CN103473544A - Robust human body feature rapid extraction method - Google Patents
Robust human body feature rapid extraction method Download PDFInfo
- Publication number
- CN103473544A CN103473544A CN2013101596012A CN201310159601A CN103473544A CN 103473544 A CN103473544 A CN 103473544A CN 2013101596012 A CN2013101596012 A CN 2013101596012A CN 201310159601 A CN201310159601 A CN 201310159601A CN 103473544 A CN103473544 A CN 103473544A
- Authority
- CN
- China
- Prior art keywords
- gradient
- passage
- human body
- image
- extracting method
- 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.)
- Pending
Links
Images
Abstract
The invention discloses a robust human body feature rapid extraction method. The method comprises a step of inputting an original image and calculating an image gradient to obtain a gradient image; a step of quantifying a step size according to the direction, performing directional division on the gradient image to obtain channels in different directions; a step of quantifying a step size according to the space and performing space division on each channel; a step of rotating each channel in a certain angle to obtain a channel; a step of using an integrogram to rapidly calculate a feature describer for every channel; a step of cascading the feature describers of all channels to form a group of feature describers; and a step of, according to different step size adjusting parameters, repeating the above steps until a preset number of groups of feature describers having different descriptive abilities are generated. The robust human body feature rapid extraction method prevents the introduction of noises, and significantly improves the efficiency of human body feature extraction.
Description
Technical field
The present invention relates to computer vision and area of pattern recognition, more particularly, relate to a kind of characteristics of human body's method for expressing of isotropic multiple hierachical description.
Background technology
Human detection refers to position and the big or small process of determining all human bodies in defeated people's image or video sequence.As a gordian technique in the visual analysis of human motion, because its widespread use demand in intelligent monitoring and vehicle-mounted supplementary security system and obtained the attention of more and more researchers and research institution.Effectively feature extracting method can significantly promote the robustness reduce false alarm of human body detector.Therefore, how extracting effective feature is described and becomes the key factor that affects the human detection performance somatic data.
According to the difference of characteristic type, can be divided into feature and the feature based on gradient and the method based on multi-feature fusion of intensity-based.The difference existed for human appearance, much results of study show, the human body expression way based on gradient has better robustness than the expression way of intensity-based for variations such as illumination.And relatively recent a lot of results of study show, the method based on the gradient statistical information has better robustness for translation and the rotation at edge.For example: Lowe has proposed the constant descriptor of famous yardstick (Scale-Invariant Feature Transform, SIFT) and has carried out object detection; Mikolajczyk etc. have proposed position-direction histogram feature; Wu etc. have set forth the thought of the multi-level expression way based on the statistic abstraction degree.
Because human body has more degree of freedom, characteristic quantity is larger, although the method for the statistical nature based on gradient can obtain performance preferably, computation complexity is relatively high, often because computing velocity is crossed, is difficult to reach slowly requirement of real-time.In addition, if the yardstick of selected characteristic area is different with shape in characteristic extraction procedure, some tiny structures may be exaggerated by normalized mode, thereby introduce noise.
Summary of the invention
The object of the invention is to, in above-mentioned Human Detection, somatic data is difficult to carry out registration and the lower problem of counting yield, has proposed a kind of characteristics of human body's rapid extracting method of robust.
The technical solution that realizes the object of the invention is: a kind of characteristics of human body's rapid extracting method of robust, and the method comprises the following steps:
1) input original image, the computed image gradient obtains gradient image;
2) according to the direction quantization step
, the gradient image travel direction is divided, obtain the different passage of direction
;
3) according to the space quantization step-length
, to each passage
carry out spatial division;
6) Feature Descriptor of all passages is connected in series, forms a stack features descriptor.
7) regulate parameter according to different step-lengths
with
, repeating step 2) ~ step 6), until generate the Feature Descriptor with different descriptive powers of preset group number.
In said method, in described step 1), original image can be gray level image or coloured image, if coloured image is converted to gray level image;
In said method, described step 2) comprise following concrete steps:
21) the compute gradient angle is the tangential angle of each pixel of gradient image;
22) adopt the direction quantization step
the gradient angle is quantized, gradient image is divided into to the passage of different directions
, the angle of each passage
, the number of passage
=
;
23) for
, after retaining those and quantizing, the gradient angle is
pixel, other pixel sets to 0;
In said method, described step 22) in
be directly proportional to the rotation robustness of Gradient Features, be inversely proportional to orientation determination.
In said method, in described step 3), space is divided and is referred to, for
, utilize one group of tangential angle to be
spacing be
parallel lines be divided into a plurality of divisions;
In said method, in described step 3)
be directly proportional to the translation robustness of Gradient Features, be inversely proportional to location positioning.
In said method, in described step 4) for passage
on the point
, be mapped to passage
on point be
.
with
relation as follows:
(2)
(4)
(5)
In said method, described step 5) comprises following concrete steps:
51) at each passage
in, generate a rectangular characteristic window;
52), in characteristic window, calculate the gradient intensity of each division;
53) select to there is the division of greatest gradient intensity and calculate its Feature Descriptor;
In said method, described step 51) characteristic window can be random the generation, also can generate regularly according to a fixed step size.
In said method, described step 52) middle gradient intensity employing integrogram calculating of dividing.
Will
brief note is
, establish passage
the integrogram at place is designated as
, establish and be divided into arbitrarily rectangle
, wherein (
) mean the point in the rectangle upper left corner,
mean the wide and high of rectangle, divide
gradient intensity adopt following formula to calculate:
(7)
In said method, described step 53) in for passage
, the division that has greatest gradient intensity in characteristic window is denoted as
, its Feature Descriptor is the heterogeneous vector of one 7 dimension
, meaning and the account form of each component of this vector are as follows:
A)
be
index value, this index value is normalized to
, wherein
it is the gradient intensity of dividing arbitrarily;
C)
be the standard deviation of the gradient intensity of all divisions, calculate with following formula:
, wherein
;
D)
with
be
the average of the position of interior all non-zero pixels points, by following formula, calculate:
,
, wherein,
the center of representation feature window,
representation feature window wide and high, "
", "
" and "
" be
,
with
write a Chinese character in simplified form,
,
with
employing formula (7) is calculated.
E)
with
be
the position distribution of interior all non-zero pixels points, along the standard deviation of gradient and normal orientation, is calculated by following formula:
,
, wherein, "
" and "
" be
with
write a Chinese character in simplified form,
with
employing formula (7) is calculated.
The present invention compared with prior art, its remarkable advantage: characteristics of human body's rapid extracting method of a kind of robust that the present invention proposes, because the yardstick of the division of all direction passages in characteristic extraction procedure is equal, tiny structure in little division can not be exaggerated by normalized mode, thereby has avoided the introducing noise.In addition, isotropic feature, because all divisions in characteristic window are all the rectangles of vertical direction, can be used the integrogram fast method to calculate, and has improved significantly the efficiency that the characteristics of human body extracts.
The accompanying drawing explanation
Fig. 1 feature extraction process flow diagram of the present invention.
Fig. 2 is transformed into original image the process of gradient image.
The process that Fig. 3 divides the gradient image travel direction.
The process of Fig. 4 spatial division quick calculated characteristics.
Embodiment
Integrated operation flow process of the present invention is as shown in Fig. 1.Below in conjunction with accompanying drawing, body embodiment of the present invention is described in further detail.
step 1:the input original image, the computed image gradient obtains gradient image.
Data source of the present invention is gray level image or coloured image.If coloured image can first convert gray level image to, then carry out subsequent treatment.
For input picture
, as shown in figure 21, a tlv triple for each pixel on image
mean, here
with
the position that means pixel,
the gray-scale value that means pixel, used wave filter
to image
compute gradient, produce gradient image 22.
step 2:according to the direction quantization step
, the gradient image travel direction is divided, obtain the passage of different directions.
The set direction quantization step
for
, to the gradient angle
quantized, gradient image is divided into to the passage that several directions are different, here, the number of passage is designated as
,
=
=9, the angle of each passage is designated as
,
, angle is
passage is designated as
.For
, only have the rear gradient angle of those quantifications to be
pixel be retained, other pixel is set to 0.As shown in Figure 3,31,32,33 ... 39 mean that respectively angle is
,
,
...... passage.
Known according to the direction division methods, parameter
control the orientation determination of Gradient Features: when
when value is larger, the orientation determination of the feature of reservation is less, and the rotation robustness is larger; Otherwise the orientation determination of the feature retained is larger, the rotation robustness is less.
step 3:according to the space quantization step-length
, gradient image is carried out to spatial division.
For each passage
, utilize tangential direction to be
, spacing is
one group of parallel lines be divided into a plurality of divisions.Take Figure 33 passage as example, and spatial division as shown in figure 41.
