possible to use a shutter that sequentially blocks each projection line one by one, but there are disadvantages in that it is structurally complicated and inferior in efficiency of vanishing point determination.

In the above description, according to the present invention, two of the projection points PP having the predetermined magnification angle, that is, the multi-projection light in which the projection angle PA of each projection light P is defined on the surface of the three-dimensional object 10 are formed. It has been described above whether the three-dimensional surface shape of the object 10 is recognized from the dimensional image.

The present invention can be summarized in the following series of processes, which are shown in a flowchart in FIG.

The three-dimensional recognition method through two-dimensional image acquisition of the multi-projection light according to the present invention is largely implemented through four steps.

The first step is a step of projecting a plurality of projection light beams P having a preset projection angle PA with respect to the center projection line PLc to the three-dimensional object 10. Here, the plurality of projection light beams P may be formed by projecting a single light source onto the diffraction grating 110 having a preset dot pattern, and being separated into a plurality of light, in particular, an equal interval based on the center projection line PLc. The diffraction grating 110 may be designed to form a projection pattern of a lattice array or a concentric array.

In the second step, the pattern of the projection points PP formed on the three-dimensional object 10 is displayed in the two-dimensional image of the light receiving sensor 200 with the center acquisition line CLc positioned on the same plane with respect to the center projection line PLc. Acquiring a plurality of acquisition points (CP) knowing each coordinate value through the detection surface 210. Even if the center projection line PLc and the center acquisition line CLc located on the same plane do not coincide with each other, the angle formed between the center projection line PLc and the center acquisition line CLc based on the center projection point PPc is determined. If you set below 20 °, you can expect about 94% accuracy.

Next, after the third step of determining the one-to-one correspondence between the plurality of projection lights P and the acquisition point CP, the acquisition angle CA and the focal length of the projection angle PA and the acquisition point CP are determined. f) performing a fourth step of calculating each acquisition distance CD from the geometric relationship therebetween.

In the fourth step, the acquisition distance CD is calculated using the above-described equations (1) and (2), and the first calculated distance (CD) is calculated as the origin distance (OD) through the equation (3). Step 4-1 of converting may be further performed.

On the other hand, the third step may determine the one-to-one correspondence of each projection light P and the acquisition point CP by measuring and matching the gradient characteristics of the measurable property values applied to the projection lights P at the acquisition points CP. have.

Alternatively, in the first step, when the plurality of projection light beams P forms a projection pattern of an equidistant lattice array or an equiaxed concentric array based on the center projection line PLc, the third step may be based on the center projection line PLc. As a result, the acquisition points CP may correspond one-to-one in correspondence with the arrangement order of the series of projection lights P existing on the radially extending line.

On the other hand, the present invention may further include a step 2-1 of comparing the number of projection light (P) and the number of acquisition points (CP) to determine whether there is a missing acquisition point (CP), the second If it is determined that the acquisition point CP lost in step -1 exists, the second operation of adjusting the direction of the center projection line PLc such that the number of acquisition points CP is equal to the number of projection light beams P is performed. It may further comprise two steps.

Instead of adjusting the direction of the center projection line PLc, a part of the plurality of projection lights P is blocked by the light blocking unit 120, and whether there is an acquisition point CP that does not disappear among the acquisition points CP. It is also possible to apply the 2-2 'step of repeatedly performing the process of determining the projection light P corresponding to the lost acquisition point CP for all the projection lights P by determining.

Here, step 2-2 'simultaneously blocks a series of projection light beams P having the same radial distance with respect to the center projection line PLc among the projection light beams P of the projection pattern at the same time with the light blocking unit 120, The projection light P corresponding to the lost acquisition point CP is determined by determining whether there is an acquisition point CP that does not disappear among the acquisition points CP corresponding to the radial arrangement order of the blocked projection lights P. ) May be repeated for all the projection light beams (P).

In particular, the light blocking unit 120 may start blocking sequentially from a series of projection light beams P having the largest radial distance or a series of projection light beams P having the smallest radial distance. The furnace may be an optical shutter, which is a liquid crystal glass 124 in which some regions are converted to non-transparent to light by application of a physical shutter or a current.

In this case, when the projection pattern is an isometric concentric array, the physical shutter may apply the structure of the camera shutter. If the projection pattern is an equally spaced grid array, the physical shutters have the shapes of "b" and "a" and the center projection line PLc. It may be configured as a pair of shutters 122 moving in a diagonal direction toward.

On the other hand, the light blocking upper part for determining the lost acquisition point CP can also be used as an auxiliary means for specifying the projection angle PA. The light blocking part 120 is positioned in a preset positional relationship with respect to the projection focus Pf. If the sequential blocking of the light blocking unit 120 is performed in a preset pattern, the projection angle PA of the projection light P immediately before being blocked according to the blocking phase with respect to the projection focus Pf at a certain point can be known. Because there is.

In addition, the present invention calculates the origin distance (OD) information for a plurality of acquisition points (CP) to grasp the spatial relationship with the three-dimensional object 10, as well as the origin distance (OD) for a plurality of acquisition points (CP) The surface shape of the three-dimensional object 10 may be calculated by curved fitting the information, which is performed in a separate fifth step.

