CN104143077B - Pedestrian target search method and system based on image - Google Patents
Pedestrian target search method and system based on image Download PDFInfo
- Publication number
- CN104143077B CN104143077B CN201310169245.2A CN201310169245A CN104143077B CN 104143077 B CN104143077 B CN 104143077B CN 201310169245 A CN201310169245 A CN 201310169245A CN 104143077 B CN104143077 B CN 104143077B
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- sequence
- pedestrian target
- histogram
- image
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 claims abstract description 42
- 230000000875 corresponding Effects 0.000 claims abstract description 41
- 230000011218 segmentation Effects 0.000 claims abstract description 31
- 238000000605 extraction Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000000034 method Methods 0.000 claims description 7
- 238000009825 accumulation Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 6
- 239000000284 extract Substances 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 2
- 238000004040 coloring Methods 0.000 description 2
- 230000001131 transforming Effects 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive Effects 0.000 description 1
- 230000037237 body shape Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000001186 cumulative Effects 0.000 description 1
- 230000001419 dependent Effects 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000002045 lasting Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006011 modification reaction Methods 0.000 description 1
- 238000004454 trace mineral analysis Methods 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000004642 transportation engineering Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Abstract
A kind of pedestrian target search method based on image, including：Pedestrian target sequence and corresponding prospect sequence are obtained from raw video image；Calculate the jth frame R of pedestrian target sequence ii,jWeights, the pixel of the corresponding prospect sequence of all frames of pedestrian target sequence i according to weights is added up, obtains absolute region histogram H1With fuzzy region histogram H2, calculate pedestrian target histogram Hf, and calculate pedestrian target histogram H using geodesic distancefThe distance between, pedestrian target sequence is ranked up according to the size of geodesic distance.So as to avoid the problem of segmentation dress ornament, the prospect for the range of motion target that detecting and tracking comes out is extracted in video.Its dress ornament information can be efficiently extracted out by the prospect for adding up each image.Therefore, the robustness of pedestrian detection in video is improved based on background segment detection, dress ornament segmentation is avoided to be effectively improved the accuracy of pedestrian retrieval.In addition, also provide a kind of pedestrian target searching system based on image.
Description
Technical field
The present invention relates to image processing techniques, more particularly to a kind of pedestrian target search method based on image and are
System.
Background technology
With the arrival of Internet era, image retrieval technologies are widely developed and apply.In particular with intelligence
The development and application of traffic, image retrieval technologies are also applied to therewith in intelligent transportation analysis.It is widely distributed in modern city
Camera has caused traditional traffic analysis and Pedestrians and vehicles tracking to become more simple and convenient.But due to monitor video quantity
It is excessively huge, it is difficult often to carry out a large amount of monitor video of trace analysis by manpower in face of the case of burst or traffic accident.
The pedestrian target image that current pedestrian retrieval method is generally detected using individual, divides pedestrian target
It cuts, is then retrieved by the use of the pedestrian's dress ornament being partitioned into as feature.Pedestrian detection technology in still image these years some
It breaks through, for example, coping with general normal pedestrian's scene, but cannot have fine robustness in actual video.Dress ornament point
The accuracy cut has a great impact to retrieval.The segmentation of the dress ornament of pedestrian is studied in other methods, but due to
The diversity of pedestrian's posture cause dress ornament segmentation can not time cost that is too accurate or accurately dividing very much it is too big.Also have
Detect using the moving target side information in video and divide pedestrian, but its do not use pedestrian dummy distinguish pedestrian with
Other moving targets such as automobile just with judging the methods of fringe density, have larger False Rate.
Invention content
Based on this, provide it is a kind of it is high detection, low flase drop the pedestrian target search method based on image.
A kind of pedestrian target search method based on image, includes the following steps：
Pedestrian target sequence and corresponding prospect sequence are obtained from raw video image；
The pixel of the corresponding prospect sequence of all frames of pedestrian target sequence i according to the weights is added up, is obtained
Obtain absolute region histogram H1With fuzzy region histogram H2,
Calculate pedestrian target histogram Hf, and calculate pedestrian target histogram H using geodesic distancefThe distance between, according to
The size of geodesic distance is ranked up pedestrian target sequence；Wherein, x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents color
Color is preferential；Color component is uniformly divided into HBINS interval, and gray value V is less than threshold value Tg for black region, gray areas [Tg, 1]
It is uniformly divided into VBINS interval.
Pedestrian target sequence and corresponding prospect sequence are obtained from raw video image in one of the embodiments,
Step includes：
Using the foreground and background of mixed Gauss model segmentation pedestrian target；
Count the window W={ w included in the corresponding prospect of the frame there may be pedestrian targeti, wherein, it is included in window
Foreground area account for window area more than half, and window is tall and big in the pixel of Hmin=60, is wider than the pixel of Wmin=30；
The intersection of window is merged；
Using the windows detecting pedestrian target based on gradient orientation histogram after merging；
Pedestrian target sequence is obtained into line trace to the pedestrian target detected using based on the method for tracking target of study；
Using the pedestrian target sequence got, prospect is obtained in the background segment figure corresponding from original video picture
Sequence.
