KR101087863B1 - Method of deciding image boundaries using structured light, record medium for performing the same and apparatus for image boundary recognition system using structured light - Google Patents

Abstract

Disclosed are a method of determining a boundary of an image using a structured light pattern capable of improving the accuracy of edge detection, and a recording medium on which the method is recorded and a boundary recognition system of an image using the structured light pattern. First, image data based on a reference pattern, a signal pattern, and an inverted signal pattern in which the signal pattern is inverted is obtained, and the reflection index of the object surface is estimated based on the obtained data, and the signal pattern is normalized based on the reflection index. The inverted incident light signal normalizing the incident light signal and the inverted signal pattern may be calculated, and the boundary between the signal pattern and the inverted signal pattern may be detected using the intersection of the normalized incident light signal and the inverted incident light signal. Therefore, the accuracy of boundary detection can be improved.

Description

Method of determining the boundary of an image using a structured light pattern, a recording medium on which such a method is recorded, and a boundary recognition system of an image using a structured light pattern TECHNICAL OF DECIDING IMAGE BOUNDARIES USING STRUCTURED LIGHT, RECORD MEDIUM FOR PERFORMING THE SAME AND APPARATUS FOR IMAGE BOUNDARY RECOGNITION SYSTEM USING STRUCTURED LIGHT}

The present invention relates to a method for determining a boundary of an image, and more particularly, a method for determining a boundary of an image using a structured light pattern for projecting a pattern with a projection means and obtaining a projected pattern with the image capturing means. An image boundary recognition system using the recorded recording medium and the structured light pattern.

Three-dimensional cameras using structured light are variants of stereo cameras. Unlike stereo cameras that use two identical cameras, one of them is a projection means such as a beam projector. Has a configuration replaced with In a three-dimensional camera system using structured light, an object is irradiated with a pattern using a projection means, and an object with the pattern is photographed by using an image capturing means such as a camera, and then processed by the photographed pattern to obtain three-dimensional information. do.

Conventional stereo camera systems passively use only features from an image, while structured optical camera systems actively use patterns projected from projection means as feature points, resulting in high processing speed and spatial resolution. Due to the advantages described above, structural optical camera systems are widely used for modeling or recognizing objects, three-dimensional measurement, industrial inspection, reverse engineering, and the like.

In particular, in the field of intelligent robotics, conventional stereo camera systems are simple and simple, such as home service robots, when a large amount of three-dimensional data is required for the modeling of a workspace, or when there is not enough property information on an object and no color change of the background. The use of structural optical camera systems is further needed because it is unable to provide adequate three-dimensional information from the environment.

The accuracy of three-dimensional data in a structured optical camera system depends on how accurately the pattern projected by the projector is determined from the image taken through the camera. For this reason, various patterns have been proposed to find a more robust and accurate relationship between the pattern projected from the projector and the image acquired by the camera.

However, the problem of how to accurately find the boundary of the pattern in order to improve the accuracy of image reconstruction in the structured optical camera system has not been of interest.

An object of the present invention is to provide a method for determining the boundary of an image using a structured light pattern that can improve the accuracy of the edge detection.

Another object of the present invention is to provide a recording medium having recorded thereon a program for performing a method for determining the boundary of an image using a structured light pattern which can improve the accuracy of edge detection.

In addition, another object of the present invention is to provide a boundary recognition system of an image using a structured light pattern for performing a method of determining the boundary of the image using the structured light pattern.

Technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, a method of determining a boundary of an image using a structured light pattern includes a reference pattern, a signal pattern, and an inverted signal pattern inverting the signal pattern. A data acquisition step of acquiring based image data, estimating a reflection index of the surface of the object based on the acquired data, an incidence light signal normalizing the signal pattern and an inversion signal pattern normalized based on the reflection index Calculating an incident light signal; And detecting a boundary between the signal pattern and the inverted signal pattern by using an intersection point of the normalized incident light signal and the inverted incident light signal. The data acquiring step may include: projecting a reference pattern onto the object and acquiring reference data from the captured image of the projected reference pattern; projecting the signal pattern and the inverted signal pattern onto the object; And acquiring signal data from the image capturing the inverted signal pattern. The reference pattern is composed of one of white light and black light, and the signal pattern may be composed of white and black light. Reference data may be obtained based on data modeling the reference pattern, reflection index, ambient light, blur and noise included in capturing the reference pattern, respectively. The signal data may be obtained based on data modeling the signal pattern, the blur included in the projection of the signal pattern, the reflection index, the ambient light, the blur included in the capture of the pattern, and the noise. The signal data may be obtained based on data modeling the inversion signal pattern, the blur included in the projection of the inverted signal pattern, the reflection index, the ambient light, the blur included in the capture of the inverted signal pattern, and the noise. The estimating of the reflection index of the surface of the object based on the obtained data may be determined based on a result of taking a derivative of the difference between the obtained reference data. Estimating the reflection index of the surface of the object based on the obtained data may include determining a position to correct the boundary detection error based on a result of taking a derivative for the difference between the obtained reference data. . Estimating the reflection index of the surface of the object based on the obtained data may be obtained by applying a Richardson-Lucy deconvolution algorithm to the difference between the obtained reference data. . The calculating of the incident light signal normalizing the signal pattern and the inverted incident light signal normalizing the inverted signal pattern based on the reflection index may include calculating the signal data obtained from the image capturing the signal pattern and the image capturing the reference pattern. Calculating the incident light signal by dividing the difference of the obtained reference data by the reflection index, and obtaining the signal data obtained from the image capturing the inverted signal pattern and the image capturing the reference pattern. And calculating the inverted incident light signal by dividing a value obtained by deconvolving a difference of reference data by the reflection index. In the method of determining the boundary of an image using the structured light pattern, when the reflectivity of the surface is constant, the boundary is formed by using an intersection point of the brightness signal of the image capturing the signal pattern and the brightness signal of the image capturing the inversion signal pattern. It may further comprise the step of detecting.

Further, in order to achieve another object of the present invention, a program of instructions that can be executed by a digital processing apparatus that performs recognition of a 3D image according to an aspect of the present invention is tangibly implemented, and read by the digital processing apparatus. The recording medium which records a program which can be used is a data acquisition step of acquiring image data based on a reference pattern projected on an object, a signal pattern, and an inversion signal pattern in which the signal pattern is inverted, based on the acquired data. Estimating a reflection index, calculating an incident light signal normalizing the signal pattern and an inverted incident light signal normalizing the inversion signal pattern based on the reflection index, and using an intersection point of the normalized incident light signal and the inverted incident light signal Of the signal pattern and the inversion signal pattern A program for executing the steps of detecting the system can be recorded.

In addition, the boundary recognition system of the image using the structured light pattern according to an aspect of the present invention for achieving another object of the present invention projection means for projecting a reference pattern, a signal pattern and an inverted signal pattern inverted the signal pattern Image capturing means for providing an image capturing the projected reference pattern, the signal pattern, and the inverted signal pattern, respectively; acquiring data corresponding to each of the reference pattern, the signal pattern, and the image capturing the inverted signal pattern; And estimating a reflection index of the surface of the object based on the obtained data, calculating an incident light signal normalizing the signal pattern and an inverted incident light signal normalizing the inversion signal pattern based on the reflection index, and the normalized incident light. The signal pattern and the inverted scene using an intersection point of a signal and the inverted incident light signal It may include an image processing unit for detecting a boundary of the pattern.

According to the method for determining the boundary of an image using the structured light pattern described above, and the boundary recognition system of the image using the recording medium and the structured light pattern on which the method is recorded, the reflection index of the object is calculated and the incident light is calculated and then applied to the pattern. The boundary is detected using the intersection of the brightness signal of the incident light corresponding to the corresponding incident light and the inversion pattern.

Therefore, the accuracy of the edge detection can be improved, and through this, the accuracy of the 3D image reconstruction can be improved.

1 is a graph for explaining an error of boundary detection in a structured light-based three-dimensional image recognition method.
2 is a conceptual diagram for explaining mathematical modeling of a pattern projected from a projector.
3 is a conceptual diagram for describing a pattern including blur.
4 is a conceptual diagram illustrating the intensity of reflected light according to the reflection index of the surface.
FIG. 5 is a flowchart illustrating a modeling process of a projected pattern and a captured pattern in a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention.
6 is a graph showing simulation results for confirming the influence of the reflection index.
7 is a flowchart illustrating a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention.
FIG. 8 is a conceptual diagram for explaining a reflection checking process illustrated in FIG. 7.
9 is a graph for explaining a process of detecting a boundary when the reflectivity is constant.
10 is a conceptual diagram illustrating a process of detecting a boundary when the reflectivity is not constant.
11 illustrates a result of applying a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention.
12 illustrates a result of applying a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention to 3D image reconstruction.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals are used for like elements in describing each drawing.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when a component is referred to as being "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Hereinafter, with reference to the accompanying drawings, it will be described in detail a preferred embodiment of the present invention. Hereinafter, the same reference numerals are used for the same components in the drawings, and duplicate descriptions of the same components are omitted.

Hereinafter, the structured light-based three-dimensional image recognition method according to an embodiment of the present invention has a structure including a projector as a projection means, a camera as an image capturing means, and an image processing unit for processing a captured image to detect a boundary. It is assumed to be performed in a light-based three-dimensional image recognition system.

FIG. 1 is a graph illustrating an error of boundary detection in a structured light-based three-dimensional image recognition method, and illustrates a case in which a surface of an object has a chessboard pattern for simplicity.

FIG. 1A is a graphical representation of data that constitutes a specific row of a captured image where the edge of the projected stripe pattern is adjacent to the border of a black or white square on the object's surface, and FIG. FIG. 1B is a graph showing data of the same row as that of FIG. 1A in an inversion pattern in which the pattern shown in FIG. 1A is inverted. In addition, FIG. 1C shows data for the same row of the pattern and the inversion pattern shown in FIGS. 1A and 1B as a graph.

In the graphs of FIGS. 1A to 1C, the horizontal axis represents pixel coordinates of a captured image, and the vertical axis represents brightness of the pixel. In addition, state 1 (1 state) projects white light from the projector, for example, represents the brightness signal of the image captured by the camera, and state 0 (0 state) projects black light from the projector (or, And no brightness) of the image captured by the camera. In addition, the Edge of Stripe represents a brightness signal of a pattern consisting of alternating white and black light projected through the projector.

Referring to (c) of FIG. 1, it can be seen that the intersection point of the projected pattern and the inverted projected inverted pattern exists in the transition region of state 1 or state 0, and thus the boundary in the restored image. Will be in the form of zigzag. From the above experimental results, it can be seen that the main cause of the error in boundary prediction is the reflection index of the projected object surface (particularly, the surface that changes from black to white).

Hereinafter, a mathematical model for reducing an error of boundary detection in a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention will be described.

FIG. 2 is a conceptual diagram for explaining mathematical modeling of a pattern projected by a projector. FIG. 2 (a) shows a projected pattern, and FIG. 2 (b) shows a step edge model corresponding to the projected pattern. Indicates.

As shown in (a) of FIG. 2, the pattern projected by the projector may be modeled by the step function s (x) as shown in Equation 1, and the step edge as shown in FIG. ) Can be represented as a model. Here, the stair edge model is a model of a specific row of the projected pattern image.

An ideal projection pattern as shown in FIG. 2A can be represented by a stair edge model as shown in FIG. 2B, but the projection pattern shown in FIG. Blur may be included through the lens provided in the projector in the process of projecting through.

