CN107527357A - Face datection positioning and method for real time tracking in Violent scene - Google Patents

Abstract

The present invention provides Face datection positioning and method for real time tracking in a kind of Violent scene, and this method includes the light stream vector amplitude for obtaining a pixel in monitoring video frame；Judge whether monitoring video frame is Violent scene according to light stream vector amplitude.The center O in face tracking region in initial Violent scene monitoring video frame is calculated according to the colour of skin average in YCbCr space and colour of skin variance⁰.According to the center O in face tracking region in initial Violent scene monitoring video frame⁰The center O in face tracking region in calculated for subsequent Violent scene monitoring video frame^m+1, finally according to the center O in face tracking region^m+1To obtain the face tracking region of m+1 frame Violent scene frame of video, detection positioning and the real-time tracking of face in Violent scene are realized.

Description

Face detection positioning and real-time tracking method in violent scenes

technical field

The invention relates to the field of video and image processing, in particular to a face detection, positioning and real-time tracking method in a violent scene.

Background technique

Face detection, positioning and real-time tracking in violent scenes have important applications in criminal monitoring occasions such as community access control systems and shopping mall video systems. At present, under ideal conditions, frontal face detection has achieved satisfactory results. However, in complex backgrounds, due to factors such as multi-pose, occlusion, and illumination, the success rate of face detection and real-time tracking is low.

Current mainstream recognition methods (such as LGBP, neural network, PCA) are all based on static images for face recognition, which cannot be applied to face detection, positioning and tracking in violent scenes in video surveillance, and lack a mechanism for real-time tracking of multiple objects. Its usefulness is limited. During the movement of the human body (such as pushing and wrestling), the face will vibrate greatly, which will cause changes in the posture characteristics of the face, which will reduce the accuracy of detection and tracking.

Contents of the invention

In order to overcome the problem that the existing technology cannot recognize faces in violent scenes, the present invention provides a face detection, positioning and real-time face recognition in violent scenes that can accurately detect, locate and track faces in violent scenes and has strong real-time performance. tracking method.

In order to achieve the above object, the present invention provides a face detection positioning and real-time tracking method in a violent scene, the method comprising:

Step 1, obtain the optical flow vector magnitude of each pixel in the surveillance video frame:

Among them, (u _i,j,t ,v _i,j,t ) is the optical flow of pixel p(i,j,t), where (i,j) is the position of the pixel in the surveillance video frame, and t is the video frame sequence index;

Step 2, judging the violent scene, when the following conditions are met, the surveillance video frame is characterized as a violent scene surveillance video frame:

Wherein, Th is the optical flow judgment threshold, and N is the number of pixels in one frame of image;

Step 3, converting the violent scene monitoring video frame from RGB space to YCbCr color space and establishing an image color histogram;

Step 4, calculate the center position O ⁰ of the face tracking area in the initial violent scene monitoring video frame according to the skin color mean value and skin color variance in the YCbCr space;

Step ⁵ , according to the center position O of the face tracking area in the initial violent scene monitoring video frame, use the following formula to obtain the center position O ^m+1 of the face tracking area in the follow-up violent scene monitoring video frame:

Wherein, k=1...L, m≡0...t, L is the histogram segment index, t is the video frame sequence index, O ^m is the center position of the face tracking area in the current violent scene monitoring video frame, O ^m+1 is the center position of the face tracking area in the next violent scene monitoring video frame, I'(i, j) is the Cb brightness of other pixels in the violent scene monitoring video frame, δ is a Dirac function, and c(i, j) is The number of histogram segments where the pixel p(i, j, t) is located, N1 is the number of pixels in the vertical direction of the image, and N2 is the number of pixels in the horizontal direction of the image;

Step 6, use the edge detection operator to obtain the critical pixels of the face tracking area centered on O ^m+1 , and carry out curve fitting to the critical pixels to form the face tracking area of the m+1th violent scene video frame until The judgment conditions for violent scenes are no longer met.

According to an embodiment of the present invention, the optical flux (u _i,j,t ,v _i,j,t ) is calculated in a differential manner, and the calculation formula is as follows:

where I is the image brightness of pixel p(i,j,t), is the horizontal brightness gradient, is the longitudinal brightness gradient, is the brightness gradient on the time axis.

According to an embodiment of the present invention, the determination of the center position O of the face tracking area in the initial violent scene monitoring video frame adopts the following steps ^:

First, calculate the difference S(I(i,j)) between the Cb brightness and skin color of each pixel in the initial violent scene monitoring video frame:

Among them, μ is the mean value of skin color, and σ is the variance of skin color;

Secondly, judge the face area. When S(I(i,j))>Th_skin, the representative pixel p(i,j,t) belongs to the face area, where Th_skin is the difference threshold of skin color;

Finally, extract the face area, extract critical pixels according to the relationship between S(I(i,j)) and Th_skin and perform curve fitting to form the face tracking area in the initial violent scene monitoring video frame and obtain the initial violent scene monitoring video frame The center position O ⁰ of the inner face tracking area.

According to an embodiment of the present invention, adopt following formula to convert monitoring video frame from RGB space to YCbCr color space:

Y＝0.257*R+0.564*G+0.098*B+16 (7)

Cb＝-0.148*R-0.291*G+0.439*B+128 (8)

Cr＝0.439*R-0.368*G-0.071*B+128 (9)

Among them, Y represents the brightness, Cb reflects the difference between the blue component and the brightness of the RGB input, and Cr reflects the difference between the red component and the brightness of the RGB input.

According to an embodiment of the present invention, an elliptic curve fitting method is used to perform curve fitting on critical pixels so as to form a face tracking area.

According to an embodiment of the present invention, in step six, the edge detection operator is any one of local difference operator, Sobel operator or Canny operator.

In summary, the face detection, positioning and real-time tracking method in the violent scene provided by the present invention adopts the improved optical flow method and curve fitting to carry out the face detection, positioning and tracking method, which can overcome the difficulty of face detection caused by human body movement. Pinpoint and track difficult issues. Further, the method for face detection, positioning and real-time tracking in violent scenes provided by the present invention simplifies the algorithm as much as possible on the basis of satisfying the accuracy of detection, positioning and tracking of faces. This algorithm greatly improves the real-time performance of location tracking and has certain robustness to face deformation; it is convenient to implement in embedded systems such as ARM microcontrollers or single-chip microcomputers, and can be well compatible with existing access control monitoring systems.

In order to make the above and other objects, features and advantages of the present invention more comprehensible, preferred embodiments are described below in detail with accompanying drawings.

Description of drawings

FIG. 1 is a flow chart of a face detection, positioning and real-time tracking method in a violent scene provided by an embodiment of the present invention.

detailed description

As shown in Figure 1, the face detection and positioning and real-time tracking method in the violent scene provided by this embodiment adopts the improved optical flow method to detect the face in the violent scene. The optical flow vector magnitude m _i,j,t of pixels.

Among them, (u _i,j,t ,v _i,j,t ) is the optical flow of pixel p(i,j,t), where (i,j) is the position of the pixel in the surveillance video frame, and t is the video frame sequence index.

The optical flow (u _i,j,t ,v _i,j,t ) optical flow (u _i,j,t ,v _i,j,t ) is calculated by difference, and the calculation formula is as follows:

where I is the image brightness of pixel p(i,j,t), is the horizontal brightness gradient, is the longitudinal brightness gradient, is the brightness gradient on the time axis.

The traditional optical flow method is calculated in a differential way, using the basic equation of optical flow and adding certain constraints to obtain a dense optical flow field. This method has a very large amount of calculation, poor real-time performance, and is difficult to use. For tracking fast-moving or shaking faces in violent scenes, it cannot be used for face recognition and tracking. This embodiment first adopts the difference method to replace the differential in the traditional optical flow method, and then uses the optical flow vector magnitude m _i,j,t with a simple calculation method as the judgment parameter of the violent scene, which has high detection accuracy and the calculation method Simple, low requirements on the microprocessor in the access control system, compatible with the access control system in the community or office building. Furthermore, a large number of experiments show that the detection of violent scenes using optical flow vector magnitude has good robustness.

Step S20 is executed after the optical flow vector magnitude m _{i ,j,t} is obtained, and whether the current surveillance video frame is a violent scene is judged according to the optical flow vector magnitude mi,j,t.

When the magnitude of the optical flow vector of the surveillance video frame satisfies formula (2), it indicates that the surveillance video frame is a violent scene surveillance video frame.

Wherein, Th is the optical flow judgment threshold, and N is the number of pixels in one frame of image.

After judging that the current video frame is a violent scene surveillance video frame, it is necessary to recognize faces in the violent scene. Step S30 is executed to convert the violent scene monitoring video frame from RGB space to YCbCr color space and establish an image color histogram.

The formula for converting RGB space to YCbCr color space is as follows:

Y＝0.257*R+0.564*G+0.098*B+16 (7)

Cb＝-0.148*R-0.291*G+0.439*B+128 (8)

Cr=0.439*R-0.368*G-0.071*B+128 (9).

Y represents brightness, Cb reflects the difference between the blue component and brightness of RGB input, and Cr reflects the difference between the red component and brightness of RGB input.

Execute step S40, calculate the center position O ⁰ of the face tracking area in the initial violent scene monitoring video frame according to the skin color mean value and skin color variance in the YCbCr space, and the specific calculation method is as follows:

First, calculate the difference S(I(i,j)) between the Cb brightness and skin color of each pixel in the initial violent scene video frame:

Among them, μ is the mean value of skin color, and σ is the variance of skin color;

Secondly, judge the face area. When S(I(i,j))>Th_skin, the representative pixel p(i,j,t) belongs to the face area, where Th_skin is the difference threshold of skin color;

Finally, extract the face area, extract the critical pixels according to the relationship between S(I(i,j)) and Th_skin and perform curve fitting to form the face tracking area in the initial violent scene video frame and obtain the human face in the initial violent scene video frame The center position O ⁰ of the face tracking area.

Execute step ^S50 after obtaining the center position O of the face tracking area in the initial ^violent scene monitoring video frame, use the following formula to obtain the follow-up violent scene monitoring video frame according to the center position O of the face tracking area in the initial violent scene monitoring video frame The center position O ^m+1 of the inner face tracking area:

Wherein, k=1...L, m≡0...t, L is the histogram segment index, t is the video frame sequence index, O ^m is the center position of the face tracking area in the current violent scene monitoring video frame, O ^m+1 is the center position of the face tracking area in the next violent scene monitoring video frame, I'(i,j) is the Cb brightness of other pixels in the violent scene monitoring video frame, δ is the Dirac function, c(i,j) is the number of histogram segments where the pixel p(i, j, t) is located, N1 is the number of pixels in the vertical direction of the image, and N2 is the number of pixels in the horizontal direction of the image.

Step S60, take Om ⁺¹ as the center and use the edge detection operator to obtain the critical pixels of the face tracking area, and carry out curve fitting to the critical pixels to form the face tracking area of the m+1th violent scene video frame, Until the judging conditions for violent scenes are no longer met, that is, Complete face recognition in surveillance video frames of all violent scenes.

The face detection and positioning and real-time tracking method in the violent scene provided by the present embodiment utilizes the center position of the face tracking area in the previous violent scene monitoring video frame to obtain the center position of the face tracking area in the latter violent scene monitoring video frame, The point-to-point calculation method has a small amount of calculation, so it can greatly improve the real-time performance of face tracking, and then use the edge detection operator to form the center area of the face according to the skin color of the face (skin color is an important feature of the face). Very good detection accuracy.

In this embodiment, the edge detection operator is a local difference operator. However, the present invention does not make any limitation thereto. In other embodiments, other edge detection operators such as Sobel operator or Canny operator can be used.

The face is basically elliptical and the parameters of ellipse fitting are relatively simple. Therefore, in this embodiment, an elliptic curve fitting method is used when forming a face tracking area in a violent scene monitoring video frame. However, the present invention does not make any limitation thereto. In other embodiments, other curves with higher matching accuracy to the contour of the human face can be used for fitting. The higher the accuracy of curve fitting, the corresponding amount of calculation it needs will also increase. This embodiment adopts the method of ellipse fitting to balance the fitting accuracy and the amount of calculation, so that it can be better applied to applications with limited computing power. in the access control system.

In summary, the face detection, positioning and real-time tracking method in the violent scene provided by the present invention adopts the improved optical flow method and curve fitting to carry out the face detection, positioning and tracking method, which can overcome the difficulty of face detection caused by human body movement. Pinpoint and track difficult issues. Further, the method for face detection, positioning and real-time tracking in violent scenes provided by the present invention simplifies the algorithm as much as possible on the basis of satisfying the accuracy of detection, positioning and tracking of faces. This algorithm greatly improves the real-time performance of location tracking and has certain robustness to face deformation; it is convenient to implement in embedded systems such as ARM microcontrollers or single-chip microcomputers, and can be well compatible with existing access control monitoring systems.

Although the present invention has been disclosed above by preferred embodiments, it is not intended to limit the present invention. Any skilled person can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection shall be subject to the scope of protection required by the claims.

Claims (6)

1. A face detection positioning and real-time tracking method in a violent scene, characterized in that it comprises: Step 1, obtain the optical flow vector magnitude of each pixel in the surveillance video frame: <mrow><msub><mi>m</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi><mo>,</mo><mi>t</mi></mrow></msub><mo>=</mo><msqrt><mrow><msubsup><mi>u</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi><mo>,</mo><mi>t</mi></mrow><mn>2</mn></msubsup><mo>+</mo><msubsup><mi>v</mi><mrow><mi>i</mi><mo>,</mo><mi>i</mi><mo>,</mo><mi>t</mi></mrow><mn>2</mn></msubsup></mrow></msqrt><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>1</mn><mo>)</mo></mrow></mrow> Among them, (u _i,j,t ,v _i,j,t ) is the optical flow of pixel p(i,j,t), where (i,j) is the position of the pixel in the surveillance video frame, and t is the video frame sequence index; Step 2, judging the violent scene, when the following conditions are met, the surveillance video frame is characterized as a violent scene surveillance video frame: <mrow><mfrac><mrow><munder><mo>&amp;Sigma;</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow></munder><msup><mrow><mo>(</mo><msub><mi>m</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi><mo>,</mo><mi>t</mi><mo>+</mo><mn>1</mn></mrow></msub><mo>-</mo><msub><mi>m</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi><mo>,</mo><mi>t</mi></mrow></msub><mo>)</mo></mrow><mn>2</mn></msup></mrow><mi>N</mi></mfrac><mo>&gt;</mo><mo>=</mo><mi>T</mi><mi>h</mi><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>2</mn><mo>)</mo></mrow></mrow> Wherein, Th is the optical flow judgment threshold, and N is the number of pixels in one frame of image; Step 3, converting the violent scene monitoring video frame from RGB space to YCbCr color space and establishing an image color histogram; Step 4, calculate the center position 0° of the face tracking area in the initial violent scene monitoring video frame according to the skin color mean value and the skin color variance in the YCbCr space; Step 5, according to the center position O° of the face tracking area in the initial violent scene monitoring video frame, use the following formula to obtain the center position O ^m+1 of the face tracking area in the follow-up violent scene monitoring video frame: <mrow><msup><mi>O</mi><mrow><mi>m</mi><mo>+</mo><mn>1</mn></mrow></msup><mo>=</mo><mfrac><mrow><munderover><mi>&amp;Sigma;</mi><mrow><mi>j</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>N</mi><mn>1</mn></mrow></munderover><munderover><mi>&amp;Sigma;</mi><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>N</mi><mn>2</mn></mrow></munderover><mo>|</mo><msup><mi>I</mi><mo>&amp;prime;</mo></msup><mrow><mo>(</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mo>)</mo></mrow><mo>-</mo><mi>I</mi><mrow><mo>(</mo><msup><mi>O</mi><mi>m</mi></msup><mo>)</mo></mrow><mo>|</mo><msub><mi>Q</mi><mi>k</mi></msub></mrow><mrow><munderover><mi>&amp;Sigma;</mi><mrow><mi>k</mi><mo>=</mo><mn>1</mn></mrow><mi>L</mi></munderover><msub><mi>Q</mi><mi>k</mi></msub></mrow></mfrac><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>3</mn><mo>)</mo></mrow></mrow> <mrow><msub><mi>Q</mi><mi>k</mi></msub><mo>=</mo><munderover><mo>&amp;Sigma;</mo><mrow><mi>j</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>N</mi><mn>1</mn></mrow></munderover><munderover><mo>&amp;Sigma;</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>N</mi><mn>2</mn></mrow></munderover><mi>exp</mi><mrow><mo>(</mo><mo>-</mo><mfrac><mn>1</mn><mn>2</mn></mfrac><mo>|</mo><mi>I</mi><mo>(</mo><msup><mi>O</mi><mi>m</mi></msup><mo>)</mo><mo>-</mo><msup><mi>I</mi><mo>&amp;prime;</mo></msup><mo>(</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mo>)</mo><msup><mo>|</mo><mn>2</mn></msup><mo>)</mo></mrow><mi>&amp;delta;</mi><mrow><mo>(</mo><mi>c</mi><mo>(</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mo>)</mo><mo>-</mo><mi>k</mi><mo>)</mo></mrow><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>4</mn><mo>)</mo></mrow></mrow> Wherein, k=1...L, m≡0...t, L is the histogram segment index, t is the video frame sequence index, O ^m is the center position of the face tracking area in the current violent scene monitoring video frame, O ^m+1 is the center position of the face tracking area in the next surveillance video frame, I'(i,j) is the Cb brightness of other pixels in the violent scene surveillance video frame, δ is the Dirac function, and c(i,j) is the pixel The number of histogram segments where p(i, j, t) is located, N1 is the number of pixels in the vertical direction of the image, and N2 is the number of pixels in the horizontal direction of the image; Step 6, use the edge detection operator to obtain the critical pixels of the face tracking area centered on O ^m+1 , and carry out curve fitting to the critical pixels to form the face tracking area of the m+1th violent scene video frame until The judgment conditions for violent scenes are no longer met. 2. The face detection and positioning and real-time tracking method in the violent scene according to claim 1, wherein the optical flow (u _{i, j, t} , v _{i, j, t} ) is calculated in a differential manner, and the calculation The formula is as follows: <mrow><mfenced open = "[" close = "]"><mtable><mtr><mtd><msub><mi>u</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi><mo>,</mo><mi>t</mi></mrow></msub></mtd></mtr><mtr><mtd><msub><mi>v</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi><mo>,</mo><mi>t</mi></mrow></msub></mtd></mtr></mtable></mfenced><mo>=</mo><mfenced open = "[" close = "]"><mtable><mtr><mtd><msup><mrow><mo>(</mo><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>x</mi></mrow></mfrac><mo>)</mo></mrow><mn>2</mn></msup></mtd><mtd><mrow><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>x</mi></mrow></mfrac><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>y</mi></mrow></mfrac></mrow></mtd></mtr><mtr><mtd><mrow><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>x</mi></mrow></mfrac><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>y</mi></mrow></mfrac></mrow></mtd><mtd><msup><mrow><mo>(</mo><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>y</mi></mrow></mfrac><mo>)</mo></mrow><mn>2</mn></msup></mtd></mtr></mtable></mfenced><mfenced open = "[" close = "]"><mtable><mtr><mtd><mrow><mo>-</mo><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>x</mi></mrow></mfrac><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>t</mi></mrow></mrow></mrow>mfrac></mrow></mtd></mtr><mtr><mtd><mrow><mo>-</mo><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>y</mi></mrow></mfrac><mfrac><mrow><mo>&amp;part;</mo><mi>I</mi></mrow><mrow><mo>&amp;part;</mo><mi>t</mi></mrow></mfrac></mrow></mtd></mtr></mtable></mfenced><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>5</mn><mo>)</mo></mrow></mrow> where I is the image brightness of pixel p(i,j,t), is the horizontal brightness gradient, is the longitudinal brightness gradient, is the brightness gradient on the time axis. 3. face detection positioning and real-time tracking method in the violent scene according to claim 1, it is characterized in that, the determination of the central position O of the face tracking area in the initial violent scene monitoring video frame adopts the following steps ^: First, calculate the difference S(I(i,j)) between the Cb brightness and skin color of each pixel in the initial violent scene monitoring video frame: <mrow><mi>S</mi><mrow><mo>(</mo><mi>I</mi><mo>(</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mo>)</mo><mo>)</mo></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><msqrt><mrow><mn>2</mn><mi>&amp;pi;</mi></mrow></msqrt><mi>&amp;sigma;</mi></mrow></mfrac><msup><mi>e</mi><mrow><mo>-</mo><mfrac><msup><mrow><mo>(</mo><mi>I</mi><mo>(</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mo>)</mo><mo>-</mo><mi>&amp;mu;</mi><mo>)</mo></mrow><mn>2</mn></msup><mrow><mn>2</mn><msup><mi>&amp;sigma;</mi><mn>2</mn></msup></mrow></mfrac></mrow></msup><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>6</mn><mo>)</mo></mrow></mrow> Among them, μ is the mean value of skin color, and σ is the variance of skin color; Secondly, judge the face area. When S(I(i,j))>Th_skin, the representative pixel p(i,j,t) belongs to the face area, where Th_skin is the difference threshold of skin color; Finally, extract the face area, extract critical pixels according to the relationship between S(I(i,j)) and Th_skin and perform curve fitting to form the face tracking area in the initial violent scene monitoring video frame and obtain the initial violent scene monitoring video frame The center position O° of the inner face tracking area. 4. face detection location and real-time tracking method in the violent scene according to claim 1, it is characterized in that, adopt following formula to convert monitoring video frame to YCbCr color space from RGB space: Y＝0.257*R+0.564*G+0.098*B+16 (7) Cb＝-0.148*R-0.291*G+0.439*B+128 (8) Cr＝0.439*R-0.368*G-0.071*B+128 (9) Among them, Y represents the brightness, Cb reflects the difference between the blue component and the brightness of the RGB input, and Cr reflects the difference between the red component and the brightness of the RGB input. 5. The face detection, positioning and real-time tracking method in a violent scene according to claim 1 or 3, characterized in that, the critical pixels are curve-fitted by means of elliptic curve fitting to form a face tracking area. 6. face detection location and real-time tracking method in the violent scene according to claim 1, it is characterized in that, in step 6, described edge detection operator is local difference operator, Sobel operator or Canny operator any kind.