CN109086682A - A kind of intelligent video black smoke vehicle detection method based on multi-feature fusion - Google Patents

A kind of intelligent video black smoke vehicle detection method based on multi-feature fusion Download PDF

Info

Publication number
CN109086682A
CN109086682A CN201810754422.6A CN201810754422A CN109086682A CN 109086682 A CN109086682 A CN 109086682A CN 201810754422 A CN201810754422 A CN 201810754422A CN 109086682 A CN109086682 A CN 109086682A
Authority
CN
China
Prior art keywords
vehicle
black smoke
feature
image
distance
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810754422.6A
Other languages
Chinese (zh)
Other versions
CN109086682B (en
Inventor
路小波
陶焕杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to CN201810754422.6A priority Critical patent/CN109086682B/en
Publication of CN109086682A publication Critical patent/CN109086682A/en
Application granted granted Critical
Publication of CN109086682B publication Critical patent/CN109086682B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
    • G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention discloses a kind of intelligent video black smoke vehicle detection methods based on multi-feature fusion, include the following steps: that (1) Utilization prospects detection algorithm extracts moving target from traffic surveillance videos, and identify vehicle target；(2) vehicle tail position is detected using integral projection and filtering technique；(3) statistical nature, frequency domain character and some manual features, fusion for extracting tailstock portion rear area form a feature vector；(4) classified using BP network classifier to mentioned feature vector, identify black smoke frame, to further identify black smoke vehicle.The present invention can be improved robustness, more effectively detect black smoke vehicle.