3 is a conceptual diagram for describing a projection pattern including a blur. FIG. 3A illustrates a mathematical modeling process of a projection pattern including a blur, and FIG. 3B illustrates a projection pattern including a blur. Indicates.

의 컨볼루션으로 모델링할 수 있고, 수학식 2와 같이 표현할 수 있다.Referring to FIG. 3, the pattern f _b (x) including the blur generated by the projector is a step function s (x) and a Gaussian blur kernel as shown in (a) of FIG. 3.

It can be modeled by the convolution of and can be expressed as Equation 2.

는 표준 편차를 의미한다. 또한, 수학식 2에서 프로젝터의 가우시안 블러 커널은 수학식 3과 같이 표현할 수 있다.In Equation 2,

Means standard deviation. In addition, in Equation 2, the Gaussian blur kernel of the projector may be expressed as Equation 3.

When the pattern is irradiated onto the surface of a certain object through the projector, the intensity (or energy) of the light reflected from the surface of the object is changed according to the reflection index of the surface of the object to which the pattern is irradiated.

4 is a conceptual diagram illustrating the intensity of reflected light according to the reflection index of the surface.

Referring to FIG. 4, when light sources having the same intensity α are irradiated onto surfaces of objects having different reflection indices R1 and R2 and R1> R2, the light reflected from the surface is larger in the reflection index. The light α ㅧ R1 reflected from it has a greater intensity than the light α ㅧ R2 reflected from the other area (ie α ㅧ R1> α ㅧ R2).

In order to consider the influence of the reflection index of the surface as described above, the reflection index R (x) at the position x of the surface of the object is defined as in Equation 4.

When the projector irradiates the pattern on the surface of the object, the reflected light changing according to the reflection index R (x) of the surface may be expressed by Equation 5.

When capturing the pattern projected by the projector through the camera in the boundary recognition system of the image using the structured light pattern, the camera captures the reflection amount and the ambient light A (x) of the projected pattern. The two elements include at least one blur by the camera lens during the capture process. Here, the blur included in the captured image by the camera lens may be modeled using a convolution of a Gaussian blur kernel as shown in Equation 6.

는 카메라 렌즈의 가우시안 블러 커널을 의미하며, W(x)는 캡처과정에서 포함되는 AWGN(Additive White Gaussian Noise)을 의미한다. 또한, 카메라 렌즈의 가우시안 블러 커널은 수학식 7과 같이 표현할 수 있다.In Equation 6, f _c (x) means a signal captured by the camera,

Denotes a Gaussian blur kernel of the camera lens, and W (x) denotes AWGN (Additive White Gaussian Noise) included in the capture process. In addition, the Gaussian blur kernel of the camera lens may be expressed as Equation (7).

FIG. 5 is a flowchart illustrating a modeling process of a projected pattern and a captured pattern in a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention.

Referring to FIG. 5, the pattern projected by the projector is modeled as a step function as shown in Equation 1 (step 110), the blur generated in the projector is modeled by a Gaussian blur kernel, and the step function and convolut modeled in step 110 are modeled. The projection pattern is modeled by taking into account the effects of blur caused by the projector as shown in Equation 2 (step 120).

Then, as shown in Equation 5, the image of the reflection index of the surface of the object irradiated by the projector is considered (step 130).

In addition, after taking into account the ambient light included in the reflection amount of the projection pattern when capturing by the camera (step 140), modeling the blur generated by the camera lens (step 150), and then modeling in step 140 The captured pattern is then convolved to model the captured pattern as shown in Equation 6 (step 160).

FIG. 6 is a graph showing simulation results for confirming the influence of a reflection index, showing brightness intensity of a captured image for a projected predetermined pattern and an inversion pattern of the predetermined pattern. In FIG. 6, the brightness of one specific row is graphed in the captured image of the projection pattern and the inverted projection pattern to confirm the influence of the reflection index at the intersection point.

