KR20120090630A - Apparatus and method for measuring 3d depth by an infrared camera using led lighting and tracking hand motion - Google Patents

Abstract

PURPOSE: A distance measuring/hands motion tracing apparatus and a method thereof are provided to by an infrared ray camera which uses LED(Light Emitting Diode) light are provided to accurately detect hands by using distance information which is measured through single infrared camera. CONSTITUTION: A storing unit(130) stores an instruction code corresponding to motion variable. A camera(110) generates an image by photographing hands of a user. An outline detecting unit(120) generates an outline image including the outline line of hands included in the image. A pattern detecting unit(140) calculates motion variable corresponding to the outline image and extracts the instruction code. The pattern detecting unit transmits the instruction code to an external device and operates the external device according to the instruction code.

Description

FIELD AND METHOD FOR MEASURING 3D DEPTH BY AN INFRARED CAMERA USING LED LIGHTING AND TRACKING HAND MOTION

The present invention relates to a method for measuring distance and a motion detection using a combination of a single infrared camera and LED light having different wavelengths, and more particularly, to measure distance based on skin detection and to detect hand movement and hand shape change. It is about technology to do.

Various input methods are used to manipulate numerous electronic products. Representative input methods include buttons and a touch screen. However, buttons and touch screens are mounted on electronic products or remote controllers, and users have to press or touch buttons and touch screens directly.

Accordingly, a method of sensing a human body motion, which is a new input method, and performing an input according to the detected motion type has been proposed.

Hands in the human body can express various movements. Therefore, if the hand movement is accurately detected, various inputs can be performed only by the movement of the hand. However, the hand can express various movements, but its shape is complicated, which makes it difficult to be accurately recognized by general motion recognition algorithms. .

A general method of detecting a hand is a model-based detection method. The model-based method compares the 3D or 2D cardboard model with the shape of the hand with the shape of the actual hand, and selects the model that most closely resembles the shape with the actual hand, and estimates the shape of the user's hand as the shape of the hand model. That's how. However, the model-based detection method has a high computational complexity, which makes it difficult to detect motion in real time.

In addition, there is a method of detecting a hand using the skin color of the hand. However, since the skin color is affected by the illumination a lot, it is difficult to detect the hand due to the large fluctuation depending on the environment.

The present invention provides skin detection using infrared LED light and distance measurement using a single infrared camera by controlling the intensity of the LED light, an effective hand detection method based on the same, and motion detection to track and recognize the correct hand shape in the detected hand. To provide a device.

According to an aspect of the invention, the storage unit for storing the instruction code corresponding to the operation variable and the viewpoint variable; A camera unit generating a photographed image by photographing a user's hand; An outline detection unit configured to generate an outline image including an outline of the hand included in the captured image; And a pattern detector for calculating the motion variable and the viewpoint variable corresponding to the contour image and extracting the calculated operation variable and the command code corresponding to the viewpoint variable from the storage. .

The pattern detection unit may transmit the command code to an external device to control the external device to operate according to the command code.

The pattern detector may calculate cylindrical coordinate system coordinates corresponding to the contour image, and convert the cylindrical coordinate system coordinates into the motion variable and the viewpoint variable.

The camera unit detects reflected light of infrared rays having different wavelengths to generate a first skin detection image and a second skin detection image, and the motion detection apparatus is configured to detect a difference in pixel values between the first skin detection image and the second skin detection image. A skin detector for generating skin region information including coordinates corresponding to a pixel having a greater than or equal to a predetermined value; And a distance detector configured to generate hand region information including coordinates corresponding to coordinates in pixels corresponding to a predetermined range of pixel values among pixels corresponding to the skin region information.

The camera unit includes a first illumination module for projecting infrared rays of a first predetermined wavelength; And a second illumination module for projecting infrared rays of a second predetermined wavelength.

The contour detector may extract a pixel corresponding to a coordinate corresponding to the hand region information from the captured image, and generate the contour image including the extracted pixel.

