KR100974538B1 - Method for location estimating of finger tip - Google Patents

Abstract

The present invention relates to a method for measuring a fingertip position, and has a characteristic object of providing a fingertip position measuring method for correcting the position of the fingertip by setting an expected range of the fingertip.

The present invention for achieving this object, (a) a first step of detecting the skin color region using a normalized RGB color model; (b) a second process of calculating an actual hand area using a blob coloring algorithm of the blob function; And (c) a third process of extracting the highest point of the area of the hand; Characterized in that it comprises a.

According to the present invention as described above, by setting the expected range of the fingertip to correct the position of the fingertip, there is an effect that can measure the position of the more accurate fingertip when the finger tracking in the bending finger state.

Fingertip, Bending Finger, Position Measurement

Description

Fingertip Positioning Method {METHOD FOR LOCATION ESTIMATING OF FINGER TIP}

The present invention relates to a method for measuring a finger tip position, and more particularly, in order to solve the problem of measuring an incorrect finger tip during finger tracking, setting a predicted range of the finger tip to correct the position of the finger tip It is about a method.

Visual Interaction using finger tracking is a new UI (User Interface) method that generates an event according to the movement and gesture of a user while continuously following a finger.

Finger tracking first detects a fingertip and then performs tracking. In general, fingertip detection extracts skin color information and then extracts the most significant point through a blob coloring algorithm of the blob function. However, if this method is applied to determine the top point of the contour as a fingertip and tracking, the user does not detect the exact fingertip when the user bends the finger at the time of tracking and finds only the topmost point of the skin color information. Occurs.

The present invention has been made in view of the above problems, and has a characteristic object of providing a fingertip position measuring method for correcting the position of the fingertip by setting an expected range of the fingertip.

In order to achieve the technical problem, the present invention relates to a fingertip position measurement method, (a) a first step of detecting the skin color region using a normalized RGB color model; (b) a second process of calculating an actual hand area using a blob coloring algorithm of the blob function; And (c) a third process of extracting the highest point of the area of the hand; Characterized in that it comprises a.

In addition, the step (c), (c-1) measuring the angle of the fingertip through a single camera; (c-2) determining whether the position of the fingertip measured through the step (c-1) exists at 0 ° to 89 °; And (c-3) setting the fingertip to be located at 45 ° and extracting it as the highest point when the determination result of step (c-2) is present at 0 ° to 89 °; Characterized in that it comprises a.

In addition, when the determination result of the step (c-2) is not present in 0 ° to 89 °, (c-4) it is determined that the finger tip is present in the 90 ° to 180 °, the finger tip to the measurement angle Setting to be located and extracting the highest point; It characterized in that it further comprises.

Further, in the process (a), the red color value is set to 30 to 60, the green color value is set to 25 to 35.

And, in the step (b), the blob coloring algorithm is characterized by distinguishing each area by performing an iterative operation of changing to the same color whenever different colors are found to be the same area.

According to the present invention as described above, by setting the expected range of the fingertip to correct the position of the fingertip, there is an effect that can measure the position of the more accurate fingertip when the finger tracking in the bending finger state.

Specific features and advantages of the present invention will become more apparent from the following detailed description based on the accompanying drawings. In the meantime, when it is determined that the detailed description of the known functions and configurations related to the present invention may unnecessarily obscure the subject matter of the present invention, it should be noted that the detailed description is omitted.

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

The fingertip position measuring method according to the present invention will be described with reference to FIGS. 1 to 5 as follows.

As a method of obtaining an existing finger-tip, a hand region is first detected using skin color information. Next, the skin contour is obtained through the blob coloring algorithm of the blob function, and the topmost point is selected as the finger tip.

In detail, a method of extracting skin color information is known to use three color spaces, namely, an RGB color area, an HSV color area, and a YCbCr color area.

The RGB area is a color space represented by using three color values of red, green, and blue, and is generally changed to a normalized form.

The basic formula for creating a normalized RGB form is:

r = R / R + G + B

g = G / R + G + B

b = B / R + G + B

Here, since b color does not have any important information, it is usually omitted and skin color information is extracted using two colors, r and g.

The HSV region, on the other hand, provides color with intuitive values based on the idea of density, saturation, and lightness. Hue defines dominant colors such as red, green, purple and yellow. Saturation refers to the degree of richness of color for a region in proportion to brightness. Contrast is also related to color brightness.

The YCbCr color model is not a color signal but a color expression method based on luminance Y and color difference signals Cb and Cr. Since the human eye is more sensitive to brightness than color, it is a model for processing using chrominance signals.

In the present invention, the normalized RGB color method was used because the change in the surface direction of the light source among the three color methods was used, and the skin color region was set by experiments with a red color value of 30 to 60 and a green color value of 25 to 35. It was. Here, the red color values 30-60 and the green color values 25-35 are the values in the general RGB method, and this value is used to change to the normalized RGB format. The formula for creating a normalized RGB form is r = R / R + G + B, g = G / R + G + B, b = B / R + G + B as described above.

The blob coloring algorithm extracts the skin color region using the normalized RGB color model, binarizes the extracted region and calculates the sum of the pixels.

According to the study of automatic extraction of liver region from CT image using blob coloring, the blob coloring algorithm is as follows.

Given binary image

Let the initial Color [k] = 0

Scan the original image from left to right and top to bottom (raster Scan)

For k = ImageSize

If Image [C] = 255 then

If Image [U] = 255 and Image [L] = 0 then

    Color [C] = Color [U]

If Image [U] = 0 and Image [L] = 255 then

    Color [C] = Color [L]

If Image [U] = 0 and Image [L] = 0 then

    Color [C] = K ++, new Color

If Image [U] = 255 and Image [L] = 255 then

If Color [U] is not equal Color [L] then

    Color [C] = Color [L]

Repeat

The blob coloring algorithm divides each area by iteratively changing the same color whenever different colors are found in the same area. That is, when a pixel differs from the color of its neighbors, it finds which index is the closest to that color. For example, if it's a pink pixel, the pixel value indicates that it's closer to red among black and red. Will be. At this time, if the operation of changing the pixel value to be the same as the adjacent color continuously, there is an effect that the separation between the surrounding and the object becomes certain.

The fingertip is the tip of the finger. After measuring the skin color information through the normalized RGB color method, we saw that the actual hand area can be calculated through the blob function.

The fingertip is usually determined by the point at the top of the area obtained through the blob function. However, this method is problematic for bending fingers in determining fingertips. According to the users' tendency, the fingertips are not easily determined due to the two conditions of using 2D image without the position and depth information of the webcam.

First, the webcam used by ordinary PC users is on the monitor. The average person looking at it does not look down at the monitor from a high position. Therefore, the webcam naturally faces the user or looks down slightly. When the user raises his hand to be visible on the webcam at that position, the hand is naturally facing up, and the finger is also facing up, so it is appropriate to use the fingertip at the top finger.

Secondly, the webcam uses 2D video without depth information. If you do not fingertip the top point through the blob function, but determine the fingertip of a point other than the topmost point, you may not be sure whether the point is actually the tip of the finger. This is because the blob function is not a function of catching only the fingertips, but a function of finding the skin color information obtained through normalized RGB through the blob coloring algorithm. In other words, when a pixel in a region is different from an adjacent surrounding color, the blob function compares the adjacent pixel data and changes it to be equal to an adjacent binary image value to perform binarization between the object and the object.

Figure 1 shows the appearance of measuring the wrong finger tip when bending. As shown in the drawing, the upper fingertip is extracted without the actual fingertip.

As shown in FIG. 1, the problem caused by the fingertip is in fact difficult to solve with a webcam using a 2D image. The most common way to solve this problem is to use two or more webcams. However, using two or more cameras can not only be costly in terms of cost, but also difficult to determine where to position the cameras. I can not say.

The present invention proposes a method of partially correcting the problem through the tendency of users by using only one camera.

In fact, general users do not bend more than a certain angle even when they are extended or bent their fingers when using the finger tracking. The range of angles at which the fingers are bent ranges from 0 ° to 180 °, with 0 ° when fully bent and 180 ° when fully bent.

Among them, the camera that processes only 2D image can grab the fingertip from 90 ° to 180 °, but it cannot recognize from 0 ~ 89 °, so the area to be solved is when the finger bends 0 ~ 89 °. If you can solve this problem, you can solve the fingertip problem.

In the present invention, in order to solve this problem, if a finger enters between 0 and 89 °, the part is set at 45 ° which is the center of the area. Since the movement range of the fingertip is very small from 0 to 89 °, the error range is very small even if the expected range is set to be located at about 45 °.

declare

FT: Finger-tip

FL: Length of Most High Finger

begin

Function BlobColoring ()

If Blob-> MaxHeight> FL / 2

FT = Blob-> MaxHeight

Else

FT = FL / 4

end.

The above algorithm shows a fingertip correction algorithm in a bending finger state, and FIG. 2 shows that the problem caused by the bending finger is improved through the corrected algorithm.

Based on the general contents of the method for measuring the fingertip position as described above, the overall flow of the method for measuring the fingertip position according to the present invention will be described with reference to FIG. 3 as follows.

As shown in FIG. 3, the fingertip position measuring method according to the present invention includes a first process (S10) of detecting a skin color region by changing to a normalized RGB format using a predetermined red color value and a green color value. If pixels of different colors are found in the skin color region, the binary image is separated by comparing the adjacent pixels based on the binary image value of the pixel and changing to the same color as the adjacent binary image value. A second process (S20) of calculating the area of the hand and a third process (S30) of extracting the highest point of the hand area are made.

Referring to FIG. 4, the third process (S30) is described in more detail. Based on image data input from a single camera, the angle of the fingertip is measured through a PC terminal (S32), and the measurement is performed in step S32. It is determined whether the angle of the fingertip is 0 ° to 89 ° or 90 ° to 180 ° (S34).

As a result of the determination in step S34, when the angle of the fingertip is 0 ° to 89 °, the fingertip is set to 45 ° and extracted as the highest point (S36).

On the other hand, if the determination result of step S34 is not 0 ° to 89 °, it is determined that the present exists in 90 ° to 180 °, and set the fingertip is located at the measurement angle and is extracted to the highest point (S38).

Experiments to evaluate the performance of the proposed algorithm were performed on a PC equipped with an AMD 64 X2 Dual Core Processor 3800+ chip. I used Windows XP Professional as the operating system and Visual C ++ 6.0 as the compiler. The data used in the experiment was input via a webcam with a resolution of 320 * 240. For the performance evaluation, the existing algorithm and the correction algorithm of the bending finger state were repeatedly performed 50 times in the same environment.

As illustrated in FIG. 5, the horizontal axis represents the number of experiments and the vertical axis represents the number of successes according to the number of experiments. Experimental results show that the accurate fingertip could not be measured when the existing algorithm of the bending finger state was used. However, when the correction algorithm was used, about 65% of the fingertip could be measured.

As described above and described with reference to a preferred embodiment for illustrating the technical idea of the present invention, the present invention is not limited to the configuration and operation as shown and described as described above, it is a deviation from the scope of the technical idea It will be understood by those skilled in the art that many modifications and variations can be made to the invention without departing from the scope of the invention. Accordingly, all such suitable changes and modifications and equivalents should be considered to be within the scope of the present invention.

1 is an exemplary view showing a finger tip incorrectly measured when bending a finger.

2 is an exemplary view showing a corrected fingertip in accordance with the present invention.

3 is an overall flow chart of a fingertip position measuring method according to the present invention.

4 is a detailed flowchart of a third process of extracting the highest point of the hand region according to the present invention;

Figure 5 is a graph showing the measurement evaluation of the fingertip measurement evaluation using the conventional algorithm and the fingertip using the algorithm according to the present invention.

Claims (5)

In the finger tip position measurement method, (a) a first step of detecting a skin color region by changing to a normalized RGB format using a preset red color value and a green color value; (b) When pixels of different colors are found in the detected skin color region, the pixels are separated from neighboring objects by comparing them with adjacent pixels based on the binary image value of the pixel and changing them to the same color as the adjacent binary image value. A second process of calculating a region of the hand by performing binarization; And (c) a third process of extracting the highest point of the area of the hand; Including; Step (c) is, (c-1) measuring the angle of the fingertip through the PC terminal based on the image data received from the single camera; (c-2) determining whether the angle of the fingertip recognized through the step (c-1) is 0 ° to 89 °; And (c-3) setting the fingertip to be located at 45 ° and extracting the highest point when the determination result of step (c-2) is 0 ° to 89 °; Fingertip position measurement method comprising a. delete The method of claim 1, If the determination result of step (c-2) is not 0 ° to 89 °, (c-4) determining that the fingertip is present at 90 ° to 180 °, setting the fingertip to be located at the measurement angle, and extracting the most significant point; Fingertip position measurement method characterized in that it further comprises. The method of claim 1, In the above (a) process, Fingertip position measurement method, characterized in that for setting the red color value 30 to 60, Green color value to 25 to 35. delete

Images