According to the embodiment of the present invention having the above configuration, the projection points formed by the plurality of projection light on the surface of the three-dimensional object is approximately from the two-dimensional pattern of the acquisition points captured on the image detection surface of the light receiving sensor. It becomes possible to grasp the typical shape. However, since the embodiments described above can be modified by those skilled in the art without departing from the spirit or principle of the present invention, the scope of the present invention is defined by the appended claims and their equivalents. Should

The present invention is useful in a variety of fields that require three-dimensional recognition, such as technical vehicles requiring recognition of three-dimensional characteristics of the search target, such as drones, missiles, as well as toys, cars, and robots used in everyday life. Can be used.

Claims

A first step of projecting a plurality of projection light beams having a projection angle with respect to the center projection line set in advance to a three-dimensional object;

Acquiring a pattern of projection points formed on the three-dimensional object as a plurality of acquisition points that know each coordinate value through the two-dimensional image detection surface of the light receiving sensor while the center acquisition line is located on the same plane with respect to the center projection line. Second step;

A third step of determining a one-to-one correspondence between the plurality of projection lights and an acquisition point; And

A fourth step of calculating each acquisition distance from a geometric relationship between the projection angle and an acquisition angle and a focal length of the acquisition point;

Three-dimensional recognition method through the two-dimensional image acquisition of the multi-projection light comprising a.
The method of claim 1,

In the fourth step, the acquisition distance is calculated by using Equation (1) and Equation (2) below.

Expression (1)

..... Equation (2)

Where L is the acquisition distance, f is the focal length, α is the projection angle, and β is the acquisition angle
The method of claim 2,

And a 4-1 step of converting the acquisition distance into an origin distance through Equation (3) below.

Expression (3)

Where D is the origin distance
The method of claim 1,

In the second step, when the center projection line and the center acquisition line do not coincide, the angle between the center projection line and the center acquisition line based on the center projection point is 20 ° or less through 2D image acquisition of the multi-projection light. 3D recognition method.
The method of claim 3,

And a fifth step of calculating a surface shape of the three-dimensional object by curvedly fitting the origin distance information of the plurality of acquisition points, wherein the two-dimensional image acquisition of the multi-projection light is performed.
The method of claim 1,

And wherein said third step is achieved by measuring and matching the gradient characteristics of measurable property values imparted to said projection light at said acquisition point and mutually matching.
The method of claim 1,

In the first step, the plurality of projection light beams form a projection pattern of an equally spaced grid array or an equilateral concentric array based on the center projection line.

The third step includes acquiring one-to-one correspondence of the acquisition points corresponding to an arrangement order of a series of projection lights existing on a line extending radially about the center projection line. 3D recognition method.
The method of claim 7, wherein

Wherein the projection pattern is made by projecting a single light source onto a diffraction grating having a preset dot pattern.
The method of claim 1,

And comparing the number of projection light beams with the number of acquisition points to determine whether there is a missing acquisition point. 2-3D recognition method according to the two-dimensional image acquisition of the multi-projection light.
The method of claim 9,

If it is determined that the acquisition point lost in the step 2-1 is present, further comprising the step 2-2 of adjusting the direction of the center projection line so that the number of the acquisition point is equal to the number of the projection light. 3D recognition method through 2D image acquisition of multi-projection light.
The method of claim 7, wherein

And comparing the number of the projection light beams with the number of acquisition points to determine whether there is a missing acquisition point.

If it is determined that the acquisition point lost in step 2-1 is present, a part of the plurality of projection lights is blocked by the light blocking unit, and it is determined whether or not there is an acquisition point that does not disappear among the acquisition points. And 2-2 'repeating the process of determining the projection light for all the projection light.
The method of claim 11,

Step 2-2 ′ simultaneously blocks a series of projection light beams having the same radial distance from the projection light beams of the projection pattern with the light blocking unit, and sequentially arranges the blocked projection light beams in the radial direction. 2D image acquisition of the multi-projection light, wherein the process of determining the projection light corresponding to the lost acquisition point is repeated for all the projection lights by determining whether there is an acquisition point that does not disappear among the acquisition points corresponding to 3D recognition method through
The method of claim 12,

The light shielding unit starts sequential blocking from a series of projection lights having the largest radial distance or a series of projection lights having the smallest radial distance. .
The method according to claim 11, wherein

The light blocking unit is a physical shutter or an optical shutter which is a liquid crystal glass in which a part of the area is converted to non-transmissive to light by the application of a current, 3D recognition method through the two-dimensional image acquisition of the multi-projection light.
The method of claim 14,

And the physical shutter has a structure of a camera shutter when the projection pattern is a conformal concentric array. 3D recognition method through 2D image acquisition of multi-projection light.
The method of claim 14,

When the projection pattern is an equally spaced grid array, the physical shutter is a pair of shutters having a shape of "b" and "a" and moving in a diagonal direction toward the center projection line. Three-dimensional recognition method through image acquisition.
The method of claim 11,

The sequential blocking of the light blocking unit is performed in a preset pattern, but when the light blocking unit is positioned in a preset positional relationship with respect to the projection focus and sequential blocking is performed for each projection light, the projection focus at a certain point of time. And specifying or confirming a projection angle of the projection light immediately before being blocked according to the blocking phase for the three-dimensional recognition method by acquiring two-dimensional images of the multi-projection light.
The method of claim 1,

The projection light is a three-dimensional recognition method through the two-dimensional image acquisition of the multi-projection light, characterized in that the laser light having a wavelength distinct from the wavelength of the light present around the three-dimensional object.