JP2013149228A - Position detector and position detection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide position detection technology for measuring a size of a detection object on an image with a resolution higher than a pitch of a pixel equivalent to the resolution of an image when extracting an outline of the object and measuring the size of the object from a vertical or horizontal outline.SOLUTION: A position detector comprises: an outline template that is preliminarily generated and stored; a camera for taking images of a measured object; a matching calculation part for applying a matching calculation to the images of the measured object taken by the camera, while moving the outline template; and a position detection part for detecting a position most coincident with the template as a position of the measured object, from a result of the calculation by the matching calculation part. In the position detector, the position detection part calculates an outline position to a subpixel as a horizontally (or vertically) adjacent function of multiple matching calculation results, with respect to at least one of the positions in X and Y directions of the measured object.

Description

  The present invention relates to a technique for detecting the position of an object in an image with a resolution higher than the pixel pitch, which is the resolution of the image.

  There is a technique for detecting which key on a virtual keyboard is pressed by detecting the position of a finger placed on a desk. Patent Document 1 discloses a technique for measuring the width of a finger from the left and right contours of a finger up to the decimal point in pixel units. As described above, there are many technical fields in which the position and size of an object need to be measured to the image resolution or less.

  Japanese Patent No. 48468871

  In the case of Patent Document 1, in order to determine the width of a finger, first, the left and right contours of the finger are extracted, and then the width of the left and right contours at each Y coordinate is measured to calculate the average value. . Thus, calculation is complicated and processing time is required.

  An object of the present invention is to extract the contour of an object in an image, and to measure the width of the left and right contours (including the top and bottom contours) accurately and in a short time to the resolution of the image or less.

  In order to solve the above problems, the position detection device of the present invention includes a contour template created and stored in advance, a camera that acquires an image of the object to be measured, and the object to be measured acquired by the camera. Position detection for detecting a position that most closely matches the template as a position of the object to be measured from a result calculated by the matching calculation unit and a matching calculation unit that performs matching calculation while moving the contour template In the position detection device constituted by a portion, the position detection portion contours at least one position in the X direction and the Y direction of the object to be measured as a function of a plurality of matching calculation results close to the left and right (or up and down). Is measured up to a subpixel.

  By configuring in this way, it is possible to determine that the object corresponds to the template with the highest match from the matching result with the contour template of a plurality of objects, and measure the position of the contour to the sub-pixel. be able to.

  The position detection apparatus of the present invention is characterized in that the size of the object to be measured is measured up to the subpixel by measuring both sides of the contour of the object to be measured.

  The position detection apparatus of the present invention also stores a storage unit that measures and stores the size of the object to be measured at the initial position, a result of measuring the size of the object to be measured at the current position, and the storage unit. The change of the distance from the camera of the object to be measured is measured by comparing the size of the object to be measured at the initial position.

  Further, the contour template of the present invention is created from a contour image of an object to be measured at an initial position.

  An example of the object to be measured according to the present invention is a user's finger placed on a desk, and the position detection unit measures a moving distance from the initial position of the user's finger placed on the desk. And

  With this configuration, it is possible to accurately determine which key on the virtual keyboard has been pressed. In addition, it is highly useful for operating a virtual mouse that requires a higher resolution than that of a virtual keyboard.

  By accurately detecting the position and size of the fingers on the desk, the operation of the virtual keyboard and virtual mouse can be easily realized. Similarly, it can be widely used in the field of image recognition where it is necessary to detect the position and size of an object on an image with high accuracy.

It is explanatory drawing of operation of a virtual mouse. It is explanatory drawing of operation of a virtual mouse. It is explanatory drawing of operation of a virtual mouse. It is explanatory drawing of operation of a virtual mouse. It is explanatory drawing of the outline of a finger, and an outline template. It is explanatory drawing of the outline of a finger, and an outline template. It is explanatory drawing of the outline of a finger, and an outline template. It is explanatory drawing of the outline of a finger, and an outline template. It is a figure which shows the pattern of a correlation matching result. It is an example of the flowchart of a back-and-forth movement distance calculation. It is an example of the hardware block diagram of a position detection apparatus.

Embodiments of the present invention will be described below with reference to the drawings.
1A, 1B, 1C, and 1D are explanatory diagrams of the operation of the virtual mouse. FIG. 1A shows a user 101, a right hand 102 for operating a mouse pointer, and a forefinger 103 in an image 100 photographed by a camera placed close to a desk. FIG. 1B shows the initial position and size of the finger by bringing only the index finger 104 of the right hand into contact with the desk and resting. FIG. 1C is a diagram in which the index finger 105 of the right hand has moved to the front right on the desk and the size of the finger on the image has increased. FIG. 1D shows a display screen 106 provided separately and the position of the displayed mouse pointer. The initial position 107 of the mouse pointer is associated with the initial position and size of the right index finger 104 in FIG. 1B, and the position 108 of the mouse pointer after movement corresponds to the position and size after movement of the right index finger 105 in FIG. 1C. Attached.

  The size of the index finger on the image changes by moving the right hand index finger back and forth, and the mouse pointer is moved back and forth accordingly. Therefore, the mouse pointer can be moved back and forth with good resolution by detecting the width of the finger with a resolution higher than the pixel pitch on the image. Thus, the present invention provides a technique that can be widely used in a technical field in which the performance of an apparatus is determined depending on how much the position of an object can be detected with a resolution higher than the pixel pitch on the image.

  2A, 2B, 2C, and 2D are explanatory diagrams of a finger contour and a contour template. The right contour 110 of the finger in FIG. 2A is cut out and shown in FIG. 2B. FIG. 2B shows each pixel 112 on the camera image and the outline 111 of the finger in an overlapping manner. The contour 111 of the finger is detected over a plurality of pixels of the X coordinate. In the figure, an example of about 3 pixels is shown. There are various reasons for extending to a plurality of pixels. Due to the cylindrical shape of the finger, there are optical system factors such as a gradual change in the reflection of light at the contour and a defocus of the camera.

  FIG. 2C shows a change in luminance by taking out the pixel row 113 in FIG. 2B. In FIG. 2C, the horizontal axis represents the X coordinate, the vertical axis represents the luminance of each pixel, and each point represents one pixel. In this example, the brightness on the outside of the finger outline is low and the inside is high. The pixel 114 has the largest differential value in the X direction, but the luminance of the left and right pixels also changes.

  FIG. 2D is an example of a contour template. FIG. 2B is an extracted pixel having an effective luminance change, and each pixel 115 is an absolute value of a differential value in the X direction of the original image 112 and has a different value for each pixel. In the figure, a 4 × 4 contour template is shown, but in the case of a finger contour, more pixels are effective as contour information in the Y direction, and a template of several tens of pixels is used. An outline template is usually created with a rectangular area, but in the case of a finger outline, the outline is slanted, and peripheral images such as nails become noise, so that the effective differentiation as a finger outline as shown in FIG. 2D It is effective to use a deformed template consisting only of pixels having values.

  An image in which each of the pixels of the original image to be detected is replaced with the absolute value of the X-direction differential (hereinafter referred to as a differential image) while the contour template thus created is moved pixel by pixel in the X-axis direction and Y-axis direction When the correlation matching calculation is performed, the maximum value is obtained at the position of the finger contour image. The correlation matching calculation can be obtained by the following formula 1.

  Here, T (i, j) is a contour template, indicating that it has i pixels in the X-axis direction and j pixels in the Y direction. I (x, y) is a differential image. Since both the contour template and the differential image have a plurality of consecutive significant differential values in the X-axis direction, (x-1, y) and (x + 1, y) adjacent to (x, y) where the correlation matching result is the maximum value. y) also provides a significant correlation matching result.

  FIG. 3 is a diagram showing a pattern of a correlation matching result. In (1) of FIG. 3, the pixel number N is the maximum value, and the correlation matching result at the X coordinates on both sides is the same value. In this case, the X coordinate of the contour is N. In (5) of FIG. 3, this is shifted to the right by one pixel, and the X coordinate of the contour in this case is N + 1. In FIG. 3 (3), pixel numbers N and N + 1 have the same maximum value, and the X coordinate of the contour in this case is N + 1/2. As can be seen from (2) and (4) in FIG. 3 which are intermediate patterns of these, the X coordinate of the contour can identify a minute coordinate difference by these patterns. Formula 2 is an example of a formula for obtaining the X coordinate of the contour from the pattern of FIG.

  Here, the coordinates (x, y) are the coordinates at which the result of the correlation matching calculation is maximized, and the range i to be added is the range where the contour template exists on the left and right with the coordinates (x, y) as the center. By this calculation, an average coordinate with the weight R as the correlation matching value R (x + i, y) in the range i to be added is obtained. In principle, the resolution of the contour position can be obtained up to the level of the AD conversion resolution of the image.

  Although the X coordinate of the right contour of the finger can be detected with high accuracy as described above, the width of the finger can be measured with high accuracy by detecting the left contour of the finger in the same manner. In addition, by accurately measuring the width of the finger at the initial position and the width of the finger after movement, the change in the relative distance between the finger and the camera can be calculated, and the mouse pointer can be moved with high resolution accordingly. Can be made.

  FIG. 4 is an example of a flowchart for calculating the longitudinal movement distance. The distance of the object on the image to be moved back and forth with respect to the camera is calculated. Correlation matching calculation is performed while shifting the right contour template one pixel at a time in the X-axis direction and the Y-axis direction with respect to the coordinates where the detection target may exist in the differential image (step S1). Similarly, correlation matching calculation is also performed for the left contour template (step S2). When the detection object moves back and forth with respect to the camera, there is usually an upper and lower limit for the change in size on the image. Therefore, in steps S1 and S2, a plurality of contour candidates are extracted in order from the maximum value of the matching calculation result, and in step S3, a pair of contours allowed as the size of the target object is extracted from the left and right contour candidates, Among them, the largest sum of the left and right correlation matching calculated values is selected and determined as the contour of the object.

  Accurate coordinates of the determined right and left contours are calculated according to Equation 2 (step S4). Since the correlation matching value itself has already been calculated in step 1 and step 2, it is only necessary to perform weight calculation in step 4. Next, the width between the left and right contours is obtained from the difference between the left and right contour coordinates (step S5). This determines the size of the measurement object.

  From the comparison of the size at the initial position of the measurement object stored in advance and the size on the differential image this time, the distance of the front and rear distance from the camera of the measurement object is calculated by a predetermined calculation (step) S6).

  FIG. 5 is an example of a hardware block diagram of the position detection device 120. An image captured from the camera 121 is processed by the processor 122 and stored in the RAM 125. 4 is realized by the software program 124 written in the ROM 123. The detected position information of the object is appropriately displayed on the display 126.

DESCRIPTION OF SYMBOLS 100 Image photographed with camera 101 User 102 Operated right hand 103, 104, 105 Index finger 106 Display screen 107 Initial position of mouse pointer 108 Position of mouse pointer after movement 110 Right contour of finger 111 Contour of finger 112 On camera image Each pixel 113 Pixel row 114 Pixel 115 Pixel 120 Position detection device 121 Camera 122 Processor 123 ROM
124 program 125 RAM
126 display

Claims (5)

Contour templates created and stored in advance,
A camera for acquiring an image of the object to be measured;
A matching calculation unit that performs a matching calculation while moving the contour template with respect to an image of the object to be measured acquired by the camera;
From the result calculated by the matching calculation unit, a position detection unit that detects the position most consistent with the template as the position of the object to be measured;
In a position detection device comprising:
The position detecting unit measures the position of the contour to a sub-pixel as a function of a plurality of matching calculation results close to the left and right (or up and down) for at least one of the X direction and the Y direction of the object to be measured. Detection device.   The position detection apparatus according to claim 1, wherein the position of the object to be measured is measured by measuring the positions of both sides of the contour of the object to be measured. A storage unit for measuring and storing the size of the object to be measured at the initial position;
By comparing the result of measuring the size of the object to be measured at the current position with the size of the object to be measured at the initial position stored in the storage unit, the change in the distance of the object to be measured from the camera can be determined. The position detecting device according to claim 2, further comprising a moving distance measuring unit for measuring.   The position detection apparatus according to claim 3, wherein the contour template is created from a contour image of an object to be measured at an initial position.   The measurement object is a user's finger placed on a desk, and the moving distance measuring unit measures a change in a distance of the user's finger placed on the desk from the camera. Item 5. The position detection device according to Item 3 or 4.

Images