CN103839038A - People counting method and device - Google Patents
People counting method and device Download PDFInfo
- Publication number
- CN103839038A CN103839038A CN201210485382.2A CN201210485382A CN103839038A CN 103839038 A CN103839038 A CN 103839038A CN 201210485382 A CN201210485382 A CN 201210485382A CN 103839038 A CN103839038 A CN 103839038A
- Authority
- CN
- China
- Prior art keywords
- head region
- people
- image
- depth
- pixel
- 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
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the field of video monitoring, and especially relates to a people counting method and device. The objective of the invention is to solve the problem that the existing head detection algorithm only utilizes head contour information, so round objects which are similar to the head shape, the head shadow and the like are easy to be determined as the head region by mistake, so that the head region determination accuracy can be improved. According to the embodiments of the invention, the following method is adopted, the method is that a depth value in a corresponding depth map is acquired through a head region detected in an inputted image, the acquired depth value is compared with a depth threshold value of a preset real head region, and the detected region is judged whether to be a real head region according to the comparison result. Three-dimensional information is adopted, and the depth value of the detected head region is compared with the depth threshold value of the set head region, so possibility that the round objects in the image which are similar to the head shape, the head shadow and the like are easy to be determined as the head region by mistake due to the reason that only the head contour information is utilized can be reduced, and the head region determination accuracy can be improved, so that the people counting accuracy can be improved.
Description
Technical field
The present invention relates to field of video monitoring, relate in particular to a kind of method and device of demographics.
Background technology
The today of day by day improving in information system management level, carry out the volume of the flow of passengers for huge places of flow of the people such as supermarket, market, station, banks and estimate in real time, the volume of the flow of passengers statistics such as passenger flow distributional analysis, degree of crowding estimation become the effective way that first-hand background information is provided for public domain management.
General passenger number statistical system capable, is that video camera is fixed on to certain commanding elevation, and the import and exports such as alignment lens market, station carry out a series of processing to the image sequence obtaining, thereby counts the flow of the people in this region in certain period.Most number of people detection algorithm is all to realize by the method for machine learning, take the machine learning method based on Boosting as example, machine is by the study of Boosting Algorithm for Training, how study identifies number of people profile information, be trained to as number of people detecting device, then image is input on the detecting device training, scan differentiation by number of people profile information, obtain people's head region, people's head region is preserved and exported, finally provide a number flow under current scene in conjunction with trace information.
The subject matter of number of people detection algorithm is at present, just utilize the profile information of the number of people, more therefore round shape object and the number of people shade etc. similar to number of people shape is easily people's head region by wrong differentiation in image, thereby reduced the accuracy rate of demographics.
Summary of the invention
The embodiment of the present invention provides a kind of method and device of demographics, and to solve current number of people detection algorithm, the profile information of the number of people just utilizing judges people's head region, and the accuracy rate of number of people regional determination is not high, causes the inaccurate problem of demographics.
The embodiment of the present invention provides a kind of method of demographics, comprising:
The image synchronously obtaining according to two cameras, determines corresponding depth map;
Any piece image in the described synchronous images obtaining is carried out to number of people detection, determine everyone head region detecting in this width image;
For a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region;
The number of any piece image in the synchronous images obtaining described in determining according to the quantity of real people's head region;
Wherein, the image section that two cameras obtain respectively or all overlapping.
The embodiment of the present invention provides a kind of device of demographics, comprising:
Depth map acquisition module, for the image synchronously obtaining according to the camera of two, determines the depth map that a width is corresponding;
Surveyed area acquisition module, carries out number of people detection for any piece image to the described synchronous images obtaining, and determines everyone head region detecting in this width image;
Discrimination module, be used for for a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region;
Demographics module, the number of any piece image of the synchronous images obtaining described in determining for the quantity of the real people's head region of basis;
Wherein, the image section that two cameras obtain respectively or all overlapping.
The embodiment of the present invention adopts the image synchronously obtaining according to two cameras, determines the depth map that input picture is corresponding; Any piece image in the synchronous images obtaining is carried out to number of people detection, determine everyone head region detecting in this width image; For a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region; The method of the number of any piece image in the synchronous images obtaining described in determining according to the quantity of real people's head region, reduce the profile information that just utilizes the number of people, the possibility that easily round shape object and number of people shade etc. similar to number of people shape in image is mistaken for to people's head region, has improved the accuracy rate of number of people regional determination and the accuracy rate of demographics.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of a kind of method of demographics in the embodiment of the present invention;
Fig. 2 is the template schematic diagram that calculates the gradient intensity of this image different directions in the embodiment of the present invention;
Fig. 3 is the schematic flow sheet of the specific implementation method of demographics in the embodiment of the present invention;
Fig. 4 is the device schematic diagram of a kind of demographics in the embodiment of the present invention.
Embodiment
The embodiment of the present invention adopts the image synchronously obtaining according to two cameras, determines the depth map that input picture is corresponding; Any piece image in the synchronous images obtaining is carried out to number of people detection, determine everyone head region detecting in this width image; For a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region; The method of the number of any piece image in the synchronous images obtaining described in determining according to the quantity of real people's head region.Owing to utilizing three-dimensional information, the depth threshold of real people's head region of the depth information of the people's head region detecting and setting is compared, reduce the profile information that just utilizes the number of people, the possibility that easily round shape object and number of people shade etc. similar to number of people shape in image is mistaken for to people's head region, has improved the accuracy rate of number of people regional determination and the accuracy rate of demographics.
Below in conjunction with Figure of description, the embodiment of the present invention is described in further detail.
As shown in Figure 1, the method for a kind of demographics of the embodiment of the present invention, comprises the following steps:
Wherein, the image section that two cameras obtain respectively or all overlapping.
Preferably, in step 101, the two width images that synchronously obtain according to two parallel cameras arranged side by side, determine the depth map that a width is corresponding.In the embodiment of the present invention, carry out demographics as example take two parallel cameras arranged side by side, the method that plural camera carries out demographics is similar with it, does not repeat them here.
Preferably, in step 101, adopt following method to obtain depth map:
The two width images that two parallel cameras are obtained simultaneously, carry out images match calculating, obtain the optimum matching image of the two width images that two parallel cameras obtain simultaneously; By the parallax of each pixel calculating by coupling in optimum matching image, be converted into depth value, the pixel value in optimum matching image is replaced with to corresponding depth value, just obtain the depth map of this matching image.
Wherein, the acquisition methods of depth image is not limited only to the method for foregoing description, and the acquisition methods of other depth images, as obtained depth map by structured light, is also applicable to the present invention.
Wherein, the algorithm of images match includes but not limited to following method: Region Matching Algorithm, global registration algorithm.
Take Region Matching Algorithm as example, the concrete grammar that obtains depth map is as follows below:
A) calculate respectively the distinction of each pixel of the two width images that two parallel cameras obtain simultaneously, computing method are as follows:
Wherein, g
cthe average gray of central area, g
pbe the average gray in eight fields, p represents eight fields of 0-7.
In the two width images that two parallel cameras obtain simultaneously, the eigenwert f of each pixel of left-side images
leftrepresent the eigenwert f of each pixel of image right
rightrepresent.
Two parallel cameras, include but not limited to following parallel mode: parallel up and down, left and right is parallel.The present invention is in two modes that camera left and right is parallel, describes the method applied in the present invention, and the image-region that two parallel cameras obtain is identical.Other parallel modes of two parallel cameras, the mode parallel with two camera left and right is identical, does not repeat them here.
B) as required, calculate successively the value of minimum mean square distance SSD (u, v) under the different degrees of overlapping of left and right two width images, finally obtain the optimum matching image of two width images.The computing method of simple square error SSD (u, v) are as follows:
Wherein, f
left(i, j) refers to the distinction value of left-side images (i, j) point, f
rigt(i, j) refer in image right and left-side images (i, j) some same position (i, j) point distinction value.
Calculate SSD (u under the different degrees of overlapping of left and right two width images, v) value, can suppose that two width images have N row pixel overlapping, SSD (u when calculating N row pixel is overlapping, v) value, after the value of the overlapping SSD (u, v) of N row pixel has been calculated, calculate the value of the overlapping SSD (u, v) of N+X row pixel; The whole SSD (u, v) calculating is compared, and SSD (u, v) value is less, and the matching degree of two width images is better, and SSD (u, v) is worth to the optimum matching image of minimum matching image as two width images.Wherein N, X is positive integer.
As: suppose that in two width images, every width image has 40 pixel columns, calculate respectively two width image one row pixels SSD (u when overlapping, v) value, the value of ten row pixels SSD (u, v) when overlapping, 20 row pixels are SSD (u when overlapping, v) value, the value of 30 row pixels SSD (u, v) when overlapping, and the complete value of SSD (u, v) when overlapping of two width images; Wherein 30 row pixels SSD (u when overlapping, v) value minimum, by 30 row pixels of the two width doublings of the image, 5 row pixels adjacent with overlapping 30 row pixels left sides in left-side images, the optimum matching image of the 5 row pixels common composition two width images adjacent with 30 overlapping row pixel right sides in image right.
Preferably, according to the optimum matching of the two width images that obtain simultaneously, the method for obtaining depth map is as follows:
First according to images match, obtain the corresponding relation between two width images; Then corresponding relation and the principle of triangulation of utilizing two width images, calculate disparity map; Obtain the point in each pixel corresponding three-dimensional space according to projective transformation matrix, then calculate three dimensions and put the distance of baseline, the inverse of this distance is exactly the depth value of this pixel, depth value corresponding to the pixel value of matching image changes into, has just obtained depth map.Wherein baseline is exactly the line of corresponding point position difference in the two width images that obtain while calculating parallax.
Preferably, in step 102, any piece image in the two width images that obtain is carried out to number of people detection, while determining all people's head region detecting in this width image, can be according to existing any one number of people detection algorithm, realize the number of people and detect, as, the methods such as AdaBoost, SVM, Bayes and artificial neural network, this is no longer going to repeat them.
In step 102, the people's head region detecting refers to the people head region definite according to existing any one number of people detection algorithm, in the embodiment of the present invention, the depth threshold of depth value corresponding the people's head region detecting and predefined real people's head region further need to be compared, thereby determine whether the people's head region detecting is real people's head region, or round shape object and the number of people shade similar to number of people shape.
Take AdaBoost method as example, set forth the preparation method of the people's head region detecting below.
Any piece image in the two width images that selection is obtained, calculates the gradient intensity of this image different directions, uses 4 templates in Fig. 2, respectively input picture is carried out to convolution, calculate the gradient intensity image of 4 directions, 4 gradient intensity figure are calculated simultaneously, obtain integrogram.
Select the optional position of this width image, set big or small window as initial detection window using one, use sorter to differentiate the image in this window, determine whether window area is the people's head region detecting; After this windows detecting has been differentiated, move detection window according to moving direction and the moving step length set, continue to use sorter to differentiate.
Wherein, use sorter to sentence method for distinguishing to the image in detection window, for adopting eigenwert to differentiate; In the time that eigenwert is greater than predefined eigenvalue threshold, judge that this window area is as the people's head region detecting; Otherwise determinating area is the non-people's head region detecting.The computing method of eigenwert are as follows:
Wherein W
irepresent the weight that each region is corresponding, sum
irepresent direction gradient intensity corresponding to each region and, can realize fast by inquiring about the above-mentioned integrogram calculating.
Wherein, eigenvalue threshold is definite in the time that this sorter is trained, and in the time that sorter adopts identical training method, the eigenvalue threshold obtaining is identical.
Preferably, select the optional position of this width image, set big or small window as initial detection window using one, use sorter to differentiate the image in this window; After this width image discriminating completes, carry out convergent-divergent by setting big or small window according to zoom factor, use the image in the detection window of sorter after to convergent-divergent to differentiate; After this width image discriminating completes, repeat to carry out convergent-divergent by setting big or small window according to zoom factor, the step that image in detection window after using sorter to convergent-divergent is differentiated, when wide or the tall and big wide and higher position in image of detection window stops convergent-divergent, and finish the differentiation to this width image.
Take detection window as 20*20, zoom factor is 1.15 below, and step-length 2 pixels are example, determines that the concrete grammar of people's head region to be detected is as follows.
The detection window of a 20*20 is moved to the upper left corner of this width image, use sorter to differentiate the image in detection window, after current detection window has been differentiated, judge whether current window place pixel column has all detected, if all detected, detection window moves to next pixel column, proceeds to detect; If this detection window place pixel column has not detected, detection window 2 pixel columns that move right, proceed to detect.In the time that this width image detection is complete, by zoom factor 1.15 amplification detection windows, repeat operation above, in the time that the size of detection window is greater than this width image wide or high, process ends.
Preferably, in step 103, for a people's head region detecting, the depth threshold of the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map and predefined real people's head region is compared, judge that according to comparative result whether this people's head region detecting is that real people's head region has two kinds of methods, is introduced respectively below.
The depth value of method one, the band of position is the depth value of the pixel in the band of position;
For a people's head region detecting, by the depth value of the each pixel in the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region respectively, the pixel that corresponding depth value is greater than to the depth threshold of predefined real people's head region is preserved, and the quantity of the pixel of statistics preservation, if the quantity of the pixel of preserving is greater than the pixel quantity of real people's head region of setting, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.
Wherein, the depth threshold of predefined real people's head region can arrange as required, because people has certain height in real world, therefore, people also has certain degree of depth on depth map, if the depth value of most of pixel of the people's head region detecting is less, can judge that this region is as overhead nearer object, can judge that this people's head region detecting is fict people's head region.When in people's head region that this detects, while there is the larger situation of the depth value of partial pixel, whether the quantity that needs judgement to be greater than the pixel of the depth threshold of predefined real people's head region is not less than the pixel quantity of real people's head region, as the pixel quantity of people's head region being greater than to 1/4 of pixel quantity in surveyed area, just determine that this people's head region detecting is real people's head region, otherwise, judge that this people's head region detecting is as fict people's head region.
The depth value of method two, the band of position is the mean value of the depth value of the part or all of pixel of the band of position
For a people's head region detecting, by the mean value of the depth value of the part or all of pixel in the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, if the mean value of the depth value of corresponding part or all of pixel is greater than the depth threshold of predefined real people's head region, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.
Wherein, when the depth value of the band of position is the mean value of depth value of the partial pixel of the band of position, can be according to predefined system of selection, select the partial pixel of the people's head region detecting of this width image, also can select at random the partial pixel of the people's head region detecting of this width image, the mean value of the depth value of the partial pixel of people's head region that calculating detects, the mean value of the depth value of partial pixel and the depth threshold that presets real people's head region are compared, judge whether this people's head region detecting is real people's head region.
As shown in Figure 3, be the specific implementation method of demographics in the embodiment of the present invention.
Step 301: two parallel cameras synchronously obtain two width images;
Step 302: the optimum matching image that obtains two width images according to Region Matching Algorithm;
Step 303: according to the parallax of the pixel of two width images match in optimum matching image, obtain the depth value of optimum matching image pixel;
Step 304: according to the depth value that obtains optimum matching image pixel, determine the depth map of present frame, then perform step 306;
Step 305: the whole people's head region that detect of determining any piece image in the two width images that two parallel cameras synchronously obtain;
Step 306: for a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region, if real people's head region, execution step 307, otherwise, execution step 308.
Step 307: real people's head region is preserved;
Step 308: whether whole people's head region that judgement detects have all detected, if complete, perform step 309, otherwise return to step 306;
Step 309: the number of determining any piece image in this synchronous images obtaining according to the quantity of real people's head region.
Wherein, in step 301, two parallel cameras, for the same area, synchronously obtain two width images.
In step 302, the computing method of the optimum matching image of the two width images that obtain have multiple, include but not limited to following computing method: Region Matching Algorithm, global registration algorithm.
In step 304, if present image is the first two field picture, the pixel value of the optimum matching image of acquisition is replaced to corresponding depth value, just obtain the depth map of present image; If present image is not the first two field picture,, by the depth value of present image, compare with the depth value of former frame depth map same position, if not identical, the depth value of former frame depth map same position is replaced with to the depth value of present image, obtain the depth map of present image.
In step 305, determine that the method for people's head region of detecting has a lot, may be used in the embodiment of the present invention, the people's head region detecting definite, this is no longer going to repeat them.
In step 306, for a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, have two kinds of methods, specifically, referring to described in step 103, this is no longer going to repeat them.
Based on same inventive concept, a kind of device of demographics is also provided in the embodiment of the present invention, because the principle that this device is dealt with problems is similar to the method for a kind of demographics of the embodiment of the present invention, therefore the enforcement of this device can be referring to the enforcement of method, repeats part and repeat no more.
As shown in Figure 4, be the device of a kind of demographics of the embodiment of the present invention, comprising:
Depth map acquisition module 401, for the image synchronously obtaining according to two cameras, determines the depth map that a width is corresponding;
Region to be detected acquisition module 402, carries out number of people detection for any piece image to the synchronous images obtaining, everyone head region of determining that all in this width image detect;
Preferably, depth map acquisition module 401, specifically for the two width images that synchronously obtain according to two parallel cameras arranged side by side, determines the depth map that a width is corresponding.In the embodiment of the present invention, carry out demographics as example take two parallel cameras arranged side by side, it is similar with it that plural camera carries out the device of demographics, do not repeat them here.
Wherein, depth map acquisition module 401 specifically for, the two width images that two parallel cameras are obtained simultaneously, carry out images match calculating, obtain the optimum matching image of the two width images that two parallel cameras obtain simultaneously.In optimum matching image there is certain displacement difference (being parallax) in the pixel of each coupling on two width images, the parallax of the pixel of each coupling is converted into the depth value of this matched pixel.Its circular can, referring to described in step 101, not repeat them here.
In surveyed area acquisition module 402, can be according to existing any one number of people detection algorithm, realize the number of people and detect, do not repeat them here.
In discrimination module 403, judge whether the people's head region detecting is that real people's head region has two kinds of differentiation algorithms, is described respectively below.
The depth value of method one, the band of position is the depth value of the pixel in the band of position;
For a people's head region detecting, by the depth value of the each pixel in the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region respectively, the pixel that corresponding depth value is greater than to the depth threshold of predefined real people's head region is preserved, and the quantity of the pixel of statistics preservation, if the quantity of the pixel of preserving is greater than the pixel quantity of real people's head region of setting, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.Concrete grammar, referring to step 103, does not repeat them here.
The depth value of method two, the band of position is the mean value of the depth value of the part or all of pixel of the band of position
For a people's head region detecting, by the mean value of the depth value of the part or all of pixel in the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, if the mean value of the depth value of corresponding part or all of pixel is greater than the depth threshold of predefined real people's head region, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.Concrete grammar, referring to step 103, does not repeat them here.
Preferably, this device also comprises image collection module 410, stores for the two width images that two parallel cameras are obtained simultaneously.
Those skilled in the art should understand, embodiments of the invention can be provided as method, system or computer program.Therefore, the present invention can adopt complete hardware implementation example, completely implement software example or the form in conjunction with the embodiment of software and hardware aspect.And the present invention can adopt the form at one or more upper computer programs of implementing of computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) that wherein include computer usable program code.
The present invention is with reference to describing according to process flow diagram and/or the block scheme of the method for the embodiment of the present invention, equipment (system) and computer program.Should understand can be by the flow process in each flow process in computer program instructions realization flow figure and/or block scheme and/or square frame and process flow diagram and/or block scheme and/or the combination of square frame.Can provide these computer program instructions to the processor of multi-purpose computer, special purpose computer, Embedded Processor or other programmable data processing device to produce a machine, the instruction that makes to carry out by the processor of computing machine or other programmable data processing device produces the device for realizing the function of specifying at flow process of process flow diagram or multiple flow process and/or square frame of block scheme or multiple square frame.
These computer program instructions also can be stored in energy vectoring computer or the computer-readable memory of other programmable data processing device with ad hoc fashion work, the instruction that makes to be stored in this computer-readable memory produces the manufacture that comprises command device, and this command device is realized the function of specifying in flow process of process flow diagram or multiple flow process and/or square frame of block scheme or multiple square frame.
These computer program instructions also can be loaded in computing machine or other programmable data processing device, make to carry out sequence of operations step to produce computer implemented processing on computing machine or other programmable devices, thereby the instruction of carrying out is provided for realizing the step of the function of specifying in flow process of process flow diagram or multiple flow process and/or square frame of block scheme or multiple square frame on computing machine or other programmable devices.
Although described the preferred embodiments of the present invention, once those skilled in the art obtain the basic creative concept of cicada, can make other change and modification to these embodiment.So claims are intended to be interpreted as comprising preferred embodiment and fall into all changes and the modification of the scope of the invention.
Obviously, those skilled in the art can carry out various changes and modification and not depart from the spirit and scope of the present invention the present invention.Like this, if within of the present invention these are revised and modification belongs to the scope of the claims in the present invention and equivalent technologies thereof, the present invention is also intended to comprise these changes and modification interior.
Claims (12)
1. a method for demographics, is characterized in that, the method comprises:
The image synchronously obtaining according to two cameras, determines the depth map that input picture is corresponding;
Any piece image in the described synchronous images obtaining is carried out to number of people detection, determine everyone head region detecting in this width image;
For a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region;
The number of any piece image in the synchronous images obtaining described in determining according to the quantity of real people's head region;
Wherein, the image that two cameras obtain respectively needs partly or entirely overlapping.
2. the method for claim 1, is characterized in that, the image synchronously obtaining according to two cameras is determined corresponding depth map, comprising:
According to two picked-up direction images that parallel camera synchronously obtains side by side, determine corresponding depth map.
3. method as claimed in claim 1 or 2, is characterized in that, the depth value of the described band of position is the depth value of the pixel in the band of position;
Judge that according to comparative result whether this people's head region detecting is real people's head region, comprising:
The depth value of each pixel in the band of position is compared with the depth threshold of predefined real people's head region respectively;
If corresponding depth value is greater than the quantity that the quantity of the pixel of the depth threshold of predefined real people's head region is greater than setting, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.
4. method as claimed in claim 1 or 2, is characterized in that, the mean value of the depth value of the part or all of pixel that the depth value of the described band of position is the described band of position;
Judge that according to comparative result whether this people's head region detecting is real people's head region, comprising:
By in definite depth map with this width image detection to the mean value of depth value and the depth threshold of predefined real people's head region of part or all of pixel of the identical band of position of people's head region compare, if the mean value of the depth value of corresponding part or all of pixel is greater than predefined real number of people regional depth threshold value, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.
5. method as claimed in claim 1 or 2, is characterized in that, the image synchronously obtaining according to two cameras is determined corresponding depth map, comprising:
The two width images that synchronously obtain according to two cameras, carry out images match by the synchronous images obtaining and obtain matching image;
Determine the depth value of the each pixel in matching image;
The pixel value of the each pixel in matching image is converted into depth value corresponding to this pixel, obtains depth image.
6. method as claimed in claim 1 or 2, is characterized in that, any piece image in the described synchronous images obtaining is carried out to number of people detection, comprising:
Any piece image in the described synchronous images obtaining is changed into characteristic pattern, this characteristic pattern is carried out to number of people detection.
7. a device for demographics, is characterized in that, this device comprises:
Depth map acquisition module, for the two width images that obtain according to two cameras simultaneously, determines corresponding depth map;
Surveyed area acquisition module, carries out number of people detection for any piece image to the described synchronous images obtaining, and determines everyone head region detecting in this width image;
Discrimination module, be used for for a people's head region detecting, by the depth value of the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region, judge according to comparative result whether this people's head region detecting is real people's head region;
Demographics module, the number of any piece image of the synchronous images obtaining described in determining for the quantity of the real people's head region of basis;
Wherein, the image section that two cameras obtain respectively or all overlapping.
8. device as claimed in claim 7, is characterized in that, depth map acquisition module specifically for:
According to two picked-up direction images that parallel camera synchronously obtains side by side, determine corresponding depth map.
As claimed in claim 7 or 8 device, it is characterized in that, described discrimination module specifically for:
For a people's head region detecting, by the depth value of each pixel in the band of position identical with the people's head region detecting of this width image in definite depth map, compare with the depth threshold of predefined real people's head region respectively; If corresponding depth value is greater than the quantity that the quantity of the pixel of the depth threshold of predefined real people's head region is greater than setting, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.
As claimed in claim 7 or 8 device, it is characterized in that, described discrimination module specifically for:
For a people's head region detecting, by the mean value of the depth value of part or all of pixel in the band of position identical with the people's head region detecting of this width image in definite depth map, starve to such an extent that depth threshold compares with predefined real people's head region; If the mean value of the depth value of corresponding part or all of pixel is greater than the depth threshold of predefined real people's head region, judge that this people's head region detecting is real people's head region; Otherwise, judge that this people's head region detecting is not real people's head region.
11. as claimed in claim 7 or 8 device, it is characterized in that, described depth map acquisition module specifically for:
The image synchronously obtaining according to two cameras, carries out images match by the synchronous images obtaining and obtains matching image; Determine the depth value of the each pixel in matching image; The pixel value of the each pixel in matching image is converted into depth value corresponding to this pixel, obtains depth image.
12. as claimed in claim 7 or 8 device, it is characterized in that, described surveyed area acquisition module specifically for:
Any piece image in the described synchronous images obtaining is changed into characteristic pattern, this characteristic pattern is carried out to number of people detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210485382.2A CN103839038A (en) | 2012-11-23 | 2012-11-23 | People counting method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210485382.2A CN103839038A (en) | 2012-11-23 | 2012-11-23 | People counting method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103839038A true CN103839038A (en) | 2014-06-04 |
Family
ID=50802517
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210485382.2A Pending CN103839038A (en) | 2012-11-23 | 2012-11-23 | People counting method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103839038A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104809787A (en) * | 2015-04-23 | 2015-07-29 | 中电科安（北京）系统集成有限公司 | Intelligent passenger flow volume statistics device based on camera |
| CN105291111A (en) * | 2015-11-27 | 2016-02-03 | 深圳市神州云海智能科技有限公司 | Patrol robot |
| CN106127812A (en) * | 2016-06-28 | 2016-11-16 | 中山大学 | A kind of passenger flow statistical method of non-gate area, passenger station based on video monitoring |
| CN106295557A (en) * | 2016-08-05 | 2017-01-04 | 浙江大华技术股份有限公司 | A kind of method and device of crowd density estimation |
| CN106485735A (en) * | 2015-09-01 | 2017-03-08 | 南京理工大学 | Human body target recognition and tracking method based on stereovision technique |
| CN106504227A (en) * | 2016-09-26 | 2017-03-15 | 深圳奥比中光科技有限公司 | Demographic method and its system based on depth image |
| CN106530310A (en) * | 2016-10-25 | 2017-03-22 | 深圳大学 | Pedestrian counting method and device based on human head top recognition |
| CN104103077B (en) * | 2014-07-29 | 2017-04-19 | 浙江宇视科技有限公司 | Human head detecting method and human head detecting device |
| CN106778617A (en) * | 2016-12-17 | 2017-05-31 | 天津工业大学 | A kind of sleep monitor method and system based on acquisition technology |
| CN107481255A (en) * | 2017-08-08 | 2017-12-15 | 浙江大华技术股份有限公司 | A kind of determination method and device of personnel amount |
| CN108069307A (en) * | 2016-11-14 | 2018-05-25 | 杭州海康威视数字技术股份有限公司 | The method and device that a kind of number in elevator is counted |
| CN108701211A (en) * | 2015-12-28 | 2018-10-23 | 罗伯特·博世有限公司 | For detecting, tracking, estimating and identifying the system based on depth sense occupied in real time |
| CN110555419A (en) * | 2019-09-09 | 2019-12-10 | 江苏慧眼数据科技股份有限公司 | passenger flow counting method based on binocular stereo vision |
| CN110781748A (en) * | 2019-09-24 | 2020-02-11 | 上海商米科技集团股份有限公司 | IPC-based image processing method and camera |
| CN112712619A (en) * | 2020-12-18 | 2021-04-27 | 中国平安人寿保险股份有限公司 | Method and device for managing and controlling passenger flow of subway station, computer equipment and storage medium |
| CN113963318A (en) * | 2021-12-22 | 2022-01-21 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1731456A (en) * | 2005-08-04 | 2006-02-08 | 浙江大学 | Bus passenger traffic statistical method based on stereoscopic vision and system therefor |
| CN101976353A (en) * | 2010-10-28 | 2011-02-16 | 北京智安邦科技有限公司 | Statistical method and device of low density crowd |
| CN102376242A (en) * | 2010-08-06 | 2012-03-14 | 乐金显示有限公司 | Stereoscopic image display and method for driving the same |
| CN102521646A (en) * | 2011-11-11 | 2012-06-27 | 浙江捷尚视觉科技有限公司 | Complex scene people counting algorithm based on depth information cluster |
-
2012
- 2012-11-23 CN CN201210485382.2A patent/CN103839038A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1731456A (en) * | 2005-08-04 | 2006-02-08 | 浙江大学 | Bus passenger traffic statistical method based on stereoscopic vision and system therefor |
| CN102376242A (en) * | 2010-08-06 | 2012-03-14 | 乐金显示有限公司 | Stereoscopic image display and method for driving the same |
| CN101976353A (en) * | 2010-10-28 | 2011-02-16 | 北京智安邦科技有限公司 | Statistical method and device of low density crowd |
| CN102521646A (en) * | 2011-11-11 | 2012-06-27 | 浙江捷尚视觉科技有限公司 | Complex scene people counting algorithm based on depth information cluster |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104103077B (en) * | 2014-07-29 | 2017-04-19 | 浙江宇视科技有限公司 | Human head detecting method and human head detecting device |
| CN104809787A (en) * | 2015-04-23 | 2015-07-29 | 中电科安（北京）系统集成有限公司 | Intelligent passenger flow volume statistics device based on camera |
| CN106485735A (en) * | 2015-09-01 | 2017-03-08 | 南京理工大学 | Human body target recognition and tracking method based on stereovision technique |
| CN105291111A (en) * | 2015-11-27 | 2016-02-03 | 深圳市神州云海智能科技有限公司 | Patrol robot |
| CN108701211B (en) * | 2015-12-28 | 2023-09-12 | 罗伯特·博世有限公司 | Depth sensing based system for detecting, tracking, estimating and identifying occupancy in real time |
| CN108701211A (en) * | 2015-12-28 | 2018-10-23 | 罗伯特·博世有限公司 | For detecting, tracking, estimating and identifying the system based on depth sense occupied in real time |
| CN106127812B (en) * | 2016-06-28 | 2018-10-12 | 中山大学 | A kind of passenger flow statistical method of the non-gate area in passenger station based on video monitoring |
| CN106127812A (en) * | 2016-06-28 | 2016-11-16 | 中山大学 | A kind of passenger flow statistical method of non-gate area, passenger station based on video monitoring |
| CN106295557B (en) * | 2016-08-05 | 2019-09-17 | 浙江大华技术股份有限公司 | A kind of method and device of crowd density estimation |
| CN106295557A (en) * | 2016-08-05 | 2017-01-04 | 浙江大华技术股份有限公司 | A kind of method and device of crowd density estimation |
| CN106504227A (en) * | 2016-09-26 | 2017-03-15 | 深圳奥比中光科技有限公司 | Demographic method and its system based on depth image |
| CN106504227B (en) * | 2016-09-26 | 2019-01-15 | 深圳奥比中光科技有限公司 | Demographic method and its system based on depth image |
| CN106530310B (en) * | 2016-10-25 | 2018-02-23 | 深圳大学 | A kind of pedestrian count method and device based on the identification of human body overhead |
| WO2018076392A1 (en) * | 2016-10-25 | 2018-05-03 | 深圳大学 | Pedestrian statistical method and apparatus based on recognition of parietal region of human body |
| CN106530310A (en) * | 2016-10-25 | 2017-03-22 | 深圳大学 | Pedestrian counting method and device based on human head top recognition |
| JP2018535457A (en) * | 2016-10-25 | 2018-11-29 | シェンチェン ユニバーシティー | Statistical method and apparatus for passersby based on identification of human head top |
| CN108069307A (en) * | 2016-11-14 | 2018-05-25 | 杭州海康威视数字技术股份有限公司 | The method and device that a kind of number in elevator is counted |
| CN106778617A (en) * | 2016-12-17 | 2017-05-31 | 天津工业大学 | A kind of sleep monitor method and system based on acquisition technology |
| CN107481255A (en) * | 2017-08-08 | 2017-12-15 | 浙江大华技术股份有限公司 | A kind of determination method and device of personnel amount |
| CN107481255B (en) * | 2017-08-08 | 2019-12-20 | 浙江大华技术股份有限公司 | Method and device for determining number of people |
| CN110555419A (en) * | 2019-09-09 | 2019-12-10 | 江苏慧眼数据科技股份有限公司 | passenger flow counting method based on binocular stereo vision |
| CN110555419B (en) * | 2019-09-09 | 2023-05-26 | 江苏慧眼数据科技股份有限公司 | Passenger flow counting method based on binocular stereoscopic vision |
| CN110781748A (en) * | 2019-09-24 | 2020-02-11 | 上海商米科技集团股份有限公司 | IPC-based image processing method and camera |
| CN112712619A (en) * | 2020-12-18 | 2021-04-27 | 中国平安人寿保险股份有限公司 | Method and device for managing and controlling passenger flow of subway station, computer equipment and storage medium |
| CN113963318A (en) * | 2021-12-22 | 2022-01-21 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
| CN113963318B (en) * | 2021-12-22 | 2022-03-25 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103839038A (en) | People counting method and device | |
| Jafari et al. | Real-time RGB-D based people detection and tracking for mobile robots and head-worn cameras | |
| CN111487641B (en) | Method and device for detecting object by using laser radar, electronic equipment and storage medium | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| Benenson et al. | Stixels estimation without depth map computation | |
| CN103310194B (en) | Pedestrian based on crown pixel gradient direction in a video shoulder detection method | |
| CN103177247B (en) | A kind of object detection method merging various visual angles information | |
| CN104517095B (en) | A kind of number of people dividing method based on depth image | |
| CN106022345B (en) | A kind of high voltage isolator state identification method based on Hough forest | |
| CN102982334B (en) | The sparse disparities acquisition methods of based target edge feature and grey similarity | |
| CN105260749B (en) | Real-time target detection method based on direction gradient binary pattern and soft cascade SVM | |
| CN103208008A (en) | Fast adaptation method for traffic video monitoring target detection based on machine vision | |
| CN104751146B (en) | A kind of indoor human body detection method based on 3D point cloud image | |
| CN104318263A (en) | Real-time high-precision people stream counting method | |
| CN105457908A (en) | Sorting and quick locating method and system for small-size glass panels on basis of monocular CCD | |
| CN105760846A (en) | Object detection and location method and system based on depth data | |
| CN102855758A (en) | Detection method for vehicle in breach of traffic rules | |
| Maier et al. | Vision-based humanoid navigation using self-supervised obstacle detection | |
| CN103593679A (en) | Visual human-hand tracking method based on online machine learning | |
| CN108196680B (en) | Robot vision following method based on human body feature extraction and retrieval | |
| CN103021183A (en) | Method for detecting regulation-violating motor vehicles in monitoring scene | |
| CN104063692A (en) | Method and system for pedestrian positioning detection | |
| CN102609724A (en) | Method for prompting ambient environment information by using two cameras | |
| CN103150552A (en) | Driving training management method based on people counting | |
| CN110889347A (en) | Density traffic flow counting method and system based on space-time counting characteristics |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140604 |