Described the step of merging the intersection of window, includes in one of the embodiments,：
It is same set S by the window indicia that any two has coincidencei；
Judge set SiIfAnd there are i ≠ j ∧ wi1∩wi2≠ φ, then merge SiWith Sj；
This step is repeated until all set for including intersection window are all merged；
With area it is minimum include set SiThe rectangle T of all windowsiTo represent set SiIn all window, weight
It is new to form window set T={ Ti}。
The absolute region histogram H in one of the embodiments,1With fuzzy region histogram H2Calculating step packet
It includes：
Pedestrian target image is transformed into hsv color space；
If gray value V the ＜ Tg, H of pedestrian target image pixel1[1]=H1[1]+1；
If grey scale pixel value V >=Tg and intensity value Sg＜ S ＜ Sc, then fuzzy region histogram H is calculated2；Wherein,
Fuzzy region histogram calculation includes color part and gray portion：
The pixel of the corresponding prospect sequence of all frames by pedestrian target sequence i in one of the embodiments,
It adds up according to the weights, obtains absolute region histogram H1With fuzzy region histogram H2The step of include：
By absolute region histogram H1With fuzzy region histogram H2Zero setting；
Calculate the jth frame prospect F of pedestrian target sequence ii,jInner ellipse, will partly be put between inner ellipse and rectangle
For background；
By the jth frame R of pedestrian target sequence ii,jCorresponding foreground part pixel carries out being added to jth according to the weights
Frame Ri,jAbsolute region histogram H1With fuzzy region histogram H2, accumulated value is weight (j), H1[index]=H1
[index]+weight (j) or H2[index]=H2[index]+weight (j), until all frames are all counted in pedestrian target sequence
It calculates and completes；
Preserve absolute region histogram H1With fuzzy region histogram H2Color characteristic histogram for sequence i.
The above-mentioned pedestrian target search method based on image is had by the target following based on study and foreground segmentation
There is the pedestrian detection of robustness as a result, and then obtaining pedestrian target sequence and prospect sequence.Pedestrian target is obtained by calculating again
The color characteristic histogram of sequence carries out feature extraction and matching, finally according to formula meter to the dress ornament color of pedestrian target
Calculate pedestrian target histogram Hf, and obtain pedestrian target histogram H using geodesic distancefThe distance between, according to geodesic distance
Size is ranked up pedestrian target sequence.So as to obtain the retrieval result arranged with correlation.The above-mentioned row based on image
People's target retrieval method avoids segmentation dress ornament this problem, but extracts a series of fortune that detecting and tracking comes out in video
The prospect of moving-target.Its dress ornament information can be efficiently extracted out by the prospect for adding up each image.Therefore, based on background point
The pedestrian detection cut improves the robustness of pedestrian detection in video, and dress ornament segmentation is avoided to be effectively improved the standard of pedestrian retrieval
True property.
In addition, also provide it is a kind of it is high detection, low flase drop the pedestrian target searching system based on image.
A kind of pedestrian target searching system based on image, which is characterized in that more including image collection module, pedestrian's sequence
Histogram feature computing module, pedestrian's sequence signature computing module and geodesic distance sorting module；The more Nogatas of pedestrian's sequence
Figure feature calculation module is connect respectively with described image acquisition module and pedestrian's sequence signature computing module, pedestrian's sequence
Row feature calculation module is also connect with the geodesic distance sorting module；
Described image acquisition module is used to obtain pedestrian target sequence and corresponding prospect sequence from raw video image；
The more histogram feature computing modules of pedestrian's sequence are used for using formula
The more histogram feature computing modules of pedestrian's sequence be additionally operable to by all frames of pedestrian target sequence i it is corresponding before
The pixel of scape sequence adds up according to the weights, obtains absolute region histogram H1With fuzzy region histogram H2,
Pedestrian's sequence signature computing module is used for according to formula
Calculate pedestrian target histogram Hf；Wherein, x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents colored preferential；Face
Colouring component is uniformly divided into HBINS interval, and it is black region that gray value V, which is less than threshold value Tg, and gray areas [Tg, 1] is uniformly divided into
VBINS interval；
The geodesic distance sorting module is used to calculate pedestrian target histogram H using geodesic distancefThe distance between, it presses
Pedestrian target sequence is ranked up according to the size of geodesic distance.
Described image acquisition module includes foreground segmentation module, pedestrian target detection module in one of the embodiments,
And pedestrian target sequential extraction procedures module；
The foreground segmentation module respectively with the pedestrian target detection module and the pedestrian target sequential extraction procedures module
Connection；
The foreground segmentation module is used for using foreground and background of the mixed Gauss model segmentation per frame pedestrian target；
The pedestrian target detection module includes pedestrian position module and the pedestrian detection module based on HOG according to a preliminary estimate；
Module exists the pedestrian position for counting to include in the corresponding foreground template window of the frame according to a preliminary estimate
Window W={ the w of pedestrian targeti, wherein, the foreground area included in window account for window area more than half, and the height of window
More than the pixel of Hmin=60, it is wider than the pixel of Wmin=30, and there will be the windows of coincidence to merge in accordance with the following steps：
It is same set S by the window indicia that any two has coincidencei；
Judge set SiIfAnd there are i ≠ j ∧ wi1∩wi2≠ φ, then merge SiWith Sj；
It repeats to judge until all set for including intersection window are all merged；
By area it is minimum include set SiThe rectangle T of all windowsiTo represent set SiIn all window, weight
It is new to form window set T={ Ti}。
The pedestrian detection module based on HOG is carried out at the pedestrian position obtained according to a preliminary estimate using HOG features
Pedestrian detection.
The pedestrian target sequential extraction procedures module includes pedestrian tracking module and prospect sequence in one of the embodiments,
Acquisition module；
The pedestrian tracking module is used to use the pedestrian tracking technology based on learning method（TLD）To the pedestrian detected
Target obtains pedestrian target sequence into line trace；
The prospect retrieval module is used to extract corresponding prospect sequence in pedestrian's target sequence；
The more histogram feature computing modules of pedestrian's sequence include color space conversion mould in one of the embodiments,
Block and single-frame images color histogram computing module；
The color space conversion module is used to pedestrian target image being transformed into hsv color space；
The single-frame images color histogram computing module includes following calculating step：
If gray value V the ＜ Tg, H of pedestrian target image pixel1[1]=H1[1]+1；
If grey scale pixel value V >=Tg and intensity value Sg＜ S ＜ Sc, then fuzzy region histogram H is calculated2；Wherein,
Fuzzy region histogram calculation includes color part and gray portion：
The more histogram feature computing modules of pedestrian's sequence further include pedestrian's sequential color in one of the embodiments,
Histogram accumulates computing module, and pedestrian's sequential color histogram accumulation computing module includes step is calculated as below：
By absolute region histogram H1With fuzzy region histogram H2Zero setting；
Calculate the jth frame prospect F of pedestrian target sequence ii,jInner ellipse, will partly be put between inner ellipse and rectangle
For background；
By the jth frame R of pedestrian target sequence ii,jCorresponding foreground part pixel carries out being added to jth according to the weights
Frame Ri,jAbsolute region histogram H1With fuzzy region histogram H2, accumulated value is weight (j), H1[index]=H1
[index]+weight (j) or H2[index]=H2[index]+weight (j), until all frames are all counted in pedestrian target sequence
It calculates and completes；
Preserve absolute region histogram H1With fuzzy region histogram H2Color characteristic histogram for sequence i.
The above-mentioned pedestrian target searching system based on image is had by the target following based on study and foreground segmentation
There is the pedestrian detection of robustness as a result, and then obtaining pedestrian target sequence and prospect sequence.Pedestrian target is obtained by calculating again
The color characteristic histogram of sequence carries out feature extraction and matching, finally according to formula meter to the dress ornament color of pedestrian target
Calculate pedestrian target histogram Hf, and obtain pedestrian target histogram H using geodesic distancefThe distance between, according to geodesic distance
Size is ranked up pedestrian target sequence.So as to obtain the retrieval result arranged with correlation.The above-mentioned row based on image
People's object retrieval system avoids segmentation dress ornament this problem, but extracts a series of fortune that detecting and tracking comes out in video
The prospect of moving-target.Its dress ornament information can be efficiently extracted out by the prospect for adding up each image.Therefore, based on background point
The pedestrian detection cut improves the robustness of pedestrian detection in video, and dress ornament segmentation is avoided to be effectively improved the standard of pedestrian retrieval
True property.
Description of the drawings
Fig. 1 is the flow chart of the pedestrian target search method based on image；
Fig. 2（a）Pedestrian target sequence diagram for acquisition；
Fig. 2（b）For the corresponding prospect sequence diagram of pedestrian target sequence；
Fig. 2（c）The schematic diagram of interference is removed for inner ellipse；
The distribution schematic diagram of Fig. 3 color histograms；
Fig. 4 is weight function schematic diagram；
Fig. 5 is the schematic diagram of the pedestrian target searching system based on image.
Specific embodiment
As shown in Figure 1, the flow chart for the pedestrian target search method based on image.
A kind of pedestrian target search method based on image, includes the following steps：
Step S110 obtains pedestrian target sequence and corresponding prospect sequence from raw video image.Specifically, it detects
Pedestrian target in video obtains pedestrian's sequence to the pedestrian target that detection obtains into line trace, and according to the pedestrian target
Sequence location obtains prospect sequence from the corresponding background segment figure of raw video image.Such as Fig. 2（a）It is shown, to obtain
Pedestrian target sequence, Fig. 2（b）For the corresponding prospect sequence of pedestrian target sequence.
Step S110 is specifically included：
1. using the foreground and background of mixed Gauss model segmentation pedestrian target.
Mixed Gauss model uses K（Essentially 3 to 5）A Gauss model characterizes the spy of each pixel in image
Sign updates mixed Gauss model, with each pixel and mixed Gauss model in present image after the acquisition of new frame image
Matching judges that the point is background dot if success, is otherwise foreground point.
2. possible pedestrian position according to a preliminary estimate specifically, includes the following steps：
(1) count the frame and correspond to the window W={ w included in prospect there may be pedestrian targeti, wherein, it is included in window
Foreground area account for window area more than half, and window is tall and big in the pixel of Hmin=60, is wider than the pixel of Wmin=30.It carries out
The single-frame images of pedestrian target detection is usually to be extracted from multiple image, it is preferable that is from 15 frame image contracts.
(2) it is same set S by the window indicia that any two has coincidencei。
IfAnd there are i ≠ j ∧ wi1∩wi2≠ φ, then merge SiWith Sj；Repeat this step
Until all set for including intersection window are all merged.
(4) with area it is minimum include set SiThe rectangle T of all windowsiTo represent set SiIn all window,
Re-form window set T={ Ti}。
3. using the windows detecting pedestrian target based on gradient orientation histogram after merging.
Image gradient direction histogram is a kind of iamge description for solving human body target detection, and this method uses gradient side
To histogram（Histogram of Oriented Gradients, abbreviation HOG）Feature expresses human body, extracts the outer of human body
Shape information and movable information form abundant feature set.In the present embodiment, after using gradient orientation histogram combining data detection
Pedestrian target included in window.Used grader threshold value is -0.5.
4. pedestrian target sequence is obtained into line trace to the pedestrian target detected using based on the method for tracking target of study
Row.
Learn detection (Tracking-Learning-Detection, TLD) algorithm using tracking to grow target
The lasting tracking of phase, to the target in dynamic image sequence into line trace.TLD can carry out target to continue tracking, even if
Under visible ray tracking failure also can making up by infrared image, so as to make tracking effect more accurate.
Using the pedestrian target sequence got, in the background segment figure corresponding from original video picture, obtain respectively
Take pedestrian target sequence and prospect sequence.
Step S130, by the pixel of the corresponding prospect sequence of all frames of pedestrian target sequence i according to the weights into
Row is cumulative, obtains absolute region histogram H1With fuzzy region histogram H2, by the absolute region histogram H after accumulation1With obscuring
Region histogram H2It is denoted as the color characteristic histogram of pedestrian target sequence i.
Coloured image is converted into HSV space from rgb space, according to the visual characteristic of human eye, according to color saturation value
Color is divided into 3 regions, saturation degree is more than ScColored region, saturation degree be less than SgGray areas and saturation degree at
InBetween fuzzy region, colored region and gray areas are referred to as Accurate color region.Colored region is only examined
Consider its color component H values, its gray component V values are only considered for gray areas, and fuzzy region then considers its color point simultaneously
Amount and gray component value；In addition, for the too small gray component value of gray value all as the black in gray scale.
Rgb color pattern is a kind of color standard of industrial quarters, is by leading to red (R), green (G), blue (B) three colors
The variation in road and their mutual superpositions obtain miscellaneous color.
HSV（Also it is HSB）It is two kinds of related expressions to rgb color space midpoint, in description than RGB more accurately
Color contact is perceived, and is calculated simple.H refers to hue（Form and aspect）, S refer to saturation（Saturation degree）, L refer to lightness（It is bright
Degree）, V refers to value (tone), B refers to brightness（Lightness）.
Form and aspect（H）It is the essential attribute of color, is exactly usually described color designation, such as red, yellow.
Saturation degree（S）Refer to the purity of color, higher color is purer, low then gradually graying, takes the numerical value of 0-100%.Tone
（V）, brightness（L）Take 0-100%.
Points of the HSV color description in cylindrical-coordinate system, the central shaft value of this cylinder be from the black of bottom to
The white at top and in the grey for being among them, the angle around this axis corresponds to " form and aspect ", and the distance to this axis corresponds to
In " saturation degree ", and correspond to " brightness " along the height of this axis, " tone " or " lightness ".
HSV（Form and aspect, saturation degree, tone）Conceptually it is considered the inverted cone of color（Stain on lower vertex,
White is in the upper bottom surface center of circle）.Because HSV is the simple transformation for the RGB that equipment relies on,（H, s, l）Or（H, s, v）Triple defines
Color dependent on used specific red, green and blue " additive primaries ".Each unique RGB equipment is along with one
A unique HSV space.But（H, s, l）Or（H, s, v）Triple is being constrained to specific rgb space.For example, sRGB when
Time reforms into explicitly.
HSV models are a kind of nonlinear transformations of primaries pattern.
Color component in HSV space is uniformly divided into HBINS interval, it is preferable that color space 0-360 degree is averaged
It is divided into HBINS=100 interval.It is black region that gray value V, which is less than Tg=0.05, and gray areas [Tg, 1] is uniformly divided into VBINS
A interval, it is preferable that VBINS=100.The interim color histogram of image includes 2 parts：Accurate color region is corresponding absolutely
To region histogram H1, the length of 1+VBINS+HBINS；And the fuzzy region histogram H in the corresponding region of fuzzy color2,
Length is VBINS+HBINS, as shown in Figure 3.The absolute region histogram H in Accurate color region1In gray areas be 1+
VBINS, color region HBINS are colored, and color alignment is followed successively by red, yellow, green, blue and red, are handed in color
Fork point, new color is formed by each color proportion.The fuzzy region histogram H in fuzzy color region2Gray areas
For VBINS, the color alignment of color region HBINS is consistent.The absolute region histogram H in Accurate color region1With final face
Color Histogram HfIt is absolute region histogram H1With fuzzy region histogram H2The sum of, distribution of color and Accurate color region
Histogram distribution is consistent, length 1+VBINS+HBINS.Final color histogram is defined as：
Wherein x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents colored preferential.
After the completion of color histogram definition, then absolute region histogram H is calculated1With fuzzy region histogram H2, it is specific to count
Calculating step is：
1. pedestrian target image is transformed into hsv color space.
If 2. gray value V the ＜ Tg, H of pedestrian target image pixel1[1]=H1[1]+1。
If 5. grey scale pixel value V >=Tg and intensity value Sg＜ S ＜ Sc, then fuzzy region histogram H is calculated2；Its
In, fuzzy region histogram calculation includes color part and gray portion.
Preferably, the threshold value S of color regionc=0.15, Sg=0.05
Based on absolute region histogram H1With fuzzy region histogram H2, then color characteristic histogram HfCalculating step be：
1. by absolute region histogram H1With fuzzy region histogram H2Zero setting.
2. take i-th of pedestrian target sequenceAnd corresponding prospect sequenceWherein NiSequence length for i-th of pedestrian target sequence；Pedestrian's mesh is calculated using the following formula
Sequence is marked per the corresponding weights f of frame image, and pedestrian target sequence i is weighted, takes T=Ni*2/3.As shown in figure 4, for T
With accumulated value weight (j) function schematic diagrames.
Calculate the jth frame prospect F of pedestrian target sequencei,jInner ellipse, will be partly set between inner ellipse and rectangle
Background.The interference in image, such as Fig. 2 can be removed using inner ellipse（c）It is shown.
3. by the jth frame R of pedestrian target sequence ii,jCorresponding foreground part pixel carries out being accumulate to according to the weights
J frames Ri,jAbsolute region histogram H1With fuzzy region histogram H2, accumulated value is weight (j), H1[index]=H1
[index]+weight (j) or H2[index]=H2[index]+weight (j), until all frames are all counted in pedestrian target sequence
It calculates and completes.