FIG. 6A illustrates a case where a reflection index is constant (for example, when the surface of the object is uniform), and FIG. 6B illustrates a case where the reflection index is not constant (for example, When the surface is not uniform).

Looking at the location of the intersection and x = 201 in Figure 6 (a) and (b) it can be seen that the position of the intersection changes according to the change of the reflection index, through which the simulation results shown in Figure 6 It can be seen that it can be applied to the data obtained in the real environment as shown in.

을 제거할 필요가 있음을 의미한다.In the structured light based image boundary recognition, it can be seen that the edge detection error occurs due to the reflection index and the blur caused by the camera as described above. In order to improve the accuracy of edge detection, Gaussian blur kernel modeling the influence of reflection index R (x) and camera blur

This means that it needs to be removed.

을 제거하기 위해서는 리차드슨-루시(Rechardson-Lucy) 디컨볼루션(deconvolution) 알고리즘을 사용할 수 있다. 또한, 반사지수 R(x)는 후술하는 참조 데이터(reference data)를 이용하여 제거할 수 있다.Gaussian Blur Kernel by Camera

To eliminate this, we can use the Rechardson-Lucy deconvolution algorithm. In addition, the reflection index R (x) can be removed using reference data described later.

First, reference data of state 1 may be expressed as in Equation 8.

In Equation 8, s ₁ (x) = H means that the projected pattern is state 1 (i.e., project white light), and f ₁ (x) is a reference corresponding to the image that captured the pattern of projected state 1 Means data.

In addition, the reference data of the state 0 may be expressed as in Equation (9).

In Equation 9, s ₀ (x) = L means that the projected pattern is state 0 (that is, no light is projected), and f ₀ (x) is the reference corresponding to the image that captured the pattern of state 0. Means data.

In addition, a projection pattern composed of white and black light and data captured by the projected pattern may be expressed as in Equation (10).

In Equation 10, s _s (x) means modeling the projected pattern as shown in (a) of Figure 2, f _s (x) is the data corresponding to the image captured the projected pattern data Means.

In addition, equation (11) can be obtained using equations (8) and (9). Equation 11 represents the signal component of the light projected from the projector excluding the influence of the object. In Equation 11, AGWN W ₁ (x) and W ₀ (x) may not be identical to each other, but may be simplified as shown in Equation 12 by omitting the term (W ₁ (x) -W ₀ (x)). .

Using the Richardson-Lucy deconvolution method, the reflection index R (x) component included in the captured data may be calculated from the above Equation 12 as shown in Equation 13.

)은 수학식 14와 같이 반사지수 R(x)로 나누어 정규화하여 산출할 수 있고, 줄무늬 패턴의 가장자리(edge of stripe)가 반사지수에 의해 영향을 받는 경우 수학식 14를 이용하여 물체 표면에 입사된 입사광(incident light)을 복원할 수 있고, 교차점에 의한 경계를 예측할 수 있다.In addition, the amount of light irradiated to the surface of the object (ie, the amount of incident light;

) Can be normalized by dividing by the reflection index R (x) as shown in Equation 14, and when the edge of the stripe pattern is affected by the reflection index, it is incident on the object surface using Equation 14 The incident light can be restored and the boundary due to the intersection can be predicted.

FIG. 7 is a flowchart illustrating a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention, and FIG. 8 is a conceptual diagram for explaining a reflection checking process shown in FIG. 7. 9 is a graph illustrating a process of detecting a boundary when the reflectivity is constant, and FIG. 10 is a conceptual diagram illustrating a process of detecting the boundary when the reflectivity is not constant.

A method of determining the boundary of an image using the structured light pattern illustrated in FIG. 7 uses a structured light pattern including a projector for projecting a pattern, a camera for capturing the projected pattern, and an image processing unit for detecting the boundary from the captured pattern. It can be executed in the boundary recognition system of the image.

7 to 10, first, the image boundary recognition system checks the reflectivity of the object surface based on reference data of state 1 and state 0 (step 210).

Specifically, the projector projects white light (ie, the pattern of state 1) and the camera provides an image that captures the projected white light, and the image processor is in reference to state 1 corresponding to the image that captured the white light. Acquire it. In addition, the projector projects black light (ie, pattern of state 0) (or does not project light) and the camera provides an image that captures the projected black light, and the image processing unit emits the black light. Obtain reference data of state 0 corresponding to the captured image. Thereafter, the image processor calculates a difference between the reference data of state 1 and state 0 using Equation 12 described above.

Here, if the reflection index R (x) is a constant (R) (that is, R (x) = R), the convolution of the constant and the Gaussian function is expressed as It may be expressed as in Equation 15.

Alternatively, assuming that the reflection index R (x) is not a constant, the absolute value of the partial derivative of Equation 12 has a non-constant value as shown in Equation 16.

As described above, the image processor may determine the reflectivity of the surface of the object based on a result of taking a derivative for the difference between the reference data of state 1 and state 0 or the difference between the reference data of state 1 and state 0. That is, when the derivative value of Equation 12 is 0, it may be determined that the reflectivity is constant. When the derivative value of Equation 12 is not 0, it may be determined that the reflectivity is not constant.

FIG. 8 is a conceptual diagram illustrating the reflectivity checking process of step 210 illustrated in FIG. 7. For convenience of description, FIG. 8 illustrates a case where the reflectivity is confirmed based on a specific row (red horizontal line) in the captured image. As shown in (a) of FIG. 8, the image processor first obtains reference data from Equation 8 and Equation 9 from the captured images of State 1 and State 0, and then refers to State 1 for the specific row. When the absolute value of the derivative for the difference between the data and the reference data of state 0 is obtained, the peak value is displayed according to the change in light intensity in the image captured in the derivative graph as shown in FIG. do. This means that the positions corresponding to the peak values in the captured image are the positions where a boundary detection error is likely to occur, and the boundary of the image using the structured light pattern according to an embodiment of the present invention is a position requiring correction of the boundary detection error. Means locations to which the method of determining is to be applied.

Referring back to FIG. 7, when the image processor determines the reflectivity of the object surface after checking the reflectivity as described above (step 220), the image processor captures the projected signal pattern and the inverted signal pattern projected by inverting the pattern. The boundary is detected using the intersection point between the two signals in the graph showing the brightness signal of the image (step 230).

Specifically, when the reflectivity is constant, as shown in FIG. 9A, first, a signal pattern composed of white and black and an inverted signal pattern in which the signal pattern is inverted are projected, and then each projected pattern is captured to capture a specific row. After displaying the intensity of brightness with respect to the graph, as shown in (b) of FIG. 9, the intersection of the two graphs is detected and the position of the detected intersection is determined as the boundary.

Alternatively, if it is determined in step 220 that the reflectivity is not constant, the image processor first calculates the reflection index R (x) (step 240). Here, the image processor may calculate the reflection index R (x) by using the Richardson-Lucy deconvolution algorithm as shown in Equation 13 with respect to Equation 12 described above.

Thereafter, the image processing unit includes reference data f ₁ (x) obtained from the captured state 1 image, reference data f ₀ (x) obtained from the captured state 0 image, as shown in Equation 14, The incident light is calculated using the data f _s (x) obtained from the captured signal pattern image and the inverted signal pattern image (step 250).

Thereafter, the image processor finds an intersection point of the graph of the incident light amount corresponding to the captured signal pattern image and the graph of the incident light amount corresponding to the captured inverted signal pattern image and detects the boundary as the intersection point (step 260).

