CN104299005A - Head detection method and system - Google Patents
Head detection method and system Download PDFInfo
- Publication number
- CN104299005A CN104299005A CN201310295570.3A CN201310295570A CN104299005A CN 104299005 A CN104299005 A CN 104299005A CN 201310295570 A CN201310295570 A CN 201310295570A CN 104299005 A CN104299005 A CN 104299005A
- Authority
- CN
- China
- Prior art keywords
- gradient
- local binary
- binary patterns
- people
- directionless
- 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.)
- Withdrawn
Links
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses a head detection method and system. The method comprises the following steps: extracting the orientation gradient based local binary pattern characteristics of an input image to be detected; and according to a head classifier, judging the extracted local binary pattern characteristics, and outputting a judgment result, wherein a head classifier acquisition process comprises the following steps: extracting the orientation gradient based local binary pattern characteristics from a sample set which comprises head positive samples and head negative samples, carrying out training learning on the extracted local binary pattern characteristics in the sample set to obtain the head classifier. The invention adopts the orientation gradient based local binary pattern (LBP) characteristics, so that the edge and outline information of the head can be presented through an orientation gradient, and in addition, on the basis, the LBP characteristics are added to present the local texture information of the head to obtain a better head detection effect so as to detect the head in a video image. The invention can be used for human tracking, event detection, human traffic statics and the like, and exhibits an important meaning for realizing automatic monitoring.
Description
Technical field
The application relates to video monitoring and technical field of image processing, is specifically related to a kind of people's head inspecting method and a kind of human head detection system.
Background technology
In field of intelligent video surveillance, be an important research contents to the Intelligent Measurement of the number of people and statistics.But the gray difference of the number of people of different attitude is comparatively large, such as, from number of people front, side, the back side and top shooting number of people gray difference can be very large, make to realize the number of people and detect more difficult.Therefore, the feature how extracting robust is the importance affecting number of people detection and statistics accuracy rate and real-time.
Summary of the invention
The application provides a kind of people's head inspecting method and a kind of human head detection system, adopts the LBP feature based on direction gradient in the method/system.
According to the first aspect of the application, the application provides a kind of people's head inspecting method, comprising: to the local binary patterns feature of the image zooming-out to be detected inputted based on direction gradient; According to number of people sorter, the local binary patterns feature extracted is adjudicated, export court verdict; The acquisition of described number of people sorter comprises: by extracting the local binary patterns feature based on direction gradient from the sample set comprising the positive sample of the number of people and number of people negative sample; Training study is carried out to the local binary patterns feature of the sample set extracted, obtains number of people sorter.
Further, the extraction of the described local binary patterns feature based on direction gradient comprises: gradient map forming step, to the image compute gradient of input, obtains the gradient map comprising gradient magnitude and gradient direction; Directional diagram forming step, is divided into N number of interval by described gradient direction, and N is positive integer, carries out interpolation according to gradient direction to described gradient magnitude, obtains direction gradient figure; Cumulative chart forming step, is decomposed into N number of directionless gradient map by described direction gradient figure, carries out accumulation gradient amplitude statistics, form the directionless gradient map of N number of accumulation to each described directionless gradient map; Independently accord with calculation procedure, in the directionless gradient map of each described accumulation, calculate local binary patterns descriptor, form N number of independent local binary patterns feature; Feature forming step, by the cascade of described N number of independent local binary patterns feature, forms the local binary patterns feature based on direction gradient.
Preferably, in described gradient map forming step, adopt difference operator to calculate gradient direction and the gradient magnitude of each pixel in the image of input, form gradient map; In described directional diagram forming step, described interpolation is one-line interpolation, the gradient magnitude that each pixel in described direction gradient figure stores adjacent two intervals and is assigned with separately.
Preferably, in described cumulative chart forming step, in i-th directionless gradient map, the value of pixel is the interval gradient magnitude of this pixel i-th of belonging to, if this pixel does not belong to described i-th interval, then the value of this pixel is that 0, i is positive integer and is less than or equal to N by tax; For each described directionless gradient map, scan by the pane location of pre-sizing, and using the value of the gradient magnitude sum on the pane location of pre-sizing as the central pixel point of this pane location, thus the directionless gradient map of the accumulation forming this directionless gradient map.
Preferably, the gradient magnitude sum on the pane location of described pre-sizing adopts formula A+D-B-C to calculate, and wherein A, B, C and D represent the integrated value of the pixel on four summits of described pane location respectively.
Preferably, in described independent symbol calculation procedure, for the directionless gradient map of each described accumulation, the calculating of its independent local binary patterns feature comprises:
By directionless for described accumulation gradient map piecemeal;
Calculate the local binary patterns descriptor of each described piecemeal;
Described local binary patterns descriptor is normalized;
By the local binary patterns descriptor cascade of all described piecemeals, form independent local binary patterns feature.
According to the second aspect of the application, the application provides a kind of human head detection system, comprising: characteristic extracting module, for the image zooming-out to be detected of input based on the local binary patterns feature of direction gradient; Classification judging module, for adjudicating the local binary patterns feature extracted according to number of people sorter, export court verdict, wherein, the acquisition of described number of people sorter comprises: by extracting the local binary patterns feature based on direction gradient from the sample set comprising the positive sample of the number of people and number of people negative sample, training study is carried out to the local binary patterns feature of the sample set extracted, obtains number of people sorter.
Further, described characteristic extracting module comprises: gradient map forming unit, for the gradient direction and the gradient magnitude that adopt difference operator to calculate each pixel in the image of input, forms gradient map, pattern forming unit, for described gradient direction is divided into N number of interval, N is positive integer, according to gradient direction, one-line interpolation is carried out to described gradient magnitude, obtain direction gradient figure, the gradient magnitude that each pixel in described direction gradient figure stores adjacent two intervals and is assigned with separately, cumulative chart forming unit, for described direction gradient figure is decomposed into N number of directionless gradient map, , wherein in i-th directionless gradient map, the value of pixel is the interval gradient magnitude of this pixel i-th of belonging to, if this pixel does not belong to described i-th interval, then the value of this pixel is 0 by tax, i is positive integer and is less than or equal to N, for each described directionless gradient map, scan by the pane location of pre-sizing, and using the value of the gradient magnitude sum on the pane location of pre-sizing as the central pixel point of this pane location, thus the directionless gradient map of the accumulation forming this directionless gradient map, independently accord with computing unit, in the directionless gradient map of each described accumulation, calculate local binary patterns descriptor, form N number of independent local binary patterns feature, feature forming unit, for by the cascade of described N number of independent local binary patterns feature, forms the local binary patterns feature based on direction gradient.
The beneficial effect of the application is: adopt the local binary patterns feature based on direction gradient, the edge of the number of people and profile information are characterized by direction gradient, and add local binary patterns feature on this basis to characterize the local grain information of the number of people, make number of people Detection results better, and then the number of people in video image can be detected, may be used for human body tracking, event detection, people flow rate statistical etc., significant for realizing automatic monitoring.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of people's head inspecting method of a kind of embodiment of the application;
Fig. 2 is the schematic flow sheet of extraction based on the LBP feature of direction gradient of a kind of embodiment of the application;
Fig. 3 is that the directionless gradient profile of a kind of embodiment of the application becomes the process schematic accumulating directionless gradient map;
Fig. 4 is the process schematic of the independent LBP feature of formation of a kind of embodiment of the application;
Fig. 5 is the structural representation of the human head detection system of a kind of embodiment of the application.
Embodiment
Number of people detection algorithm generally comprises characteristics extraction and sorter and to classify two parts.Characteristics extraction refers to and extract effective data in order to judge whether have target in image in piece image.The quality of characteristics extraction algorithm is directly connected to the accuracy of detection algorithm.Classifier algorithm comprises training and detects two parts.Training part by extracting eigenwert to sample set, and with these eigenwert training classifiers; Whether detecting portion, for image to be detected, extracts its eigenwert, then classifies to the eigenwert extracted with the sorter trained, judge in image containing detecting target.
After considering in number of people detection algorithm the various features adopted, the each embodiment of the application is by HOG(Histogram of Oriented Gradient, histograms of oriented gradients) advantage of feature and LBP(Local Binary Pattern, local binary patterns) operator combines, obtain the LBP feature based on direction gradient, use it for the number of people and detect.
By reference to the accompanying drawings the present invention is described in further detail below by embodiment.
Embodiment 1:
As shown in Figure 1, people's head inspecting method of the present embodiment comprises the steps:
Step S111, to the LBP feature of the image zooming-out to be detected inputted based on direction gradient;
Step S113, adjudicates the LBP feature based on direction gradient obtained according to number of people sorter, and wherein number of people sorter is obtained by training, and this training comprises the steps:
Step S121, to inputting the sample set that is made up of the positive sample of the number of people and number of people negative sample, extracts its LBP feature based on direction gradient; Wherein, the positive sample of the number of people can comprise from number of people front, side, the back side and top shooting number of people image, namely cover the true number of people image of different attitude, different hair, the different cap of band, and number of people negative sample can comprise the image that such as landscape, animal, word etc. do not comprise arbitrarily the number of people.
Step S123, carries out training study to the LBP eigenwert based on direction gradient of sample set;
Step S125, obtains number of people sorter by the result of training study.
Step S115, according to the court verdict of step S113, exports the number of people testing result of image to be detected.
For step S121 and step S111, before extracting the LBP feature based on direction gradient, existing digital image processing techniques can be adopted to carry out the process such as image denoising, enhancing.For step S123, when carrying out training study to the eigenwert of sample set, the correlation technique of mode identification technology can be adopted to realize, such as, adopt support vector machine (SVM, Support Vector Machine) or Adaboost algorithm or Fisher arbiter etc. to train.Can adjudicate according to training in step S123 the categorised decision obtained in step S113, exporting according to court verdict in step sl 15.Step S123, S125, S113, S115 specifically can realize with reference to existing relevant image procossing and mode identification technology.
The training sample size of supposing to carry out number of people detection is width*height pixel, about in step S121 and step S111 based on the extraction of the LBP feature of direction gradient, present embodiments provide flow process as shown in Figure 2, specifically comprise the steps S201 ~ S206:
Step S201, calculates gradient direction and the gradient magnitude of each pixel in the image of input, forms gradient map.
In the present embodiment, according to difference operator [-1,0,1], formula (1) and (2) are utilized to calculate the gradient direction θ (x of each pixel (x, y) in the image of input, y) with gradient magnitude G (x, y), gradient map is formed.
G
x(x,y)=I(x+1,y)-I(x-1,y) （1）
G
y(x,y)=I(x,y+1)-I(x,y-1)
θ(x,y)=π/2+arctan(G
y(x,y)/G
x(x,y))
Wherein, G
x(x, y) and G
y(x, y) horizontal direction gradient and the vertical gradient at pixel (x, y) place is represented respectively, I (x+1, y), I (x-1, y), I (x, y+1), I (x, y-1) represent pixel (x+1 respectively, y), (x-1, y), the gray-scale value of (x, y+1), (x, y-1).
In other embodiments, other boundary operator can also be used as Roberts operator, Prewitt operator, Sobel operator etc.
Step S202, by gradient direction decile, generates direction gradient figure.
In this step, gradient direction is evenly divided into N number of interval (bin), N is positive integer, then the direction of certain gradient is dropped in certain bin and is namely determined to belong to this bin, then carry out one-line interpolation according to gradient direction to gradient magnitude, then gradient map becomes the gradient map of two passages, is referred to as direction gradient figure, in direction gradient figure, the gradient magnitude that each location of pixels stores two adjacent bin and is assigned with separately.N=3 in a kind of realization.
Step S203, is decomposed into N number of directionless gradient map by direction gradient figure.
In this step, direction gradient figure is divided into N number of directionless gradient map, and wherein, in i-th directionless gradient map, the value of pixel is the gradient magnitude of i-th bin that this pixel belongs to, if this pixel does not belong to this i-th bin, then the value of pixel is composed is 0.
Step S204, carries out accumulation gradient amplitude statistics to each directionless gradient map, forms the directionless gradient map of N number of accumulation.
In this step, for each directionless gradient map, carry out accumulation gradient amplitude statistics, by size be cell_width*cell_height pixel pane location on gradient magnitude and as the value of this grid unit center pixel, so just form the directionless gradient map of accumulation.
In a kind of specific implementation, for ensureing resolution, pane location generally selects the size of 3*3 ~ 9*9 pixel, such as 7*7.
Integrogram is adopted to calculate with raising speed in another specific implementation, as described in Figure 3, the value of the pixel (the solid black circle in diagram) shown in the figure of the rightmost side be in the pane location (black in diagram adds the rectangle frame that thick lines are formed) of the cell_width*cell_height pixel size confined in leftmost diagram all gradient magnitudes and, and the pixel shown in the figure of the rightmost side is positioned at the center of this pane location, middle schematic diagram represents that integrogram calculates, in this figure, the value of pixel is integrated value, in pane location, all gradient magnitudes calculates with available A+D-B-C, wherein A, B, C and D represents the integrated value of four summit pixels (the solid black circle in diagram) of pane location respectively.
Step S205, in the directionless gradient map of each accumulation, calculates LBP descriptor, forms N number of independent LBP feature LBP
i(p), i=1,2 ... N.
The detailed process of this step is as follows:
Directionless for each accumulation gradient map is divided into the piecemeal of block_width*block_height pixel size, it can be even division, also exist overlapping between piecemeal that can be adjacent, calculate the LBP descriptor of each piecemeal, then normalized, finally by the LBP descriptor cascade of all piecemeals, form independent LBP feature.
A kind of realize in adopt LBP operator be R
1 8, and use uniform pattern (Uniform Pattern).Symbol R
1 8the radius in the local neighborhood region of expression LBP collection apparatus is 1 pixel,
And centered by central pixel point, radius be 1 pixel annulus on the number of equally distributed surrounding pixel point be 8, after this operator scanning piece image, obtain the R of view picture figure
1 8neighborhood LBP feature coding.Be described below with formula:
Wherein, i represents that this LBP descriptor calculates in i-th directionless gradient map of accumulation, I
pand I
cjrepresent pixel p and surrounding pixel C thereof in the directionless histogram of each accumulation respectively
jvalue.Threshold tau in f (x) is greater than 0 one less value.
The mode of normalized can be the normalization mode that L1-Norm, L1-Sqrt and L2-Norm etc. are conventional.
As shown in Figure 4, left hand view is R in the directionless gradient map of accumulation to the schematic diagram of said process
1 8the schematic diagram of operator, pixel P represents central pixel point, C
j, j=0,1,2 ..., 7 represent 8 pixels around pixel P.Value due to pixel each in figure be by pixel gradient values all on the pane location centered by it and gained, as shown in the right part of flg of Fig. 4, wherein rectangular area represents the corresponding pane location of each pixel.
Step S206, by the LBP feature descriptor cascade in directionless for each accumulation gradient map, forms final feature descriptor, is described below with formula:
LBP(p)={LBP
1(p),...LBP
i(p)}i=1,2...N
LBP feature is a kind of textural characteristics, all shows good robustness in recognition of face, pedestrian detection field.LBP operator utilizes the difference of center pixel and neighborhood territory pixel to determine coded system, in extraction center pixel process, too responsive to noise.LBP operator focuses on outstanding image-region variations in detail simultaneously, can cause higher fallout ratio when background environment change is violent.Therefore for reducing LBP operator to the susceptibility of noise, we propose the individual pixel of (2r+1) × (2r+1) to obtain the value of accumulated value as center pixel, then utilize R
1 8pattern (namely radius is 1, and all pixel numbers are 8) extracts LBP operator.In addition, HOG feature is that pedestrian detection is the most conventional, there is the feature of robustness simultaneously very much, its thought is that the presentation of object in piece image and shape can be described well by the directional spreding at image pixel intensities gradient or edge, its implementation first image is divided into little pane location, then gradient direction or the edge orientation histogram of each pixel in pane location is gathered, finally altogether just can constitutive characteristic descriptor these set of histograms.The present embodiment is when design feature, take into full account LBP characteristic sum HOG feature superiority-inferiority, the LBP feature proposed based on direction gradient carries out number of people detection, the edge of the number of people and profile information are characterized by direction gradient, and add LBP feature on this basis to characterize the local grain information of the number of people, make number of people Detection results better.
Embodiment 2:
As shown in Figure 5, the present embodiment provides a kind of number of people pick-up unit, comprising: characteristic extracting module and classification judging module.Characteristic extracting module is used for the local binary patterns feature of the image zooming-out to be detected inputted based on direction gradient; Classification judging module, for adjudicating the local binary patterns feature extracted according to number of people sorter, export court verdict, wherein, the acquisition of number of people sorter comprises: by extracting the local binary patterns feature based on direction gradient from the sample set comprising the positive sample of the number of people and number of people negative sample, training study is carried out to the local binary patterns feature of the sample set extracted, obtains number of people sorter.
Wherein, characteristic extracting module comprises: gradient map forming unit, pattern forming unit, cumulative chart forming unit, independent symbol computing unit and feature forming unit.Gradient map forming unit, for the gradient direction and the gradient magnitude that adopt difference operator to calculate each pixel in the image of input, forms gradient map, pattern forming unit, for gradient direction is divided into N number of interval, N is positive integer, carries out one-line interpolation according to gradient direction to gradient magnitude, obtain direction gradient figure, the gradient magnitude that each pixel in direction gradient figure stores adjacent two intervals and is assigned with separately, cumulative chart forming unit, for direction gradient figure is decomposed into N number of directionless gradient map, , wherein in i-th directionless gradient map, the value of pixel is the interval gradient magnitude of this pixel i-th of belonging to, if this pixel does not belong to i-th interval, then the value of this pixel is 0 by tax, i is positive integer and is less than or equal to N, for each directionless gradient map, scan by the pane location of pre-sizing, and using the value of the gradient magnitude sum on the pane location of pre-sizing as the central pixel point of this pane location, thus the directionless gradient map of the accumulation forming this directionless gradient map, independently accord with computing unit, in the directionless gradient map of each accumulation, calculate local binary patterns descriptor, form N number of independent local binary patterns feature, feature forming unit, for by the cascade of N number of independent local binary patterns feature, forms the local binary patterns feature based on direction gradient.
The realization of each module and unit can associated description in reference example 1 above, does not repeat at this.
It will be appreciated by those skilled in the art that, in above-mentioned embodiment, all or part of step of various method can be carried out instruction related hardware by program and completes, this program can be stored in a computer-readable recording medium, and storage medium can comprise: ROM (read-only memory), random access memory, disk or CD etc.
Above content is in conjunction with concrete embodiment further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, some simple deduction or replace can also be made.
Claims (10)
1. people's head inspecting method, is characterized in that, comprising:
To the local binary patterns feature of the image zooming-out to be detected inputted based on direction gradient;
According to number of people sorter, the local binary patterns feature extracted is adjudicated, export court verdict;
The acquisition of described number of people sorter comprises: by extracting the local binary patterns feature based on direction gradient from the sample set comprising the positive sample of the number of people and number of people negative sample; Training study is carried out to the local binary patterns feature of the sample set extracted, obtains number of people sorter.
2. people's head inspecting method as claimed in claim 1, it is characterized in that, the extraction of the described local binary patterns feature based on direction gradient comprises:
Gradient map forming step, to the image compute gradient of input, obtains the gradient map comprising gradient magnitude and gradient direction;
Directional diagram forming step, is divided into N number of interval by described gradient direction, and N is positive integer, carries out interpolation according to gradient direction to described gradient magnitude, obtains direction gradient figure;
Cumulative chart forming step, is decomposed into N number of directionless gradient map by described direction gradient figure, carries out accumulation gradient amplitude statistics, form the directionless gradient map of N number of accumulation to each described directionless gradient map;
Independently accord with calculation procedure, in the directionless gradient map of each described accumulation, calculate local binary patterns descriptor, form N number of independent local binary patterns feature;
Feature forming step, by the cascade of described N number of independent local binary patterns feature, forms the local binary patterns feature based on direction gradient.
3. people's head inspecting method as claimed in claim 2, is characterized in that,
In described gradient map forming step, adopt difference operator to calculate gradient direction and the gradient magnitude of each pixel in the image of input, form gradient map;
In described directional diagram forming step, described interpolation is one-line interpolation, the gradient magnitude that each pixel in described direction gradient figure stores adjacent two intervals and is assigned with separately.
4. people's head inspecting method as claimed in claim 2, is characterized in that, in described cumulative chart forming step,
In i-th directionless gradient map, the value of pixel is the interval gradient magnitude of this pixel i-th of belonging to, if this pixel does not belong to described i-th interval, then the value of this pixel is that 0, i is positive integer and is less than or equal to N by tax;
For each described directionless gradient map, scan by the pane location of pre-sizing, and using the value of the gradient magnitude sum on the pane location of pre-sizing as the central pixel point of this pane location, thus the directionless gradient map of the accumulation forming this directionless gradient map.
5. people's head inspecting method as claimed in claim 4, it is characterized in that, gradient magnitude sum on the pane location of described pre-sizing adopts formula A+D-B-C to calculate, and wherein A, B, C and D represent the integrated value of the pixel on four summits of described pane location respectively.
6. people's head inspecting method as claimed in claim 2, is characterized in that, in described independent symbol calculation procedure, for the directionless gradient map of each described accumulation, the calculating of its independent local binary patterns feature comprises:
By directionless for described accumulation gradient map piecemeal;
Calculate the local binary patterns descriptor of each described piecemeal;
Described local binary patterns descriptor is normalized;
By the local binary patterns descriptor cascade of all described piecemeals, form independent local binary patterns feature.
7. people's head inspecting method as claimed in claim 6, is characterized in that,
Described piecemeal for directionless for described accumulation gradient map evenly to be divided, or exists overlapping between adjacent piecemeal;
Described local binary patterns descriptor adopts the R of uniform pattern
1 8operator;
The computing formula of described independent local binary patterns feature is:
Wherein, LBP
ip () represents that the central pixel point of i-th directionless gradient map of accumulation is the local binary patterns descriptor of p, I
pand I
cjrepresent described central pixel point p and surrounding pixel C thereof respectively
jvalue, τ is threshold value and is greater than 0.
8. people's head inspecting method as claimed in claim 2, it is characterized in that, in described feature forming step, the described local binary patterns feature formula based on direction gradient is described as:
LBP(p)={LBP
1(p),...LBP
i(p)}，i=1,2...N
Wherein LBP (p) represents the described local binary patterns feature based on direction gradient, LBP
ip () represents that the central pixel point of i-th directionless gradient map of accumulation is the independent local binary patterns feature of p.
9. a human head detection system, is characterized in that, comprising:
Characteristic extracting module, for the image zooming-out to be detected of input based on the local binary patterns feature of direction gradient;
Classification judging module, for adjudicating the local binary patterns feature extracted according to number of people sorter, export court verdict, wherein, the acquisition of described number of people sorter comprises: by extracting the local binary patterns feature based on direction gradient from the sample set comprising the positive sample of the number of people and number of people negative sample, training study is carried out to the local binary patterns feature of the sample set extracted, obtains number of people sorter.
10. people's head inspecting method as claimed in claim 9, it is characterized in that, described characteristic extracting module comprises:
Gradient map forming unit, for the gradient direction and the gradient magnitude that adopt difference operator to calculate each pixel in the image of input, forms gradient map;
Pattern forming unit, for described gradient direction is divided into N number of interval, N is positive integer, according to gradient direction, one-line interpolation is carried out to described gradient magnitude, obtain direction gradient figure, the gradient magnitude that each pixel in described direction gradient figure stores adjacent two intervals and is assigned with separately;
Cumulative chart forming unit, for described direction gradient figure is decomposed into N number of directionless gradient map, , wherein in i-th directionless gradient map, the value of pixel is the interval gradient magnitude of this pixel i-th of belonging to, if this pixel does not belong to described i-th interval, then the value of this pixel is 0 by tax, i is positive integer and is less than or equal to N, for each described directionless gradient map, scan by the pane location of pre-sizing, and using the value of the gradient magnitude sum on the pane location of pre-sizing as the central pixel point of this pane location, thus the directionless gradient map of the accumulation forming this directionless gradient map,
Independently accord with computing unit, in the directionless gradient map of each described accumulation, calculate local binary patterns descriptor, form N number of independent local binary patterns feature;
Feature forming unit, for by the cascade of described N number of independent local binary patterns feature, forms the local binary patterns feature based on direction gradient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310295570.3A CN104299005A (en) | 2013-07-15 | 2013-07-15 | Head detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310295570.3A CN104299005A (en) | 2013-07-15 | 2013-07-15 | Head detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104299005A true CN104299005A (en) | 2015-01-21 |
Family
ID=52318726
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310295570.3A Withdrawn CN104299005A (en) | 2013-07-15 | 2013-07-15 | Head detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104299005A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919889A (en) * | 2015-12-25 | 2017-07-04 | 株式会社日立制作所 | The method and apparatus detected to the number of people in video image |
| CN108205658A (en) * | 2017-11-30 | 2018-06-26 | 中原智慧城市设计研究院有限公司 | Detection of obstacles early warning system based on the fusion of single binocular vision |
| CN110136100A (en) * | 2019-04-16 | 2019-08-16 | 华南理工大学 | The automatic classification method and device of CT sectioning image |
| CN111008589A (en) * | 2019-12-02 | 2020-04-14 | 杭州网易云音乐科技有限公司 | Face key point detection method, medium, device and computing equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004924A (en) * | 2010-11-03 | 2011-04-06 | 无锡中星微电子有限公司 | Human head detection system and method |
| CN102043953A (en) * | 2011-01-27 | 2011-05-04 | 北京邮电大学 | Real-time-robust pedestrian detection method aiming at specific scene |
| CN102663409A (en) * | 2012-02-28 | 2012-09-12 | 西安电子科技大学 | Pedestrian tracking method based on HOG-LBP |
-
2013
- 2013-07-15 CN CN201310295570.3A patent/CN104299005A/en not_active Withdrawn
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004924A (en) * | 2010-11-03 | 2011-04-06 | 无锡中星微电子有限公司 | Human head detection system and method |
| CN102043953A (en) * | 2011-01-27 | 2011-05-04 | 北京邮电大学 | Real-time-robust pedestrian detection method aiming at specific scene |
| CN102663409A (en) * | 2012-02-28 | 2012-09-12 | 西安电子科技大学 | Pedestrian tracking method based on HOG-LBP |
Non-Patent Citations (1)
| Title |
|---|
| 车志富: "基于支持向量机的行人检测", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919889A (en) * | 2015-12-25 | 2017-07-04 | 株式会社日立制作所 | The method and apparatus detected to the number of people in video image |
| CN108205658A (en) * | 2017-11-30 | 2018-06-26 | 中原智慧城市设计研究院有限公司 | Detection of obstacles early warning system based on the fusion of single binocular vision |
| CN110136100A (en) * | 2019-04-16 | 2019-08-16 | 华南理工大学 | The automatic classification method and device of CT sectioning image |
| CN111008589A (en) * | 2019-12-02 | 2020-04-14 | 杭州网易云音乐科技有限公司 | Face key point detection method, medium, device and computing equipment |
| CN111008589B (en) * | 2019-12-02 | 2024-04-09 | 杭州网易云音乐科技有限公司 | Face key point detection method, medium, device and computing equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102722712B (en) | Multiple-scale high-resolution image object detection method based on continuity | |
| CN104050481B (en) | Multi-template infrared image real-time pedestrian detection method combining contour feature and gray level | |
| CN102682287B (en) | Pedestrian detection method based on saliency information | |
| CN103048329B (en) | A kind of road surface crack detection method based on active contour model | |
| CN105740758A (en) | Internet video face recognition method based on deep learning | |
| CN104318225B (en) | Detection method of license plate and device | |
| CN105913040A (en) | Real time double cameras provided pedestrian detection system for use under scotopic vision conditions | |
| CN105260749B (en) | Real-time target detection method based on direction gradient binary pattern and soft cascade SVM | |
| CN104268596B (en) | A kind of Car license recognition device and its detection method of license plate and system | |
| CN104077577A (en) | Trademark detection method based on convolutional neural network | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN109376740A (en) | A kind of water gauge reading detection method based on video | |
| CN104268528A (en) | Method and device for detecting crowd gathered region | |
| CN104978567A (en) | Vehicle detection method based on scenario classification | |
| CN103310194A (en) | Method for detecting head and shoulders of pedestrian in video based on overhead pixel gradient direction | |
| CN103745197B (en) | A kind of detection method of license plate and device | |
| CN105447859A (en) | Field wheat aphid counting method | |
| CN106371148A (en) | Millimeter wave image-based human body foreign substance detection method and system | |
| CN104268595B (en) | general object detection method and system | |
| CN107480585A (en) | Object detection method based on DPM algorithms | |
| CN103065163B (en) | A kind of fast target based on static images detects recognition system and method | |
| CN104008404B (en) | Pedestrian detection method and system based on significant histogram features | |
| CN103106414A (en) | Detecting method of passer-bys in intelligent video surveillance | |
| CN104299005A (en) | Head detection method and system | |
| CN104123714A (en) | Optimal target detection scale generation method in people flow statistics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 518055 Baoan, Guangdong, Shiyan with Beverly science and Technology Innovation Park 8-6 Applicant after: SHENZHEN ZHENBANG INTELLIGENT TECHNOLOGY CO., LTD. Address before: 518000 Guangdong, Nanshan District Province, Dragon Ball Road, Longjing, the second industrial zone A building, building B, building 3, floor, building C, building 3, building five, building 3, building Applicant before: Shenzhen Zhenbang Industry Co., Ltd. |
|
| COR | Change of bibliographic data | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20150121 |