CN102194235A - Motion detection system and method based on gradient direction angle - Google Patents

Abstract

The invention relates to a motion detection system and a motion detection method based on a gradient direction angle. The system comprises a gradient direction angle quantifying module and a motion sensing module, wherein the gradient direction angle quantifying module is used for acquiring the gradient direction angle of each pixel in a received video frame image and quantifying the gradient direction angle to obtain the quantified value of the gradient direction angle of each pixel in the video frame image; and the motion sensing module is used for receiving the quantified value of the gradient direction angle of each pixel in the video frame image, comparing the quantified value with a quantified value of the gradient direction angle of the corresponding pixel in a last frame image, performing analysis and statistic processing on the comparison result, and thus detecting whether a motional target exists in the video frame image. By the method, the adverse effect of the flicker interference on the motion detection is effectively resisted and the warning error rate of the system is greatly reduced; and the method can be extensively used in the flicker-interference-resistant motion detection system.

Description

Movement detection systems and method based on the gradient direction angle

Technical field

The present invention relates to image processing techniques, relate in particular to movement detection systems and method.

Background technology

Along with popularizing of video monitoring system, the mobile detection technology has obtained increasing application.Video mobile detection camera will be reported to the police dexterously and monitoring combines, it is except being undertaken the picture control by televisor or monitor, can also pass through front end embedded intelligence analysis module, the vision signal that monitors is carried out intellectual analysis, in case awaring image content changes, it will send alarm sound or prompt tone, can also export relevant device work such as interlock signal triggering video recorder simultaneously.

Yet around under the brighter situation of ambient brightness, the camera time shutter can shorten, will introduce flicker interference such as (horizontal scroll bar schlieren pictures) in the sequence of video images that obtain this moment, the mobile detection effect is impacted, as easy as rolling off a log causing reported by mistake or fails to report.

The image gradient deflection is a kind ofly to rotate the characteristic quantity that keeps relative stability under the situation at complex illumination environment and target, therefore the gradient direction angle is incorporated into interference problems such as will effectively solving anti-flicker image in the moving object detection system.

Summary of the invention

The characteristics that the present invention has utilized the image gradient deflection can keep relative stability under the complex illumination environment have solved the moving sensing system wrong report that is brought by flicker image interference etc. and have failed to report problem.

In first aspect, the invention provides a kind of movement detection systems of the anti-horizontal scroll bar schlieren picture based on the gradient direction angle, this system comprises gradient direction angle quantization modules and mobile detection module.

This gradient direction angle quantization modules is used for obtaining the gradient direction angle of its each pixel of video frame images that receives, and quantification treatment is done at each gradient direction angle, thereby obtains the gradient direction angle quantized value of each pixel in this video frame images.This mobile detection module receives the gradient direction angle quantized value of each pixel in this video frame images, and with this quantized value and respective pixel the gradient direction angle quantized value in the former frame image compares, again this comparative result is done and analyzed and statistics, thereby obtain moving target.

In second aspect, the invention provides a kind of method for testing motion of the anti-horizontal scroll bar schlieren picture based on the gradient direction angle, this method is at first obtained the gradient direction angle of each pixel in its video frame images that receives, and each gradient direction angle done quantification treatment, thereby obtain the gradient direction angle quantized value of each pixel in this video frame images.Receive the gradient direction angle quantized value of each pixel in this video frame images then, and the gradient direction angle quantized value in the former frame image compares with this quantized value and respective pixel, this comparative result is done analyzed and statistics again, thereby obtain moving target.

In one embodiment of the invention, gradient direction angle quantization modules is obtained each pixel gradient angle in the video image by the Suo Beier edge detection operator.

In another embodiment of the present invention, the mobile detection module deducts the gradient direction angle quantized value of respective pixel in the former frame image with each pixel gradient deflection quantized value in the current frame image, and relatively with this difference and difference threshold Th2, add up the pixel quantity NUMC of this difference again greater than threshold value Th2, and this pixel quantity NUMC greater than amount threshold Th3 situation under, judge in this video image to have moving target.

The characteristics that the present invention has utilized the image gradient deflection can keep relative stability in the complex illumination environment, change by gray-value variation in the video image being converted to gradient direction angle quantized value, thereby the problems such as flicker image interference that exist in the complex illumination environment have been solved effectively, therefore can obtain the moving target in the video image more exactly, to finish the motion target detection function, reduced the movement detection systems false alarm rate greatly.

Description of drawings

Below with reference to accompanying drawings specific embodiments of the present invention is described in detail, in the accompanying drawings:

Fig. 1 is the moving object detection system block diagram based on the anti-flicker image at gradient direction angle of one embodiment of the invention;

Fig. 2 is a 3*3 area pixel distribution schematic diagram;

Fig. 3 is that the gradient direction angle of Δ θ=π/8 quantizes synoptic diagram.

Embodiment

Fig. 1 is the moving object detection system block diagram based on the anti-flicker image at gradient direction angle of one embodiment of the invention, and this system comprises video acquisition module 110, gradient direction angle quantization modules 120, mobile detection module 130.

Video acquisition module 110 is used to gather its video image on every side, and this video image is done digitized processing.

Gradient direction angle quantization modules 120 is used to receive described digitized video image, and obtain the gradient direction angle of each pixel in this video image by SOBEL (Suo Beier) edge detection operator, again quantification treatment is carried out at this gradient direction angle, thereby obtain the gradient direction angle quantized value of each pixel in this video image.

Particularly, Fig. 2 is a 3*3 area pixel distribution schematic diagram, and wherein, each grid is represented a pixel, A1, A2 ... A9 represents the relevant position pixel respectively, and the SOBEL operator detects template S1, S2 and be,

$S_1=\left[\begin{array}{ccc}-1& 0& 1\\ -2& 0& 2\\ -1& 0& 1\end{array}\right];$ $S_2=\left[\begin{array}{ccc}-1& -2& -1\\ 0& 0& 0\\ 1& 2& 1\end{array}\right]---\left(1\right)$

Be example with pixel A 5 below, elaborate pixel gradient deflection acquisition methods and pixel gradient deflection quantization method.

Suppose a1, a2 ... a9 is respectively pixel A 1, A2 ... the gray-scale value of A9, then pixel A 5 horizontal edge value f _h(a5) be,

$f_h\left(a5\right)=\left[\begin{array}{ccc}a1& a2& a3\\ a4& a5& a6\\ a7& a8& a9\end{array}\right]\·{S_1}^{\′}---\left(2\right)$

Pixel a5 vertical edge value f _v(a5) be,

$f_v\left(a5\right)=\left[\begin{array}{ccc}a1& a2& a3\\ a4& a5& a6\\ a7& a8& a9\end{array}\right]\·{S_2}^{\′}---\left(3\right)$

Wherein, S ₁' be array S ₁Transposition.

S1 value substitution formula (2) is obtained pixel A 5 horizontal edge value f _h(a5) be,

f _h(a5)＝(-1)×a ₁+0×a ₂+1×a ₃+(-2)×a ₄+0×a ₅+2×a ₆+(-1)×a ₇+0×a ₈+1×a ₉。

S2 value substitution formula (2) is obtained pixel A 5 vertical edge value f _v(a5) be,

f _v(a5)＝(-1)×a ₁+(-2)×a ₂+(-1)×a ₃+0×a ₄+0×a ₅+0×a ₆+1×a ₇+2×a ₈+1×a ₉。

Among Fig. 2, the direction of arrow is pixel A 5 supposition gradient directions, its angle value θ _A5Computing method are θ _A5=tan ^-1[f _v(a5)/f _h(a5)].

Because arctan function tan ^-1Result of calculation is in [pi/2, pi/2] scope, and the actual gradient deflection is in [0,2 π] scope, therefore needs comprehensive utilization θ value and gray scale derived indice to determine the gradient direction angle of pixel A 5.Below to angle value θ _A5Carry out quantification treatment,

$Q_{a5}=\left\{\begin{array}{cc}\mathrm{INT}\left(\frac{{\mathrm{\θ}}_{a5}}{\mathrm{\Δ\θ}}\right)& \left|f_h\left(a5\right)\right|+\left|f_v\left(a5\right)\right|>\mathrm{Th}1\\ N& \mathrm{otherwise}\end{array}\right.---\left(4\right)$

Wherein, Q _A5Quantized value for pixel A 5; Th1 is for suppressing the threshold value of low contrast point and noise, and Th1 value choosing method is to select high threshold for use when the template texture is very obvious, otherwise then selects low threshold value for use; Δ θ is a quantization step, and this quantization step Δ θ has determined quantified precision, and this quantization step Δ θ value is smaller or equal to π/8; INT asks whole computing; N is a constant, and N=2 π/Δ θ.

In one embodiment of the invention, suppressing low contrast point and noise threshold value Th1 value is 30, and quantization step Δ θ value is π/8.

Fig. 3 is that the gradient direction angle of Δ θ=π/8 quantizes synoptic diagram.| f _h(a5) |+| f _v(a5) | under＞Th1=30 the situation,

Particularly,

Work as θ _A5=0 o'clock, Q _A5=0;

......

Work as θ _A5During=pi/2,

......

Work as θ _A5During=π,

......

Therefore, exist | f _h(a5) |+| f _v(a5) | under＞Th1=30 the situation, pixel A 5 gradient direction angle quantized values are between [0,15].

| f _h(a5) |+| f _v(a5) | under≤Th1=30 the situation,

This moment, pixel A 5 gradient direction angle quantized values were 16.

Since the gradient direction angle be a kind of under the complex illumination environment and target rotate the characteristic quantity that still keeps relative stability under the situation, therefore the present invention is by changing the conversion that converts gradient direction angle quantized value to each grey scale pixel value in the video image, this quantized value situation of change of statistics and analysis is again disturbed the moving object detection adverse effect thereby resisted the flicker that is caused by illumination etc. effectively.Set forth how each pixel gradient deflection quantized value situation of change of statistics and analysis of mobile detection module 130 below, and how to obtain moving target based on this statistic analysis result.

Mobile detection module 130 is used for the gradient direction angle quantized value of each pixel of receiver, video image, and compare with respective pixel gradient direction angle quantized value in this quantized value and the former frame image, again this comparative result is done and analyzed and statistics, thereby obtain moving target.

Particularly, transportable frame is surveyed the gradient direction angle quantized value of each pixel in module 130 receiver, video images and this video image, judge two field picture headed by current this video image that receives whether then, if first two field picture, then with the gradient direction angle quantized value of each pixel in this video image reference gradient deflection (this reference gradient deflection real-time update) as this pixel.

If surveying module 130, transportable frame judges that current this video image that receives is not first two field picture, then the gradient direction angle quantized value of respective pixel in the gradient direction angle quantized value of each pixel in the current frame video image and the former frame video image is done poor (promptly with reference gradient deflection quantized value do poor), to obtain the variation of each pixel gradient deflection quantized value in the video image, difference threshold Th2 with this difference and default compares again, this difference greater than difference threshold Th2 situation under, judge that this difference respective pixel belongs to moving target collection of pixels S _Obj

Then, the module 130 of surveying transportable frame continues difference described in the whole frame video image of statistics greater than the pixel quantity NUMC of threshold value Th2, compare with this quantity NUMC and default amount threshold Th3 again, if this quantity NUMC is greater than amount threshold Th3, illustrate that then moving target is enough big, therefore there is moving target in this video image, and by this moving target collection of pixels S _ObjThe target of forming is moving target, otherwise then is noise.

In sum, because the gradient direction angle is stablized under the complex illumination environment, is difficult for changing and can disturb by preferably anti-flicker image, therefore the present invention is converted to the variation of gradient direction angle quantized value with the variation of gray-scale value, thereby has solved the problems such as flicker image interference that exist in the complex illumination environment.That is to say, even in the complex illumination environment, (have interference such as flicker), can accurately obtain moving target in the video image by system of the present invention or method, to finish motion detection function, and obtain moving target in the video image by detecting gray-value variation in the classic method, can't resist system's wrong report that flicker image interference etc. brings effectively and fail to report problem.

Obviously, under the prerequisite that does not depart from true spirit of the present invention and scope, the present invention described here can have many variations.Therefore, the change that all it will be apparent to those skilled in the art that all should be included within the scope that these claims contain.The present invention's scope required for protection is only limited by described claims.

Claims (12)

1. One kind based on the gradient direction angle movement detection systems, it is characterized in that, comprise gradient direction angle quantization modules (120) and mobile detection module (130);

   This gradient direction angle quantization modules (120) is used for obtaining the gradient direction angle of its each pixel of video frame images that receives, and quantification treatment is done at each gradient direction angle, thereby obtains the gradient direction angle quantized value of each pixel in this video frame images;

   This mobile detection module (130) receives the gradient direction angle quantized value of each pixel in the described video frame images, and with this quantized value and respective pixel the gradient direction angle quantized value in the former frame image compares, again this comparative result is done and analyzed and statistics, and then detect whether there is moving target in this video image.
2. 2. a kind of movement detection systems based on the gradient direction angle as claimed in claim 1 is characterized in that, comprises video acquisition module (110), and this video acquisition module (110) is used to gather its video image on every side, and this video image is done digitized processing.
3. 3. a kind of movement detection systems based on the gradient direction angle as claimed in claim 1 is characterized in that, described gradient direction angle quantization modules (120) is obtained each pixel gradient deflection in the described video image by the Suo Beier edge detection operator.
4. 4. a kind of movement detection systems based on the gradient direction angle as claimed in claim 3 is characterized in that, gray-scale value is the pixel gradient deflection θ of a5 _A5Be θ _A5=tan ^-1[f _v(a5)/f _h(a5)];

   Wherein, f _h(a5) be the horizontal edge value of this pixel, f _v(a5) be the vertical edge value of this pixel.
5. 5. a kind of movement detection systems based on the gradient direction angle as claimed in claim 4 is characterized in that, described pixel level rim value f _h(a5) be,

   $f_h\left(a5\right)=\left[\begin{array}{ccc}a1& a2& a3\\ a4& a5& a6\\ a7& a8& a9\end{array}\right]\·{S_1}^{\′}$

   And described pixel vertical edge value f _v(a5) be,

   $f_v\left(a5\right)=\left[\begin{array}{ccc}a1& a2& a3\\ a4& a5& a6\\ a7& a8& a9\end{array}\right]\·{S_2}^{\′}$

   Wherein, a1, a2, a3, a4, a6, a7, a8, a9 are that gray-scale value is the adjacent position pixel corresponding grey scale value of a5 pixel, S ₁' be vectorial S ₁Transposition, S ₂' be vectorial S ₂Transposition, and

   $S_1=\left[\begin{array}{ccc}-1& 0& 1\\ -2& 0& 2\\ -1& 0& 1\end{array}\right]$ $S_2=\left[\begin{array}{ccc}-1& -2& -1\\ 0& 0& 0\\ 1& 2& 1\end{array}\right].$
6. 6. a kind of movement detection systems based on the gradient direction angle as claimed in claim 4 is characterized in that, described pixel gradient deflection quantized value Q _A5For,

   $Q_{a5}=\left\{\begin{array}{cc}\mathrm{INT}\left(\frac{{\mathrm{\θ}}_{a5}}{\mathrm{\Δ\θ}}\right)& \left|f_h\left(a5\right)\right|+\left|f_v\left(a5\right)\right|>\mathrm{Th}1\\ N& \mathrm{otherwise}\end{array}\right.$

   Wherein, Th1 is for suppressing the threshold value of low contrast point and noise, and Δ θ is a quantization step, and INT is for asking whole computing, and N is constant and N=2 π/Δ θ.
7. 7. a kind of movement detection systems based on the gradient direction angle as claimed in claim 6 is characterized in that, described threshold value Th1 choosing method is for to select high threshold for use when the template texture is obvious, otherwise then selects low threshold value for use.
8. 8. a kind of movement detection systems based on the gradient direction angle as claimed in claim 6 is characterized in that, described quantization step Δ θ value is smaller or equal to π/8.
9. 9. a kind of movement detection systems as claimed in claim 1 based on the gradient direction angle, it is characterized in that, described mobile detection module (130) deducts the gradient direction angle quantized value of respective pixel in the former frame image with each pixel gradient deflection quantized value in the current frame image, and relatively with this difference and difference threshold Th2, add up the pixel quantity NUMC of described difference again greater than threshold value Th2, and this pixel quantity NUMC greater than amount threshold Th3 situation under, judge in this video image to have moving target.
10. 10. the method for testing motion based on the gradient direction angle is characterized in that, comprise,

    Step a, the gradient direction angle that obtains each pixel in its video frame images that receives, and each gradient direction angle done quantification treatment, thus obtain the gradient direction angle quantized value of each pixel in this video frame images;

    Step b, receive the gradient direction angle quantized value of each pixel in the described video frame images, and with this quantized value and respective pixel the gradient direction angle quantized value in the former frame image compares, this comparative result is done analyzed and statistics again, and then detect whether there is moving target in the video image.
11. 11. a kind of method for testing motion based on the gradient direction angle as claimed in claim 10 is characterized in that step a comprises the step of obtaining each pixel gradient deflection in the described video frame images by the Suo Beier edge detection operator.
12. 12. a kind of method for testing motion as claimed in claim 10 based on the gradient direction angle, it is characterized in that, step b comprise with each pixel gradient deflection quantized value in the current frame image deduct respective pixel in the former frame image gradient direction angle quantized value, and relatively with this difference and difference threshold Th2, add up the pixel quantity NUMC of described difference again greater than threshold value Th2, and this pixel quantity NUMC greater than amount threshold Th3 situation under, judge in this video image to have moving target.

Images