CN102214308B - Pedestrian detecting method and system - Google Patents
Pedestrian detecting method and system Download PDFInfo
- Publication number
- CN102214308B CN102214308B CN 201110127360 CN201110127360A CN102214308B CN 102214308 B CN102214308 B CN 102214308B CN 201110127360 CN201110127360 CN 201110127360 CN 201110127360 A CN201110127360 A CN 201110127360A CN 102214308 B CN102214308 B CN 102214308B
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- view data
- exp
- value
- proper vector
- 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
Images
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention is suitable for the technical field of image processing and provides a pedestrian detecting method, a pedestrian detecting system and a pedestrian detecting device. The method comprises the following steps of: receiving input image data and acquiring the characteristic vector of the image data; calculating the characteristic value d(x) of the input image data according to a formula d(x)=1g((exp(a*f(x))+exp(a*g(x)))/a), wherein f(x)=w.x; g(x)=u.x; w and u are preset vector constants; a is a preset constant, '.' is a vector dot product operation symbol; and x is the characteristicvector of the acquired image data; and comparing the calculated characteristic value of the image data with a preset threshold value, and outputting information whether a moving object is a pedestrian. The problems that the pedestrian cannot be adaptively detected by the conventional pedestrian detecting system when the pedestrian is sheltered and a detecting effect is poor are solved, so that the pedestrian can be detected by adaptively selecting a corresponding detecting function when the pedestrian exists in the input image data, and a universal, robust and precise pedestrian detecting method is provided for users.
Description
Technical field
The invention belongs to technical field of image processing, relate in particular to a kind of pedestrian detection method, system.
Background technology
Pedestrian detection is the important branch of object detection, is the forward position direction and the study hotspot that receive much concern of computer vision field in recent years.In fields such as video monitoring, automobile safety system, shopping behavioural analyses important application is arranged.
Yet, since some intrinsic characteristics of human body, the complicacy of application scenarios, and influencing each other between person to person or the human and environment is so that the detection and tracking of human body become the difficult point of studying in the computer vision research field.Whether when the pedestrian that existing pedestrian detecting system is blocked in detection and the pedestrian who is not blocked, can't be blocked according to the pedestrian and take adaptive detection mode to carry out pedestrian's detection, be subject to larger impact so that it detects performance, the detection effect is poor.
Summary of the invention
The purpose of the embodiment of the invention is to provide a kind of pedestrian detection method, system, being intended to solve because whether existing pedestrian detecting system can't be blocked according to the pedestrian takes adaptive detection mode to carry out pedestrian's detection, be subject to larger impact so that it detects performance, detect the poor problem of effect.
The embodiment of the invention is achieved in that a kind of pedestrian detection method, and described method comprises the steps:
Receive the view data of input, obtain the proper vector of view data;
According to formula
d(x)=lg((exp(a*f(x))+exp(a*g(x)))/a)
Calculate the eigenwert d (x) of the view data of input, f (x)=wx wherein, g (x)=ux, w and u are the dot product sign of operation for default vectorial constant, a for default constant 10, x is the proper vector of the view data obtained;
Whether the eigenwert of the view data that calculates and default threshold value, exporting described moving object is pedestrian's information if being compared;
Wherein, the value of described w and u arranges by following manner:
Head, shoulder and whole body to pedestrian in the view data that comprises the pedestrian that collects mark;
The background image data that collects and the view data that comprises the pedestrian are stored in the view data training set of setting up in advance;
Set in advance the initial value of vectorial constant w and u, take described training set as field of definition, obtain function lg ((1+exp (a* (1-y*d (x))))/a) value of w and u when obtaining minimum value by the optimal value computing module, wherein a is default constant, x is the proper vector of the view data of input, y equals 1 or-1, when y=1, and d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a), x
1Be the proper vector of the pedestrian's that marks among the x whole body, x
2For the pedestrian's that marks among the x head and the proper vector of shoulder, when y=-1, d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a);
The value of w and u is set to the function lg (value of (1+exp (a* (1-y*d (x))))/when a) obtaining minimum value w and u.
Another purpose of the embodiment of the invention is to provide a kind of pedestrian detecting system, and described system comprises:
The proper vector acquiring unit is used for receiving the view data of inputting, and obtains the proper vector of view data;
The eigenwert computing unit, be used for (the eigenwert d (x) of the view data of (exp (a*f (x))+exp (a*g (x)))/a) calculating input according to formula d (x)=lg, f (x)=wx wherein, g (x)=ux, w and u are default vectorial constant, a is the dot product sign of operation for default constant 10, and x is the proper vector of the view data obtained; And
Information output unit compares for the eigenwert of the view data that will calculate and default threshold value, and whether export described moving object is pedestrian's information;
Wherein, the value of described w and u arranges by following manner:
Head, shoulder and whole body to pedestrian in the view data that comprises the pedestrian that collects mark;
The background image data that collects and the view data that comprises the pedestrian are stored in the view data training set of setting up in advance;
Set in advance the initial value of vectorial constant w and u, take described training set as field of definition, obtain function lg ((1+exp (a* (1-y*d (x))))/a) value of w and u when obtaining minimum value by the optimal value computing module, wherein a is default constant, x is the proper vector of the view data of input, y equals 1 or-1, when y=1, and d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a), x
1Be the proper vector of the pedestrian's that marks among the x whole body, x
2For the pedestrian's that marks among the x head and the proper vector of shoulder, when y=-1, d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a);
The value of w and u is set to the function lg (value of (1+exp (a* (1-y*d (x))))/when a) obtaining minimum value w and u.
The embodiment of the invention is by receiving the proper vector of the view data of inputting, ((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input according to formula d (x)=lg, the eigenwert of the view data that calculates and default threshold value are compared, whether the output movement object is pedestrian's information, whether solved existing pedestrian detecting system can't be blocked according to the pedestrian and take adaptive detection mode to carry out pedestrian's detection, detect the poor problem of effect, thereby in the view data of input when having the pedestrian by selecting adaptively corresponding detection function to detect the pedestrian, for the user provides a kind of general, healthy and strong, accurate pedestrian detection method.
Description of drawings
Fig. 1 is the realization flow figure of the pedestrian detection method that provides of first embodiment of the invention;
Fig. 2 is the realization flow figure of the pedestrian detection method that provides of second embodiment of the invention;
Fig. 3 is the structural drawing of the pedestrian detecting system that provides of third embodiment of the invention;
Fig. 4 is the structural drawing of the pedestrian detecting system that provides of fourth embodiment of the invention.
Embodiment
In order to make purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, is not intended to limit the present invention.
The embodiment of the invention is by receiving the proper vector of the view data of inputting, ((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input according to formula d (x)=lg, the eigenwert of the view data that calculates and default threshold value are compared, whether the output movement object is pedestrian's information, whether solved existing pedestrian detecting system can't be blocked according to the pedestrian and take adaptive detection mode to carry out pedestrian's detection, detect the poor problem of effect, thereby in the view data of input when having the pedestrian by selecting adaptively corresponding detection function to detect the pedestrian, for the user provides a kind of general, healthy and strong, accurate pedestrian detection method.
Below in conjunction with specific embodiment specific implementation of the present invention is described in detail:
Embodiment one:
Fig. 1 shows the realization flow of the pedestrian detection method that first embodiment of the invention provides, and details are as follows:
In step S101, receive the view data of input, obtain the proper vector of view data.
In embodiments of the present invention, in the time need to carrying out pedestrian detection to the image of input, can carry out pre-service to image in advance, for example, gray level image to input carries out size normalization, avoid affecting follow-up processing because of the distortion of image, stretching by gray scale strengthens picture contrast, by cutting apart of background and object in the binary conversion treatment realization image.Adopt dynamic thresholding method to determine the crucial threshold value of image binaryzation, the adaptive neighborhood method of average removal of images interference that the use band is revised and noise etc., thus image blurring, the crooked or damaged situation that image causes because of reasons such as weather or acquisition angles solved.
In step S 102, ((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input according to formula d (x)=lg, f (x)=wx wherein, g (x)=ux, w and u are default vectorial constant, a is the dot product sign of operation for default constant, and x is the proper vector of the view data obtained.
In embodiments of the present invention, adopt function softmax (f (x), g (x)) calculates the eigenwert of the view data of input, softmax (f (x) wherein, g (x))=lg ((exp (a*f (x))+exp (a*g (x)))/a), f (x)=wx, be the pedestrian detection function based on whole body information, g (x)=ux be one based on the pedestrian detection function of head and shoulder, w and u are default vectorial constant, a is default constant, be the dot product sign of operation, x is the proper vector of the view data obtained, but because softmax function softening output valve, poor between reducing to be worth so that the computing formula of d (x) can be selected corresponding detection function (based on whole body information or based on a shoulder information) adaptively, detects the pedestrian in the input picture exactly.
In step S103, the eigenwert of the view data that calculates and default threshold value are compared, whether output image data comprises pedestrian's information.
In embodiments of the present invention, the eigenwert of the view data that calculates and default threshold value are compared, when the eigenwert of the view data that calculates during greater than default threshold value, output image data comprises pedestrian's information, when the eigenwert of the view data that calculates was not more than default threshold value, output image data did not comprise pedestrian's information.In concrete implementation process, can mark out according to user's the demand information pedestrian to the view data that comprises the pedestrian, thus prompting user more intuitively, or be used for pedestrian's tracking and statistics etc.
Embodiment two:
In embodiments of the present invention, for the eigenwert d (x) that determines view data in the pedestrian detection process=lg (w that relates in (exp (a*f (x))+exp (a*g (x)))/a) and value of u, should train this function in advance, obtain the value of w and u.
Fig. 2 shows the realization flow of the pedestrian detection method that second embodiment of the invention provides, and details are as follows:
In step S201, head, shoulder and the whole body of pedestrian in the view data that comprises the pedestrian that collects marked.
Exist in the invention process, acquisition of image data is trained detection function being used in advance, determine the value of vector parameter w and u, the image that gathers should comprise background image and the image that includes pedestrian's whole body, and the quantity of image can gather according to the user image of respective numbers to the accuracy requirement that detects.Head, shoulder and whole body comprising pedestrian in the view data of pedestrian's whole body are marked, specifically can be by marking manually or automatically, when manually marking, at first mark out the encirclement frame of pedestrian's head, based on certain ratio, head is surrounded frame expand, as the encirclement frame of head-shoulders, further head is surrounded the encirclement frame that frame is expanded into whole body.Also can adopt corresponding algorithm identified to go out pedestrian's head, shoulder and whole body, realize the mark of pedestrian's head, shoulder and whole body, concrete mask method need not limit the present invention at this.
In step S203, the background image data that collects and the view data that comprises the pedestrian stored into set up in advance in the view data training set.
In step S203, set in advance the initial value of vector parameter w and u, take training set as field of definition, obtain function softmax (0 by the optimal value computing module, 1-y*d (x)) value of w and u when obtaining minimum value, wherein softmax (0,1-y*d (x))=lg ((1+exp (a* (1-y*d (x))))/a), a is default constant, x is the proper vector of the view data of input, y equals 1 or-1, when y=1, and d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a), x
1Be the proper vector of the pedestrian's that marks among the x whole body, x
2For the pedestrian's that marks among the x head and the proper vector of shoulder, when y=-1, d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a).
In embodiments of the present invention, set in advance the initial value of vector parameter w and u, with the input of the image in the training set as function, obtain function lg ((1+exp (a* (1-y*d (x))))/a) value of w and u when obtaining minimum value by the optimal value computing module.Particularly, after receiving the image that comprises pedestrian's whole body of input, obtain whole Characteristic of Image vector x, and according to markup information or according to the value (y=1) of variable y, obtain head and shoulder characteristic of correspondence vector x among the x
2, and pedestrian's whole body characteristic of correspondence vector x
1, according to formula d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a) find the solution the value of d (x), when y=-1, illustrate that this image is not for comprising pedestrian's image, ((exp (a*f (x))+exp (a*g (x)))/a) finds the solution the value of d (x) according to formula d (x)=lg, thereby in function lg ((1+exp (a* (1-y*d (x))))/obtain when a) obtaining minimum value value of w and u, more preferably, a can be made as 10, in specific implementation process, the optimal value computing module can adopt degree of passing descent algorithm, genetic algorithm or neural network etc., at this not in order to limit the present invention.
In step S204, the value of w and u value is set to the function lg (value of (1+exp (a* (1-y*d (x))))/when a) obtaining minimum value.
In embodiments of the present invention, with ((1+exp (a* (1-y*d (x))))/w that obtains when a) obtaining minimum value and the value of u are made as the value of parameter w and u at function lg by the optimal value computing module among the step S203.
In step S205, the employing aspect ratio is that the window of 8:3 scans the image data information that synthetic image is corresponding to input picture to be detected.
In embodiments of the present invention, after the training of finishing detection function, obtained the value of w and u, thereby can realize the detection to input picture to be detected, particularly, can adopt typical aspect ratio is that the window of 8:3 scans the image data information that synthetic image is corresponding to input picture to be detected.In addition, also can before detection, detection function be tested, determine according to test result whether also needs are trained function, adjust the value of w and u, thereby obtain more high-precision pedestrian detection effect.
In step S206, receive the view data of input, obtain the proper vector of view data.
In step S207, ((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input according to formula d (x)=lg, f (x)=wx wherein, g (x)=ux, w and u are default vectorial constant, a is the dot product sign of operation for default constant, and x is the proper vector of the view data obtained.
In step S208, the eigenwert of the view data that calculates and default threshold value are compared, export the information whether described view data comprises the pedestrian.
One of ordinary skill in the art will appreciate that all or part of step that realizes in above-described embodiment method is to come the relevant hardware of instruction to finish by program, described program can be stored in the computer read/write memory medium, described storage medium is such as ROM/RAM, disk, CD etc.
Embodiment three:
Fig. 3 shows the structure of the pedestrian detecting system that third embodiment of the invention provides, and for convenience of explanation, only shows the part relevant with the embodiment of the invention.
This pedestrian detecting system can be for the Video Monitoring Terminal with pedestrian detection function, in video monitor etc., it can be the software unit that runs in these Video Monitoring Terminals, also can be used as independently, suspension member is integrated in these Video Monitoring Terminals or runs in the application system of these Video Monitoring Terminals, wherein:
Proper vector acquiring unit 31 receives the view data of input, obtains the proper vector of view data.
In embodiments of the present invention, in the time need to carrying out pedestrian detection to the image of input, can carry out pre-service to image in advance, for example, gray level image to input carries out size normalization, avoid affecting follow-up processing because of the distortion of image, stretching by gray scale strengthens picture contrast, by cutting apart of background and object in the binary conversion treatment realization image.Adopt dynamic thresholding method to determine the crucial threshold value of image binaryzation, the adaptive neighborhood method of average removal of images interference that the use band is revised and noise etc., thus image blurring, the crooked or damaged situation that image causes because of reasons such as weather or acquisition angles solved.
((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input to eigenwert computing unit 32 according to formula d (x)=lg, f (x)=wx wherein, g (x)=ux, w and u are default vectorial constant, a is default constant, be the dot product sign of operation, x is the proper vector of the view data obtained.
In embodiments of the present invention, adopt function softmax (f (x), g (x)) calculates the eigenwert of the view data of input, softmax (f (x), g (x))=lg ((exp (a*f (x))+exp (a*g (x)))/a), f (x)=wx wherein, be used for the pedestrian detection function based on whole body information, g (x)=ux, be used for the pedestrian detection function based on head and shoulder, w and u are default vectorial constant, a is default constant, be the dot product sign of operation, x is the proper vector of the view data obtained, but because softmax function softening output valve, poor between reducing to be worth, so that the computing formula of d (x) can be selected corresponding detection function (based on whole body information or based on a shoulder information) adaptively, detect exactly the pedestrian in the input picture.
In embodiments of the present invention, the eigenwert of the view data that calculates and default threshold value are compared, when the eigenwert of the view data that calculates during greater than default threshold value, output image data comprises pedestrian's information, when the eigenwert of the view data that calculates was not more than default threshold value, output image data did not comprise pedestrian's information.In concrete implementation process, can mark out according to user's the demand information pedestrian to the view data that comprises the pedestrian, thus prompting user more intuitively, or be used for pedestrian's tracking and statistics etc.
Embodiment four:
Fig. 4 shows the structure of the pedestrian detecting system that fourth embodiment of the invention provides, and for convenience of explanation, only shows the part relevant with the embodiment of the invention.
Pedestrian's head, shoulder and whole body mark in the view data that comprises the pedestrian that 41 pairs of unit of mark collect.
Parameter value acquiring unit 43 sets in advance the initial value of vector parameter w and u, take training set as field of definition, obtain function lg ((1+exp (a* (1-y*d (x))))/a) value of w and u when obtaining minimum value by the optimal value computing module, wherein a is default constant, x is the proper vector of the view data of input, y equals 1 or-1, when y=1, and d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a), x
1Be the proper vector of the pedestrian's that marks among the x whole body, x
2For the pedestrian's that marks among the x head and the proper vector of shoulder, when y=-1, d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a).
The value of parameter value setting unit 44 w and u value is set to parameter value acquiring unit 43 in function lg ((1+exp (a* (1-y*d (x))))/w that obtains when a) obtaining minimum value and value of u.
Image data information generation unit 45 employing aspect ratios are that the window of 8:3 scans the image data information that synthetic image is corresponding to input picture to be detected.
Proper vector acquiring unit 46 receives the view data of input, obtains the proper vector of view data.
((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input to eigenwert computing unit 47 according to formula d (x)=lg, f (x)=wx wherein, g (x)=ux, w and u are default vectorial constant, a is default constant, be the dot product sign of operation, x is the proper vector of the view data obtained.
The embodiment of the invention is by to d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a) train, the value of w and u when obtaining it and obtaining minimum value, thereby for the user provides an effective detection function, when receiving input image data to be detected, obtain this Characteristic of Image vector, ((exp (a*f (x))+exp (a*g (x)))/a) calculates the eigenwert d (x) of the view data of input according to formula d (x)=lg, the eigenwert of the view data that calculates and default threshold value are compared, whether the output movement object is pedestrian's information, whether solved existing pedestrian detecting system can't be blocked according to the pedestrian and take adaptive detection mode to carry out pedestrian's detection, detect the poor problem of effect, thereby in the view data of input when having the pedestrian by selecting adaptively corresponding detection function to detect the pedestrian, for the user provides a kind of general, healthy and strong, accurate pedestrian detection method.
The above only is preferred embodiment of the present invention, not in order to limiting the present invention, all any modifications of doing within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.
Claims (5)
1. a pedestrian detection method is characterized in that, described method comprises the steps:
Receive the view data of input, obtain the proper vector of view data;
According to formula
d(x)=lg((exp(a*f(x))+exp(a*g(x)))/a)
Calculate the eigenwert d (x) of the view data of input, f (x)=wx wherein, g (x)=ux, w and u are the dot product sign of operation for default vectorial constant, a for default constant 10, x is the proper vector of the view data obtained;
The eigenwert of the view data that calculates and default threshold value are compared, export the information whether described view data comprises the pedestrian;
Wherein, the value of described w and u arranges by following manner:
Head, shoulder and whole body to pedestrian in the view data that comprises the pedestrian that collects mark;
The background image data that collects and the view data that comprises the pedestrian are stored in the view data training set of setting up in advance;
Set in advance the initial value of vectorial constant w and u, take described training set as field of definition, obtain function lg ((1+exp (a* (1-y*d (x))))/a) value of w and u when obtaining minimum value by the optimal value computing module, wherein a is default constant, x is the proper vector of the view data of input, y equals 1 or-1, when y=1, and d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a), x
1Be the proper vector of the pedestrian's that marks among the x whole body, x
2For the pedestrian's that marks among the x head and the proper vector of shoulder, when y=-1, d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a);
The value of w and u is set to the function lg (value of (1+exp (a* (1-y*d (x))))/when a) obtaining minimum value w and u.
2. the method for claim 1 is characterized in that, the view data of described reception input is obtained before the step of proper vector of view data, and described method also comprises:
The employing aspect ratio is that the window of 8:3 scans the image data information that synthetic image is corresponding to input picture to be detected.
3. the method for claim 1 is characterized in that, described eigenwert with the view data that calculates compares with default threshold value, exports the step whether described view data comprise pedestrian's information and is specially:
When the eigenwert of the view data that calculates during greater than default threshold value, export the information that described view data comprises the pedestrian;
When the eigenwert of the view data that calculates is not more than default threshold value, export the information that described view data does not comprise the pedestrian.
4. a pedestrian detecting system is characterized in that, described system comprises:
The proper vector acquiring unit is used for receiving the view data of inputting, and obtains the proper vector of view data;
The eigenwert computing unit, be used for (the eigenwert d (x) of the view data of (exp (a*f (x))+exp (a*g (x)))/a) calculating input according to formula d (x)=lg, f (x)=wx wherein, g (x)=ux, w and u are default vectorial constant, a is the dot product sign of operation for default constant 10, and x is the proper vector of the view data obtained; And
Information output unit compares for the eigenwert of the view data that will calculate and default threshold value, exports the information whether described view data comprises the pedestrian;
Wherein, the value of described w and u arranges by following manner:
Head, shoulder and whole body to pedestrian in the view data that comprises the pedestrian that collects mark;
The background image data that collects and the view data that comprises the pedestrian are stored in the view data training set of setting up in advance;
Set in advance the initial value of vectorial constant w and u, take described training set as field of definition, obtain function lg ((1+exp (a* (1-y*d (x))))/a) value of w and u when obtaining minimum value by the optimal value computing module, wherein a is default constant, x is the proper vector of the view data of input, y equals 1 or-1, when y=1, and d (x)=lg ((exp (a*f (x
1))+exp (a*g (x
2)))/a), x
1Be the proper vector of the pedestrian's that marks among the x whole body, x
2For the pedestrian's that marks among the x head and the proper vector of shoulder, when y=-1, d (x)=lg ((exp (a*f (x))+exp (a*g (x)))/a);
The value of w and u is set to the function lg (value of (1+exp (a* (1-y*d (x))))/when a) obtaining minimum value w and u.
5. system as claimed in claim 4 is characterized in that, described system also comprises:
The image data information generation unit, be used for adopting aspect ratio is that the window of 8:3 scans the image data information that synthetic image is corresponding to input picture to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110127360 CN102214308B (en) | 2011-05-17 | 2011-05-17 | Pedestrian detecting method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110127360 CN102214308B (en) | 2011-05-17 | 2011-05-17 | Pedestrian detecting method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102214308A CN102214308A (en) | 2011-10-12 |
| CN102214308B true CN102214308B (en) | 2013-04-24 |
Family
ID=44745608
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110127360 Active CN102214308B (en) | 2011-05-17 | 2011-05-17 | Pedestrian detecting method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102214308B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975925B (en) * | 2016-05-03 | 2019-05-03 | 电子科技大学 | Partially occluded pedestrian detection method based on joint detection model |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1963839A (en) * | 2002-12-06 | 2007-05-16 | 三星电子株式会社 | Detection method and apparatus of human |
| CN101290682A (en) * | 2008-06-25 | 2008-10-22 | 北京中星微电子有限公司 | Movement target checking method and apparatus |
| CN101887518A (en) * | 2010-06-17 | 2010-11-17 | 北京交通大学 | Human body detection device and method |
-
2011
- 2011-05-17 CN CN 201110127360 patent/CN102214308B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1963839A (en) * | 2002-12-06 | 2007-05-16 | 三星电子株式会社 | Detection method and apparatus of human |
| EP1426898B1 (en) * | 2002-12-06 | 2007-05-23 | Samsung Electronics Co., Ltd. | Human detection through face detection and motion detection |
| CN101290682A (en) * | 2008-06-25 | 2008-10-22 | 北京中星微电子有限公司 | Movement target checking method and apparatus |
| CN101887518A (en) * | 2010-06-17 | 2010-11-17 | 北京交通大学 | Human body detection device and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102214308A (en) | 2011-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11120270B2 (en) | Spatiotemporal action detection method | |
| CN107886048B (en) | Target tracking method and system, storage medium and electronic terminal | |
| EP3651055A1 (en) | Gesture recognition method, apparatus, and device | |
| CN106709928B (en) | fast two-dimensional maximum inter-class variance threshold method for noisy images | |
| EP3968280A1 (en) | Target tracking method and apparatus, storage medium and electronic device | |
| CN105593903B (en) | Organism identification device, measurement device and method | |
| CN103054569A (en) | Method, device and handhold device for measuring human body heart rate based on visible image | |
| KR20180109665A (en) | A method and apparatus of image processing for object detection | |
| US9280703B2 (en) | Apparatus and method for tracking hand | |
| CN111368596B (en) | Face recognition backlight compensation method and device, readable storage medium and equipment | |
| Chen et al. | WiFace: Facial expression recognition using Wi-Fi signals | |
| Huang et al. | PhaseAnti: An anti-interference WiFi-based activity recognition system using interference-independent phase component | |
| CN110533166A (en) | It is a kind of based on when space fusion feature indoor orientation method | |
| CN113269010B (en) | Training method and related device for human face living body detection model | |
| CN105786189A (en) | Finger independent action recognition method and system based on MMG signal | |
| CN110062379A (en) | Identity identifying method based on channel state information under a kind of human body behavior scene | |
| CN109640255B (en) | Location fingerprint indoor positioning system and method | |
| CN104679967B (en) | A kind of method for judging psychological test reliability | |
| CN117648620A (en) | WIFI fall monitoring algorithm combined with human body posture assessment | |
| CN104598914A (en) | Skin color detecting method and device | |
| CN114220125A (en) | Pedestrian detection method, device, equipment and medium | |
| CN112613564A (en) | Target detection post-processing method for eliminating overlapped frames | |
| CN102214308B (en) | Pedestrian detecting method and system | |
| CN110688969A (en) | Video frame human behavior identification method | |
| Chen et al. | WiTT: Modeling and the evaluation of table tennis actions based on WIFI signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| ASS | Succession or assignment of patent right |
Owner name: XIAMEN RECONOVA INFORMATION TECHNOLOGY CO., LTD. Free format text: FORMER OWNER: ZHAN DONGHUI Effective date: 20141114 |
|
| C41 | Transfer of patent application or patent right or utility model | ||
| COR | Change of bibliographic data |
Free format text: CORRECT: ADDRESS; FROM: 518000 SHENZHEN, GUANGDONG PROVINCE TO: 361008 XIAMEN, FUJIAN PROVINCE |
|
| TR01 | Transfer of patent right |
Effective date of registration: 20141114 Address after: View of the road Xiamen city Fujian province 361008 Software Park No. 22 Room 102, unit D17 Patentee after: Xiamen Reconova Information Technology Co., Ltd. Address before: Shenzhen City, Guangdong Province, 518000 Shennan Road No. 12069, block 2914 kraft Nuodong coast Patentee before: Zhan Donghui |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Pedestrian detection method and system Effective date of registration: 20150923 Granted publication date: 20130424 Pledgee: China Co truction Bank Corp Xiamen branch Pledgor: Xiamen Reconova Information Technology Co., Ltd. Registration number: 2015350000076 |
|
| PLDC | Enforcement, change and cancellation of contracts on pledge of patent right or utility model |