Description

A kind of intelligent video black smoke vehicle detection method based on multi-feature fusion
Technical field
The present invention relates to pyrotechnics detection technique field, especially a kind of intelligent video black smoke car test based on multi-feature fusion Survey method.
Background technique
Accelerate the construction of motor vehicle blowdown monitor supervision platform, keypoint treatment heavy-duty diesel vehicle and high emission vehicle in region.It is heavy The usual performance of diesel vehicle and high emission vehicle is that Vehicular exhaust hole emits dense black smoke, we are normally referred to as black smoke vehicle. The black smoke tail gas of black smoke vehicle discharge not only pollutes air, also damage human health.Therefore how research is effectively detected black smoke vehicle It is significantly.
The method of current detection black smoke vehicle can be divided into three categories:
(1) conventional method.For example, reports, the inspection of regular road, night inspection manual video monitoring.Conventional method is often Extensive work personnel can be expended, and due to the sharp increase of vehicle guaranteeding organic quantity, traffic it is busy etc., such methods efficiency is very It is low；
(2) semi-intelligent method.Such as installation Vehicular exhaust analytical equipment, sensor detection etc..These methods are in certain journey The efficiency that black smoke car test survey is improved on degree reduces the pollution of black smoke vehicle, but the purchase and maintenance of equipment need a large amount of financial resources Support, and to each car all install tails assay device implementation it is difficult；
(3) intelligent video monitoring method.This method is automatic from magnanimity traffic surveillance videos using computer vision technique Detect black smoke vehicle.Such methods belong to remote monitor, do not block traffic, it can be achieved that whole day is on duty online, are adapted to two-way traffic It is and easy for installation with the various roads environment such as multilane, it is suitble to deploying to ensure effective monitoring and control of illegal activities on a large scale for urban road, it is easier to be formed for height The on-line monitoring network of black smoke vehicle is polluted, law enforcement efficiency is improved.But such methods are at present still in the starting stage of research.
The invention proposes a kind of intelligent video monitoring method, which fully considers the reality of black smoke vehicle test problems Border feature detects tailstock portion, more accurate lock candidate region using integral projection and people's filtering technique, and merges tailstock portion rear Statistical nature, frequency domain character and some manual features in region more effectively detect black smoke vehicle to further increase robustness.
Summary of the invention
Technical problem to be solved by the present invention lies in provide a kind of intelligent video black smoke car test based on multi-feature fusion Survey method, can be improved robustness, more effectively detect black smoke vehicle.
In order to solve the above technical problems, the present invention provides a kind of intelligent video black smoke car test survey side based on multi-feature fusion Method includes the following steps:
(1) Utilization prospects detection algorithm extracts moving target from traffic surveillance videos, and identifies vehicle target；
(2) vehicle tail position is detected using integral projection and filtering technique；
(3) statistical nature, frequency domain character and some manual features, fusion for extracting tailstock portion rear area form a spy Levy vector；
(4) classified using BP network classifier to mentioned feature vector, black smoke frame is identified, to further identify black Cigarette vehicle.
Preferably, the foreground detection algorithm in step (1) includes the following steps:
(11) following formula initial background I is usedback(t),
Wherein, I (t) represents t frame image, and N represents the number of image frames of initial background use；
(12) foreground target I is calculated using following formulafore(t),
βt=mean (| I (t)-Iback(t)|)
P=threshold (| I (t)-Iback(t)|,βt+ε)
Ifore(t)=dilate (erode (P))
Wherein, threshold (I, βt+ ε) it is one with βt+ ε is the Binarization methods of threshold value, and mean (I) is a calculating The algorithm of the average value of all pixels of image I, erode (I) and dilate (I) are morphologic corrosion and expansion behaviour respectively Make；
(13) background model is updated using following formula,
Wherein, threshold alpha is the adjustment factor of a control background precision；
(14) (12) are gone to step and calculate Ifore(t+1)。
Preferably, the identification vehicle target in step (1), which refers to while meeting following two criterion, is considered as vehicle Target:
Rule one: the area of moving target is greater than some threshold value；
Rule two: the length-width ratio of the boundary rectangle frame of moving target is within the scope of some.
Preferably, included the following steps: in step (2) using integral projection and filtering technique detection vehicle tail position
(21) vehicle target image I is calculatedobjHorizontal integral projection E1(x), i.e.,
Wherein, Iobj(x, y) is coordinate of the vehicle target image at point (x, y), and w is the width of vehicle target image, operation Norm () is normalization process；
(22) stochastic filtering is carried out to vehicle target image, calculates the horizontal integral projection of filtered image, i.e.,
Operating rangefilt () is stochastic filtering process；
(23) to horizontal integral projection curve E1(x) and E2(x) it is weighted fusion, i.e.,
F (x)=λ1E1(x)+λ2E2And λ (x),1+λ2=1
Wherein, λ1And λ2Respectively E1(x) and E2(x) weight coefficient；
(24) vehicle tail position coordinate x is calculated by one of following two moderear,
Wherein, Δ x is that one and tailstock portion coordinate calculate related parameter.
Preferably, the textural characteristics of the extraction tailstock portion rear area in step (3) include the following steps:
(31) the rear area I of vehicle tail position is determinedrear, which is set to start line with tailstock position, extends back 60 pixels, the wide width for being set as vehicle target；
(32) following formula zoning I is usedrearGray level co-occurrence matrixes P,
Wherein, P (i, j, d, θ) indicates that direction is picture of the gray level co-occurrence matrixes P at position (i, j) that θ pixel distance is d Element value, w and h are respectively the width and height of vehicle target image, and round () is a function, indicates to round up；
(33) Normalized Grey Level co-occurrence matrix P, obtains
(34) a series of statistical natures based on gray level co-occurrence matrixes are calculated, i.e.,
One Angular second moment (ASM) of feature, note ASM (d, θ) indicate that angle is the feature one that θ distance is d ASM,
Wherein, L × L indicates Normalized Grey Level co-occurrence matrixSize；
Two Entropy of feature (ENT), note ENT (d, θ) indicate that angle is two ENT of feature that θ distance is d,
Three Contrast of feature (CON), note CON (d, θ) indicate that angle is three CON of feature that θ distance is d,
Four Correlation of feature (COR), note COR (d, θ) indicate that angle is four COR of feature that θ distance is d,
Five Inverse difference moment (IDM) of feature, note IDM (d, θ) indicate that angle is the spy that θ distance is d Five IDM are levied,
(35) four direction θ=0 ° is used, 45 °, 90 °, 135 ° and two kinds pixel distance d=2,3 obtain different normalizings Change gray level co-occurrence matrixesTo each gray level co-occurrence matrixes, five statistical natures of ASM, ENT, CON, COR and IDM are calculated, it will Different directions, different distance five statistical natures be together in series the statistical nature based on gray level co-occurrence matrixes can be obtained.
Preferably, the frequency domain character of the extraction tailstock portion rear area in step (3) includes the following steps:
(36) by tailstock portion rear area IrearIt is bisected into 1x2 fritter, two layers of wavelet decomposition, record are carried out to each fritter The wavelet coefficient image of the horizontal direction of i-th (i=1,2.) layer, vertical direction and diagonal direction is respectively Hi, ViAnd Di；
(37) in the following way calculate i-th (i=1,2.) layer, the wavelet energy of kth (i=1,2.) fritter,
Wherein, wiAnd hiRespectively indicate HiWidth and height；
(38) frequency domain character that step (37) obtain is together in series for identifying black smoke vehicle.
Preferably, some manual features of the extraction tailstock portion rear area in step (3) include:
(1) matching degree: vehicle tail region I is calculatedrearRegion corresponding with backgroundMatching degree Fmatch, i.e.,
Wherein, Irear(i, j) indicates image IrearPixel value at position (i, j),Indicate image? Pixel value at position (i, j)；
(2) mean value: vehicle tail region I is calculatedrearPixel mean value, i.e.,
Wherein, N0For region IrearSum of all pixels；
(3) variance: vehicle tail region I is calculatedrearPixel mean value, i.e.,
(4) ratio: ratio feature F is calculated in the following wayratiO,
Wherein, h indicates distance of the tailstock portion to vehicle target boundary rectangle frame bottom, the expression tailstock portion H to present frame figure As the distance at top.
Preferably, the identification black smoke vehicle in step (4) includes the following steps:
(41) classified using trained BP network classifier to all vehicle target pictures in current frame image, If there is at least one vehicle target picture is identified as black smoke vehicle picture, then present frame is identified as black smoke frame；
(42) in per continuous 100 frame, if there is K frame is identified as black smoke frame, and K meets following formula, then it is assumed that works as forward sight There are black smoke vehicle in frequency sequence,
