CN102930287B - A kind of detection number system and method for overlooking pedestrian - Google Patents
A kind of detection number system and method for overlooking pedestrian Download PDFInfo
- Publication number
- CN102930287B CN102930287B CN201210364963.0A CN201210364963A CN102930287B CN 102930287 B CN102930287 B CN 102930287B CN 201210364963 A CN201210364963 A CN 201210364963A CN 102930287 B CN102930287 B CN 102930287B
- Authority
- CN
- China
- Prior art keywords
- sub
- block
- image
- gradient
- 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.)
- Expired - Fee Related
Links
Abstract
Present invention is disclosed a kind of detection number system and method for overlooking pedestrian, described system comprises: image unit, detecting unit, training unit, judgement and counting unit.Image unit is arranged at the upper end, top of gateway, in order to take the pedestrian passed through.Detecting unit comprises square target frame setting module, slip detection window (i.e. square box) image gradient acquisition module, any pixel gradient direction value acquisition module of image, sub-block pixel gradient directional statistics module, sub-block normalized module, Gradient direction information overlapped in series module.The present invention propose for detection number system and the method for overlooking pedestrian, substantially there is not the problem of blocking between target, thus the degree of accuracy of detection can be improved.
Description
Technical field
The invention belongs to technical field of image processing, relate to a kind of pedestrian detection number system, particularly relating to a kind of detection number system for overlooking pedestrian; Meanwhile, the invention still further relates to a kind of detection method of counting for overlooking pedestrian.
Background technology
Nowadays, the progress of science and technology at any time and the requirement of social development, need to arrange pedestrian detection number system in a lot of place.The number system of traditional sense mainly utilizes sensor to carry out, and degree of accuracy has much room for improvement.Meanwhile, along with the progressively development of image procossing counting, start to occur utilizing video camera and image processing techniques to calculate the method for pedestrian's number.
The existing method of image processing techniques detect lines people number that utilizes looks squarely pedestrian by video camera to detect, and the pedestrian detected is trunk and the four limbs of vertical hinge, and pedestrian below is easily blocked by pedestrian above, and accuracy of detection is not high.
In view of this, nowadays in the urgent need to designing a kind of new pedestrian detection number system, to improve the degree of accuracy detected.
Summary of the invention
Technical matters to be solved by this invention is: providing a kind of detection number system for overlooking pedestrian, owing to substantially there is not the problem of blocking between target, thus can improve the degree of accuracy of detection.
In addition, the present invention also provides a kind of detection method of counting for overlooking pedestrian, substantially there is not the problem of blocking between target, thus can improve the degree of accuracy of detection.
For solving the problems of the technologies described above, the present invention adopts following technical scheme:
For the detection number system overlooking pedestrian, described system comprises:
Image unit, is arranged at the upper end, top of gateway, in order to take the pedestrian passed through, uses the camera better effects if of band wide-angle lens;
Detecting unit, comprising:
-square target frame setting module, in order to set a square target frame, i.e. square box, square box is made to confine the region of the number of people at overhead view image, the size of square box is determined according to the actual conditions detecting target, change with camera setting height(from bottom) difference, the rule of delineation is: the number of people overlooked is positioned in the middle of window, surrounding is stayed at regular intervals, slip detection window is exactly this above-mentioned square box, detection window is divided into 3 × 3 totally 9 square sub-blocks, between sub-block with sub-block, has half area overlapping; If establish the length of detection window and wide be w respectively, then the length of side of square sub-block is w/2;
-slip detection window (i.e. square box) image gradient acquisition module, in order to by the greyscale image data of square-shaped frame and template [-1,0,1] and [-1,0,1]
tdo convolution, obtain any pixel of image at x direction gradient value d
x(x, y) and y direction gradient value d
y(x, y);
Any pixel gradient direction value acquisition module of-image, in order to obtain the gradient direction value θ (x, y) of any pixel (x, y) of image:
-sub-block pixel gradient directional statistics module, in order to add up the gradient direction of each sub-block pixel: calculate the gradient direction of all pixels in sub-block and drop on the interval accumulated value of 9 of-180 ° ~ 180 ° of scopes respectively, each interval is the scope at 40 ° of angles; It is band weighting coefficient that the gradient direction of pixel adds up, and namely the gradient direction value of each pixel will be multiplied by a coefficient, and this coefficient equals the absolute value sum of this pixel x direction and y direction gradient; Each sub-block, according to from the statistical value in 9 interval gradient directions, obtains a row or column vector; Namely each component of each sub-block vector is exactly that the gradient direction of this sub-block pixel drops on 9 interval statistical values respectively;
-sub-block normalized module, in order to each sub-block of normalized; If v represents not normalized vector, || v|| represents the single order norm of vector, and e is a little constant, then normalized formula is v=v/ (|| v||+e);
-Gradient direction information overlapped in series module, in order to the Gradient direction information overlapped in series successively by 9 sub-blocks inside a square detection block, forms the vector of one 81 dimension; The i.e. Feature Conversion of each moving window (i.e. square box) vector about gradient direction statistics that becomes one 81 to tie up.
As a preferred embodiment of the present invention, described system comprises further:
Training unit, comprise the sample of target in a large number in order to gather from the environment of reality, to extract and do not comprise the samples pictures of target, the sample comprising target is positive sample, and the sample not comprising target is negative sample; Described detecting unit is utilized to extract proper vector to all samples pictures; The size of samples pictures must be the same with the size of slip detection window; Then, adopt support vector machine to make sorter, by the proper vector of positive and negative samples, it is trained, obtain the parameter of sorter, namely obtain the discrimination model of sorter;
Data acquisition and detecting unit, the picture that the picture in order to utilize described detecting unit sliding window swept-volume newly to gather newly gathers scene, detects forms to set step-length interlacing or every arranging successively scanned picture with slip;
Judge and counting unit, in order to the proper vector of picture region detected after extracting sliding window, and be input in the middle of sorter and judge, judge whether containing target; If so, then count, and continue scanning by step-length; If not, then continue by step scan.
As a preferred embodiment of the present invention, described training unit also comprises greyscale image transitions module, in order to just first to convert coloured image to gray level image.
As a preferred embodiment of the present invention, described data acquisition and detecting unit slide and detect forms with step-length is 5 pixel interlacing or every arranging successively scanned picture.
As a preferred embodiment of the present invention, described data acquisition and detecting unit only detect 2 ~ 3 frame pictures each second, according to actual needs each region detecting only detection picture 1/3 ~ 1/2.
The above-mentioned method of counting of detection number system for overlooking pedestrian, described method comprises the steps:
Image pickup step: by the pedestrian of image unit shooting by setting regions;
Detecting step: comprising:
-square target frame setting procedure: set a square-shaped frame by square target frame setting module, square-shaped frame is made to confine the region of the number of people at overhead view image, the size of square-shaped frame is determined according to the actual conditions detecting target, change with camera setting height(from bottom) difference, the rule of delineation is: the number of people overlooked is positioned in the middle of window, surrounding is stayed at regular intervals, slip detection window is exactly this above-mentioned square-shaped frame, detection window is divided into 3 × 3 totally 9 little square sub blocks, between block with block, has half area overlapping; If establish the length of side of detection window to be w and h, then the length of side of little square sub-block is w/2;
-slip detection window image gradient obtaining step: by slip detection window image gradient acquisition module by the greyscale image data of square-shaped frame and template [-1,0,1] and [-1,0,1]
tdo convolution, obtain any pixel of image at x direction gradient value d
x(x, y) and y direction gradient value d
y(x, y);
Any pixel gradient direction value obtaining step of-image: the gradient direction value θ (x, y) being obtained any pixel (x, y) of image by any pixel gradient direction value acquisition module of image:
-sub-block pixel gradient directional statistics step: the gradient direction being added up each sub-block pixel by sub-block pixel gradient directional statistics module: calculate the gradient direction of all pixels in sub-block and drop on the interval accumulated value of 9 of-180 ° ~ 180 ° of scopes respectively, each interval is the scope at 40 ° of angles; It is band weighting coefficient that the gradient direction of pixel adds up, and namely the gradient direction value of each pixel will be multiplied by a coefficient, and this coefficient equals the absolute value sum of this pixel x direction and y direction gradient; Each sub-block, according to from 9 interval gradient direction statistical values, obtains a row or column vector; Namely each component of each sub-block vector is exactly that the gradient direction of this sub-block pixel drops on 9 interval statistical values respectively;
-sub-block normalized step: by each sub-block of sub-block normalized module normalized; If v represents not normalized vector, || v|| represents the single order norm of vector, and e is a little constant, then normalized formula is v=v/ (|| v||+e);
-Gradient direction information overlapped in series step: by the Gradient direction information successively overlapped in series of Gradient direction information overlapped in series module by 9 sub-blocks inside a square-shaped frame, forms the vector of one 81 dimension; The i.e. Feature Conversion of each sliding window vector about gradient direction statistics that becomes one 81 to tie up.
As a preferred embodiment of the present invention, described method comprises further:
Training step: collection from the environment of reality, extraction comprise the sample of target in a large number and do not comprise the samples pictures of target, and the sample comprising target is positive sample, and the sample not comprising target is negative sample; Described detecting unit is utilized to extract proper vector to all samples pictures; The size of samples pictures must be the same with the size of slip detection window; Then, adopt support vector machine to make sorter, by the proper vector of positive and negative samples, it is trained, obtain the parameter of sorter, namely obtain the discrimination model of sorter;
Data acquisition and detecting step: the picture that the picture utilizing described detecting unit sliding window swept-volume newly to gather newly gathers scene, detect forms to set step-length interlacing or every arranging successively scanned picture with slip;
Judge and counting step: the proper vector extracting the picture region detected after sliding window, and be input in the middle of sorter and judge, judge whether containing target; If so, then count, and continue scanning by step-length; If not, then continue by step scan.
As a preferred embodiment of the present invention, described training step also comprises greyscale image transitions step, just first converts coloured image to gray level image.
As a preferred embodiment of the present invention, in described data acquisition and detecting step, detecting forms with step-length with slip is 3 ~ 5 pixel interlacing or every arranging successively scanned picture.
As a preferred embodiment of the present invention, in described data acquisition and detecting step, only detect 2 ~ 3 frame pictures each second, detect the region only detecting picture 1/3 ~ 1/2 at every turn according to actual needs.
Beneficial effect of the present invention is: the present invention propose for detection number system and the method for overlooking pedestrian, substantially there is not the problem of blocking between target, thus the degree of accuracy of detection can be improved.
The present invention is directed the external appearance characteristic overlooking the homogeneous characteristic sum edge contour of pedestrian head field color designs: head zone color is homogeneous, means that Grad is low, and the Grad of head edge then reflects the change of target wheel profile.Based on the gradient direction distribution of regional area, embody its importance by the number of times repeating to add up.As the region of core, comprise the edge of the number of people, the distribution characteristics of its gradient direction repeats at most, have 4 times; The feature at four angles only occurs 1 time; Remainder occurs 2 times.That is, the gradient direction statistical nature of middle sub-image is crucial.Such design is the target signature highlighting detection, weakens again the impact of periphery background.
Accompanying drawing explanation
Fig. 1 is slip detection window inner structure and feature schematic diagram.
Fig. 2 is the process flow diagram of the inventive method detecting step.
Fig. 3 is the process flow diagram of the inventive method.
Embodiment
The preferred embodiments of the present invention are described in detail below in conjunction with accompanying drawing.
Embodiment one
Present invention is disclosed a kind of detection number system for overlooking pedestrian, under the present invention utilizes monocular cam to overlook state, the carrying out of pedestrian being detected and being counted, being mainly used in the statistics to various passenger flow.
Camera is arranged on the upper end, top of public passage gateway, and camera takes the pedestrian passed through downwards, and the largest benefit of putting camera is like this exactly substantially there is not the problem of blocking between target.Camera is general apart from 3 ~ 10 meters, ground, is a kind of pedestrian detection of closer distance.With traditional sense look squarely compared with pedestrian detection, although detect congruence, detect clarification of objective obviously different, general pedestrian is trunk and the four limbs of vertical hinge, and the pedestrian in overlooking is the agglomerate of movement one by one.
Passenger flow detection method in this paper, target is identified with head feature, the region of the number of people at overhead view image is confined by a square-shaped frame, the size of square-shaped frame is determined according to the actual conditions (can change with camera setting height(from bottom) difference) detecting target, the rule of delineation is: the number of people overlooked is positioned in the middle of window, surrounding is stayed at regular intervals, as shown in Figure 1, slip detection window is exactly this above-mentioned square-shaped frame, detection window is divided into 3 × 3 totally 9 little square sub blocks, between block with block, has half area overlapping.If establish the length of side of detection window to be w, then the length of side of little square sub blocks is w/2.The view field of overlooking the number of people just should comprise the area of that little square block of (or substantially comprising) middle.
Described system comprises: image unit, detecting unit, training unit, judgement and counting unit.
Image unit is arranged at the upper end, top of gateway, in order to take the pedestrian passed through; Image unit can be video camera.
Detecting unit comprises: any pixel gradient direction value acquisition module of square target frame setting module, slip detection window image gradient acquisition module, image, sub-block pixel gradient directional statistics module, sub-block normalized module, Gradient direction information overlapped in series module.
Square-shaped frame setting module is in order to set a square-shaped frame, square-shaped frame is made to confine the region of the number of people at overhead view image, the size of square-shaped frame is determined according to the actual conditions detecting target, change with camera setting height(from bottom) difference, the rule of delineation is: the number of people overlooked is positioned in the middle of window, and surrounding is stayed at regular intervals, and slip detection window is exactly this above-mentioned square-shaped frame, detection window is divided into 3 × 3 totally 9 little square sub blocks, between sub-block with sub-block, has half area overlapping; If establish the length of detection window and wide be w respectively, then the length of side of little square sub-block is w/2.
Slip detection window image gradient acquisition module is in order to by the greyscale image data of square-shaped frame and template [-1,0,1] and [-1,0,1]
tdo convolution, obtain any pixel of image at x direction gradient value d
x(x, y) and y direction gradient value d
y(x, y).
Any pixel gradient direction value acquisition module of image is in order to obtain the gradient direction value θ (x, y) of any pixel (x, y) of image:
Sub-block pixel gradient directional statistics module is in order to add up the gradient direction of each sub-block pixel: calculate the gradient direction of all pixels in sub-block and drop on the interval accumulated value of 9 of-180 ° ~ 180 ° of scopes respectively, each interval is the scope at 40 ° of angles; It is band weighting coefficient that the gradient direction of pixel adds up, and namely the gradient direction value of each pixel will be multiplied by a coefficient, and this coefficient equals the absolute value sum of this pixel x direction and y direction gradient; Each sub-block, according to the statistical value of self 9 interval gradient direction, obtains a row or column vector; Namely each component of the corresponding vector of each sub-block is exactly that this sub-block gradient direction drops on 9 interval statistical values respectively.
Sub-block normalized module is in order to each sub-block of normalized; If v represents not normalized vector, || v|| represents the single order norm of vector, and e is a little constant, then normalized formula is v=v/ (|| v||+e).
Gradient direction information overlapped in series module, in order to the Gradient direction information overlapped in series successively by 9 sub-blocks inside a square-shaped frame, forms the vector of one 81 dimension; The i.e. Feature Conversion of each sliding window vector about gradient direction statistics that becomes one 81 to tie up.
Training unit comprises the sample of target in order to gather, to extract a large amount of (crossing a thousand sheets) from the environment of reality and does not comprise the samples pictures of target, and the sample comprising target is positive sample, and the sample not comprising target is negative sample; Described detecting unit is utilized to extract proper vector to all samples pictures; The size of samples pictures must be the same with the size of slip detection window; Then, adopt support vector machine to make sorter, by the proper vector of positive and negative samples, it is trained, obtain the parameter of sorter, namely obtain the discrimination model of sorter.Native system process be gray level image, if gathering is that coloured image just first converts gray level image to, image size is 320 × 240 pixels or less.
The picture that data acquisition and detecting unit newly gather scene in order to the picture that sliding window swept-volume newly gathers, detects forms to set step-length (if step-length is 3 ~ 5 pixels) interlacing or every arranging successively scanned picture with slip.For reducing calculated amount, only detect 2 ~ 3 frame pictures each second, according to actual needs each region detecting only detection picture 1/3 ~ 1/2.
Judge and the proper vector of picture region of counting unit in order to detect after extracting sliding window, and be input in the middle of sorter and judge, judge whether containing target; If so, then count, and continue scanning by step-length; If not, then continue by step scan.
Be described above the detection number system that the present invention is directed to and overlook pedestrian, the present invention, while announcement said system, also discloses a kind of above-mentioned method of counting of detection number system for overlooking pedestrian; Refer to Fig. 2, described method comprises:
Image pickup step: by the pedestrian of image unit shooting by setting regions.
Detecting step: comprising:
-square-shaped frame setting procedure: set a square-shaped frame by square-shaped frame setting module, square-shaped frame is made to confine the region of the number of people at overhead view image, the size of square-shaped frame is determined according to the actual conditions detecting target, change with camera setting height(from bottom) difference, the rule of delineation is: the number of people overlooked is positioned in the middle of window, surrounding is stayed at regular intervals, slip detection window is exactly this above-mentioned square-shaped frame, detection window is divided into 3 × 3 totally 9 little square block, between block with block, has half area overlapping; If establish the length of detection window and wide be w respectively, then the length of side of little square block is w/2;
-slip detection window image gradient obtaining step: by slip detection window image gradient acquisition module by the greyscale image data of square-shaped frame and template [-1,0,1] and [-1,0,1]
tdo convolution, obtain any pixel of image at x direction gradient value d
x(x, y) and y direction gradient value d
y(x, y);
Any pixel gradient direction value obtaining step of-image: the gradient direction value θ (x, y) being obtained any pixel (x, y) of image by any pixel gradient direction value acquisition module of image:
-sub-block pixel gradient directional statistics step: the gradient direction being added up each sub-block pixel by sub-block pixel gradient directional statistics module: calculate the gradient direction of all pixels in sub-block and drop on the interval accumulated value of 9 of-180 ° ~ 180 ° of scopes respectively, each interval is the scope at 40 ° of angles; It is band weighting coefficient that the gradient direction of pixel adds up, and namely the gradient direction value of each pixel will be multiplied by a coefficient, and this coefficient equals the absolute value sum of this pixel x direction and y direction gradient; Each sub-block, according to the statistical value of self 9 interval gradient direction, obtains a row or column vector; Namely each component of the corresponding vector of each sub-block is exactly that this sub-block gradient direction drops on 9 interval statistical values respectively;
-sub-block normalized step: by each sub-block of sub-block normalized module normalized; If v represents not normalized vector, || v|| represents the single order norm of vector, and e is a little constant, then normalized formula is v=v/ (|| v||+e);
-Gradient direction information overlapped in series step: by the Gradient direction information successively overlapped in series of Gradient direction information overlapped in series module by 9 sub-blocks inside a square-shaped frame, forms the vector of one 81 dimension; The i.e. Feature Conversion of each sliding window vector about gradient direction statistics that becomes one 81 to tie up.
Refer to Fig. 3, whole method of the present invention comprises:
[training step] gathers from the environment of reality, extraction comprises the sample of target in a large number and do not comprise the samples pictures of target, and the sample comprising target is positive sample, and the sample not comprising target is negative sample; Described detecting unit is utilized to extract proper vector to all samples pictures; The size of samples pictures must be the same with the size of slip detection window; Then, adopt support vector machine to make sorter, by the proper vector of positive and negative samples, it is trained, obtain the parameter of sorter, namely obtain the discrimination model of sorter;
The picture that the picture that [data acquisition and detecting step] utilizes described detecting unit sliding window swept-volume newly to gather newly gathers scene, detects forms to set step-length interlacing or every arranging successively scanned picture with slip;
The proper vector of the picture region detected after [judge and counting step] extracts sliding window, and be input in the middle of sorter and judge, judge whether containing target; If so, then count, and continue scanning by step-length; If not, then continue by step scan.
In sum, the present invention propose for detection number system and the method for overlooking pedestrian, substantially there is not the problem of blocking between target, thus the degree of accuracy of detection can be improved.
The present invention is directed the external appearance characteristic overlooking the homogeneous characteristic sum edge contour of pedestrian head field color designs: head zone color is homogeneous, means that Grad is low, and the Grad of head edge then reflects the change of target wheel profile.Based on the gradient direction distribution of regional area, embody its importance by the number of times repeating to add up.As the region of core, comprise the edge of the number of people, the distribution characteristics of its gradient direction repeats at most, have 4 times; The feature at four angles only occurs 1 time; Remainder occurs 2 times.That is, the gradient direction statistical nature of middle sub-image is crucial.Such design is the target signature highlighting detection, weakens again the impact of periphery background.
Here description of the invention and application is illustrative, not wants by scope restriction of the present invention in the above-described embodiments.Distortion and the change of embodiment disclosed are here possible, are known for the replacement of embodiment those those of ordinary skill in the art and the various parts of equivalence.Those skilled in the art are noted that when not departing from spirit of the present invention or essential characteristic, the present invention can in other forms, structure, layout, ratio, and to realize with other assembly, material and parts.When not departing from the scope of the invention and spirit, can other distortion be carried out here to disclosed embodiment and change.
Claims (8)
1. for the detection number system overlooking pedestrian, it is characterized in that, described system comprises:
Image unit, is arranged at the upper end, top of gateway, in order to take the pedestrian passed through downwards;
Detecting unit, comprising:
-square target frame setting module, in order to set a square target frame, i.e. square box, square box is made to confine the region of the number of people at overhead view image, the size of square box is determined according to the actual conditions detecting target, change with camera setting height(from bottom) difference, the rule of delineation is: the number of people overlooked is positioned in the middle of window, surrounding is stayed at regular intervals, slip detection window is exactly this above-mentioned square box, detection window is divided into 3 × 3 totally 9 little square sub blocks, between sub-block with sub-block, has half area overlapping; If establish the length of detection window and wide be w respectively, then the length of side of square sub-block is w/2;
-slip detection window image gradient acquisition module, in order to by the greyscale image data of square box and template [-1,0,1] and [-1,0,1]
tdo convolution, obtain any pixel of image at x direction gradient value d
x(x, y) and y direction gradient value d
y(x, y);
Any pixel gradient direction value acquisition module of-image, in order to obtain the gradient direction value θ (x, y) of any pixel (x, y) of image:
-sub-block pixel gradient directional statistics module, in order to add up the gradient direction of each sub-block pixel: calculate the gradient direction of all pixels in sub-block and drop on the interval accumulated value of 9 of-180 ° ~ 180 ° of scopes respectively, each interval is the scope at 40 ° of angles; The cumulative of pixel gradient direction is band weighting coefficient, and namely the gradient direction value of each pixel will be multiplied by a coefficient, and this coefficient equals the absolute value sum of this pixel x direction and y direction gradient; Each sub-block drops on 9 interval statistical values according to grading itself direction, obtains a row or column vector: namely each component of the corresponding vector of each sub-block is exactly that this sub-block gradient direction drops on 9 interval statistical values respectively;
-sub-block normalized module, in order to each sub-block of normalized; If v represents not normalized vector, || v|| represents the single order norm of vector, and e is a little constant, then normalized formula is v=v/ (|| v||+e);
-Gradient direction information overlapped in series module, in order to the Gradient direction information overlapped in series successively by 9 sub-blocks inside a square box, forms the vector of one 81 dimension; The i.e. Feature Conversion of each slip detection window vector about gradient direction statistics that becomes one 81 to tie up;
Training unit, comprise the sample of target in a large number in order to gather from the environment of reality, to extract and do not comprise the samples pictures of target, the sample comprising target is positive sample, and the sample not comprising target is negative sample; Described detecting unit is utilized to extract proper vector to all training sample pictures; The size of samples pictures must be the same with the size of slip detection window; Then, adopt support vector machine to make sorter, by the proper vector of positive and negative samples, it is trained, obtain the parameter of sorter, namely obtain the discrimination model of sorter;
Data acquisition and detecting unit, in order to the picture scanning utilizing the slip detection window of described detecting unit newly to gather scene, with slip detection window to set step-length interlacing or every arranging successively scanned picture;
Judge and counting unit, in order to extract the proper vector of the picture region that slip detection window detects, and be input in the middle of sorter and judge, judge whether containing target; If so, then count, and continue scanning by step-length; If not, then continue by step scan.
2. the detection number system for overlooking pedestrian according to claim 1, is characterized in that:
Native system only need use gray level image process, if collection is coloured image, described training unit also comprises greyscale image transitions module, in order to convert coloured image to gray level image in advance.
3. the detection number system for overlooking pedestrian according to claim 1, is characterized in that:
Described data acquisition and detecting unit slip detection window are 3 ~ 5 pixel interlacing or every arranging successively scanned picture with step-length.
4. the detection number system for overlooking pedestrian according to claim 3, is characterized in that:
Described data acquisition and detecting unit only detect 2 ~ 3 frame pictures each second, according to actual needs each region detecting only detection picture 1/3 ~ 1/2.
5. the method for counting of detection number system for overlooking pedestrian according to claim 1, it is characterized in that, described method comprises the steps:
Image pickup step: by the pedestrian of image unit shooting by setting regions;
Detecting step: comprising:
-square target frame setting procedure: set a square target frame by square target frame setting module, square box is made to confine the region of the number of people at overhead view image, the size of square box is determined according to the actual conditions detecting target, change with camera setting height(from bottom) difference, the rule of delineation is: the number of people overlooked is positioned in the middle of window, surrounding is stayed at regular intervals, slip detection window is exactly this above-mentioned square box, detection window is divided into 3 × 3 totally 9 little square sub blocks, between sub-block with sub-block, has half area overlapping; If establish the length of detection window and wide be w respectively, then little square block-length of side is w/2;
-slip detection window image gradient obtaining step: by slip detection window image gradient acquisition module by the greyscale image data of detection block and template [-1,0,1] and [-1,0,1]
tdo convolution, obtain any pixel of image at x direction gradient value d
x(x, y) and y direction gradient value d
y(x, y);
Any pixel gradient direction value obtaining step of-image: the gradient direction value θ (x, y) being obtained any pixel (x, y) of image by any pixel gradient direction value acquisition module of image:
-sub-block pixel gradient directional statistics step: the gradient direction being added up each sub-block pixel by sub-block pixel gradient directional statistics module: calculate the gradient direction of all pixels in sub-block and drop on the interval accumulated value of 9 of-180 ° ~ 180 ° of scopes respectively, each interval is the scope at 40 ° of angles; It is band weighting coefficient that the gradient direction of pixel adds up, and namely the gradient direction value of each pixel will be multiplied by a coefficient, and this coefficient equals the absolute value sum of this pixel x direction and y direction gradient; Each sub-block, according to from the statistical value in 9 interval gradient directions, obtains a row or column vector; Namely each component of the corresponding vector of each sub-block is exactly that this sub-block gradient direction drops on 9 interval statistical values respectively;
-sub-block normalized step: by each sub-block of sub-block normalized module normalized; If v represents not normalized vector, || v|| represents the single order norm of vector, and e is a little constant, then normalized formula is v=v/ (|| v||+e);
-Gradient direction information overlapped in series step: by the Gradient direction information successively overlapped in series of Gradient direction information overlapped in series module by 9 sub-blocks inside a square box, forms the vector of one 81 dimension; The i.e. Feature Conversion of each slip detection window vector about gradient direction statistics that becomes one 81 to tie up;
Training step: collection from the environment of reality, extraction comprise the sample of target in a large number and do not comprise the samples pictures of target, and the sample comprising target is positive sample, and the sample not comprising target is negative sample; Described detecting unit is utilized to extract proper vector to all samples pictures; The size of samples pictures must be the same with the size of slip detection window; Then, adopt support vector machine to make sorter, by the proper vector of positive and negative samples, it is trained, obtain the parameter of sorter, namely obtain the discrimination model of sorter;
Data acquisition and detecting step: utilize the picture scanning that the slip detection window of described detecting unit newly gathers scene, with slip detection window to set step-length interlacing or every arranging successively scanned picture;
Judge and counting step: the proper vector extracting the picture region that slip detection window detects, and be input in the middle of sorter and judge, judge whether containing target; If so, then count, and continue scanning by step-length; If not, then continue by step scan.
6. method of counting according to claim 5, is characterized in that:
Described training step also comprises greyscale image transitions step, converts coloured image to gray level image in advance.
7. method of counting according to claim 5, is characterized in that:
In described data acquisition and detecting step, be 5 pixel interlacing or every arranging successively scanned picture with slip detection window with step-length.
8. method of counting according to claim 7, is characterized in that:
In described data acquisition and detecting step, only detect 2 ~ 3 frame pictures each second, detect the region only detecting picture 1/3 ~ 1/2 at every turn according to actual needs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210364963.0A CN102930287B (en) | 2012-09-26 | 2012-09-26 | A kind of detection number system and method for overlooking pedestrian |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210364963.0A CN102930287B (en) | 2012-09-26 | 2012-09-26 | A kind of detection number system and method for overlooking pedestrian |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102930287A CN102930287A (en) | 2013-02-13 |
| CN102930287B true CN102930287B (en) | 2015-09-02 |
Family
ID=47645084
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210364963.0A Expired - Fee Related CN102930287B (en) | 2012-09-26 | 2012-09-26 | A kind of detection number system and method for overlooking pedestrian |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102930287B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103400142B (en) * | 2013-05-14 | 2016-06-01 | 上海交通大学 | A kind of pedestrian counting method |
| CN104239905A (en) * | 2013-06-17 | 2014-12-24 | 上海盖普电梯有限公司 | Moving target recognition method and intelligent elevator billing system having moving target recognition function |
| CN103473953B (en) * | 2013-08-28 | 2015-12-09 | 奇瑞汽车股份有限公司 | A kind of pedestrian detection method and system |
| CN103559478B (en) * | 2013-10-07 | 2018-12-04 | 唐春晖 | Overlook the passenger flow counting and affair analytical method in pedestrian's video monitoring |
| CN103559492A (en) * | 2013-11-12 | 2014-02-05 | 公安部第三研究所 | Car logo recognition device and method |
| CN104020848A (en) * | 2014-05-15 | 2014-09-03 | 中航华东光电（上海）有限公司 | Static gesture recognizing method |
| CN105046316B (en) * | 2015-05-12 | 2018-08-03 | 中山大学 | A kind of two-way pedestrian counting method of laser returned based on Gaussian process |
| CN106951885A (en) * | 2017-04-08 | 2017-07-14 | 广西师范大学 | A kind of people flow rate statistical method based on video analysis |
| CN109670519B (en) * | 2017-10-13 | 2023-09-26 | 佳能株式会社 | Image processing apparatus and image processing method |
| CN108388883A (en) * | 2018-03-16 | 2018-08-10 | 广西师范大学 | A kind of video demographic method based on HOG+SVM |
| CN109059863B (en) * | 2018-06-29 | 2020-09-22 | 大连民族大学 | Method for mapping track point vector of head-up pedestrian to two-dimensional world coordinate system |
| CN111899180B (en) * | 2019-05-05 | 2023-11-17 | 上海闻通信息科技有限公司 | Image key pixel direction positioning method |
| CN110472593B (en) * | 2019-08-20 | 2021-02-09 | 重庆紫光华山智安科技有限公司 | Training image acquisition method, model training method and related device |
| CN111402632B (en) * | 2020-03-18 | 2022-06-07 | 五邑大学 | Risk prediction method for pedestrian movement track at intersection |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1426898A2 (en) * | 2002-12-06 | 2004-06-09 | Samsung Electronics Co., Ltd. | Human detection through face detection and motion detection |
| CN101655910A (en) * | 2008-08-21 | 2010-02-24 | 索尼（中国）有限公司 | Training system, training method and detection method |
-
2012
- 2012-09-26 CN CN201210364963.0A patent/CN102930287B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1426898A2 (en) * | 2002-12-06 | 2004-06-09 | Samsung Electronics Co., Ltd. | Human detection through face detection and motion detection |
| CN101655910A (en) * | 2008-08-21 | 2010-02-24 | 索尼（中国）有限公司 | Training system, training method and detection method |
Non-Patent Citations (1)
| Title |
|---|
| 黄朝露."基于机器视觉的客流量统计技术研究".《中国优秀硕士学位论文全文数据库(电子期刊)》.2012,第2章至第5章. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102930287A (en) | 2013-02-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102930287B (en) | A kind of detection number system and method for overlooking pedestrian | |
| CN103500322B (en) | Automatic lane line identification method based on low latitude Aerial Images | |
| CN102006425B (en) | Method for splicing video in real time based on multiple cameras | |
| CN102163284B (en) | Chinese environment-oriented complex scene text positioning method | |
| CN105740809B (en) | A kind of highway method for detecting lane lines based on Airborne camera | |
| CN102132323B (en) | System and method for automatic image straightening | |
| CN103886308B (en) | A kind of pedestrian detection method of use converging channels feature and soft cascade grader | |
| CN104881662B (en) | A kind of single image pedestrian detection method | |
| CN104077577A (en) | Trademark detection method based on convolutional neural network | |
| CN102622584B (en) | Method for detecting mask faces in video monitor | |
| CN102663357A (en) | Color characteristic-based detection algorithm for stall at parking lot | |
| CN106127137A (en) | A kind of target detection recognizer based on 3D trajectory analysis | |
| CN104050481B (en) | Multi-template infrared image real-time pedestrian detection method combining contour feature and gray level | |
| CN106548182A (en) | Based on deep learning and the causal analytic pavement distress survey method and device of master | |
| CN104408746B (en) | A kind of passenger flow statistical system based on depth information | |
| CN103017869A (en) | Water level measuring system and method based on digital image processing | |
| CN104537651B (en) | Proportion detecting method and system for cracks in road surface image | |
| CN103179315A (en) | Continuous video image processing scanner and scanning method for paper documents | |
| CN112084928A (en) | Road traffic accident detection method based on visual attention mechanism and ConvLSTM network | |
| CN102749034B (en) | Railway switch gap offset detection method based on image processing | |
| CN106709952A (en) | Automatic calibration method of display screen | |
| CN102155933B (en) | Method for measuring galloping of transmission conductor on the basis of video difference analysis | |
| CN101867729B (en) | Method for detecting news video formal soliloquy scene based on features of characters | |
| CN104021385A (en) | Video subtitle thinning method based on template matching and curve fitting | |
| CN102682293B (en) | Method and system for identifying salient-point mould number of revolution-solid glass bottle based on images |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150902 Termination date: 20180926 |