4. preserve absolute region histogram H1With fuzzy region histogram H2Color characteristic histogram for sequence i.
Step S140, according to formula
Calculate pedestrian target histogram Hf, and calculate pedestrian target histogram H using geodesic distancefThe distance between, according to
The size of geodesic distance is ranked up pedestrian target sequence；Wherein, x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents color
Color is preferential；Color component is uniformly divided into HBINS=100 interval, and gray value V is less than threshold value Tg=0.05 for black region, gray scale
Region [Tg, 1] is uniformly divided into VBINS=10 interval.
Geodesic distance EMD（Earth Mover's Distance）For calculating the distance between histogram, when between feature
When the distance of (bin and bin) can be acquired using ground distance, made of Earth Mover's Distance similar
Calculating can obtain more accurate result.
The above-mentioned pedestrian target search method based on image is had by the target following based on study and foreground segmentation
There is the pedestrian detection of robustness as a result, and then obtaining pedestrian target sequence and prospect sequence.Pedestrian target is obtained by calculating again
The color characteristic histogram of sequence carries out feature extraction and matching, finally according to formula meter to the dress ornament color of pedestrian target
Calculate pedestrian target histogram Hf, and obtain pedestrian target histogram H using geodesic distancefThe distance between, according to geodesic distance
Size is ranked up pedestrian target sequence.So as to obtain the retrieval result arranged with correlation.The above-mentioned row based on image
People's target retrieval method avoids segmentation dress ornament this problem, but extracts a series of fortune that detecting and tracking comes out in video
The prospect of moving-target.Its dress ornament information can be efficiently extracted out by the prospect for adding up each image.Therefore, based on background point
The pedestrian detection cut improves the robustness of pedestrian detection in video, and dress ornament segmentation is avoided to be effectively improved the standard of pedestrian retrieval
True property.
As shown in figure 5, a kind of pedestrian target searching system based on image, including image collection module 500, pedestrian's sequence
More histogram feature computing modules 540, pedestrian's sequence signature computing module 550 and geodesic distance sorting module 560；The pedestrian
The more histogram feature computing modules 540 of sequence calculate mould with described image acquisition module 500 and pedestrian's sequence signature respectively
Block 550 connects, and pedestrian's sequence signature computing module 550 is also connect with the geodesic distance sorting module 560.
Image collection module 500 is used to obtain pedestrian target sequence and corresponding prospect sequence from raw video image.
The more histogram feature computing modules 540 of pedestrian's sequence are used for using formula
The more histogram feature computing modules 540 of pedestrian's sequence are additionally operable to correspond to all frames of pedestrian target sequence i
The pixel of prospect sequence add up according to the weights, obtain absolute region histogram H1With fuzzy region histogram H2,
Pedestrian's sequence signature computing module 550 is used for according to formula
Calculate pedestrian target histogram Hf；Wherein, x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents colored preferential；Face
Colouring component is uniformly divided into HBINS interval, and it is black region that gray value V, which is less than threshold value Tg, and gray areas [Tg, 1] is uniformly divided into
VBINS interval；
The geodesic distance sorting module 560 is used to calculate pedestrian target histogram H using geodesic distancefBetween away from
From being ranked up according to the size of geodesic distance to pedestrian target sequence.
Image collection module 500 includes foreground segmentation module 510, pedestrian target detection module 520 and pedestrian target sequence
Extraction module 530.
The foreground segmentation module 510 carries respectively with the pedestrian target detection module 520 and the pedestrian target sequence
Modulus block 530 connects.
The foreground segmentation module 510 is used for using foreground and background of the mixed Gauss model segmentation per frame pedestrian target.
The pedestrian target detection module 520 includes pedestrian position module 522 and the pedestrian detection based on HOG according to a preliminary estimate
Module 524.
The pedestrian position according to a preliminary estimate for counting to include in the corresponding foreground template window of the frame deposit by module 522
In the window W={ w of pedestrian targeti, wherein, the foreground area that is included in window account for window area more than half, and window
It is tall and big to be wider than the pixel of Wmin=30 in the pixel of Hmin=60, and there will be the windows of coincidence to merge in accordance with the following steps：
It is same set S by the window indicia that any two has coincidencei；
Judge set SiIfAnd there are i ≠ j ∧ wi1∩wi2≠ φ, then merge SiWith Sj；
It repeats to judge until all set for including intersection window are all merged；
By area it is minimum include set SiThe rectangle T of all windowsiTo represent set SiIn all window, weight
It is new to form window set T={ Ti}。
The pedestrian detection module 524 based on HOG be using HOG features at the pedestrian position obtained according to a preliminary estimate into
Row pedestrian detection.
Pedestrian target sequential extraction procedures module 530 includes pedestrian tracking module 532 and prospect retrieval module 534.
The pedestrian tracking module 532 is used to use the pedestrian tracking technology based on learning method（TLD）To what is detected
Pedestrian target obtains pedestrian target sequence into line trace；
The prospect retrieval module 534 is used to extract corresponding prospect sequence in pedestrian's target sequence.
The more histogram feature computing modules 540 of pedestrian's sequence include color space conversion module 542 and single-frame images color
Histogram calculation module 544.
The color space conversion module 542 is used to pedestrian target image being transformed into hsv color space.
The single-frame images color histogram computing module 544 includes following calculating step：
If gray value V the ＜ Tg, H of pedestrian target image pixel1[1]=H1[1]+1；
If grey scale pixel value V >=Tg and intensity value Sg＜ S ＜ Sc, then fuzzy region histogram H is calculated2；Wherein,
Fuzzy region histogram calculation includes color part and gray portion：
The more histogram feature computing modules 540 of pedestrian's sequence further include pedestrian's sequential color histogram accumulation computing module
546, pedestrian's sequential color histogram accumulation computing module 546 includes step is calculated as below：
By absolute region histogram H1With fuzzy region histogram H2Zero setting；
Calculate the jth frame prospect F of pedestrian target sequence ii,jInner ellipse, will partly be put between inner ellipse and rectangle
For background；
By the jth frame R of pedestrian target sequence ii,jCorresponding foreground part pixel carries out being added to jth according to the weights
Frame Ri,jAbsolute region histogram H1With fuzzy region histogram H2, accumulated value is weight (j), H1[index]=H1
[index]+weight (j) or H2[index]=H2[index]+weight (j), until all frames are all counted in pedestrian target sequence
It calculates and completes；
Preserve absolute region histogram H1With fuzzy region histogram H2Color characteristic histogram for sequence i.
The above-mentioned pedestrian target searching system based on image extracts pedestrian target by extracting the pedestrian target in video
Feature, rapidly in this video or other videos retrieval with similar features pedestrian target.Utilize the ladder for extracting pedestrian
Spend direction histogram（HOG）Then feature distinguishes pedestrian and non-pedestrian using these features.Pedestrian is movement in video
, it can be by general pedestrian detection algorithm using this effective information（Such as HOG）More robust realization in video.
The above-mentioned pedestrian target searching system based on image is had by the target following based on study and foreground segmentation
There is the pedestrian detection of robustness as a result, and then obtaining pedestrian target sequence and prospect sequence.Pedestrian target is obtained by calculating again
The color characteristic histogram of sequence carries out feature extraction and matching, finally according to formula meter to the dress ornament color of pedestrian target
Calculate pedestrian target histogram Hf, and obtain pedestrian target histogram H using geodesic distancefThe distance between, according to geodesic distance
Size is ranked up pedestrian target sequence.So as to obtain the retrieval result arranged with correlation.The above-mentioned row based on image
People's object retrieval system avoids segmentation dress ornament this problem, but extracts a series of fortune that detecting and tracking comes out in video
The prospect of moving-target.Its dress ornament information can be efficiently extracted out by the prospect for adding up each image.Therefore, based on background point
The pedestrian detection cut improves the robustness of pedestrian detection in video, and dress ornament segmentation is avoided to be effectively improved the standard of pedestrian retrieval
True property.
Embodiment described above only expresses the several embodiments of the present invention, and description is more specific and detailed, but simultaneously
Cannot the limitation to the scope of the claims of the present invention therefore be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the guarantor of the present invention
Protect range.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of pedestrian target search method based on image, which is characterized in that include the following steps：
Pedestrian target sequence and corresponding prospect sequence are obtained from raw video image；
Using formulaCalculate every frame image R of the pedestrian target sequence ii,jPower
Value, T=Ni*2/3；NiFor the sequence length of i-th of pedestrian target sequence, T is sequence length；
The pixel of the corresponding prospect sequence of all frames of pedestrian target sequence i according to the weights is added up, is obtained exhausted
To region histogram H1With fuzzy region histogram H2,
According to formula
Calculate pedestrian target histogram Hf, and calculate pedestrian target histogram H using geodesic distancefThe distance between, according to geodetic
The size of distance is ranked up pedestrian target sequence；Wherein, x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents colored excellent
First；Color component is uniformly divided into HBINS interval, and it is black region that gray value V, which is less than threshold value Tg, and gray areas [Tg, 1] is uniform
It is divided into VBINS interval.
2. the pedestrian target search method according to claim 1 based on image, which is characterized in that from raw video image
The step of middle acquisition pedestrian target sequence and corresponding prospect sequence, includes：
Using the foreground and background of mixed Gauss model segmentation pedestrian target；
Count the window W={ w for corresponding to and including in prospect there are pedestrian target in window in the framei, wherein, it is included in window
Foreground area account for window area more than half, and window is tall and big in Hmin=60 pixels, is wider than Wmin=30 pixels；
The intersection of window is merged；
Using the windows detecting pedestrian target of the method based on gradient orientation histogram (HOG) after merging；
Pedestrian target sequence is obtained into line trace to the pedestrian target detected using based on the method for tracking target of study；
Using the pedestrian target sequence got, prospect sequence is obtained in the background segment figure corresponding from original video picture
Row.
3. the pedestrian target search method according to claim 2 based on image, which is characterized in that the weight by window
The step of part merges is closed to include：
It is same set S by the window indicia that any two has coincidencei；
Judge set SiIfwi2∈Sj, and have i ≠ j ∧ wi1∩wi2≠ φ, then merge SiWith Sj；Repeat this
Step is until all set for including intersection window are all merged；
With area it is minimum include set SiThe rectangle T of all windowsiTo represent set SiIn all window, shape again
Into window set T={ Ti}。
4. the pedestrian target search method according to claim 1 based on image, which is characterized in that the absolute region is straight
Side figure H1With fuzzy region histogram H2Calculating step include：
Pedestrian target image is transformed into hsv color space；
If gray value V the ＜ Tg, H of pedestrian target image pixel1[1]=H1[1]+1；
If the gray value V >=Tg and intensity value S≤S of pedestrian target image pixelg, then the gray value of pixel is calculated, is indexedH1[index]=H1[index]+1；
If pedestrian target image pixel gray level value V >=Tg and intensity value S >=Sc, then the color value of pixel is calculated, is indexedH1[index]=H1[index]+1；
If grey scale pixel value V >=Tg and intensity value Sg＜ S ＜ Sc, then fuzzy region histogram H is calculated2；Wherein, it obscures
Region histogram calculating includes color part and gray portion：
Gray portion：H2[index]=H2[index]+1；
Chrominance section：H2[index]=H2[index]+1。
5. the pedestrian target search method according to claim 1 based on image, which is characterized in that described by pedestrian target
The pixel of the corresponding prospect sequence of all frames of sequence i adds up according to the weights, obtains absolute region histogram H1
With fuzzy region histogram H2The step of include：
By absolute region histogram H1With fuzzy region histogram H2Zero setting；
Calculate the jth frame prospect F of pedestrian target sequence ii,jInner ellipse, the back of the body will be partly set between inner ellipse and rectangle
Scape；
By the jth frame R of pedestrian target sequence ii,jCorresponding foreground part pixel carries out being added to jth frame R according to the weightsi,j
Absolute region histogram H1With fuzzy region histogram H2, accumulated value is weight (j), H1[index]=H1[index]+
Weight (j) or H2[index]=H2[index]+weight (j), until all frames all calculate completion in pedestrian target sequence；
Preserve absolute region histogram H1With fuzzy region histogram H2Color characteristic histogram for sequence i.