K>α
Wherein, α is an adjustment factor for controlling recall rate and accurate rate.
The invention has the benefit that (1) improves law enforcement efficiency, traditional artificial monitoring black smoke vehicle inefficiency is made up not Foot；Intelligent video monitoring method proposed by the present invention is examined automatically from magnanimity traffic surveillance videos using computer vision technique Black smoke vehicle is surveyed, video related data uploads environmental protection administration automatically, while retaining the licence plate of black smoke vehicle, cross vehicle place, spend the vehicle time Equal evidences；This method belongs to remote monitor, does not block traffic, it can be achieved that whole day is on duty online, is adapted to two-way traffic and Duo Che The various roads environment such as road, and it is easy for installation, it is suitble to deploying to ensure effective monitoring and control of illegal activities on a large scale for urban road, it is easier to be formed black for high pollution The on-line monitoring network of cigarette vehicle improves law enforcement efficiency；(2) rate of false alarm is reduced；Technical solution proposed by the present invention is thrown using integral Shadow and filtering technique detect tailstock portion, to reduce the candidate region of black smoke identification, on the other hand, this technology has merged tailstock portion Statistical nature, frequency domain character and some manual features of rear area etc., further increase robustness, reduce rate of false alarm, keep away The erroneous detection as caused by leaf shaking, white clouds movement etc. is exempted from.
Detailed description of the invention
Fig. 1 is method flow schematic diagram of the invention.
Fig. 2 be the invention detects that vehicle target schematic diagram.
Fig. 3 (a) be the invention detects that a non-black smoke vehicle and its projection blend curve F (x) schematic diagram.
Fig. 3 (b) be the invention detects that a black smoke vehicle and its projection blend curve F (x) schematic diagram.
Fig. 4 is the schematic diagram of vehicle tail position of the present invention rear area.
Fig. 5 is the schematic diagram of the matching degree feature in manual features of the present invention.
Fig. 6 is the schematic diagram of the ratio characteristic in manual features of the present invention.
Specific embodiment
The present invention provides a kind of intelligent video black smoke vehicle detection method based on multi-feature fusion, flow chart such as Fig. 1 institute Show, specifically follow the steps below:
Step 1: Utilization prospects detection algorithm extracts moving target from traffic surveillance videos, and identifies vehicle target；
Step 2: detecting vehicle tail position using integral projection and filtering technique；
Step 3: the statistical nature, frequency domain character and some manual features, fusion for extracting tailstock portion rear area form one A feature vector；
Step 4: being classified using BP network classifier to mentioned feature vector, black smoke frame is identified, to further know Other black smoke vehicle.
Foreground detection algorithm in the step 1 uses following process:
Step 1.1: using following formula initial background Iback(t),
Wherein, I (t) represents t frame image, and N represents the number of image frames of initial background use；
Step 1.2: foreground target I is calculated using following formulafore(t),
βt=mean (| I (t)-Iback(t)|)
P=threshold (| I (t)-Iback(t)|,βt+ε)
Ifore(t)=dilate (erode (P))
Wherein, threshold (I, βt+ ε) it is one with βt+ ε is the Binarization methods of threshold value, and mean (I) is a calculating The algorithm of the average value of all pixels of image I .erode (I) and dilate (I) are morphologic corrosion and expansion behaviour respectively Make；
Step 1.3: background model is updated using following formula,
Wherein, threshold alpha is the adjustment factor of a control background precision；
Step 1.4: going to step 1.2 calculating Ifore(t+1)。
Identification vehicle target in the step 1, which refers to while meeting following two criterion, is considered as vehicle mesh Mark:
Rule one: the area of moving target is greater than some threshold value；
Rule two: the length-width ratio of the boundary rectangle frame of moving target is within the scope of some.
Fig. 2 shows the vehicle target testing result to a certain frame.
In the step 2 includes following process using integral projection and filtering technique detection vehicle tail position:
Step 2.1: calculating vehicle target image IobjHorizontal integral projection E1(x), i.e.,
Wherein, Iobj(x, y) is coordinate of the vehicle target image at point (x, y), and w is the width of vehicle target image, operation Norm () is normalization process；
Step 2.2: stochastic filtering being carried out to vehicle target image, calculates the horizontal integral projection of filtered image, i.e.,
Operating rangefilt () is stochastic filtering process；
Step 2.3: to horizontal integral projection curve E1(x) and E2(x) it is weighted fusion, i.e.,
F (x)=λ1E1(x)+λ2E2And λ (x),1+λ2=1
Wherein, λ1And λ2Respectively E1(x) and E2(x) weight coefficient；
Fig. 3 (a) shows a non-black smoke vehicle and its projection blend curve F (x), and Fig. 3 (b) shows a black smoke vehicle with its throwing Shadow blend curve F (x), it can be seen that the abscissa at the right groove of curve is exactly equal to the ordinate of vehicle tail.
Step 2.4: vehicle tail position coordinate x is calculated by one of following two moderear,
Or
Wherein, Δ x is that one and tailstock portion coordinate calculate related parameter.
The textural characteristics of extraction tailstock portion rear area in the step 3 include following process:
Step 3.1: determining the rear area I of vehicle tail positionrear, which is set to start line with tailstock position, backward Extend 60 pixels, the wide width for being set as vehicle target；
Step 3.2: using following formula zoning IrearGray level co-occurrence matrixes P,
Wherein, P (i, j, d, θ) indicates that direction is picture of the gray level co-occurrence matrixes P at position (i, j) that θ pixel distance is d Element value, w and h are respectively the width and height of vehicle target image, and round () is a function, indicates to round up；
Step 3.3: Normalized Grey Level co-occurrence matrix P is obtained
Step 3.4: calculating a series of statistical natures based on gray level co-occurrence matrixes, i.e.,
One Angular second moment (ASM) of feature, note ASM (d, θ) indicate that angle is the feature one that θ distance is d ASM,
Wherein, L × L indicates Normalized Grey Level co-occurrence matrixSize；
Two Entropy of feature (ENT), note ENT (d, θ) indicate that angle is two ENT of feature that θ distance is d,
Three Contrast of feature (CON), note CON (d, θ) indicate that angle is three CON of feature that θ distance is d,
Four Correlation of feature (COR), note COR (d, θ) indicate that angle is four COR of feature that θ distance is d,
Five Inverse difference moment (IDM) of feature, note IDM (d, θ) indicate that angle is the spy that θ distance is d Five IDM are levied,
Step 3.5: using four direction θ=0 °, 45 °, 90 °, 135 ° and two kinds pixel distance d=2,3 obtain different Normalized Grey Level co-occurrence matrixTo each gray level co-occurrence matrixes, it is special to calculate five statistics of ASM, ENT, CON, COR and IDM Sign, five statistical natures of different directions, different distance are together in series, and it is special that the statistics based on gray level co-occurrence matrixes can be obtained Sign.
The frequency domain character of extraction tailstock portion rear area in the step 3 includes following process:
Step 3.6: by tailstock portion rear area IrearIt is bisected into 1x2 fritter, two layers of wavelet decomposition is carried out to each fritter, The wavelet coefficient image for recording the horizontal direction of i-th (i=1,2.) layer, vertical direction and diagonal direction is respectively Hi, ViAnd Di；
Step 3.7: i-th (i=1,2.) layer of calculating in the following way, the wavelet energy of kth (i=1,2.) fritter,
Wherein, wiAnd hiRespectively indicate HiWidth and height；
Step 3.8: the frequency domain character that step 3.7 is obtained is together in series for identifying black smoke vehicle.
Some manual features of extraction tailstock portion rear area in the step 3 include:
(1) matching degree: vehicle tail region I is calculatedrearRegion corresponding with backgroundMatching degree Fmatch, i.e.,
Wherein, Irear(i, j) indicates image IrearPixel value at position (i, j),Indicate imageIn place Set the pixel value at (i, j)；
Fig. 5 shows the schematic diagram of the matching degree feature in manual features.
(2) mean value: vehicle tail region I is calculatedrearPixel mean value, i.e.,
Wherein, N0For region IrearSum of all pixels；
(3) variance: vehicle tail region I is calculatedrearPixel mean value, i.e.,
(4) ratio: ratio feature F is calculated in the following wayratio,
Wherein, h indicates distance of the tailstock portion to vehicle target boundary rectangle frame bottom, the expression tailstock portion H to present frame figure As the distance at top；
Fig. 6 shows the schematic diagram of the ratio characteristic in manual features.
Identification black smoke vehicle in the step 4 includes the following steps:
Step 4.1: all vehicle target pictures in current frame image being carried out using trained BP network classifier Classification, if there is at least one vehicle target picture is identified as black smoke vehicle picture, then present frame is identified as black smoke frame；
Step 4.2: in per continuous 100 frame, if there is K frame is identified as black smoke frame, and K meets following formula, then it is assumed that when There are black smoke vehicles in preceding video sequence.
K>α
Wherein, α is an adjustment factor for controlling recall rate and accurate rate.

Claims (8)

1. a kind of intelligent video black smoke vehicle detection method based on multi-feature fusion, which comprises the steps of:
(1) Utilization prospects detection algorithm extracts moving target from traffic surveillance videos, and identifies vehicle target；
(2) vehicle tail position is detected using integral projection and filtering technique；
(3) extract the statistical nature of tailstock portion rear area, frequency domain character and some manual features, fusion formed a feature to Amount；
(4) classified using BP network classifier to mentioned feature vector, black smoke frame is identified, to further identify black smoke Vehicle.
2. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (1) the foreground detection algorithm in includes the following steps:
(11) following formula initial background I is usedback(t),
Wherein, I (t) represents t frame image, and N represents the number of image frames of initial background use；
(12) foreground target I is calculated using following formulafore(t),
βt=mean (| I (t)-Iback(t)|)
P=threshold (| I (t)-Iback(t)|,βt+ε)
Ifore(t)=dilate (erode (P))
Wherein, threshold (I, βt+ ε) it is one with βt+ ε is the Binarization methods of threshold value, and mean (I) is a calculating image The algorithm of the average value of all pixels of I, erode (I) and dilate (I) are morphologic corrosion and expansive working respectively；
(13) background model is updated using following formula,
Wherein, threshold alpha is the adjustment factor of a control background precision；
(14) (12) are gone to step and calculate Ifore(t+1)。
3. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (1) the identification vehicle target in, which refers to while meeting following two criterion, is considered as vehicle target:
Rule one: the area of moving target is greater than some threshold value；
Rule two: the length-width ratio of the boundary rectangle frame of moving target is within the scope of some.
4. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (2) included the following steps: in using integral projection and filtering technique detection vehicle tail position
(21) vehicle target image I is calculatedobjHorizontal integral projection E1(x), i.e.,
Wherein, Iobj(x, y) is coordinate of the vehicle target image at point (x, y), and w is the width of vehicle target image, operates norm () is normalization process；
(22) stochastic filtering is carried out to vehicle target image, calculates the horizontal integral projection of filtered image, i.e.,
Operating rangefilt () is stochastic filtering process；
(23) to horizontal integral projection curve E1(x) and E2(x) it is weighted fusion, i.e.,
F (x)=λ1E1(x)+λ2E2And λ (x),1+λ2=1
Wherein, λ1And λ2Respectively E1(x) and E2(x) weight coefficient；
(24) vehicle tail position coordinate x is calculated by one of following two moderear,
Or
Wherein, Δ x is that one and tailstock portion coordinate calculate related parameter.
5. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (3) textural characteristics of the extraction tailstock portion rear area in include the following steps:
(31) the rear area I of vehicle tail position is determinedrear, which is set to start line with tailstock position, and extend back 60 pictures Element, the wide width for being set as vehicle target；
(32) following formula zoning I is usedrearGray level co-occurrence matrixes P,
Wherein, P (i, j, d, θ) indicates that direction is pixel of the gray level co-occurrence matrixes P at position (i, j) that θ pixel distance is d Value, w and h are respectively the width and height of vehicle target image, and round () is a function, indicates to round up；
(33) Normalized Grey Level co-occurrence matrix P, obtains
(34) a series of statistical natures based on gray level co-occurrence matrixes are calculated, i.e.,
One Angular second moment (ASM) of feature, note ASM (d, θ) indicate that angle is one ASM of feature that θ distance is d,
Wherein, L × L indicates Normalized Grey Level co-occurrence matrixSize；
Two Entropy of feature (ENT), note ENT (d, θ) indicate that angle is two ENT of feature that θ distance is d,
Three Contrast of feature (CON), note CON (d, θ) indicate that angle is three CON of feature that θ distance is d,
Four Correlation of feature (COR), note COR (d, θ) indicate that angle is four COR of feature that θ distance is d,
Five Inverse difference moment (IDM) of feature, note IDM (d, θ) indicate that angle is the feature five that θ distance is d IDM,
(35) four direction θ=0 ° is used, 45 °, 90 °, 135 ° and two kinds pixel distance d=2,3 obtain different normalizing ashings Spend co-occurrence matrixTo each gray level co-occurrence matrixes, five statistical natures of ASM, ENT, CON, COR and IDM are calculated, it will be different Direction, different distance five statistical natures be together in series the statistical nature based on gray level co-occurrence matrixes can be obtained.
6. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (3) frequency domain character of the extraction tailstock portion rear area in includes the following steps:
(36) by tailstock portion rear area IrearIt is bisected into 1 × 2 fritter, two layers of wavelet decomposition, record i-th are carried out to each fritter (i=1,2.) the wavelet coefficient image of the horizontal direction of layer, vertical direction and diagonal direction is respectively Hi, ViAnd Di；
(37) in the following way calculate i-th (i=1,2.) layer, the wavelet energy of kth (i=1,2.) fritter,
Wherein, wiAnd hiRespectively indicate HiWidth and height；
(38) frequency domain character that step (37) obtain is together in series for identifying black smoke vehicle.
7. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (3) some manual features of the extraction tailstock portion rear area in include:
(1) matching degree: vehicle tail region I is calculatedrearRegion corresponding with backgroundMatching degree Fmatch, i.e.,
Wherein, Irear(i, j) indicates image IrearPixel value at position (i, j),Indicate imageIn position Pixel value at (i, j)；
(2) mean value: vehicle tail region I is calculatedrearPixel mean value, i.e.,
Wherein, N0For region IrearSum of all pixels；
(3) variance: vehicle tail region I is calculatedrearPixel mean value, i.e.,
(4) ratio: ratio feature F is calculated in the following wayratio,
Wherein, h indicates distance of the tailstock portion to vehicle target boundary rectangle frame bottom, H expression tailstock portion to current frame image top The distance in portion.
8. intelligent video black smoke vehicle detection method based on multi-feature fusion as described in claim 1, which is characterized in that step (4) the identification black smoke vehicle in includes the following steps:
(41) classified using trained BP network classifier to all vehicle target pictures in current frame image, if There is at least one vehicle target picture to be identified as black smoke vehicle picture, then present frame is identified as black smoke frame；
(42) in per continuous 100 frame, if there is K frame is identified as black smoke frame, and K meets following formula, then it is assumed that current video sequence There are black smoke vehicle in column,
K>α
Wherein, α is an adjustment factor for controlling recall rate and accurate rate.
CN201810754422.6A 2018-07-11 2018-07-11 Intelligent video black smoke vehicle detection method based on multi-feature fusion Active CN109086682B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810754422.6A CN109086682B (en) 2018-07-11 2018-07-11 Intelligent video black smoke vehicle detection method based on multi-feature fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810754422.6A CN109086682B (en) 2018-07-11 2018-07-11 Intelligent video black smoke vehicle detection method based on multi-feature fusion

Publications (2)

Publication Number Publication Date
CN109086682A true CN109086682A (en) 2018-12-25
CN109086682B CN109086682B (en) 2021-07-27

Family

ID=64837559

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810754422.6A Active CN109086682B (en) 2018-07-11 2018-07-11 Intelligent video black smoke vehicle detection method based on multi-feature fusion

Country Status (1)

Country Link
CN (1) CN109086682B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112580401A (en) * 2019-09-29 2021-03-30 杭州海康威视数字技术股份有限公司 Vehicle detection method and device
CN112699801A (en) * 2020-12-30 2021-04-23 上海船舶电子设备研究所(中国船舶重工集团公司第七二六研究所) Fire identification method and system based on video image
CN112784709A (en) * 2021-01-06 2021-05-11 华南理工大学 Remote multi-target efficient detection and identification method
CN113378629A (en) * 2021-04-27 2021-09-10 阿里云计算有限公司 Method and device for detecting abnormal vehicle in smoke discharge
CN113657305A (en) * 2021-08-20 2021-11-16 深圳技术大学 Video-based intelligent detection method for blackness level of black smoke vehicle and Ringelmann

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2312475A1 (en) * 2004-07-09 2011-04-20 Nippon Telegraph and Telephone Corporation Sound signal detection and image signal detection
CN103116746A (en) * 2013-03-08 2013-05-22 中国科学技术大学 Video flame detecting method based on multi-feature fusion technology
CN103871077A (en) * 2014-03-06 2014-06-18 中国人民解放军国防科学技术大学 Extraction method for key frame in road vehicle monitoring video
CN106022231A (en) * 2016-05-11 2016-10-12 浙江理工大学 Multi-feature-fusion-based technical method for rapid detection of pedestrian

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2312475A1 (en) * 2004-07-09 2011-04-20 Nippon Telegraph and Telephone Corporation Sound signal detection and image signal detection
CN103116746A (en) * 2013-03-08 2013-05-22 中国科学技术大学 Video flame detecting method based on multi-feature fusion technology
CN103871077A (en) * 2014-03-06 2014-06-18 中国人民解放军国防科学技术大学 Extraction method for key frame in road vehicle monitoring video
CN106022231A (en) * 2016-05-11 2016-10-12 浙江理工大学 Multi-feature-fusion-based technical method for rapid detection of pedestrian

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112580401A (en) * 2019-09-29 2021-03-30 杭州海康威视数字技术股份有限公司 Vehicle detection method and device
CN112699801A (en) * 2020-12-30 2021-04-23 上海船舶电子设备研究所(中国船舶重工集团公司第七二六研究所) Fire identification method and system based on video image
CN112784709A (en) * 2021-01-06 2021-05-11 华南理工大学 Remote multi-target efficient detection and identification method
CN112784709B (en) * 2021-01-06 2023-06-20 华南理工大学 Efficient detection and identification method for remote multiple targets
CN113378629A (en) * 2021-04-27 2021-09-10 阿里云计算有限公司 Method and device for detecting abnormal vehicle in smoke discharge
CN113657305A (en) * 2021-08-20 2021-11-16 深圳技术大学 Video-based intelligent detection method for blackness level of black smoke vehicle and Ringelmann
CN113657305B (en) * 2021-08-20 2023-08-04 深圳技术大学 Video-based intelligent detection method for black smoke vehicle and ringeman blackness level

Also Published As

Publication number Publication date
CN109086682B (en) 2021-07-27

Similar Documents

Publication Publication Date Title
CN109086682A (en) A kind of intelligent video black smoke vehicle detection method based on multi-feature fusion
CN109670404B (en) Road ponding image detection early warning method based on hybrid model
Bossu et al. Rain or snow detection in image sequences through use of a histogram of orientation of streaks
CN109977782B (en) Cross-store operation behavior detection method based on target position information reasoning
Zheng et al. A novel vehicle detection method with high resolution highway aerial image
CN100545867C (en) Aerial shooting traffic video frequency vehicle rapid checking method
CN101315701B (en) Movement destination image partition method
CN103077423B (en) To run condition detection method based on crowd's quantity survey of video flowing, local crowd massing situation and crowd
CN104978567B (en) Vehicle checking method based on scene classification
US20030053659A1 (en) Moving object assessment system and method
CN106204640A (en) A kind of moving object detection system and method
WO2003003309A1 (en) Method for monitoring a moving object and system regarding same
CN102903124A (en) Moving object detection method
CN105574488A (en) Low-altitude aerial infrared image based pedestrian detection method
KR102282800B1 (en) Method for trackig multi target employing ridar and camera
CN103810703B (en) A kind of tunnel based on image procossing video moving object detection method
CN103400120B (en) Video analysis-based bank self-service area push behavior detection method
CN108765453B (en) Expressway agglomerate fog identification method based on video stream data
CN109191492B (en) Intelligent video black smoke vehicle detection method based on contour analysis
Kongurgsa et al. Real-time intrusion—detecting and alert system by image processing techniques
CN110659546A (en) Illegal booth detection method and device
CN111783700A (en) Automatic recognition early warning method and system for road foreign matters
Tao et al. Smoky vehicle detection based on range filtering on three orthogonal planes and motion orientation histogram
CN109325426A (en) A kind of black smoke vehicle detection method based on three orthogonal plane space-time characteristics
CN115223106A (en) Sprinkler detection method fusing differential video sequence and convolutional neural network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant