CN105938624A - Fouling monitoring device - Google Patents

Abstract

The invention provides a fouling monitoring device comprising (1) an image acquisition module which is used for acquiring video monitoring images; (2) an image preliminary processing module which is used for performing moving target preliminary detection processing on original video image sequence information and outputting effective video image sequence information including a moving target; (3) a filtering module which is used for receiving the effective video image sequence information and performing filtering processing on the background shape of the effective video image sequence information; (4) a background modeling module which is used for establishing a background model after filtering, wherein the background model is arranged to be composed of N shape context histograms which have weight and represent edge points; (5) a background subtraction module which is used for performing attributive classification on the edge points of a current frame of image and subtracting the edge points belonging to the background; and (6) a feature matching module. False detection of the moving target caused by background noise and camera jittering can be reduced to the largest extent and the real-time performance is great.

Description

A kind of foul supervising device

Technical field

The present invention relates to athletic competition field, be specifically related to a kind of foul supervising device.

Background technology

The foul supervising device detecting moving target in correlation technique, there is problems in that (1) supervises at video Background flase drop is moving target due to the existence of background noise by control image series；(2) exist during video monitoring Slight jitter, can be moving target by background flase drop；(3) during the detection to moving target, exist computationally intensive, in real time Property difference problem and moving target can not be detected in time.

Summary of the invention

For the problems referred to above, the present invention provides a kind of foul supervising device, and this foul supervising device can farthest subtract Few flase drop to moving target caused due to background noise and DE Camera Shake, and real-time is good.

The purpose of the present invention realizes by the following technical solutions:

Provide a kind of foul supervising device, including:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein Original video image sequence information；

(2) image just processing module, is connected with image capture module, for carrying out described original video image sequence information The Preliminary detection of moving target processes and exports the effective video image sequence information including moving target；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information right The background shape of effective video image sequence information is filtered processing, at the beginning of carrying out described background shape including using wiener ripple The first-level filtering of secondary filtering is except submodule and uses again be filtered first filtered background shape two grades of gaussian filtering Filter submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, described background model Set by N number of Weight represent marginal point Shape context rectangular histogram form, background model up contour point in shape under Literary composition histogram table is shown as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,\mathrm{...}N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, the value model of N Enclose for [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent and sit for pole with background edge point x The mark center of circle, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attribute Classify and cut down the marginal point belonging to background, judging son including matching degree calculating sub module, constraints calculating sub module, attribute Module and abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge Submodule, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point with described Matching degree between the Shape context rectangular histogram of corresponding marginal point in background model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent background mould The Shape context rectangular histogram of corresponding marginal point x in type, n=1 ... N；Represent the neighborhood of marginal point x,It is used for weighing the histogrammic difference of Shape context of two edges point, The least, show that the Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating Shape context rectangular histogram and the institute of current frame image up contour point Stating the constraints between the histogrammic difference of Shape context corresponding in background model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the genus of the marginal point on current frame image Property, the marginal point on described current frame image is that the decision condition of the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold set according to background model Value；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to for rejecting The area image of moving target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and number Characteristic matching is carried out, it is determined that misconduct according to solid plate pre-in storehouse.

Preferably, described image just processing module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target Value, described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video The judgement factor of image, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating having of m frame video image in original video image sequence information Validity, the computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{\left|d-d_p\right|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents that original video image sequence is believed Video image totalframes included in breath, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than bright The totalframes of the video image of degree rate of change reference threshold, d is the target size of m frame video image, d_pJoin for target size Examine threshold value；

C, output sub-module, belong to the image information of effective video image sequence, when described current frame image for output Effective degree more than described effective degree reference threshold time, described output sub-module export described current frame image.

The invention have the benefit that

1, image just processing module is set, for described original video image sequence information carries out the preliminary inspection of moving target Survey processes and exports the effective video image sequence information including moving target, it is possible to be greatly saved memory space, improves inspection The speed surveyed；

2, filtration module is set, is filtered local shape processing, it is possible to effectively filter out environment noise, it is to avoid will make an uproar Sound flase drop is moving target；

3, background modeling module is set, uses weighting Shape context rectangular histogram to enter by the filtered background of filtration module Row modeling, farthest decreases the flase drop to moving target caused due to background noise and DE Camera Shake；

4, in modeling process, only the Shape context rectangular histogram of marginal point is calculated, be greatly saved storage sky Between, improve arithmetic speed, the real-time of system is strengthened；

5, background subtraction module is set, introduces matching degree and background is cut down by matching constraint condition, it is possible to be quickly accurate True registrates moving target, completes detection.

Accompanying drawing explanation

The invention will be further described to utilize accompanying drawing, but the embodiment in accompanying drawing does not constitute any limit to the present invention System, for those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to obtain according to the following drawings Other accompanying drawing.

Fig. 1 is the connection diagram of each module of the present invention.

Detailed description of the invention

The invention will be further described with the following Examples.

Embodiment 1

Seeing Fig. 1, the foul supervising device of the present embodiment includes:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein Original video image sequence information；

(2) image just processing module, is connected with image capture module, for carrying out described original video image sequence information The Preliminary detection of moving target processes and exports to be located at the beginning of the effective video image sequence information including moving target, described image Reason module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target Value, described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video The judgement factor of image, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating having of m frame video image in original video image sequence information Validity, the computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{\left|d-d_p\right|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents that original video image sequence is believed Video image totalframes included in breath, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than bright The totalframes of the video image of degree rate of change reference threshold, d is the target size of m frame video image, d_pJoin for target size Examine threshold value；

C, output sub-module, belong to the image information of effective video image sequence, when described current frame image for output Effective degree more than described effective degree reference threshold time, described output sub-module export described current frame image；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information right The background shape of effective video image sequence information is filtered processing, at the beginning of carrying out described background shape including using wiener ripple The first-level filtering of secondary filtering is except submodule and uses again be filtered first filtered background shape two grades of gaussian filtering Filter submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, described background model Set by N number of Weight represent marginal point Shape context rectangular histogram form, background model up contour point in shape under Literary composition histogram table is shown as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,\mathrm{...}N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, the value model of N Enclose for [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent and sit for pole with background edge point x The mark center of circle, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attribute Classify and cut down the marginal point belonging to background, judging son including matching degree calculating sub module, constraints calculating sub module, attribute Module and abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge Submodule, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point with described Matching degree between the Shape context rectangular histogram of corresponding marginal point in background model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent background mould The Shape context rectangular histogram of corresponding marginal point x in type, n=1 ... N；Represent the neighborhood of marginal point x,It is used for weighing the histogrammic difference of Shape context of two edges point, The least, show that the Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating Shape context rectangular histogram and the institute of current frame image up contour point Stating the constraints between the histogrammic difference of Shape context corresponding in background model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the genus of the marginal point on current frame image Property, the marginal point on described current frame image is that the decision condition of the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold set according to background model Value；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to for rejecting The area image of moving target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and number Characteristic matching is carried out, it is determined that misconduct according to solid plate pre-in storehouse.

The present embodiment arranges image just processing module, for described original video image sequence information is carried out moving target Preliminary detection processes and exports the effective video image sequence information including moving target, it is possible to be greatly saved memory space, Improve the speed of detection；Filtration module is set, is filtered local shape processing, it is possible to effectively filter out environment noise, keep away Exempting from noise flase drop is moving target；Background modeling module is set, uses weighting Shape context rectangular histogram to by filtration module Filtered background is modeled, farthest decrease due to background noise and DE Camera Shake cause to moving target Flase drop；In modeling process, only the Shape context rectangular histogram of marginal point is calculated, is greatly saved memory space, Improve arithmetic speed, the real-time of system is strengthened；Background subtraction module is set, introduces matching degree and matching constraint condition Background is cut down, it is possible to fast and accurately moving target is registrated, complete detection；Wherein, N value is the biggest to background It is the strongest that dynamic adapts to ability, but can take and store resource more, increases amount of calculation, and real-time also can be deteriorated, the present embodiment Value N=5, compared with the supervising device in relative skill, false drop rate reduces 1%, and arithmetic speed improves 5%.

Embodiment 2

Seeing Fig. 1, the foul supervising device of the present embodiment includes:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein Original video image sequence information；

(2) image just processing module, is connected with image capture module, for carrying out described original video image sequence information The Preliminary detection of moving target processes and exports to be located at the beginning of the effective video image sequence information including moving target, described image Reason module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target Value, described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video The judgement factor of image, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating having of m frame video image in original video image sequence information Validity, the computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{\left|d-d_p\right|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents that original video image sequence is believed Video image totalframes included in breath, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than bright The totalframes of the video image of degree rate of change reference threshold, d is the target size of m frame video image, d_pJoin for target size Examine threshold value；

C, output sub-module, belong to the image information of effective video image sequence, when described current frame image for output Effective degree more than described effective degree reference threshold time, described output sub-module export described current frame image；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information right The background shape of effective video image sequence information is filtered processing, at the beginning of carrying out described background shape including using wiener ripple The first-level filtering of secondary filtering is except submodule and uses again be filtered first filtered background shape two grades of gaussian filtering Filter submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, described background model Set by N number of Weight represent marginal point Shape context rectangular histogram form, background model up contour point in shape under Literary composition histogram table is shown as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,\mathrm{...}N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, the value model of N Enclose for [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent and sit for pole with background edge point x The mark center of circle, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attribute Classify and cut down the marginal point belonging to background, judging son including matching degree calculating sub module, constraints calculating sub module, attribute Module and abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge Submodule, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point with described Matching degree between the Shape context rectangular histogram of corresponding marginal point in background model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent background mould The Shape context rectangular histogram of corresponding marginal point x in type, n=1 ... N；Represent the neighborhood of marginal point x,It is used for weighing the histogrammic difference of Shape context of two edges point, The least, show that the Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating Shape context rectangular histogram and the institute of current frame image up contour point Stating the constraints between the histogrammic difference of Shape context corresponding in background model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the genus of the marginal point on current frame image Property, the marginal point on described current frame image is that the decision condition of the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold set according to background model Value；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to for rejecting The area image of moving target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and number Characteristic matching is carried out, it is determined that misconduct according to solid plate pre-in storehouse.

The present embodiment arranges image just processing module, for described original video image sequence information is carried out moving target Preliminary detection processes and exports the effective video image sequence information including moving target, it is possible to be greatly saved memory space, Improve the speed of detection；Filtration module is set, is filtered local shape processing, it is possible to effectively filter out environment noise, keep away Exempting from noise flase drop is moving target；Background modeling module is set, uses weighting Shape context rectangular histogram to by filtration module Filtered background is modeled, farthest decrease due to background noise and DE Camera Shake cause to moving target Flase drop；In modeling process, only the Shape context rectangular histogram of marginal point is calculated, is greatly saved memory space, Improve arithmetic speed, the real-time of system is strengthened；Background subtraction module is set, introduces matching degree and matching constraint condition Background is cut down, it is possible to fast and accurately moving target is registrated, complete detection；Wherein, N value is the biggest to background It is the strongest that dynamic adapts to ability, but can take and store resource more, increases amount of calculation, and real-time also can be deteriorated, the present embodiment Value N=6, compared with the supervising device in relative skill, false drop rate reduces 2%, and arithmetic speed improves 4.5%.

Embodiment 3

Seeing Fig. 1, the foul supervising device of the present embodiment includes:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein Original video image sequence information；

(2) image just processing module, is connected with image capture module, for carrying out described original video image sequence information The Preliminary detection of moving target processes and exports to be located at the beginning of the effective video image sequence information including moving target, described image Reason module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target Value, described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video The judgement factor of image, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating having of m frame video image in original video image sequence information Validity, the computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{\left|d-d_p\right|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents that original video image sequence is believed Video image totalframes included in breath, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than bright The totalframes of the video image of degree rate of change reference threshold, d is the target size of m frame video image, d_pJoin for target size Examine threshold value；

C, output sub-module, belong to the image information of effective video image sequence, when described current frame image for output Effective degree more than described effective degree reference threshold time, described output sub-module export described current frame image；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information right The background shape of effective video image sequence information is filtered processing, at the beginning of carrying out described background shape including using wiener ripple The first-level filtering of secondary filtering is except submodule and uses again be filtered first filtered background shape two grades of gaussian filtering Filter submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, described background model Set by N number of Weight represent marginal point Shape context rectangular histogram form, background model up contour point in shape under Literary composition histogram table is shown as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,\mathrm{...}N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, the value model of N Enclose for [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent and sit for pole with background edge point x The mark center of circle, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attribute Classify and cut down the marginal point belonging to background, judging son including matching degree calculating sub module, constraints calculating sub module, attribute Module and abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge Submodule, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point with described Matching degree between the Shape context rectangular histogram of corresponding marginal point in background model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent background mould The Shape context rectangular histogram of corresponding marginal point x in type, n=1 ... N；Represent the neighborhood of marginal point x,It is used for weighing the histogrammic difference of Shape context of two edges point, The least, show that the Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating Shape context rectangular histogram and the institute of current frame image up contour point Stating the constraints between the histogrammic difference of Shape context corresponding in background model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the genus of the marginal point on current frame image Property, the marginal point on described current frame image is that the decision condition of the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold set according to background model Value；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to for rejecting The area image of moving target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and number Characteristic matching is carried out, it is determined that misconduct according to solid plate pre-in storehouse.

The present embodiment arranges image just processing module, for described original video image sequence information is carried out moving target Preliminary detection processes and exports the effective video image sequence information including moving target, it is possible to be greatly saved memory space, Improve the speed of detection；Filtration module is set, is filtered local shape processing, it is possible to effectively filter out environment noise, keep away Exempting from noise flase drop is moving target；Background modeling module is set, uses weighting Shape context rectangular histogram to by filtration module Filtered background is modeled, farthest decrease due to background noise and DE Camera Shake cause to moving target Flase drop；In modeling process, only the Shape context rectangular histogram of marginal point is calculated, is greatly saved memory space, Improve arithmetic speed, the real-time of system is strengthened；Background subtraction module is set, introduces matching degree and matching constraint condition Background is cut down, it is possible to fast and accurately moving target is registrated, complete detection；Wherein, N value is the biggest to background It is the strongest that dynamic adapts to ability, but can take and store resource more, increases amount of calculation, and real-time also can be deteriorated, the present embodiment Value N=7, compared with the supervising device in relative skill, false drop rate reduces 3.5%, and arithmetic speed improves 4%.

Embodiment 4

Seeing Fig. 1, the foul supervising device of the present embodiment includes:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein Original video image sequence information；

(2) image just processing module, is connected with image capture module, for carrying out described original video image sequence information The Preliminary detection of moving target processes and exports to be located at the beginning of the effective video image sequence information including moving target, described image Reason module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target Value, described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video The judgement factor of image, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating having of m frame video image in original video image sequence information Validity, the computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{\left|d-d_p\right|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents that original video image sequence is believed Video image totalframes included in breath, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than bright The totalframes of the video image of degree rate of change reference threshold, d is the target size of m frame video image, d_pJoin for target size Examine threshold value；

C, output sub-module, belong to the image information of effective video image sequence, when described current frame image for output Effective degree more than described effective degree reference threshold time, described output sub-module export described current frame image；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information right The background shape of effective video image sequence information is filtered processing, at the beginning of carrying out described background shape including using wiener ripple The first-level filtering of secondary filtering is except submodule and uses again be filtered first filtered background shape two grades of gaussian filtering Filter submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, described background model Set by N number of Weight represent marginal point Shape context rectangular histogram form, background model up contour point in shape under Literary composition histogram table is shown as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,\mathrm{...}N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, the value model of N Enclose for [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent and sit for pole with background edge point x The mark center of circle, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attribute Classify and cut down the marginal point belonging to background, judging son including matching degree calculating sub module, constraints calculating sub module, attribute Module and abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge Submodule, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point with described Matching degree between the Shape context rectangular histogram of corresponding marginal point in background model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent background mould The Shape context rectangular histogram of corresponding marginal point x in type, n=1 ... N；Represent the neighborhood of marginal point x,It is used for weighing the histogrammic difference of Shape context of two edges point, The least, show that the Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating Shape context rectangular histogram and the institute of current frame image up contour point Stating the constraints between the histogrammic difference of Shape context corresponding in background model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the genus of the marginal point on current frame image Property, the marginal point on described current frame image is that the decision condition of the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold set according to background model Value；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to for rejecting The area image of moving target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and number Characteristic matching is carried out, it is determined that misconduct according to solid plate pre-in storehouse.

The present embodiment arranges image just processing module, for described original video image sequence information is carried out moving target Preliminary detection processes and exports the effective video image sequence information including moving target, it is possible to be greatly saved memory space, Improve the speed of detection；Filtration module is set, is filtered local shape processing, it is possible to effectively filter out environment noise, keep away Exempting from noise flase drop is moving target；Background modeling module is set, uses weighting Shape context rectangular histogram to by filtration module Filtered background is modeled, farthest decrease due to background noise and DE Camera Shake cause to moving target Flase drop；In modeling process, only the Shape context rectangular histogram of marginal point is calculated, is greatly saved memory space, Improve arithmetic speed, the real-time of system is strengthened；Background subtraction module is set, introduces matching degree and matching constraint condition Background is cut down, it is possible to fast and accurately moving target is registrated, complete detection；Wherein, N value is the biggest to background It is the strongest that dynamic adapts to ability, but can take and store resource more, increases amount of calculation, and real-time also can be deteriorated, the present embodiment Value N=8, compared with the supervising device in relative skill, false drop rate reduces 4%, and arithmetic speed improves 3.7%.

Embodiment 5

Seeing Fig. 1, the foul supervising device of the present embodiment includes:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein Original video image sequence information；

(2) image just processing module, is connected with image capture module, for carrying out described original video image sequence information The Preliminary detection of moving target processes and exports to be located at the beginning of the effective video image sequence information including moving target, described image Reason module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target Value, described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video The judgement factor of image, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating having of m frame video image in original video image sequence information Validity, the computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{\left|d-d_p\right|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents that original video image sequence is believed Video image totalframes included in breath, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than bright The totalframes of the video image of degree rate of change reference threshold, d is the target size of m frame video image, d_pJoin for target size Examine threshold value；

C, output sub-module, belong to the image information of effective video image sequence, when described current frame image for output Effective degree more than described effective degree reference threshold time, described output sub-module export described current frame image；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information right The background shape of effective video image sequence information is filtered processing, at the beginning of carrying out described background shape including using wiener ripple The first-level filtering of secondary filtering is except submodule and uses again be filtered first filtered background shape two grades of gaussian filtering Filter submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, described background model Set by N number of Weight represent marginal point Shape context rectangular histogram form, background model up contour point in shape under Literary composition histogram table is shown as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,\mathrm{...}N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, the value model of N Enclose for [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent and sit for pole with background edge point x The mark center of circle, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attribute Classify and cut down the marginal point belonging to background, judging son including matching degree calculating sub module, constraints calculating sub module, attribute Module and abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge Submodule, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point with described Matching degree between the Shape context rectangular histogram of corresponding marginal point in background model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent background mould The Shape context rectangular histogram of corresponding marginal point x in type, n=1 ... N；Represent the neighborhood of marginal point x,It is used for weighing the histogrammic difference of Shape context of two edges point, The least, show that the Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating Shape context rectangular histogram and the institute of current frame image up contour point Stating the constraints between the histogrammic difference of Shape context corresponding in background model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the genus of the marginal point on current frame image Property, the marginal point on described current frame image is that the decision condition of the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold set according to background model Value；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to for rejecting The area image of moving target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and number Characteristic matching is carried out, it is determined that misconduct according to solid plate pre-in storehouse.

The present embodiment arranges image just processing module, for described original video image sequence information is carried out moving target Preliminary detection processes and exports the effective video image sequence information including moving target, it is possible to be greatly saved memory space, Improve the speed of detection；Filtration module is set, is filtered local shape processing, it is possible to effectively filter out environment noise, keep away Exempting from noise flase drop is moving target；Background modeling module is set, uses weighting Shape context rectangular histogram to by filtration module Filtered background is modeled, farthest decrease due to background noise and DE Camera Shake cause to moving target Flase drop；In modeling process, only the Shape context rectangular histogram of marginal point is calculated, is greatly saved memory space, Improve arithmetic speed, the real-time of system is strengthened；Background subtraction module is set, introduces matching degree and matching constraint condition Background is cut down, it is possible to fast and accurately moving target is registrated, complete detection；Wherein, N value is the biggest to background It is the strongest that dynamic adapts to ability, but can take and store resource more, increases amount of calculation, and real-time also can be deteriorated, the present embodiment Value N=10, compared with the supervising device in relative skill, false drop rate reduces 4.2%, and arithmetic speed improves 3.5%.

Last it should be noted that, above example is only in order to illustrate technical scheme, rather than the present invention is protected Protecting the restriction of scope, although having made to explain to the present invention with reference to preferred embodiment, those of ordinary skill in the art should Work as understanding, technical scheme can be modified or equivalent, without deviating from the reality of technical solution of the present invention Matter and scope.

