JPH02279181A - System for measuring pitch of runner - Google Patents

Abstract

PURPOSE:To measure the pitch of a runner at processing speed close to real time by masking only the leg part of the runner, extracting the leg part in a picture processing part, catching the motion of a foot in the leg part according to the area change of the foot, analyzing the time change of the area by a frequency analysis part and calculating the pitch of the runner. CONSTITUTION:A pitch measuring system is composed of a picture input part 1, picture processing part 2 and frequency analysis part 3. Then the picture processing part 2 is composed of a means 21 to extract an area in specified color from an input picture, masking means 22 to define only a certain specified area as a processing object out of the extracted picture and means 23 to measure the number of picture elements with a specified density value in the masked picture. By using the pitch measuring system, based on picture data from the input picture in which the leg part of the runner to be outputted from the picture input part 1 is photographed, only the leg part of the runner is masked and extracted by the picture processing part 2 and the motion of the foot for the runner is caught according to the area change of the foot. Then the time change of the area is analyzed by the frequency analysis part 3 and the pitch of the runner is calculated. Frequency analysis can be processed only for several seconds when a personal computer and work station, etc., for example, are used.

Description

[Detailed Description of the Invention] [(Summary)] A method for measuring the pitch of a runner in marathons, etc. using images input from a television camera, etc. A method for measuring the pitch of a runner at a processing speed close to real time without human intervention. For the purpose of measurement, an image input unit, means for extracting a specific color area from the input image, a masking unit for processing only a specific area of the extracted image, and a masking unit for processing only a specific color area of the extracted image. A pitch measurement system is constituted by an image processing section comprising a means for measuring the number of pixels of a specific density value in an image, and a frequency analysis section, and using the pitch measurement system,
Based on the image data from the input image output from the image input unit that shows the legs of the runner, the image processing unit masks and extracts only the legs of the runner, and extracts the legs of the runner. The movement is captured by the change in area of the foot, and the change in area over time is
The pitch of the runner is determined by analysis by a frequency analysis section.

[Industrial application field]

The present invention relates to a method for measuring the pitch of a runner in a marathon or the like using images input from a television camera or the like.

In track and field events such as marathons, a runner's pitch provides useful data not only for exercise physiology but also for estimating expected finish time.

In particular, live broadcasts of marathons require real-time performance, so a high-speed pitch measurement system is required.

[Prior Art and Problems to be Solved by the Invention] FIG. 3 is a diagram illustrating a conventional method for measuring the pitch of a runner.

Conventionally, it has been impossible to automatically measure the pitch of runners in marathons, etc., so as shown in this figure, the video captured by a television camera 1a is recorded on a videotape recording device (VTR) lb. After recording the data, it is transported to, for example, a measurement room, where it is played back in slow motion on a display 4, where people can visually watch and take measurements.

Therefore, the videotape recording device? All processes such as rewinding the recording to the W (VTR) lb, playing it back in slow motion, measuring, and calculating the pitch from the measurement results must be performed manually. There is a problem in that it takes more than one minute just to try to find the pitch.

As described above, conventional pitch measurement methods are time-consuming, and it is virtually impossible to measure a runner's pitch quickly enough in fields that require real-time performance, such as live broadcasting. there were.

SUMMARY OF THE INVENTION In view of the above conventional drawbacks, an object of the present invention is to provide a pitch measuring method that measures a runner's pitch at a processing speed close to real time without requiring human intervention. It is something.

[Means for Solving the Problems] FIG. 1 is a diagram explaining the principle of the present invention, in which (a) shows a configuration example, (b) shows a processing flow in an image processing section, and (c)
l) and (c3) show examples of changes in foot area over time,
(c2) and (c4) show examples of power spectra of frequency analysis results.

The above problem can be solved by a runner's pitch measuring method configured as follows.

An image input unit l; means 21 for extracting a specific color region from the input image; masking means 22 for processing only a specific region of the extracted image; An image processing unit 2 comprising a means 23 for measuring the number of pixels with a specific density value and a frequency analysis unit 3 constitute a pitch measurement system, and the pitch measurement system is used to calculate the number of pixels output from the image input unit 1. Based on the image data from the input image that shows the legs of the runner, the image processing unit 2 masks and extracts only the legs of the runner, and calculates the movements of the legs of the runner. The frequency analysis unit 3 captures the change in area and calculates the time change in the area.
The configuration is such that the pitch of the runner is determined by analysis.

[Effect]

That is, according to the present invention, an input image captured by a television camera is processed by the image processing unit at the same speed as the television image (per second,
30 frames), the following image processing is performed.

First, from the input color image, the skin color part of the runner is extracted using, for example, "Color Extraction Method J" disclosed in Japanese Patent Application No. 165792/1983 filed by the applicant of the present application. .

In this color extraction method, for example, a lookup table (LUI'
) Look up the address using the idF scale of the total of 16 bits from the upper few bits of each data of the three primary colors.

Table (LOT) A is accessed and the hue, saturation, and brightness of the runner's skin are extracted as one-word data.

Next, a lookup table (LOT) is created to extract a specific skin color (white, yellow, black, etc.) area using the hue, saturation, and brightness information composed of the 8 bits (1 hide) as an address. By accessing B, an area having a specific pixel value is extracted pixel by pixel, for example, including the pixel value 'FF''.

Using this technique, the background image and the human part are separated from the input color image.

After that, a mask is applied to the lower part of the screen, that is, the legs of the person, and only the legs of the runner are extracted.

In the masked part, the pixel value is the above 'FF'
The area of the area is measured using a histogram calculation, and changes in the area of the foot over time are captured.

FIG. 1 (cl) shows the area-time change of the apparent upper leg when the background of the input image and the runner's leg can be completely separated.

As the runner's foot moves up and down, the area of the foot changes periodically.

Therefore, in the frequency analysis section, the image obtained by the image processing section is
The area change rx(t), 1 of the runner's foot, the fast Fourier transform (FFT) L, and the frequency component X(f) are determined.

The relationship between ×(t) and X (f) above is X(f)
= x(t)e−”Ldt.

Since the obtained X(f) has a real part and an imaginary part, the power spectrum IX(f)12 is
It can be obtained by ``l''=(real part)z+(fflG part)''.

As shown in FIG. 1 (cl), for example, if 3 cycles take 1 second, the power spectrum IX(f)2 is (c
2) As shown in the figure, it has a peak at 311z.

This means that the runner's foot moves up and down three times per second, so in this case, it can be determined that the runner's pitch is 3 steps/second.

However, since both feet are usually captured in the input image,
It is difficult to capture only one foot.

Therefore, in reality, the time transformation of the sum of the areas of both feet is captured. At this time, since the temporal changes in the area of the right foot and the left foot are exactly half out of phase, the waveform of the temporal change of the sum of these will be flat compared to the case of one foot. (See FIG. 1 (c3)) Furthermore, if the background of the input image and the part of one foot are incompletely separated, the area of that part becomes noise.

However, since the Fourier-transformed noise component has a −-like distribution (in other words, it is white noise), if the Fourier transform is used, it is possible to find the frequency of the pix, and the runner's You can find the pitch. (
(See FIG. 1(c4)) Image processing can be performed in real time, and the time required for the frequency analysis is sufficiently short.

Therefore, by using the pitch measuring system of the present invention, there is an effect that the pitch of a runner can be determined at high speed without human intervention.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

The above-mentioned FIG. 1 is a diagram explaining the present invention in detail, and FIG.
The figures are diagrams showing one embodiment of the present invention, in which (a) shows an example of the system configuration, and (b) schematically shows the progress of pitch measurement processing, in which input from a television camera 1a is shown. Extract a region of a specific color from the image, mask the human foot part from the extracted region, and calculate the area time of the part of the specific color (i.e., the foot) from the masked region. Means for determining the change in area, performing frequency analysis from the time change of the determined area, and determining the power spectrum are necessary means for carrying out the present invention. Note that the same reference numerals indicate the same objects throughout the figures.

Hereinafter, with reference to FIG. 1 and FIG. 2, a method for measuring the pitch of a runner according to the present invention will be explained.

First, the pitch measuring system of the present invention is constructed using a television camera 1 a + an image processing device 21 as shown in FIG. 2(a).
When photographing a runner from a television camera la, a device consisting of a computer 3 that performs frequency analysis, (1) ↑
The darkest conditions do not change significantly over time.

(2) The area of a certain region (foot) must be close enough to be measured with sufficient accuracy.

(3) Multiple runners are not included in the measurement area.

It is desirable that the following conditions are met.

The image processing device 2 of the present invention performs the following processing.

■ Color processing Using the "color extraction method J" disclosed in the above-mentioned patent application No. 165792/1982 "Video Ray 1 Color Extraction Device",
For each pixel of the input image, for example, the upper few bits of the 8 bits of the three primary colors of the pixel, red (R), green (G), and blue (B), are used to determine the hue and color of each pixel. Access the lookup table A for calculating saturation and brightness, calculate the hue, saturation, and brightness of each pixel as, for example, 1 byte data, and determine the color of each pixel based on that value. A lookup table B is accessed to extract a specific skin color part, and the background of the input image and the human are separated.

At this time, the pixel value of the extracted portion is output as, for example, FF.

■ Noise processing The salt-and-pepper noise that occurs due to the incomplete color discrimination process described in ■ above can be removed using, for example, a median filter (for example, by scanning an image with a 3 x 3 window and sorting it in ascending order of pixel values). , a filter whose median value is the pixel value at the center of the window), a spatial filter (for example,
The noise (usually, the noise has a sharp pixel value) is smoothed using a filter that scans the image in a 3 x 3 window and uses the average value of all pixels as the pixel value at the center of the window. , taking advantage of the fact that human images are planar) can be removed.

■ Mask Processing In order to measure the area of the person's body, focusing only on the feet, the mask is set so that only the lower part of the screen will be the area to be processed from now on.

(2) Area calculation The number of specific image values (the above "FF" portion) is calculated by histogram calculation for the image obtained by performing the above processes (1) to (2).

Specifically, by the histogram calculation, the number of pixels for each pixel value in the n range of object names is determined, and only the number of pixels having a pixel value of °FF' is taken out.

Processes ① to ② above are performed at video speed, and the time change in area of the foot portion is determined. (Figure 1 (cl).

(See (c3)) The computer (frequency analysis unit) 3 that performs frequency analysis calculates, based on the time change data of the runner's foot calculated by the image processing device 2, the fast Fourier transformation tA (ppr) result. Calculate the power spectrum described above. (Figure 1(c)
2) (See (c4)) Read the pitch of the runner from the calculated power spectrum.

In the runner's pitch measurement method according to the present invention, the area is measured, for example, 30 times per second (i.e., 3011
If the measurement (sampling at z) is repeated N times,
Since it is a discrete measurement, a measurement error occurs, but the pitch calculation error in this case follows the following formula. That is, when frequency analysis is performed using fast Fourier transform (FFT), the frequency f of the first data is the sampling frequency F
Then, f=(i/N)F Therefore, the pitch calculation error Δ is Δ= ((i+1)/NI F−(i/N)F=F/N In other words, the above 30 times per second In area measurement, the above error Δ
becomes 307N.

Also, since the measurement time T is T=N/30, the above error Δ is Δ-1/'T Therefore, the measurement error Δ is as shown in the table below, and the measurement time can be increased appropriately. It can be seen that this can be ignored.

In this way, the present invention extracts a specific color region from an input image from a television camera, and from the extracted region,
Mask the human foot part and from the masked area,
Find the time change in the area of the specific colored part (i.e. foot), perform frequency analysis from the time change in the determined area,
The feature is that the pitch of the runner is determined by determining the power spectrum.

[Effects of the Invention] As described above in detail, the method for measuring the pitch of a runner of the present invention can be used to measure the pitch of a runner in a marathon or the like using an image input from a television camera or the like.
an image input section, means for extracting a specific color region from the input image, a masking means for processing only a specific region of the extracted image, and a specific density in the masked image. An image processing unit consisting of a means for measuring the number of pixels of a value and a frequency analysis unit constitute a pitch measurement system, and the pitch measurement system is used to calculate the image of the runner's legs output from the image input unit. Based on the image data from the input image, the image processing unit masks and extracts only the legs of the runner, captures the movement of the runner's legs by changing the area of the legs, and calculates the area. The pitch of the runner is determined by analyzing the temporal changes of , which can be processed in a few seconds, has the effect of being able to quickly determine the pitch of a runner without human intervention.

[Brief explanation of drawings]

FIG. 1 is a diagram explaining the principle of the present invention. FIG. 2 is a diagram showing an embodiment of the present invention. FIG. 3 is a diagram illustrating a conventional pitch measuring method for a runner. It is. In the drawing, ■ is an image input unit + la is a television camera. lb is a videotape recording device (VTR). 2 is an image processing unit, or the image processing device 21 is a means for extracting a skin color area using color processing. 22 is a means for extracting the leg portion by mask processing. 23 is a means of measuring the area of the foot. 3 is a frequency analysis section or a computer. 4 is the display. ■ indicates color processing, ■ indicates noise removal, ■ indicates mask processing, and ■ indicates area calculation. (cl) Time change in the area of one leg 3 Hz r (c2) Power spectrum of (cl) Figure 1 Display A diagram explaining the conventional runner's Pino 1000-1j method

Claims (1)

[Claims] An image input unit (1), means (21) for extracting a specific color area from an input image, and a masking unit for processing only a specific area of the extracted image. (22), a means (23) for measuring the number of pixels with a specific density value in the masked image, and a frequency analysis unit (3). and using the pitch measurement system, the image processing unit (2) calculates the distance between the runner and the runner based on image data from an input image in which the legs of the runner are outputted from the image input unit (1). Only the legs of the runner are masked and extracted, the movement of the runner's feet is captured by changes in the area of the feet, and the temporal changes in the area are analyzed by the frequency analysis unit (3) to determine the runner's pitch. A method for measuring a runner's pitch, which is characterized by the following: