JP2006163662A - Device and method for recognizing number of fingers - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device and method capable of correctly recognizing the number of fingers. <P>SOLUTION: This device comprises an image pickup device 1, a means for setting an edge reference direction in a picked-up image, an edge line image acquiring means for vertically scanning an image with respect to the set edge reference direction and extracting an edge line, a means for setting a finger thickness reference value in the image, a finger region extracting means for vertically scanning the image with respect to the edge reference direction, measuring an interval between a positive edge line and a negative edge line to calculate a difference between the measured interval and a set finger thickness reference value in the case of a negative edge line which intersects a positive edge line, wherein edge strength in the edge line extracted by the edge line image acquiring means has a positive value and where the next edge line appearing on the same scanning line has negative edge strength, and extracting a region corresponding to a region between the two lines as a finger region, and the number of finger regions counting means for counting the number of the finger regions. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a finger number recognition apparatus and method for recognizing how many fingers of a hand are stretched.

A system for recognizing the shape of a hand from an image captured by a camera or the like is described in Patent Document 1 below.
In this system, first, an image captured based on skin color information of a color image is binarized, noise is removed, and a hand region is extracted. Next, the extracted hand region is converted into a thin line image, and the direction of the hand is detected from the thin line image. Next, scanning is performed in a direction perpendicular to the detected hand direction, and the run length over the hand region is measured. Next, the frequency distribution of the measured run length is obtained, and the run length at which the frequency is maximum is set as a finger width threshold value. Next, scanning is performed perpendicular to the direction of the hand, and if the run length is equal to or less than the finger width threshold, it is extracted as a finger region, and from the logical product of the thinned image and the finger region Extract the corresponding finger area. When removing noise from the binarized image, a smoothed contour line of the binarized image is extracted, and the number of fingers is estimated from the shape of the contour line. Using this estimated value, the line corresponding to noise is removed from the lines extracted as the line corresponding to the finger. As a result, a line corresponding to the finger is finally extracted.

JP 2003-346162 A

In the above conventional system, a hand region is extracted by binarization based on the luminance value of the captured original image and a fixed threshold value, and the run length, contour shape, and thinned image are extracted for the extracted hand region. Since the finger area is extracted using these, if the luminance value of the hand falls outside the assumed threshold range, for example, optical disturbance in the visible image or change in the temperature of the hand in the infrared image In such a case, it is not possible to cope with the occurrence of an error or the like, so that an accurate hand region cannot be extracted, and it may be erroneously recognized.
An object of the present invention is to provide a finger number recognition apparatus and method capable of suppressing such erroneous recognition and accurately recognizing the number of fingers.

  In order to solve the above-described problems, the present invention provides an imaging unit that captures an image of a hand, an edge reference direction setting unit that sets an edge reference direction corresponding to the orientation of the hand in the captured image, and a captured image. , Edge line image acquiring means for extracting edge lines by scanning in a direction perpendicular to the edge reference direction set by the edge reference direction setting means, and acquiring a finger line in the captured image Finger thickness reference value setting means for setting the thickness reference value, and scanning in a direction perpendicular to the set edge reference direction, the edge strength in the extracted edge line intersects with a positive edge line having a positive value, When the next edge line that appears on the same scanning line is a negative edge line having a negative edge strength, the positive edge line and the negative edge line A finger region extracting means for measuring an interval, obtaining a difference between the measured interval and a set finger thickness reference value, and extracting a corresponding region between the two lines as a finger region based on the difference; And a finger area number counting means for counting the number of the finger areas.

  According to the present invention, it is possible to provide a finger number recognition apparatus and method capable of suppressing erroneous recognition and accurately recognizing the number of fingers.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings described below, components having the same function are denoted by the same reference numerals, and repeated description thereof is omitted.
《Configuration of finger number recognition device》
FIG. 2 is a diagram showing the configuration of the finger number recognition apparatus according to the embodiment of the present invention.
In FIG. 2, 1 is an image pickup device, 2 is a computing device, 3 is a memory, and 4 is a processing device based on the number of fingers. The image capturing apparatus 1 captures an image of a hand held by a user of the finger number recognition apparatus. For example, an electronic camera. The arithmetic device 2 processes the image picked up by the image pickup device 1 and recognizes the number of fingers extended in the hand held by the user. The memory 3 stores information necessary for executing the recognition process in the arithmetic device 2. Information on the number of fingers recognized by the arithmetic device 2 is delivered to the processing device 4 based on the number of fingers. The processing device 4 based on the number of fingers performs an arbitrary process with information on the number of fingers extended by the user as input. For example, the shape of the hand (number of fingers) is used as computer input information, and various processes are executed based on the input information. Needless to say, the structure may be replaced with another form as long as the processing in this embodiment is possible.

<Processing flow>
FIG. 1 is a diagram showing an outline of the processing flow of the finger number recognition apparatus and method according to the present embodiment. The outline of the finger number recognition process will be described below with reference to FIG.
First, in S101, an image of a hand held by the user is acquired by the image pickup apparatus (see FIG. 2).
Next, in S102, an edge reference direction for extracting an edge used for recognition of the number of fingers by the arithmetic device 2, that is, the direction of the hand in the image is set for the image acquired in S101.
Next, in S103, the image acquired in S101 is scanned (searched) in a direction perpendicular to the edge reference direction set in S102, and an edge line based on the edge reference direction is extracted to create an edge line image. To do.
Next, in S104, a reference value (hereinafter referred to as a finger thickness reference value) of the thickness (width) of the finger in the image is set using the edge line image created in S103.
Next, in S105, scanning is performed in a direction perpendicular to the edge reference direction set in S102, the edge intensity in the edge line extracted in S103 intersects with a positive edge line having a positive value, and is on the same scanning line. When the edge line that appears next is a negative edge line having a negative edge strength, the interval between the positive edge line and the negative edge line is measured. Then, the difference between the measured interval and the finger thickness reference value set in S104 is obtained. If this difference is small, a finger area extraction image is created by extracting and extracting the finger area as an area corresponding to the finger between the two edge lines.
Next, in S106, the number of finger areas in the finger area extraction image created in S105 is counted.
Finally, in S107, an arbitrary process is performed based on the number of fingers, assuming that the number of finger areas counted in S106 corresponds to the number of fingers.

Hereinafter, processing contents at each processing stage in the processing flow shown in FIG. 1 will be described.
<Acquiring images>
In S101, an image of the hand held by the user is acquired using an arbitrary imaging system as shown in the image imaging device 1 of FIG. 2, and is delivered to the calculation unit 2 as a processing target image. The calculation unit 2 may perform an arbitrary noise removal process on the delivered image as necessary. The processing target image at the time when it is delivered to the calculation unit 2 is defined as I (n). Here, n = 1,..., N, the pixels of the processing target image are N pixels in total, and n corresponds to each pixel of the processing target image.

<Setting the edge reference direction>
In S102, an edge reference direction for extracting an edge line is set for the image acquired in S101 in the processes after S103. FIG. 3 is a diagram showing a processing flow in setting the edge reference direction in S102.
In the present embodiment, it is assumed that the edge reference direction is either the vertical direction or the horizontal direction of the image.
First, in S301 and S302, the frequency of vertical and horizontal edges is calculated.
In S301, the frequency of vertical edges is acquired. A vertical edge line image Ev (n) = Sv (I (n)) is acquired for the processing target image I (n) acquired in S101. Sv (•) is an arbitrary vertical edge extraction filter such as a Sobel filter. At this time, the number of pixels of | Ev (n) |> θe is assumed to be the edge frequency Hv in the vertical direction. θe is an arbitrary threshold value.
In S302, the frequency of horizontal edges is acquired. A horizontal edge line image Eh (n) = Sh (I (n)) is acquired for the processing target image I (n) acquired in S101. Sh (•) is an arbitrary horizontal edge extraction filter such as a Sobel filter. At this time, it is assumed that the number of pixels | Eh (n) |> θe is the horizontal edge frequency Hh. θe is an arbitrary threshold value.

In step S303, the vertical edge frequency Hv is compared with the horizontal edge frequency Hh. If Hv> Hh is satisfied, the process proceeds to S304. Otherwise, the process proceeds to S305.
In S304, the edge reference direction is the vertical direction, and the variable MD for identifying the edge reference direction is set to 90.
In S305, the edge reference direction is the horizontal direction, and the variable MD for identifying the edge reference direction is set to zero.
Here, the variable MD for identifying the edge reference direction corresponds to an angle when the horizontal direction of the processing target image is 0 °. Here, the horizontal direction, that is, 0 °, or the vertical direction, ie, 90 °. Therefore, in the horizontal direction, that is, in the x direction, the edge is set at MD = 0, and in the vertical direction, that is, in the y direction, MD = 90 is set.
It should be noted that a method other than the processing shown here may be used as long as it is known which direction the edge is mainly directed. Further, any direction other than the vertical and horizontal directions may be used as the edge reference direction. Furthermore, when the edge reference direction, that is, the direction of the hand on the screen is known, the MD value may be set by storing the direction in the memory 3 and referring to it.

<Acquisition of edge line image>
In S103, an edge line based on the edge reference direction set in S102 is extracted from the image acquired in S101 to create an edge line image.
The contents of the process in S103 will be described with reference to FIGS. FIG. 4 is a diagram illustrating a processing flow in edge line extraction in S103, FIGS. 5A to 5B are diagrams illustrating edge line extraction, and FIG. 6 is a diagram illustrating edge line correction.
First, in S401, the edge line image G (n) in the three directions of the edge reference direction MD obtained in S102 and the ± 45 ° direction of the edge reference direction MD with respect to the processing target image I (n) acquired in S101. ) = S (I (n)). S (•) is a filter for edge acquisition, and an arbitrary method such as a Sobel filter is used.
Next, in S402, for the edge line image G (n) created in S401, a label of 1 is added (labeled) to a pixel having an edge intensity equal to or greater than a positive threshold value, and a pixel having an intensity equal to or less than a negative threshold value. A label of 2 is added to the other pixels, and a label of 0 is added to the other pixels to obtain a ternary image T (n). That is, T (n) = 1 if G (n)> θp, T (n) = 2 if G (n) <θm, and T (n) = 0 otherwise. Here, θp and θm are arbitrary threshold values.
Next, in S403, noise removal is performed on the image T (n) ternarized in S402. An image from which noise has been removed is assumed to be T ′ (n) = M (T (n)). Here, M (•) is an arbitrary noise removal filter, and an arbitrary method such as an expansion / reduction filter may be used.

The process from S401 to S403 will be described with reference to FIGS. Here, it is assumed that the edge reference direction is the vertical direction. FIG. 5A shows a part of the input image. The imaged object 501 in FIG. 5A is a hand with two fingers extended, and it is assumed that there is a lot of noise in the image. At this time, if the edge line image in S401 is acquired by any method, the ternarization in S402 and the noise removal in S403 are performed, as a result, the edge region shown in FIG. 5B is obtained. In FIG. 5B, reference numerals 502 and 504 denote positive edge regions. When pixels belonging to these regions are n _p , G (n _p )> θp and T (n _p ) = 1. Further, reference numerals 503 and 505 denote negative edge regions. When pixels belonging to these regions are _nm , G (n _m ) <θm and T (n _m ) = 2. If the pixels belonging to other regions are n _o , T (n _o ) = 0.