For example, the image processing unit may capture an image of state 1 captured, an image of state 0 captured, a signal pattern image capturing a projected signal pattern, and the signal pattern as illustrated in FIGS. 10A and 10B. After calculating the incident light by applying Equation 14 to the inverted signal pattern image capturing the inverted signal pattern projected by inverting the (n), the incident light corresponding to the captured signal pattern image as shown in (c) of FIG. The intersection of the signal and the signal of the incident light corresponding to the captured inverted signal pattern image is detected with the corrected boundary.

FIG. 10D illustrates an incident light signal and an inverted signal pattern corresponding to a case of detecting a boundary using an intersection point of a conventional captured signal pattern and an inverted signal pattern and corresponding to the captured signal pattern according to an embodiment of the present invention. Simultaneously displaying the case of detecting the boundary using the intersection of the corresponding incident light signals, as shown in FIG. 1, the coordinates of the boundary detected according to an embodiment of the present invention and the coordinates of the boundary detected according to the conventional boundary detection method. It can be seen that there is a significant difference between them, which means that the boundary detection method according to an embodiment of the present invention can detect the boundary more robustly and accurately.

11 illustrates a result of applying a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention.

FIG. 11A illustrates an object-specific image captured by the 3D image recognition system, and FIG. 11B captures a pattern projected through a projector of the 3D camera system.

In the image illustrated in FIG. 11B, since the rectangular area 300 has a flat surface, the predicted boundary should have a straight line shape.

In the case of applying the method of detecting a boundary using only the intersection points corresponding to the conventional signal pattern and the inverted signal pattern described above, as shown in FIG. 11C, the lines displayed in the rectangular area 300 are arranged in a zigzag form. Is displayed.

However, when the method of determining the boundary of the image using the structured light pattern (that is, using the intersection point of the incident light signal) according to an embodiment of the present invention is applied, as shown in (d) of FIG. The lines shown in 300 are displayed in a straight line, and it can be seen that the zigzag lines shown in FIG. 11C are corrected.

12 illustrates a result of applying a method of determining a boundary of an image using a structured light pattern according to an embodiment of the present invention to 3D image reconstruction.

FIG. 12A illustrates an image reconstructed in three dimensions, and FIGS. 12B and 12C show an enlarged portion of an image reconstructed in three dimensions, and FIG. 12B illustrates a conventional boundary detection. The result of applying the method is shown, and FIG. 12C shows the result of applying the method of determining the boundary of an image using the structured light pattern according to an embodiment of the present invention.

As shown in (b) of FIG. 12, when detecting a boundary by using an intersection point of a signal capturing a conventional signal pattern and an inverted signal pattern, a large number of zigzag-shaped lines are included in the reconstructed image to detect an error in boundary detection. In the case of detecting a boundary using an intersection point of incident light corresponding to each of the signal pattern and the inverted signal pattern according to an embodiment of the present invention, as shown in (c) of FIG. A reconstructed image can be obtained.

In the above, the present invention has been described in detail with reference to preferred embodiments, but the present invention is not limited to the above embodiments, and various modifications and changes by those skilled in the art within the spirit and scope of the present invention. This is possible.

Claims (13)

A data obtaining step of obtaining image data based on a reference pattern, a signal pattern projected on an object, and an inverted signal pattern inverting the signal pattern;
Estimating a reflection index of the surface of the object based on the obtained data;
Calculating an incident light signal normalizing the signal pattern and an inverted incident light signal normalizing the inversion signal pattern based on the reflection index; And
And detecting a boundary between the signal pattern and the inverted signal pattern by using an intersection point of the normalized incident light signal and the inverted incident light signal. The method of claim 1, wherein the data acquisition step
Projecting a first reference pattern and a second reference pattern onto the object and obtaining first reference data and second reference data from an image of the projected first reference pattern and the second reference pattern;
Projecting the signal pattern and the inverted signal pattern onto the object and obtaining signal data from the captured image of the projected signal pattern and the inverted signal pattern. How to determine. The method of claim 2,
The first reference pattern and the second reference pattern are each composed of different light among white light and black light, and the signal pattern is composed of white and black light boundary of an image using a structured light pattern How to determine. The method of claim 2,
The first reference data and the second reference data are used to capture blur and noise included in capturing the first reference pattern and the second reference pattern, a reflection index, ambient light, the first reference pattern, and the second reference pattern. A method of determining the boundary of an image using a structured light pattern, characterized in that obtained based on the modeled data. The method of claim 2,
The signal data is obtained based on data modeling the signal pattern, the blur included in the projection of the signal pattern, the reflection index, the ambient light, the blur included in the capture of the signal pattern, and the noise. A method of determining the boundary of an image using a pattern. The method of claim 2,
The signal data may be obtained based on data modeling the inversion signal pattern, the blur included in the projection of the inverted signal pattern, the reflection index, the ambient light, and the blur and noise included in the capture of the inverted signal pattern. A method for determining the boundary of an image using a structured light pattern. The method of claim 2, wherein estimating a reflection index of the surface of the object based on the obtained data comprises:
And determining the boundary of the image using the structured light pattern, based on a result of taking a derivative on the difference between the obtained first reference data and the second reference data. The method of claim 2, wherein estimating a reflection index of the surface of the object based on the obtained data comprises:
And determining a position to correct a boundary detection error based on a result of taking a derivative on the difference between the obtained first reference data and the second reference data. How to determine. The method of claim 2, wherein estimating a reflection index of the surface of the object based on the obtained data comprises:
A boundary of an image using a structured light pattern, wherein the reflection index is obtained by applying a Richardson-Lucy deconvolution algorithm to the difference between the obtained first reference data and the second reference data. How to determine. The method of claim 1, wherein the calculating of the incident light signal normalizing the signal pattern and the inverted incident light signal normalizing the inversion signal pattern is performed based on the reflection index.
Calculating the incident light signal by dividing a value obtained by deconvolution of the difference between the signal data obtained from the image capturing the signal pattern and the reference data obtained from the image capturing the reference pattern by the reflection index; And
Calculating a reverse incident light signal by dividing a value obtained by deconvolution of a difference between signal data obtained from the image capturing the inverted signal pattern and reference data obtained from the image capturing the reference pattern, by the reflection index; A method for determining the boundary of an image using a structured light pattern. The method of claim 1, wherein the boundary of the image using the structured light pattern is determined.
If the reflectivity of the surface is constant, further comprising detecting a boundary by using an intersection point of the brightness signal of the image capturing the signal pattern and the brightness signal of the image capturing the inverted signal pattern. A method of determining the boundary of an image using a pattern. In the recording medium in which a program of instructions that can be executed by a digital processing apparatus that performs recognition of a three-dimensional image is tangibly embodied, and a program that can be read by the digital processing apparatus is recorded.
A data obtaining step of obtaining image data based on a reference pattern, a signal pattern projected on an object, and an inverted signal pattern inverting the signal pattern;
Estimating a reflection index of the surface of the object based on the obtained data;
Calculating an incident light signal normalizing the signal pattern and an inverted incident light signal normalizing the inversion signal pattern based on the reflection index; And
And recording a program using the intersection of the normalized incident light signal and the inverted incident light signal to detect a boundary between the signal pattern and the inverted signal pattern. Projection means for projecting a reference pattern, a signal pattern, and an inverted signal pattern inverting the signal pattern onto an object;
Image capturing means for providing an image capturing the projected reference pattern, the signal pattern, and the inverted signal pattern, respectively; And
Acquire data corresponding to each of the image captured by the reference pattern, the signal pattern, and the inverted signal pattern, estimate a reflection index of the object surface based on the acquired data, and calculate the signal pattern based on the reflection index. An image processor configured to calculate a normalized incident light signal and an inverted incident light signal normalized to the inverted signal pattern, and detect a boundary between the signal pattern and the inverted signal pattern by using an intersection point of the normalized incident light signal and the inverted incident light signal An image boundary recognition system using a structured light pattern.

Images