Known according to space-division method, parameter
control the location positioning of Gradient Features: when
when value is larger, the location positioning of the Gradient Features of reservation is less, and the translation robustness is larger; Otherwise the location positioning of the Gradient Features retained is larger, the translation robustness is less.
At passage
the place two field picture on point
, be mapped to passage
point on the image of place frame is
.
with
relation as follows:
(4)
Characteristic window can be random the generation, also can generate regularly according to a fixed step size.If picture traverse and highly being respectively
, the characteristic window of generation is
, wherein (
) mean the point in the window upper left corner,
the representation feature window is wide and high respectively, and characteristic window satisfies condition:
.
step 52:adopt integrogram calculated characteristics window
the gradient intensity of interior all divisions.
Will
brief note is
, establish passage
the integrogram at place is designated as
, divide and be designated as rectangular area arbitrarily
, divide
gradient intensity adopt following formula to calculate:
(7)
step 53:selection has the division of greatest gradient intensity, calculates its Feature Descriptor.
Selection has the division of greatest gradient intensity, as shown in figure 43, is denoted by
,
feature Descriptor be one 7 the dimension heterogeneous vector
, meaning and the account form of each component of this vector are as follows:
A)
be
index value, this index value is normalized to
, wherein
it is the gradient intensity of dividing arbitrarily;
C)
be the standard deviation of the gradient intensity of all divisions, calculate with following formula:
, wherein
;
D)
with
be
the average of the position of interior all non-zero pixels points, by following formula, calculate:
,
, wherein,
the center of representation feature window,
the representation feature window is wide and high respectively, "
", "
" and "
" be
,
with
write a Chinese character in simplified form,
,
with
employing formula (7) is calculated.
E)
with
be
the position distribution of interior all non-zero pixels points, along the standard deviation of gradient and normal orientation, is calculated by following formula:
,
, wherein, "
" and "
" be
with
write a Chinese character in simplified form,
with
employing formula (7) is calculated.
step 6:the Feature Descriptor of all direction passages is joined together, form final descriptor, be expressed as
.
Claims (12)
1. characteristics of human body's rapid extracting method of a robust is characterized in that step is as follows:
1) input original image, the computed image gradient obtains gradient image;
2) according to the direction quantization step
, the gradient image travel direction is divided, obtain the different passage of direction
;
6) Feature Descriptor of all passages is connected in series, forms a stack features descriptor;
2. characteristics of human body's rapid extracting method of robust according to claim 1, it is characterized in that: in described step 1), original image is gray level image or coloured image, if coloured image is converted to gray level image.
3. characteristics of human body's rapid extracting method of robust according to claim 1, is characterized in that described step 2) comprise following concrete steps:
21) the compute gradient angle is the tangential angle of each pixel of gradient image;
22) adopt the direction quantization step
the gradient angle is quantized, gradient image is divided into to the passage of different directions
, the angle of each passage
, the number of passage
=
,
=1,2
;
23) for
, after retaining those and quantizing, the gradient angle is
pixel, other pixel sets to 0.
7. characteristics of human body's rapid extracting method of robust according to claim 1 is characterized in that: in described step 4) for passage
on the point
, be mapped to passage
on point be
,
with
relation as follows:
(6)。
8. characteristics of human body's rapid extracting method of robust according to claim 1 is characterized in that described step 5) comprises following concrete steps:
52), in characteristic window, calculate the gradient intensity of each division;
53) select to there is the division of greatest gradient intensity and calculate its Feature Descriptor.
9. characteristics of human body's rapid extracting method of robust according to claim 8 is characterized in that: described step 51) characteristic window can be random the generation, also can generate regularly according to a fixed step size.
10. characteristics of human body's rapid extracting method of robust according to claim 8, is characterized in that: the gradient intensity employing integrogram calculating of dividing described step 52);
Will
brief note is
, establish passage
the integrogram at place is designated as
, establish and be divided into arbitrarily rectangle
, wherein (
) mean the point in the rectangle upper left corner,
mean the wide and high of rectangle, divide
gradient intensity adopt following formula to calculate:
(7).