In S404 and later, thinning processing is performed on the ternary image T ′ (n) from which noise has been removed in S403. For each region corresponding to the edge where T ′ (n) = 1, 2, a smooth curve with no branching is obtained for each region. As a technique for this, in the present embodiment, steps S404 to S405 are taken. Here, the case where the edge reference direction is vertical will be described.
In S404, each pixel is connected to a portion where T ′ (n) = 1 corresponding to the positive edge is connected, and another pixel is T ′ (n) = 2 corresponding to the negative edge. An image K (n) to which a label is added is obtained by using each of the connected areas as one region (clustering). K (n) is a label image in which a label value is added to each pixel corresponding to each region to which a label is added. Pixels corresponding to the positive edge region have values of K (n _p ) = 1, 2,..., Kp. Here, Kp is the number of positive edge regions. Similarly, pixels corresponding to the negative edge region have values of K (n _m ) = − 1, −2,..., −Km. Here, Km is the number of negative edge regions. As a result of adding a label, if the number of pixels belonging to a certain label is equal to or less than an arbitrary threshold, it is possible to remove noise by substituting the label of that pixel with 0, assuming that the region is caused by noise.

  Referring to FIG. 5, a label image is created by adding a label to each region in FIG. 5B as described above. A label image is shown in FIG. In FIG. 5B, the edge regions 502 and 504 and 503 and 505 have the same value, but in FIG. 5C, the edge regions 502, 504, 503, and 505 are different from each other. Label values are assigned and can be recognized as different areas.

In S405, an edge line is set. First, each region labeled in S404 is scanned in the direction perpendicular to the edge reference direction, and for each pixel in the edge reference direction, in the direction perpendicular to the edge reference direction (that is, in this case, the horizontal direction and the x direction). Temporarily set the midpoint as an edge line node. As an example, the case where the edge reference direction is the vertical direction will be described with reference to FIG.
Consider setting an edge line for the edge region 601 shown in FIG. 6 in which a certain cluster label is k in the region to which the label is added in S404. From the minimum coordinate ys in the y direction to the maximum coordinate ye, each y coordinate is scanned in the direction perpendicular to the edge reference direction, that is, in the x direction. At this time, an intermediate value xm (y) = (xr (y) between the maximum value xr (y) of the x coordinate and the minimum value xl (y) among the pixels where K (n) = k at a certain y coordinate value. ) + Xl (y)) / 2 is temporarily set as the node 602 of the edge line at the coordinate y. At this time, as shown in FIG. 6A, a node group in which the increase / decrease in the x coordinate value is significant is obtained. Subsequently, correction is performed so that xm (y) (ys <y <ye) of each node becomes smooth. An arbitrary method may be used as the correction method. For example, the x coordinate values of the upper and lower neighboring nodes are referred to, and a new x coordinate value is obtained as an average. In this manner, the node group of the smooth nodes 603 as shown in FIG. 6B is corrected, and a line connecting these nodes 603 is set as a final edge line. The pixel value on the edge line finally obtained is replaced with the label number 1 or 2 in S404, and other areas are replaced with 0 to obtain an edge line image L (n). That is, edge line images L (n) = 1, 2,..., Kp, 0, −1, −2,.

The edge regions 506, 507, 508, and 509 in FIG. 5C are obtained in a noisy shape due to the influence of noise and blurring of the edges. Therefore, when a line connecting the center coordinates of the edge regions 506 to 509 is simply used as an edge line, an edge line deviating from the original finger shape is obtained due to the influence of noise. Therefore, in S405, a line connecting the center coordinates of the edge regions 506 to 509 is simply extracted as a temporary line, and then the noise of the line is removed, so that a smooth line along the original finger shape is obtained. Extract. As a result, edge lines 510, 511, 512, and 513 as shown in FIG. As a result of edge line extraction, if the length of the line is outside an arbitrary threshold range, the pixel value belonging to that line can be erased by setting it to 0, and noise can be removed.
If one smooth edge line that passes through the vicinity of the edge peak and is not branched for one edge region can be obtained, a method other than the thinning method in this embodiment is used. It doesn't matter.

<Setting the finger thickness reference value>
In S104, a finger thickness reference value in the image is set using the edge line image created in S103.
Here, the case where the edge reference direction is vertical will be described. When the edge reference direction is horizontal, the vertical / horizontal relationship may be switched.
This will be described with reference to FIGS. FIG. 7 is a diagram for explaining extraction of the palm width, and FIG. 8 is a diagram for explaining the basis for setting the finger thickness reference value.
First, the palm width Wmax in the image is obtained.
When the input image is as shown in FIG. 7A, an edge line image as shown in FIG. 7B is obtained. The edge line image in FIG. 7B is scanned from the upper left in the direction perpendicular to the edge reference direction. On the scan line at each y coordinate, the intersection of the positive label edge line and the negative label edge if the line that intersects the positive label edge line and then intersects is a negative edge line Measure the distance to the intersection with the line. As a result of scanning the entire screen, the maximum distance Wmax is recorded as the palm width. In FIG. 7B, the distance indicated by 701 corresponds to Wmax.

Next, a finger thickness (width) reference value Wfin is obtained. Let Wfin = α × Wmax. α is an arbitrary threshold value, and generally around 0.25 is appropriate. The basis for this will be described with reference to FIG.
In general, the relative value between the palm width and the finger width is constant. If this is used, even if the size of the hand in the image is different, the width of the finger in the image can be obtained as long as the width of the palm in the image is known. Factors that cause changes in the size of the hand in the image include (1) the distance between the camera and the hand changes, and (2) there are individual differences in the size of the hand.

<Finger area extraction>
In S105, scanning is performed in a direction perpendicular to the edge reference direction set in S102, and the interval between a point on the positive edge line appearing on the scanning line and the next negative edge line appearing on the same scanning line is measured. Then, a difference between the measured interval and the finger thickness reference value set in S104 is obtained, and if the difference is small, the area between the two positive and negative edge lines is extracted as a finger area corresponding to the finger. Create an image.
The content of the finger area extraction process in S105 will be described with reference to FIG. FIG. 9 is a diagram for explaining extraction of a finger region. Assume that the edge reference direction is vertical, that is, MD = 90.
Let F (n) be the final finger region extraction image. First, F (n) is initialized to F (n) = 0 for all pixels.
Assume that an edge line image shown in FIG. 9A is obtained as a processing result in S104. This edge line image is scanned from the upper left in the direction perpendicular to the edge reference direction, and the line that intersects with the edge line of the positive label on the scanning line at each y coordinate is the negative edge. If it is a line, the distance (interval) between the intersection of the edge line of the positive label and the intersection of the edge line of the negative label is measured. Let this distance be Wp. Here, the subscript p is a value of 1, 2,..., P, and P is the total number of places between the positive edge line and the negative edge line. The measured distance Wp is compared with the finger thickness reference value Wfin, and if | Wp−Wfin | <θfin, the pixels on the pth straight line sandwiched between the positive edge line and the negative edge line are Replace with F (n) = 1. θfin is an arbitrary threshold value.

  In the areas other than the fingers, that is, the lines obtained for the palm, wrist, and arm, the difference between the distance Wp sandwiched between the positive edge line and the negative edge line and the finger thickness reference value Wfin is equal to or greater than the threshold value. Therefore, F (n) = 0 remains in the region sandwiched between these lines.

The above will be described with reference to FIG. 9. The distances 901 and 902 in FIG. 9A have a small difference from the finger thickness reference value Wfin. Therefore, the pixels on the distances 901 and 902 are 1. On the other hand, the distance 903 has a larger difference than the finger thickness reference value Wfin. Therefore, the pixels on the distance 903 are 0.
As a result of the above operation, the finger region extraction image F (n) takes a value of F (n) = 1 in the pixels corresponding to the finger region, and takes a value of F (n) = 0 in the other pixels. In FIG. 9B, reference numerals 904 and 905 denote finger areas.

<Counting finger area>
In S106, the number of finger areas Fin_Num in the finger area extraction image F (n) created in S105 is counted. The counting method may be any method such as labeling.

<Processing means based on the number of fingers>
In S107, an arbitrary process based on the number of fingers is performed assuming that the number of finger areas counted in S106 corresponds to the number of fingers. For example, the shape of the hand (number of fingers) is used as computer input information, and various processes are executed based on the input information.

  As described above, the finger number recognition device according to the present embodiment is a finger number recognition device that recognizes how many fingers of a hand are stretched, and includes an image pickup device 1 that picks up an image of a hand, Edge reference direction setting means for setting an edge reference direction corresponding to the direction of the hand in the image picked up by the image pickup apparatus 1 and edge reference set by the edge reference direction setting means in the image picked up by the image pickup apparatus 1 An edge line image acquiring means for acquiring an edge line image by scanning in a direction perpendicular to the direction, and a finger for setting a finger thickness reference value in the image captured by the image capturing apparatus 1 The thickness reference value setting means and the edge reference direction set by the edge reference direction setting means are scanned in a direction perpendicular to the edge reference direction and extracted by the edge line image acquisition means A positive edge if the edge strength at the edge of the wedge line intersects a positive edge line having a positive value and the next edge line appearing on the same scan line is a negative edge line having a negative edge strength. The distance between the line and the negative edge line is measured, and a difference between the measured interval and the finger thickness reference value set by the finger thickness reference value setting means is obtained. Based on this difference, the two It is configured to include a finger area extracting unit that extracts a corresponding area between lines as a finger area, and a finger area number counting unit that counts the number of the finger areas.

  Further, the finger number recognition method of the present embodiment is a finger number recognition method for recognizing how many fingers of a hand are stretched, and an imaging process for capturing an image of a hand, and imaging in the imaging process. In the image, an edge reference direction setting step for setting an edge reference direction corresponding to the orientation of the hand, and a direction perpendicular to the edge reference direction set in the edge reference direction setting step in the image captured in the imaging step Edge line image acquisition step of acquiring an edge line image by scanning the edge line, and a finger thickness reference value setting step of setting a finger thickness reference value in the image captured in the imaging step; Scanning in a direction perpendicular to the edge reference direction set in the edge reference direction setting step, and the edge line extracted in the edge line image acquisition step If the edge strength intersects with a positive edge line having a positive value and the next edge line appearing on the same scan line is a negative edge line having a negative edge strength, the positive edge line The interval between the negative edge line is measured, and a difference between the measured interval and the finger thickness reference value set in the finger thickness reference value setting step is obtained. Based on this difference, the two lines It is configured to include a finger region extracting step for extracting a corresponding region as a finger region, and a finger region number counting step for counting the number of finger regions.

In the above conventional system, a hand region is extracted by binarization based on the luminance value of the captured original image and a fixed threshold value, and the run length, contour shape, and thinned image are extracted for the extracted hand region. Since the finger area is extracted using these, if the luminance value of the hand falls outside the assumed threshold range, for example, optical disturbance in the visible image or change in the temperature of the hand in the infrared image In such a case, it is not possible to cope with the occurrence of an error or the like, so that an accurate hand region cannot be extracted, and it may be erroneously recognized. Also, when setting the threshold of the finger width, the frequency of the run length is measured, the run length with the maximum frequency is set as the threshold of the finger width, and the hand region having the run length of the width less than the threshold is set as the finger region. If the run length with the maximum frequency does not correspond to the width of the finger, that is, the frequency of the palm or wrist run length other than the finger is greater than the frequency of the finger run length. For example, there is a problem that misrecognition may occur when the hand region is extracted over a wide range including the wrist, or when the finger is short relative to the palm or the number of fingers is small. Further, based on the luminance value of the original image, a hand region is extracted by binarization based on a fixed threshold value, a thinned image is obtained for the extracted hand region, and the thinned line direction vector is obtained. By detecting the direction of the hand, scanning in the direction perpendicular to the detected direction of the hand, the frequency of the run length is obtained, and the run length having the maximum frequency is set as the threshold value of the finger width. Therefore, if there is a lot of noise in the image, the noise in the shape of the hand area obtained by binarization also increases, and as a result, the line of the thinned image also points in the direction along the shape of the hand. It can happen. In such a case, the direction of the hand cannot be obtained correctly, the threshold value of the finger area based on the run length cannot be obtained accurately, and an erroneous finger area can be extracted, resulting in erroneous recognition. There's a problem.
According to the present embodiment, the information used for recognition is edge information, that is, a value based on the spatial differentiation of the image, and does not use the size of the luminance value of the image itself. The number of fingers can be recognized even when the brightness of the hand region changes based on optical disturbances in the visible image as in the above-described prior art or temperature changes in the hand in the infrared image. be able to.

In addition, the finger thickness reference value setting unit sets the maximum interval among the intervals between the positive edge line and the negative edge line measured by the finger region extraction unit as a palm width value, The finger thickness reference value is set on the basis of the set palm width value. In this way, the edge line with the positive strength and the edge line with the negative strength are adjacent to each other, and the distance at which the distance between the two is the maximum is set as the palm width value. By setting the finger thickness reference value by normalizing as, it is possible to extract a wide range including the wrist as a hand area, or the fingers are short relative to the palm, the number of fingers is small, In such a case, the finger thickness reference value is not erroneously set, and as a result, the number of fingers can be recognized stably. At the same time, because it is normalized based on the palm, individual differences in hand size, changes in hand size in the image due to changes in the distance between the imaging system and the subject hand, etc. Even if this occurs, the finger thickness reference value is not set incorrectly, and as a result, the number of fingers can be recognized stably.
The edge reference direction setting means calculates edge frequencies in a plurality of directions in the image captured by the image capturing apparatus 1, compares the calculated edge frequencies, and selects the direction with the highest frequency as the edge reference. It is supposed to be a direction. In this way, the scanning direction is determined by calculating the edge frequency of each direction in the acquired image, comparing the calculated edge frequencies, and determining the direction with the highest frequency as the hand reference. By adopting a configuration in which the direction is set, even a noisy image is hardly affected by noise and stable recognition can be performed.
Further, the edge line image acquisition means creates the edge line image for the edge reference direction set by the edge reference direction setting means and three directions of ± 45 °, and each edge region of the created edge line image Edge lines that pass smoothly around the center of the are extracted. In this way, as an edge line creation method, an edge line image in which edges are calculated in the reference direction of the edge and its three directions ± 45 ° is created, and a line that smoothly passes through the center of each edge region in the edge line image is extracted. With this configuration, even when there is a lot of noise in the image, or when the contrast of the image is weak and the contour is not sharp, that is, even when the shape of the edge area of the edge line image is noisy, it is stable. Edge lines can be extracted, and as a result, a stable number of fingers can be recognized.

The embodiment described above is described in order to facilitate understanding of the present invention, and is not described in order to limit the present invention. Therefore, each element disclosed in the above embodiment includes all design changes and equivalents belonging to the technical scope of the present invention.
Further, the correspondence between each component in the claims and each component in the embodiment of the invention will be described. The image pickup apparatus 1 of the embodiment and S101 of FIG. The computing devices 2 and S102 are edge reference direction setting means, the computing devices 2 and S103 are edge line image acquisition means, the computing devices 2 and S104 are finger thickness reference value setting means, and the computing devices 2 and S105 are finger regions. The computing device 2 and S106 correspond to the extraction means, and the finger area number counting means, respectively. S101 is an imaging step, S102 is an edge reference direction setting step, S103 is an edge line image acquisition step, S104 is a finger thickness reference value setting step, S105 is a finger region extraction step, and S106 is the number of finger regions. It corresponds to each counting process.

It is a figure which shows the outline | summary of the processing flow of the finger number recognition apparatus and method of embodiment of this invention. It is a figure which shows the structure of the finger number recognition apparatus of embodiment of this invention. It is a figure which shows the processing flow in the setting of the edge reference direction of FIG. It is a figure which shows the processing flow in extraction of the edge line of FIG. It is a figure explaining extraction of the edge line of an embodiment of the invention. It is a figure explaining the correction | amendment of the edge line of embodiment of this invention. It is a figure explaining extraction of the width of the palm of an embodiment of the invention. It is a figure explaining the basis of the setting of the finger thickness reference value of the embodiment of the present invention. It is a figure explaining extraction of the finger field of an embodiment of the invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Image pick-up device 2 ... Arithmetic device 3 ... Memory 4 ... Processing device 501 based on the number of fingers ... Object to be picked 502, 504 ... Positive edge region 503, 505 ... Negative edge region 506, 507, 508, 509 ... Edge Area 510, 511, 512, 513 ... Edge line 601 ... Edge area 602, 603 ... Nodes 701, 801, 802, 803, 804, 901, 902, 903 ... Distance 904, 905 ... Finger area

Claims (5)

A finger number recognition device that recognizes how many fingers of a hand are stretched,
An imaging means for capturing an image of a hand;
Edge reference direction setting means for setting an edge reference direction corresponding to the orientation of the hand in the image captured by the imaging means;
In the image captured by the imaging means, edge line image acquisition means for extracting an edge line by scanning in a direction perpendicular to the edge reference direction set by the edge reference direction setting means, and acquiring an edge line image;
Finger thickness reference value setting means for setting a finger thickness reference value in the image captured by the imaging means;
Scan in a direction perpendicular to the edge reference direction set by the edge reference direction setting means,
A negative edge line in which the edge line in the edge line extracted by the edge line image acquisition means intersects with a positive edge line having a positive value, and the next edge line appearing on the same scanning line has a negative edge intensity. If it is, measure the interval between the positive edge line and the negative edge line,
The difference between the measured interval and the finger thickness reference value set by the finger thickness reference value setting means is obtained, and based on this difference, a region corresponding to the area between the two lines is extracted as a finger region. Region extraction means;
Finger area number counting means for counting the number of finger areas;
A finger number recognition apparatus characterized by comprising: The finger thickness reference value setting means includes:
Of the intervals between the positive edge line and the negative edge line measured by the finger region extraction means, the maximum interval is set as the palm width value,
2. The finger number recognition apparatus according to claim 1, wherein the finger thickness reference value is set on the basis of the set palm width value. The edge reference direction setting means includes
In each of the images picked up by the image pickup means, the edge frequencies in a plurality of directions are calculated,
2. The finger number recognition apparatus according to claim 1, wherein the calculated edge frequencies are compared, and the direction with the highest frequency is set as the edge reference direction. The edge line image acquisition means includes
Creating the edge line image for the edge reference direction set by the edge reference direction setting means and the three directions of ± 45 °;
4. The finger number recognition apparatus according to claim 3, wherein an edge line that smoothly passes near the center of each edge region of the created edge line image is extracted. A finger number recognition method for recognizing how many fingers of a hand are stretched,
An imaging process for capturing an image of the hand;
An edge reference direction setting step for setting an edge reference direction corresponding to the orientation of the hand in the image captured in the imaging step;
In the image captured in the imaging step, an edge line image acquisition step of extracting an edge line by scanning in a direction perpendicular to the edge reference direction set in the edge reference direction setting step, and acquiring an edge line image;
A finger thickness reference value setting step for setting a finger thickness reference value in the image captured in the imaging step;
Scan in a direction perpendicular to the edge reference direction set in the edge reference direction setting step,
A negative edge line in which the edge line in the edge line extracted in the edge line image acquisition step intersects with a positive edge line having a positive value, and the next edge line appearing on the same scanning line has a negative edge intensity. If it is, measure the interval between the positive edge line and the negative edge line,
The difference between the measured interval and the finger thickness reference value set in the finger thickness reference value setting step is obtained, and a region corresponding to the area between the two lines is extracted as a finger region based on the difference. Region extraction process;
A finger area number counting step for counting the number of finger areas;
A finger number recognition method characterized by comprising:

Images