6. a kind of pedestrian target searching system based on image, which is characterized in that how straight including image collection module, pedestrian's sequence
Square figure feature calculation module, pedestrian's sequence signature computing module and geodesic distance sorting module；The more histograms of pedestrian's sequence
Feature calculation module is connect respectively with described image acquisition module and pedestrian's sequence signature computing module, pedestrian's sequence
Feature calculation module is also connect with the geodesic distance sorting module；
Described image acquisition module is used to obtain pedestrian target sequence and corresponding prospect sequence from raw video image；
The more histogram feature computing modules of pedestrian's sequence are used for using formula
Calculate every frame image R of the pedestrian target sequence ii,jWeights, T=Ni*
2/3；NiFor the sequence length of i-th of pedestrian target sequence, T is sequence length；
The more histogram feature computing modules of pedestrian's sequence are additionally operable to the corresponding prospect sequence of all frames of pedestrian target sequence i
The pixel of row adds up according to the weights, obtains absolute region histogram H1With fuzzy region histogram H2,
Pedestrian's sequence signature computing module is used for according to formula
Calculate pedestrian target histogram Hf；Wherein, x ∈ { 0,1 }, x=0 are gradation preference, and x=1 represents colored preferential；Color point
Amount is uniformly divided into HBINS interval, and it is black region that gray value V, which is less than threshold value Tg, and gray areas [Tg, 1] is uniformly divided into VBINS
A interval；
The geodesic distance sorting module is used to calculate pedestrian target histogram H using geodesic distancefThe distance between, according to survey
The size of ground distance is ranked up pedestrian target sequence.
7. the pedestrian target searching system according to claim 6 based on image, which is characterized in that described image obtains mould
Block includes foreground segmentation module, pedestrian target detection module and pedestrian target sequential extraction procedures module；
The foreground segmentation module is connect respectively with the pedestrian target detection module and the pedestrian target sequential extraction procedures module；
The foreground segmentation module is used for using foreground and background of the mixed Gauss model segmentation per frame pedestrian target；
The pedestrian target detection module includes pedestrian position module and the pedestrian detection module based on HOG according to a preliminary estimate；
Module is interior comprising there are pedestrian targets for counting the corresponding foreground template window of the frame according to a preliminary estimate for the pedestrian position
Window W={ wi, wherein, the foreground area included in window account for window area more than half, and window is tall and big in Hmin
=60 pixels are wider than Wmin=30 pixels, and there will be the windows of coincidence to merge in accordance with the following steps：
It is same set S by the window indicia that any two has coincidencei；
Judge set SiIfwi2∈Sj, and have i ≠ j ∧ wi1∩wi2≠ φ, then merge SiWith Sj；Repetition is sentenced
Break until all set for including intersection window are all merged；
By area it is minimum include set SiThe rectangle T of all windowsiTo represent set SiIn all window, shape again
Into window set T={ Ti}；
The pedestrian detection module based on HOG is that pedestrian is carried out at the pedestrian position obtained according to a preliminary estimate using HOG features
Detection.
8. the pedestrian target searching system according to claim 6 based on image, which is characterized in that the pedestrian target sequence
Row extraction module includes pedestrian tracking module and prospect retrieval module；
The pedestrian tracking module is used for using the pedestrian tracking technology (TLD) based on learning method to the pedestrian target that detects
Into line trace, pedestrian target sequence is obtained；
The prospect retrieval module is used to extract corresponding prospect sequence in pedestrian's target sequence.