11. characteristics of human body's rapid extracting method of robust according to claim 8 is characterized in that: described step 53) for passage
, the division that has greatest gradient intensity in characteristic window is denoted as
, its Feature Descriptor is the heterogeneous vector of one 7 dimension
, meaning and the account form of each component of this vector are as follows:
A)
be
index value, this index value is normalized to
, wherein
it is the gradient intensity of dividing arbitrarily;
C)
be the standard deviation of the gradient intensity of all divisions, calculate with following formula:
, wherein
;
D)
with
be
the average of the position of interior all non-zero pixels points, by following formula, calculate:
,
, wherein,
the center of representation feature window,
representation feature window wide and high, "
", "
" and "
" be
,
with
write a Chinese character in simplified form,
,
with
employing formula (7) is calculated;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013101596012A CN103473544A (en) | 2013-04-28 | 2013-04-28 | Robust human body feature rapid extraction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013101596012A CN103473544A (en) | 2013-04-28 | 2013-04-28 | Robust human body feature rapid extraction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103473544A true CN103473544A (en) | 2013-12-25 |
Family
ID=49798388
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013101596012A Pending CN103473544A (en) | 2013-04-28 | 2013-04-28 | Robust human body feature rapid extraction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103473544A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080304750A1 (en) * | 2002-07-16 | 2008-12-11 | Nec Corporation | Pattern feature extraction method and device for the same |
CN102136066A (en) * | 2011-04-29 | 2011-07-27 | 电子科技大学 | Method for recognizing human motion in video sequence |
CN102663449A (en) * | 2012-03-12 | 2012-09-12 | 西安电子科技大学 | Method for tracing human body movement based on maximum geometric flow histogram |
CN102663391A (en) * | 2012-02-27 | 2012-09-12 | 安科智慧城市技术(中国)有限公司 | Image multifeature extraction and fusion method and system |
-
2013
- 2013-04-28 CN CN2013101596012A patent/CN103473544A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080304750A1 (en) * | 2002-07-16 | 2008-12-11 | Nec Corporation | Pattern feature extraction method and device for the same |
CN102136066A (en) * | 2011-04-29 | 2011-07-27 | 电子科技大学 | Method for recognizing human motion in video sequence |
CN102663391A (en) * | 2012-02-27 | 2012-09-12 | 安科智慧城市技术(中国)有限公司 | Image multifeature extraction and fusion method and system |
CN102663449A (en) * | 2012-03-12 | 2012-09-12 | 西安电子科技大学 | Method for tracing human body movement based on maximum geometric flow histogram |
Non-Patent Citations (2)
Title |
---|
刘亚洲: ""基于时空分析和多粒度特征表示的人体检测方法研究"", 《中国博士学位论文全文数据库 信息科技辑(月刊)》 * |
陈刚: ""基于多特征的行人检测方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109918969B (en) | Face detection method and device, computer device and computer readable storage medium | |
Akinlar et al. | Edlines: Real-time line segment detection by edge drawing (ed) | |
CN106960451B (en) | Method for increasing number of feature points of image weak texture area | |
WO2020119661A1 (en) | Target detection method and device and pedestrian detection method and system | |
US9681047B2 (en) | Image feature extraction method and system | |
WO2016082277A1 (en) | Video authentication method and apparatus | |
CN103020993B (en) | Visual saliency detection method by fusing dual-channel color contrasts | |
CN104778476B (en) | A kind of image classification method | |
CN103530619A (en) | Gesture recognition method of small quantity of training samples based on RGB-D (red, green, blue and depth) data structure | |
CN102722891A (en) | Method for detecting image significance | |
IL268215B (en) | Method for image processing for content detection | |
WO2018082308A1 (en) | Image processing method and terminal | |
CN106778768A (en) | Image scene classification method based on multi-feature fusion | |
CN103473537B (en) | A kind of target image contour feature method for expressing and device | |
CN104408445A (en) | Automatic real-time human body detecting method | |
CN103440662A (en) | Kinect depth image acquisition method and device | |
CN105139417A (en) | Method for real-time multi-target tracking under video surveillance | |
CN103646249A (en) | Greenhouse intelligent mobile robot vision navigation path identification method | |
CN103218605A (en) | Quick eye locating method based on integral projection and edge detection | |
US20180144490A1 (en) | Method, Apparatus, Storage Medium and Device for Controlled Synthesis of Inhomogeneous Textures | |
CN106600613B (en) | Improvement LBP infrared target detection method based on embedded gpu | |
CN103093226B (en) | A kind of building method of the RATMIC descriptor for characteristics of image process | |
CN108304833A (en) | Face identification method based on MBLBP and DCT-BM2DPCA | |
CN110570450B (en) | Target tracking method based on cascade context-aware framework | |
CN104050673A (en) | Depth map extraction method based on K-means image segmentation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20131225 |
|
RJ01 | Rejection of invention patent application after publication |