CN111798434A - Martial arts competition area detection method based on Ranpac model - Google Patents

Abstract

A martial arts competition area detection method based on a Ranpac model belongs to the technical field of competition area detection. The method solves the problem that the detected competition area is easy to include the non-competition area pixel points due to the poor fault tolerance of the conventional competition area detection method based on hough transformation on the non-boundary pixels. The method comprises the steps of firstly converting a color image in an RGB format into an image in an HSB format, then screening white dot pixels which possibly form the boundary of a competition area by carrying out threshold value screening and neighborhood searching on H channel data, further designing a boundary detection algorithm based on a Ranpac model to fit the boundary of the competition area, and finally processing non-boundary pixels by a ray method, thereby accurately detecting the position of the martial arts competition area.

Description

Martial arts competition area detection method based on Ranpac model

Technical Field

The invention belongs to the technical field of competition area detection, and particularly relates to a martial arts competition area detection method based on a Ranpac model.

Background

Since the region detection plays important roles in target detection and tracking, athlete motion recognition and automatic scoring in the process of sports competition, the technology becomes a hot point technology of image processing research and is applied to some popular sports competitions. For example, a football field marker line detection method (football field marker line detection [ J ] based on Hough transform, computer and digital engineering), a line detection method for detecting a tennis field (rapid tennis court ground detection [ J ] based on Hough transform, technical spread), and a general automatic sports field detection method (automatic field detection method in sports video [ J ]. computer system application). However, no detection method specially aiming at the martial arts competition area exists up to now. In addition, the three methods adopt hough transformation to detect the boundary of the sports field so as to detect the competition area, the hough transformation has poor fault tolerance to non-boundary pixels, and the detected competition area easily comprises non-competition area pixels.

Disclosure of Invention

The invention aims to solve the problem that a detected competition area is easy to include pixel points of a non-competition area due to the poor fault tolerance of the conventional competition area detection method based on hough transformation on non-boundary pixels, and provides a martial arts competition area detection method based on a Ranpac model.

The technical scheme adopted by the invention for solving the technical problems is as follows: a martial arts competition area detection method based on a Ranpac model comprises the following steps:

step one, acquiring an image containing an martial arts competition area;

step two, after the image obtained in the step one is converted into an HSB format image, white point pixel screening is carried out according to H channel data in the HSB format image;

step three, determining a whole pixel point set forming the boundary of the martial arts competition area from the white point pixels screened out in the step two according to a Ranpac model;

and step four, determining the Wushu competition area by adopting a ray method based on the whole pixel point set forming the boundary of the Wushu competition area determined in the step three.

The invention has the beneficial effects that: the invention provides a method for detecting a martial arts competition area based on a Ranpac model, which aims at the characteristics of a martial arts competition field.

Drawings

FIG. 1 is a flow chart of a method for detecting the Wushu competition area based on a Ranpac model;

FIG. 2 is a flow chart of a white point pixel screening method of the present invention;

FIG. 3 is a flow chart for determining an area for a martial arts competition using ray method;

FIG. 4a) is an image of a martial arts competition area for scene 1;

FIG. 4b) is an image of the martial arts competition area of scene 2;

FIG. 5a) is a graph of white point pixel detection results for scene 1;

FIG. 5b) is a graph of white point pixel detection results for scene 2;

fig. 6a) is a diagram of the boundary pixel detection result of scene 1;

fig. 6b) is a diagram of the boundary pixel detection result of scene 2;

FIG. 7a) is a diagram of race area detection results for scenario 1;

fig. 7b) is a competition area detection result diagram of scene 2.

Detailed Description

The first embodiment is as follows: this embodiment will be described with reference to fig. 1. The martial arts competition area detection method based on the Randac model is specifically realized by the following steps:

step one, acquiring an image containing an martial arts competition area;

step two, after the image obtained in the step one is converted into an HSB format image, white point pixel screening is carried out according to H channel data in the HSB format image;

step three, determining a whole pixel point set forming the boundary of the martial arts competition area from the white point pixels screened out in the step two according to a Ranpac model;

and step four, determining the Wushu competition area by adopting a ray method based on the whole pixel point set forming the boundary of the Wushu competition area determined in the step three.

The second embodiment is as follows: the first difference between the present embodiment and the specific embodiment is: in the first step, the format of the acquired image is an RGB format.

The invention designs a white dot pixel screening method, and white dot pixels which possibly form the boundary of a competition area are initially screened out. Since the martial arts competition area boundary is formed of a white line having a certain width, the boundary peripheral area is formed of a blue carpet. According to the condition, the invention designs a pixel screening method based on a color threshold value and a neighborhood searching method as a white point screening condition, and performs primary screening on H channel data. Where the color threshold corresponds to the color characteristic of the white point pixel and the neighborhood search method corresponds to the color characteristic of the pixels surrounding the boundary.

The third concrete implementation mode: the first difference between the present embodiment and the specific embodiment is: the specific process of the second step is as follows:

suppose P represents the set of all pixel points in the HSB format image;

P＝{p_x,y|0≤x≤M-1，0≤y≤N-1} (1)

wherein p is_x,yRepresenting pixel points in the set P, x and y respectively represent horizontal and vertical coordinates of the pixel points, and M, N respectively represent width and height of the HSB format image;

the x direction refers to the width direction of the image, and the y direction refers to the height direction of the image.

A flow chart of the white point pixel screening method is shown in fig. 2.

Let H_x,yRepresenting a pixel point p_x,ySet the threshold range of pixel screening to be [320,360%]If the pixel point p_x,yH channel gray value of_x,yIn the threshold range 320,360]In, then determine pixel point p_x,yIs a candidate point, otherwise, the pixel point p_x,yIs an invalid point;

adopting a neighborhood search method to search the pixel point p_x,ySearching in the neighborhood of (1), judging by pixel point p_x,yWhether there is an H-channel gray value in the 9 × 9 neighborhood centered is [230,250 ]]Pixels within the range, if any, pixel p_x,yIs a white dot pixel, otherwise a pixel point p_x,yFor the invalid point, the formula is:

in the formula, L_x,y1 stands for pixel p_x,yIs a white dot pixel, L _x,y0 stands for pixel p_x,yAs a null point, H_x,y∈[320,360]Representative pixel point p_x,yH channel gray value of_x,yIs taken to be [320,360 ]]Q of_x,yRepresenting by a pixel p_x,ySet of pixel points in 9 × 9 neighborhood centered, H_x+m,y+nRepresentative pixel point p_x+m,y+nH channel gray scale value of (1); m ═ 4, -3, -2, -1,0,1,2,3, 4; n ═ 4, -3, -2, -1,0,1,2,3, 4;

and after traversing each pixel point in the HSB format image, screening out all white point pixels in the HSB format image.

Some white point pixels have been screened out according to equation (2). Some of the white dot pixels belong to the martial arts competition area boundary, and the other part belongs to the background pixels. In order to further screen out area boundary pixels, the invention designs a Randac model special for detecting the boundary of the martial arts competition area, and the Randac model is used for performing straight line fitting on white point pixels. The iterative process of the Randac model is as follows:

the fourth concrete implementation mode: the first difference between the present embodiment and the specific embodiment is: the specific process of the third step is as follows:

step three, selecting any four white point pixels from the white point pixels screened in the step two to form a pixel point set I for the nth iteration⁽ⁿ⁾；

Respectively by pixel point set I⁽ⁿ⁾Each white dot pixel in the four-dimensional image is a vertex, and the vertexes are sequentially connected end to form a quadrangle;

step three and two, traversing the rest white point pixels (except the pixel point set I)⁽ⁿ⁾All white point pixels except for the white point pixels), respectively judging whether the rest white point pixels are on the quadrilateral boundary of the step three, and recording the number C of all white point pixels on the quadrilateral boundaryⁿThe set of all white point pixels on the quadrilateral boundary is B⁽ⁿ⁾；

Step three, repeating the step three to the step three, wherein when the iteration is performed for the (n + 1) th time, a pixel point set consisting of four selected white point pixels is I⁽ⁿ⁺¹⁾；

At the n +1 th iteration, the number of all white point pixels on the quadrilateral boundary is recorded as C⁽ⁿ⁺¹⁾The set of all white point pixels on the quadrilateral boundary is B⁽ⁿ⁺¹⁾；

Step three, step four, if C⁽ⁿ⁺¹⁾＞C⁽ⁿ⁾Then, I is retained⁽ⁿ⁺¹⁾、C⁽ⁿ⁺¹⁾And B⁽ⁿ⁺¹⁾Otherwise, C⁽ⁿ⁺¹⁾≤C⁽ⁿ⁾Retention of I⁽ⁿ⁾，C⁽ⁿ⁾And B⁽ⁿ⁾；

Until all cases of a pixel point set consisting of four white point pixels are traversed, a set B containing the most white point pixels is reserved^(m)，B^(m)The set of all white point pixels on the quadrilateral boundary recorded in the mth iteration is obtained;

B^(m)and I^(m)All white point pixels contained in (A) constitute the whole pixel point set of the boundary of the martial arts competition area, I^(m)And the pixel point set consists of four white point pixels selected in the mth iteration.

In this embodiment, the entire white point pixels are traversed, and whether the white point pixels are on the quadrilateral boundary is determined according to whether the coordinates of the white point are consistent with the coordinates of the boundary pixel points.

Selecting four white point pixels from the white point pixels screened in the step two, and selecting the set B containing the most white point pixels after traversing all possible situations^(m)Corresponding to I^(m)。

The fifth concrete implementation mode: this embodiment will be described with reference to fig. 3. The first difference between the present embodiment and the specific embodiment is: the specific process of the step four is as follows:

and traversing pixel points on the non-regional boundary in the HSB format image one by one, and for a pixel point on a certain non-regional boundary, if the number of the regional boundary pixel points on the horizontal scanning line of the pixel point is K and K% 2 is 0, the pixel point is a pixel point outside the martial arts competition region, otherwise, the pixel point is a pixel point inside the martial arts competition region.

Wherein, K% 2 ═ 0 represents K divided by 2 and 0. And traversing pixel points on the non-region boundary in the HSB format image one by one to obtain all pixel points in the martial arts competition region.

Analysis of Experimental results

The software platform adopted in the experimental part is a 64-bit Windows 7 operating system, and the simulation environment is Matlab 2016 a. The processor model of the hardware platform is Intel i 59400F, the memory capacity is 16GB DDR4, and the video card model is GTX 1660 ti. The simulation process performs area detection on two images containing the martial arts competition area, and the input and output of the program are both digital images with jpg suffix. Images of the martial arts competition area are shown in fig. 4a) and 4 b). The preliminarily screened white dot pixels are shown in fig. 5a) and 5b) by the white dot pixel screening method. The region boundaries fitted by the Randac model are shown in FIG. 6a) and FIG. 6 b). The martial arts competition area resulting from processing non-boundary pixels by ray method is shown in fig. 7a) and 7 b).

As can be seen from fig. 5a) and 5b), the white-point pixel screening method screens the competition area boundary pixels and some non-boundary pixels, which all satisfy the screening condition shown in formula (2). As can be seen from fig. 6a) and 6b), the rannac model can discard non-boundary pixels and leave only the boundary pixels that constitute the race area. As can be seen from fig. 7a) and 7b), the ray method can identify the entire pixels within the boundary from the boundary pixels. From the above experimental results, it can be seen that: the Wushu competition area detection method based on the Ranpac model can effectively eliminate the interference of pixels in a non-competition area, and can accurately detect the Wushu competition area.

The above-described calculation examples of the present invention are merely to explain the calculation model and the calculation flow of the present invention in detail, and are not intended to limit the embodiments of the present invention. It will be apparent to those skilled in the art that other variations and modifications of the present invention can be made based on the above description, and it is not intended to be exhaustive or to limit the invention to the precise form disclosed, and all such modifications and variations are possible and contemplated as falling within the scope of the invention.

Claims (5)

1. A martial arts competition area detection method based on a Ranpac model is characterized by comprising the following steps:

step one, acquiring an image containing an martial arts competition area;

step two, after the image obtained in the step one is converted into an HSB format image, white point pixel screening is carried out according to H channel data in the HSB format image;

step three, determining a whole pixel point set forming the boundary of the martial arts competition area from the white point pixels screened out in the step two according to a Ranpac model;

and step four, determining the Wushu competition area by adopting a ray method based on the whole pixel point set forming the boundary of the Wushu competition area determined in the step three.

2. The method for detecting martial arts competition area based on the Randac model as claimed in claim 1, wherein in the first step, the acquired image is in RGB format.

3. The method for detecting martial arts competition area based on the Randac model as claimed in claim 1, wherein the specific process of the second step is as follows:

suppose P represents the set of all pixel points in the HSB format image;

P＝{p_x,y|0≤x≤M-1，0≤y≤N-1} (1)

wherein p is_x,yRepresenting pixel points in the set P, x and y respectively represent horizontal and vertical coordinates of the pixel points, and M, N respectively represent width and height of the HSB format image;

let H_x,yRepresenting a pixel point p_x,ySet the threshold range of pixel screening to be [320,360%]If the pixel point p_x,yH channel gray value of_x,yIn the threshold range 320,360]In, then determine pixel point p_x,yIs a candidate point, otherwise, the pixel point p_x,yIs an invalid point;

adopting a neighborhood search method to search the pixel point p_x,ySearching in the neighborhood of (1), judging by pixel point p_x,yWhether there is an H-channel gray value in the 9 × 9 neighborhood centered is [230,250 ]]Pixels within the range, if any, pixel p_x,yIs a white dot pixel, otherwise a pixel point p_x,yFor the invalid point, the formula is:

in the formula, L_x,y1 stands for pixel p_x,yIs a white dot pixel, L_x,y0 stands for pixel p_x,yAs a null point, H_x,y∈[320,360]Representative pixel point p_x,yH channel gray value of_x,yIs taken to be [320,360 ]]Q of_x,yRepresenting by a pixel p_x,ySet of pixel points in 9 × 9 neighborhood centered, H_x+m,y+nRepresentative pixel point p_x+m,y+nH channel gray scale value of (1);

and after traversing each pixel point in the HSB format image, screening out all white point pixels in the HSB format image.

4. The method for detecting martial arts competition area based on the Randac model as claimed in claim 1, wherein the concrete process of the third step is as follows:

step three, selecting any four white point pixels from the white point pixels screened in the step two to form a pixel point set I for the nth iteration⁽ⁿ⁾；

Respectively by pixel point set I⁽ⁿ⁾Each white dot pixel in the four-dimensional image is a vertex, and the vertexes are sequentially connected end to form a quadrangle;

step two, traversing the rest white point pixels, respectively judging whether the rest white point pixels are on the quadrilateral boundary of the step three, and recording the number C of all white point pixels on the quadrilateral boundaryⁿThe set of all white point pixels on the quadrilateral boundary is B⁽ⁿ⁾；

Step three, repeating the step three to the step three, wherein when the iteration is performed for the (n + 1) th time, a pixel point set consisting of four selected white point pixels is I⁽ⁿ⁺¹⁾；

At the n +1 th iteration, the number of all white point pixels on the quadrilateral boundary is recorded as C⁽ⁿ⁺¹⁾The set of all white point pixels on the quadrilateral boundary is B⁽ⁿ⁺¹⁾；

Step three, step four, if C⁽ⁿ⁺¹⁾＞C⁽ⁿ⁾Then, I is retained⁽ⁿ⁺¹⁾、C⁽ⁿ⁺¹⁾And B⁽ⁿ⁺¹⁾Otherwise, C⁽ⁿ⁺¹⁾≤C⁽ⁿ⁾Retention of I⁽ⁿ⁾，C⁽ⁿ⁾And B⁽ⁿ⁾；

Until all cases of a pixel point set consisting of four white point pixels are traversed, a set B containing the most white point pixels is reserved^(m)，B^(m)The set of all white point pixels on the quadrilateral boundary recorded in the mth iteration is obtained;

B^(m)and I^(m)All white point pixels contained in (A) constitute the whole pixel point set of the boundary of the martial arts competition area, I^(m)And the pixel point set consists of four white point pixels selected in the mth iteration.

5. The method for detecting martial arts competition area based on the Randac model as claimed in claim 1, wherein the concrete process of the fourth step is as follows:

and traversing pixel points on the non-regional boundary in the HSB format image one by one, and for a pixel point on a certain non-regional boundary, if the number of the regional boundary pixel points on the horizontal scanning line of the pixel point is K and K% 2 is 0, the pixel point is a pixel point outside the martial arts competition region, otherwise, the pixel point is a pixel point inside the martial arts competition region.

Images