According to another aspect of the present invention, a method for detecting a motion of a user's hand by the motion detection device, the method comprising: photographing the user's hand to generate a photographed image; Generating a contour image including an outline of the hand included in the captured image; Calculating the motion variable and the viewpoint variable corresponding to the contour image; There is provided a motion detection method comprising extracting command codes corresponding to the operation variable and the viewpoint variable from among pre-stored command codes.

The operation detecting method may further include controlling the external device to operate according to the command code by transmitting the command code to an external device.

The calculating of the motion variable and the viewpoint variable may include calculating cylindrical coordinate system coordinates corresponding to the contour image; And converting the cylindrical coordinate system coordinates into the motion variable and the viewpoint variable.

The motion detection method may further include generating a first skin detection image and a second skin detection image by detecting reflected light with respect to infrared rays having different wavelengths; Generating skin region information including coordinates corresponding to pixels whose difference in pixel values between the first skin detection image and the second skin detection image is equal to or greater than a predetermined value; The method may further include generating hand region information including coordinates corresponding to coordinates in a pixel corresponding to a predetermined range of pixel values among the pixels corresponding to the skin region information.

The generating of the contour image may include extracting a pixel corresponding to the coordinate corresponding to the hand region information from the photographed image; And generating the contour image including the extracted pixels.

According to an embodiment of the present invention, there is an effect of accurately recognizing the movement of a complex hand that changes the viewpoint such as the rotation of the wrist.

In addition, according to an embodiment of the present invention, it is possible to measure distance information using a single infrared camera and to accurately detect a hand using distance information.

1 is a block diagram schematically illustrating a motion detection apparatus.
2 is a diagram illustrating a sample image corresponding to a motion variable and a viewpoint variable stored in a storage unit of a motion detection apparatus.
3 is a flowchart illustrating a method in which a motion detection device detects a motion of a user's finger.

The present invention may be variously modified and have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail with reference to the accompanying drawings. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

Further, in the present specification, when one component is referred to as "transmitting a signal" to another component, the one component may be directly connected to the other component to transmit a signal, but there is a particular opposite description. It is to be understood that unless otherwise, the signal may be transmitted in the intermediary with another component.

1 is a block diagram briefly illustrating a motion detection apparatus.

Referring to FIG. 1, the motion detection apparatus includes a camera unit 110, an outline detection unit 120, a storage unit 130, and a pattern detection unit 140.

The camera unit 110 is a camera that photographs a user's hand to generate an image. At this time, the camera unit 110 may be an infrared camera for effective recognition even in an environment where the brightness of the illumination changes and a dark environment. In addition, the camera unit 110 may be a general camera photographing a visible light region according to an implementation method. The camera unit 110 transmits the generated image to the contour detection unit 120.

인 적외선을 투사하고, 카메라부(110)는 반사된 적외선을 감지하여 제1 피부 검출 이미지를 생성한다. 이어서, 제2 조명모듈(112)은 파장이

인 적외선을 투사하고, 카메라부(110)는 반사된 적외선을 감지하여 제2 피부 검출 이미지를 생성한다. 카메라부(110)는 제1 피부 검출 이미지 및 제2 피부 검출 이미지를 피부 검출부(114)로 전송한다.The camera unit 110 performs additional photographing to calculate skin area and distance information. Hereinafter, an additional photographing process will be described in the camera unit 110 with reference to FIG. 2. The camera unit 110 includes a first illumination module 111 and a second illumination module 112 that project infrared rays. The first illumination module 111 and the second illumination module 112 may project infrared rays of each predetermined wavelength. For example, the first lighting module 111 has a wavelength

The infrared ray is projected, and the camera unit 110 detects the reflected infrared rays to generate a first skin detection image. Subsequently, the second illumination module 112 has a wavelength

The infrared ray is projected, and the camera unit 110 detects the reflected infrared rays to generate a second skin detection image. The camera unit 110 transmits the first skin detection image and the second skin detection image to the skin detection unit 114.

The skin detector 114 detects an area corresponding to the skin according to a pixel difference between the first skin detection image and the second skin detection image received from the camera unit 110. When the infrared wavelength changes, the skin has a large change in reflectance compared to other media. Accordingly, the skin detector 114 calculates a value obtained by subtracting pixels of the second skin detection image corresponding to the coordinates of the pixel from each pixel of the first skin detection image. The skin detector 114 selects coordinates whose subtracted value is equal to or greater than a specified value as an area corresponding to the skin. The skin detector 114 transmits the skin region information including the first skin detection image or the second skin detection image and the coordinates of each pixel corresponding to the skin to the distance detector 117.

The distance detector 117 detects a distance of pixels corresponding to each coordinate included in the skin region information in the first skin detection image or the second skin detection image. The brightness l _i of the infrared image detected by each pixel i is established in the following relationship.

[Equation 1]

와 시간 t에서의 조명의 밝기 L_t에 비례하고 픽셀 i의 거리 r_i에 반비례한다. 즉 조명의 밝기가 밝을수록 카메라에 가깝고, 카메라에서 멀어질수록 밝기가 어두워진다. 따라서, 거리 검출부(117)는 피부에 해당하는 각 픽셀 중 그 값이 미리 지정된 범위에 해당하는 픽셀을 추출한다. 거리 검출부(117)는 추출한 픽셀들에 상응하는 좌표를 포함하는 손 영역 정보를 생성하여 윤곽선 검출부(120)로 전송한다.According to Equation 1, the brightness l _i is the reflectance of each pixel i

It is proportional to the brightness L _t of the illumination at and t and inversely proportional to the distance r _i of pixel i. In other words, the brighter the light is, the closer it is to the camera, and the farther it is, the darker it is. Therefore, the distance detector 117 extracts a pixel whose value corresponds to a predetermined range among pixels corresponding to the skin. The distance detector 117 generates hand region information including coordinates corresponding to the extracted pixels, and transmits the generated hand region information to the contour detector 120.

The contour detector 120 generates a contour image extracted only an area corresponding to a hand from the image received from the camera unit 110 according to the hand region information received from the distance detector 117. The contour detector 120 transmits the contour image to the pattern detector 140.

The storage unit 130 stores the motion variable and the viewpoint variable corresponding to the coordinates on the cylindrical coordinate system corresponding to the sample image of the at least one pose hand photographed at one or more viewpoints. In addition, the storage unit 130 stores instruction codes corresponding to the combination of each operation variable and a viewpoint variable. The command code is a code for instructing a specified operation in an external device.

The pattern detector 140 extracts the command code corresponding to the contour from the contour detector 120 in the storage 130 and outputs the command code to the outside. The pattern detector 140 may calculate a viewpoint variable and an operation variable from the outline image, and extract the command code corresponding to the calculated viewpoint variable and the operation variable from the storage 130. In this case, the viewpoint variable is a value indicating the direction of the hand shown in the contour image, and the operation variable is a value indicating the shape of the hand shown in the outline image.

Hereinafter, the process of generating the viewpoint variable and the operation variable by the pattern detector 140 using the contour image will be described.

First, the pattern detector 140 accumulates a contour image in columns and displays a matrix having a size n (a predetermined constant) (hereinafter, referred to as a cumulative matrix). For example, the pattern detector 140 accumulates each row in order in an outline image having a width of 4 pixels and a length of 3 pixels to calculate a 1 × 12 cumulative matrix.

The pattern detector 140 calculates the coordinates on the cylindrical coordinate system by applying the cumulative matrix to Equation 2 below.

[Equation 2]

에서 미리 지정된 프로젝션 계수이고, j는 윤곽선 이미지의 누적 행렬이고. S^-1은 SVD(singular value decomposition)에 따라 산출된 프로젝션 계수 B에 대한 의사역행렬이다.

는 각 샘플 이미지를 원통형 좌표계 좌표이고,

는

를 최소화하는 추정된 촬영한 이미지에 대한 원통 좌표계 좌표이다. B에 대하여 singular value decomposition(SVD)를 적용하면 아래식과 같이 메트릭스를 분리할 수 있으며, 이렇게 분리된 메트릭스에 대하여 두개의 orthonormal 한 행렬이 각각 U, V이다. S는 diagonal에만 값을 가지는 매트릭스이며 S의 역함수는 diagonal 항목에서 0을 제외한 값에 대하여 역수값, 즉 값이 x라면 1/x의 값을 가지는 함수이다.Where B is the kernel space

Where is a predefined projection coefficient, j is the cumulative matrix of the contour image. S ⁻¹ is a pseudo inverse of the projection coefficient B calculated according to the singular value decomposition (SVD).

Is the cylindrical coordinate system coordinates for each sample image,

The

Is the cylindrical coordinate system coordinates for the estimated captured image that minimizes. If we apply singular value decomposition (SVD) to B, the matrix can be separated as shown in the following equation, and two orthonormal matrices for the separated matrix are U and V, respectively. S is a matrix having a value only in the diagonal, and the inverse function of S is a function having an inverse value, that is, 1 / x for a value except 0 in the diagonal item.

를 하기의 수학식 3에 적용하여 시점 변수(Circular view variation) 및 동작 변수(Non-cyclic motion variation)를 산출한다. 이 때,

는 3차원 좌표로써 (x, y, z)라고 가정한다.
The pattern detection unit 140 calculates the cylindrical coordinate coordinates

Is applied to Equation 3 below to calculate a circular view variation and a non-cyclic motion variation. At this time,

Is assumed to be (x, y, z) as three-dimensional coordinates.

&Quot; (3) "

는 시점 변수이고,

는 동작 변수이다.

및

는 0 이상 1 이하의 실수일 수 있다. At this time, Ra and Rb is a predetermined constant,

Is the point in time variable,

Is an action variable.

And

May be a real number between 0 and 1, inclusive.

At this time, the pattern detecting unit 140 calculates the viewpoint variable and the operation variable according to the above-described Equation 3 only when the calculated coordinates of the cylindrical coordinate system satisfy the conditions of Equations 4 and 5 below. Do it right away. That is, when the cylindrical coordinate system coordinates of the contour image that do not satisfy Equation 4 are calculated, the pattern detector 140 sums the squares of x and y with respect to the cylindrical coordinate system coordinates (x, y, z) of the contour image. Calculate the square root of R _est . The pattern detection unit 140 multiplies x and y by R _a / R _est to calculate (x * R _a / R _{est ,} y * R _a / R _est, z). Thereafter, the pattern detector 140 calculates a viewpoint variable and an operation variable according to the calculated coordinates (x * R _a / R _{est ,} y * R _a / R _{est ,} z).

When the cylindrical coordinate system coordinates of the contour image that do not satisfy the equation (5) are calculated, the pattern detector 140 calculates the cylindrical coordinate system coordinates of the contour image of the contour image among the coordinates of the cylindrical image corresponding to each sample image stored in the storage unit 130. Can be approximated to close coordinates. Thereafter, the pattern detector 140 may calculate a viewpoint variable and an operation variable according to the approximated coordinates.

&Quot; (4) "

[Equation 5]

The pattern detector 140 queries the storage 130 for command codes corresponding to the calculated viewpoint variables and operation variables. The pattern detector 140 receives the command code from the storage 130 and outputs the command code to the outside.

The external device may perform an operation according to the corresponding command code received from the pattern detector 140. That is, the motion detection apparatus analyzes an image generated by photographing a user's hand and outputs command codes corresponding to the shape and movement of the hand to an external device. For example, when the user moves his hand in a predetermined form for changing the TV channel, the motion detection apparatus photographs the user's hand to generate an image, and generates an outline image corresponding to the generated image. The motion detecting apparatus accumulates columns of the contour image to generate a cumulative matrix, and calculates cylindrical coordinate system coordinates corresponding to the cumulative matrix. The motion detection device calculates a viewpoint variable and an operation variable corresponding to the cylindrical coordinate system coordinates. The motion detection apparatus extracts the calculated view variable and a command code for changing a channel corresponding to the motion variable from the storage unit 130 and transmits the command code to the television. The television can change the channel according to the command code.

FIG. 2 is a diagram illustrating a sample image corresponding to an operation variable and a viewpoint variable stored in a storage unit of the motion detection apparatus.

Referring to FIG. 2, the motion variable and the viewpoint variable stored in the storage unit 130 of the motion detection apparatus may be calculated to correspond to each of a plurality of sample images having different shapes of the viewpoint and the hand as illustrated in FIG. 2; Viewpoint variable. That is, the motion variable and the viewpoint variable may be matched with respect to each of the sample images, which are contour images having different viewpoints and shapes illustrated in FIG. 2.

A process of calculating an operation variable and a viewpoint variable from a plurality of sample images is well known and will be briefly described below.

인 2차원 배열로 표현될 수 있다. 2차원 배열을 칼럼 누적에 의하여 나타낸 값을 크기가

인 1차원 벡터로 나타낼 수 있고, 이 벡터를 J라고 가정하도록 한다. 각 샘플 이미지에 상응하는 벡터를

라고 가정하면, k개의 샘플 이미지 각 샘플에 대응하는 시점 변수 u와 동작변수 v가 주어진다면, u, 와 v를 바탕으로 원통형 좌표계의 좌표 (x,y,z)가 하기의 수학식 6에 의하여 산출될 수 있다.
Each sample image illustrated in FIG. 2 is normalized to

It can be expressed as a two-dimensional array. The size of a two-dimensional array represented by column stacking

It can be expressed as a one-dimensional vector, which is assumed to be J. The vector corresponding to each sample image

Suppose that the viewpoint variable u and the motion variable v corresponding to each sample of k sample images are given, the coordinates (x, y, z) of the cylindrical coordinate system based on u, and v are expressed by Equation 6 below. Can be calculated.

&Quot; (6) "

For coordinates and image pairs in a cylindrical coordinate system, the relationship between a cylindrical embedding space and a given image can be learned by learning a known nonlinear relationship, Generalized Radial Basis Function (GRBF) or similar nonlinear mapping function. In the case of GRBF, the mapping function in the form of Equation 7 can be learned for each element of the hand-shaped sample, and the matrix like Equation 8 can be calculated by accumulating the mapping functions.

[Equation 7]

[Equation 8]

In this case, e in Equations 5 and 6 means (x, y, z) in Equation 4 which is a point of the cylindrical coordinate system. B is also the mapping parameter B, which is the projection coefficient in φ () kernel space. e _i is a cylindrical coordinate system coordinate for the i th learning data, w _i ^k is a weight for the k th pixel value for the i th learning data according to each pair of learning data and embedding coordinates thereof.

The embedding space on the cylindrical coordinate system may be set according to the matrix f (e) according to Equation 7, and the storage unit 130 of the motion detection apparatus corresponds to the coordinates of each cylindrical coordinate system included in the f (e) and the respective coordinates. Stores the command code according to the combination of the viewpoint variable and the action variable, each viewpoint variable and the action variable. Therefore, the storage unit 130 may search for and provide command codes for the viewpoint variable and the operation variable according to the request of the pattern detector 140.

3 is a flowchart illustrating a method of detecting a motion of a user's finger by the motion detection device.

Referring to FIG. 3, in operation 310, the motion detection apparatus generates an image by photographing a user's hand. For example, the motion detection apparatus generates one or more images by photographing a user's hand through the camera unit 110.

In operation 320, the apparatus for detecting motion generates a contour image by detecting contours in the image generated in operation 310.

In operation 330, the apparatus for detecting motion calculates a viewpoint variable and an operation variable according to the contour image. The process of calculating the viewpoint variable and the motion variable in the contour image is a process of calculating the cylindrical coordinate system coordinates from the contour image and converting the cylindrical coordinate system coordinates into the viewpoint variable and the motion variable as described above with reference to Equations 2 and 3 below. Can be.

The equation for calculating the viewpoint variable and the motion variable for the contour image is basically expressed based on the change of the motion variable for the continuous change between the two representative hand shapes. By applying the proposed method to the two hand shapes, it can be extended to multiple hand shapes.

In operation 340, the motion detection apparatus searches for and extracts a command code stored in the storage 130 matching the viewpoint variable and the operation variable. At this time, the storage unit 130 may store the instruction code corresponding to the viewpoint variable and the operation variable in advance.

In operation 350, the motion detection apparatus outputs the extracted command code to the outside. For example, when the motion detection device is connected to the television, and the motion detection device transmits a command code according to the motion of the user's finger to the television, the television performs the motion according to the command code. In this case, the motion detection device may be a command code designated by an external device connected to the motion detection device.

So far I looked at the center of the embodiment for the present invention. Many embodiments other than the above-described embodiments are within the claims of the present invention. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. The disclosed embodiments should, therefore, be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

Claims (11)

A storage unit for storing command codes corresponding to the operation variable and the viewpoint variable;
A camera unit generating a photographed image by photographing a user's hand;
An outline detection unit configured to generate an outline image including an outline of the hand included in the captured image; And
A pattern detector configured to calculate the motion variable and the viewpoint variable corresponding to the contour image, and extract command codes corresponding to the calculated motion variable and the viewpoint variable from the storage unit;
Motion detection device comprising a.
The method according to claim 1,
The pattern detecting unit transmits the command code to an external device, and controls the external device to operate according to the command code.
The method according to claim 1,
And the pattern detecting unit calculates cylindrical coordinate system coordinates corresponding to the contour image, and converts the cylindrical coordinate system coordinates into the motion variable and the viewpoint variable.
The method according to claim 1,
The camera unit generates a first skin detection image and a second skin detection image by detecting reflected light with respect to infrared rays having different wavelengths,
A skin detection unit generating skin region information including coordinates corresponding to pixels whose difference in pixel values between the first skin detection image and the second skin detection image is equal to or greater than a predetermined value; And
And a distance detector configured to generate hand region information including coordinates corresponding to coordinates in a pixel corresponding to a predetermined range of pixels of each pixel corresponding to the skin region information.
5. The method of claim 4,
The camera unit
A first illumination module projecting infrared rays of a first predetermined wavelength; And
And a second illumination module for projecting infrared rays of a second predetermined wavelength.
5. The method of claim 4,
And the contour detector extracts a pixel corresponding to a coordinate corresponding to the hand region information from the captured image, and generates the contour image including the extracted pixel.
In the method for detecting the motion of the user's hand in the motion detection device,
Photographing a user's hand to generate a photographed image;
Generating a contour image including an outline of the hand included in the captured image;
Calculating the motion variable and the viewpoint variable corresponding to the contour image;
Extracting an instruction code corresponding to the operation variable and the viewpoint variable among pre-stored instruction codes
Motion detection method comprising a.
The method of claim 7, wherein
Transmitting the command code to an external device, and controlling the external device to operate according to the command code.
The method of claim 7, wherein
Computing the operation variable and the viewpoint variable
Calculating cylindrical coordinate system coordinates corresponding to the contour image; And
And converting the cylindrical coordinate system coordinates into the motion variable and the viewpoint variable.

The method of claim 7, wherein
Generating a first skin detection image and a second skin detection image by detecting reflected light with respect to infrared rays having different wavelengths;
Generating skin region information including coordinates corresponding to pixels whose difference in pixel values between the first skin detection image and the second skin detection image is equal to or greater than a predetermined value;
And generating hand region information including coordinates corresponding to coordinates in pixels of which a pixel value corresponds to a predetermined range among pixels corresponding to the skin region information.
The method of claim 10,
Generating the contour image
Extracting a pixel corresponding to a coordinate corresponding to the hand region information from the captured image; And
And generating the contour image including the extracted pixels.

Images