Claims (2)

1. a foul supervising device, is characterized in that, including:

(1) image capture module, is used for gathering video monitoring image, and it is connected to video monitoring equipment and gathers therein former regard Frequently image sequence information；

(2) image just processing module, is connected with image capture module, for moving described original video image sequence information The Preliminary detection of target processes and exports the effective video image sequence information including moving target；

(3) filtration module, is connected with image just processing module, is used for receiving described effective video image sequence information and to effectively The background shape of sequence of video images information is filtered processing, and filters described background shape for the first time including using wiener ripple The first-level filtering of ripple filters except submodule and use gaussian filtering to be again filtered first filtered background shape two grades Submodule；

(4) background modeling module, is connected with filtration module, is used for setting up filtered background model, and described background model sets Being made up of the Shape context rectangular histogram representing marginal point of N number of Weight, the Shape context of background model up contour point is straight Side's figure is expressed as:

$\{w_{n,x}{\stackrel{\→}{h}}_{n,x},n\∈1,...N,w_{n,x}=\frac{\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}}{\underset{n=1}{\overset{N}{\Σ}}\[\frac{(B_{\mathrm{\θ}}+n)\×(B_l+n)}{B_{\mathrm{\θ}}\×B_l}\]}\}$

Wherein, x is background edge point coordinates, and N represents the histogrammic number of comprised Shape context, and the span of N is [5,10], w_n,xRepresent the weight that the n-th Shape context rectangular histogram is corresponding, B_lRepresent with background edge point x for polar coordinate circle The heart, radius are the circle concentric circular number according to logarithm distance foundation of R, B_θRepresent and angle of circumference division is waited number；

(5) background subtraction module, is connected with background modeling module, for the marginal point on current frame image is carried out attributive classification And cut down the marginal point belonging to background, including matching degree calculating sub module, constraints calculating sub module, attribute decision sub-module With abatement submodule, described matching degree calculating sub module, constraints calculating sub module are all connected to described attribute and judge submodule Block, described attribute decision sub-module is connected to described abatement submodule, wherein:

A, matching degree calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point and described background Matching degree between the Shape context rectangular histogram of corresponding marginal point on model, the computing formula of described matching degree is:

In formula,The Shape context rectangular histogram of the marginal point x on expression current frame image,Represent in background model The Shape context rectangular histogram of corresponding marginal point x, n=1 ... N；Represent the neighborhood of marginal point x, It is used for weighing the histogrammic difference of Shape context of two edges point,The least, show The Shape context histogram difference of two edges point is the least；

B, constraints calculating sub module, for calculating the Shape context rectangular histogram of current frame image up contour point and the described back of the body Constraints between the histogrammic difference of Shape context corresponding on scape model, constraints formula is:

C, attribute decision sub-module, described attribute decision sub-module is for judging the attribute of the marginal point on current frame image, institute Stating the decision condition that the marginal point on current frame image is the marginal point belonging to background is:

And

Wherein, T_PFor the matching degree threshold value set according to background model, T_YFor the constraints threshold value set according to background model；

D, abatement submodule, be judged to belong to the marginal point of background by attribute decision sub-module and output belongs to motion for rejecting The area image of target；

(6) characteristic matching module, is connected with abatement submodule, for by the described area image belonging to moving target and data base In pre-solid plate carry out characteristic matching, it is determined that misconduct.

A kind of foul supervising device the most according to claim 1, is characterized in that, described image just processing module includes:

A, reference effective degree set submodule, for the effective degree reference threshold of the storage video image sample containing moving target, Described effective degree represents for judging in original video image sequence information that m frame video image is whether as described effective video image The judgement factor, described effective degree reference threshold includes rate of change of brightness reference threshold and target size reference threshold；

B, actually active degree calculating sub module, for calculating the effective degree of m frame video image in original video image sequence information, The computing formula of effective degree is:

$V=L_m\×\sqrt{\left(\frac{M-m_v}{M}\right)\×\frac{|d-d_p|}{d_p}}$

Wherein, V represents effective degree, L_mBeing the rate of change of brightness of m frame video image, M represents institute in original video image sequence information The video image totalframes comprised, m=1 ... M, m_vFor rate of change of brightness in original video image sequence information more than brightness flop The totalframes of the video image of rate reference threshold, d is the target size of m frame video image, d_pFor target size reference threshold；

C, output sub-module, belong to the image information of effective video image sequence, when having of described current frame image for output When validity is more than described effective degree reference threshold, described output sub-module exports described current frame image.