9. the pedestrian target searching system according to claim 6 based on image, which is characterized in that pedestrian's sequence is more
Histogram feature computing module includes color space conversion module and single-frame images color histogram computing module；
The color space conversion module is used to pedestrian target image being transformed into hsv color space；
The single-frame images color histogram computing module includes following calculating step：
If gray value V the ＜ Tg, H of pedestrian target image pixel1[1]=H1[1]+1；
If the gray value V >=Tg and intensity value S≤S of pedestrian target image pixelg, then the gray value of pixel is calculated, is indexedH1[index]=H1[index]+1；
If pedestrian target image pixel gray level value V >=Tg and intensity value S >=Sc, then the color value of pixel is calculated, is indexedH1[index]=H1[index]+1；
If grey scale pixel value V >=Tg and intensity value Sg＜ S ＜ Sc, then fuzzy region histogram H is calculated2；Wherein, it obscures
Region histogram calculating includes color part and gray portion：
Gray portion：H2[index]=H2[index]+1；
Chrominance section：H2[index]=H2[index]+1。
10. the pedestrian target searching system according to claim 6 based on image, which is characterized in that pedestrian's sequence
More histogram feature computing modules further include pedestrian's sequential color histogram accumulation computing module, pedestrian's sequential color Nogata
Figure accumulation computing module includes step is calculated as below：
By absolute region histogram H1With fuzzy region histogram H2Zero setting；
Calculate the jth frame prospect F of pedestrian target sequence ii,jInner ellipse, the back of the body will be partly set between inner ellipse and rectangle
Scape；
By the jth frame R of pedestrian target sequence ii,jCorresponding foreground part pixel carries out being added to jth frame R according to the weightsi,j
Absolute region histogram H1With fuzzy region histogram H2, accumulated value is weight (j), H1[index]=H1[index]+
Weight (j) or H2[index]=H2[index]+weight (j), until all frames all calculate completion in pedestrian target sequence；
Preserve absolute region histogram H1With fuzzy region histogram H2Color characteristic histogram for sequence i.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310169245.2A CN104143077B (en) | 2013-05-09 | 2013-05-09 | Pedestrian target search method and system based on image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310169245.2A CN104143077B (en) | 2013-05-09 | 2013-05-09 | Pedestrian target search method and system based on image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104143077A CN104143077A (en) | 2014-11-12 |
| CN104143077B true CN104143077B (en) | 2018-07-03 |
Family
ID=51852247
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310169245.2A Active CN104143077B (en) | 2013-05-09 | 2013-05-09 | Pedestrian target search method and system based on image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104143077B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881662B (en) * | 2015-06-26 | 2019-03-08 | 北京畅景立达软件技术有限公司 | A kind of single image pedestrian detection method |
| CN105893963B (en) * | 2016-03-31 | 2019-03-08 | 南京邮电大学 | A kind of method of the best frame easy to identify of single pedestrian target in screening video |
| CN106023245B (en) * | 2016-04-28 | 2019-01-01 | 绍兴文理学院 | Moving target detecting method under the static background measured based on middle intelligence collection similarity |
| CN106250918B (en) * | 2016-07-26 | 2019-08-13 | 大连理工大学 | A kind of mixed Gauss model matching process based on improved soil-shifting distance |
| CN106504264B (en) * | 2016-10-27 | 2019-09-20 | 锐捷网络股份有限公司 | Video foreground image extraction method and device |
| CN110458045A (en) * | 2019-07-22 | 2019-11-15 | 浙江大华技术股份有限公司 | Acquisition methods, image processing method and the device of response probability histogram |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101616309A (en) * | 2009-07-16 | 2009-12-30 | 上海交通大学 | Non-overlapping visual field multiple-camera human body target tracking method |
| CN102194270A (en) * | 2011-06-02 | 2011-09-21 | 杭州电子科技大学 | Statistical method for pedestrian flow based on heuristic information |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8385599B2 (en) * | 2008-10-10 | 2013-02-26 | Sri International | System and method of detecting objects |
-
2013
- 2013-05-09 CN CN201310169245.2A patent/CN104143077B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101616309A (en) * | 2009-07-16 | 2009-12-30 | 上海交通大学 | Non-overlapping visual field multiple-camera human body target tracking method |
| CN102194270A (en) * | 2011-06-02 | 2011-09-21 | 杭州电子科技大学 | Statistical method for pedestrian flow based on heuristic information |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104143077A (en) | 2014-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104143077B (en) | Pedestrian target search method and system based on image | |
| CN105809138B (en) | A kind of road warning markers detection and recognition methods based on piecemeal identification | |
| CN104766071B (en) | A kind of traffic lights fast algorithm of detecting applied to pilotless automobile | |
| CN105678318B (en) | The matching process and device of traffic sign | |
| CN103035013A (en) | Accurate moving shadow detection method based on multi-feature fusion | |
| CN102915433B (en) | Character combination-based license plate positioning and identifying method | |
| CN103761529B (en) | A kind of naked light detection method and system based on multicolour model and rectangular characteristic | |
| CN104978567B (en) | Vehicle checking method based on scene classification | |
| CN104573685A (en) | Natural scene text detecting method based on extraction of linear structures | |
| CN103679677B (en) | A kind of bimodulus image decision level fusion tracking updating mutually based on model | |
| CN102881160B (en) | Outdoor traffic sign identification method under low-illumination scene | |
| CN105205489A (en) | License plate detection method based on color texture analyzer and machine learning | |
| CN103714181A (en) | Stratification specific figure search method | |
| CN103680145B (en) | A kind of people's car automatic identifying method based on local image characteristics | |
| CN105894503A (en) | Method for restoring Kinect plant color and depth detection images | |
| CN107301378A (en) | The pedestrian detection method and system of Multi-classifers integrated in image | |
| CN108537239A (en) | A kind of method of saliency target detection | |
| CN107491762A (en) | A kind of pedestrian detection method | |
| CN105069816B (en) | A kind of method and system of inlet and outlet people flow rate statistical | |
| CN108346160A (en) | The multiple mobile object tracking combined based on disparity map Background difference and Meanshift | |
| CN104376334A (en) | Pedestrian comparison method based on multi-scale feature fusion | |
| CN106295657A (en) | A kind of method extracting human height's feature during video data structure | |
| CN107944407A (en) | A kind of crossing zebra stripes recognition methods based on unmanned plane | |
| CN104268509B (en) | The method and system of dump truck car plate detection | |
| CN110490150A (en) | A kind of automatic auditing system of picture violating the regulations and method based on vehicle retrieval |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| C06 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| C10 | Entry into substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |