CN102222265A - Binocular vision and laterally mounted video camera-based passenger flow counting method - Google Patents
Binocular vision and laterally mounted video camera-based passenger flow counting method Download PDFInfo
- Publication number
- CN102222265A CN102222265A CN201010145863XA CN201010145863A CN102222265A CN 102222265 A CN102222265 A CN 102222265A CN 201010145863X A CN201010145863X A CN 201010145863XA CN 201010145863 A CN201010145863 A CN 201010145863A CN 102222265 A CN102222265 A CN 102222265A
- Authority
- CN
- China
- Prior art keywords
- target
- counting
- parallax
- foreground
- passenger flow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention relates to a binocular vision and laterally mounted video camera-based passenger flow counting method. The method comprises the following steps of: (1) acquiring left and right synchronous video image sequences by using left and right parallel and synchronous video cameras; pre-setting two mutually parallel rows of counting lines which are vertical to the walking direction in a video image picture; and pre-setting a parallax error on the two rows of the counting lines as a background parallax error; (2) calculating the parallax errors point by point on the dual-row counting lines; then respectively unfolding the two rows of parallax errors on a time axis to form double rows of parallax error time space diagrams; (3) extracting a foreground target based on the differential value between the parallax error of one counting line and the background parallax error so as to generate a dynamic parallax error time space diagram; (4) performing planar mapping on the foreground target to obtain the mapped image of the corresponding planar position, the grey of the mapped image corresponding to the height of the foreground target; (5) performing the region division, marking, matching and counting on the foreground target in the mapped image; and (6) repeating the steps (1) to (5) until the real-time passenger flow counting is finished. The method is suitable for the counting occasions, such as outdoor sidewalks, indoor gates with over-high roof and corridors, where the video cameras can only be mounted laterally; and the method has very high real-time performance and robustness.
Description
Technical field
The present invention is used for the passenger flow information collection at a kind of relating to, and specifically relates to a kind of passenger flow counting method based on binocular vision and the installation of video camera side.
Background technology
No matter frequent along with Increase of population and people trip is all kinds of traffic systems, or ground such as market, museum, airport harbour all are that the stream of people springs up, and it is more and more important that the collection of these stream of peoples' statistics becomes gradually.As concerning traffic system, passenger flow quantity is the main foundation that public traffic management department rationally arranges public bus network, Optimization Dispatching public transit vehicle, raising efficiency of operation; Concerning the museum of market, the passenger flow quantitative relation the number of economic benefit and the quality of service quality.Therefore the research to automatic passenger flow data acquisition method has great practical value.
The automatic passenger flow counting method that has proposed at present mainly contains based on methods such as active infrared induction, passive infrared induction, pedal pressing force sensing, video image processing.Based on the system of active infrared induction, this technology maturation, antijamming capability is strong, but no matter is to adopt single bundle or multi beam infrared light, all can not effectively solve the counting of crowded passenger flow; The thermal infrared that passive type infrared counting technology is sent by human body is counted, and can distinguish life and abiotic object, but is subject to the influence of remarkable dressing, environment temperature etc., also can't adapt to the counting of crowded passenger flow.Method of counting based on the pedal pressing force sensor is generally used for the bus occasion, but requires the passenger to get on or off the bus successively, can not crowd, and can't differentiate the turnover direction of passenger flow preferably.
Method based on video image processing technology is the most rising method of counting at present, is divided into based on monocular and binocular camera shooting two classes.Preceding a kind of method utilizes the gray scale of target, chrominance information to carry out cutting apart of moving target, but to counting the variation and the sensitivity thereof of light in the scene, shade and chaff interference are also very big to the influence that moving target extracts, and are difficult to count exactly; A kind of method in back has been utilized the three-dimensional information of moving target, the illumination variation that a kind of method exists before can solving preferably and the problem of shade, but calculated amount is bigger.Existing method of counting based on stereoscopic vision is mainly used in the occasion that install at the video camera top, as bus door, buildings gateway etc., can reduce blocking of passenger flow so to greatest extent, realizes counting exactly.But for as this class occasion of outdoor walkway, because the walkway does not have ceiling, video camera can only be fixed on the buildings or vertical rod of side, this will aggravate the influence of blocking crowded and illumination shade between the passenger flow greatly, and the counting disposal route must be handled effectively to solve these stubborn problems.
Summary of the invention
The objective of the invention is to install the problem of the passenger flow counting technology existence of occasion at camera side, a kind of passenger flow counting method based on binocular vision is provided, its is fit to carry out passenger flow counting in the scenes such as indoor gate, corridor of outdoor walkway, roof excessive height, the influence that method can eliminate serious illumination sudden change, shade, perspective distortion well, block etc.
In order to achieve the above object, technical scheme of the present invention is:
Passenger flow counting method based on binocular vision and side installation video camera, this method comprises the steps: that (1) obtains left and right two synchronization video image sequences with parallel synchronous left and right video camera, and at default two parallel to each other and vertical with the direction of travel biserial counting lines of video image picture, the locational parallax of the biserial that prestores counting line is parallax as a setting;
(2) parallax is calculated in pointwise on the biserial counting line, two row parallaxes is launched on time shaft respectively again, forms biserial parallax spacetime diagram;
(3) extract foreground target according to the parallax of a counting line and the difference of background parallax, generate the dynamical parallax spacetime diagram;
(4) above-mentioned foreground target is carried out the ground level mapping, obtain the map image of planimetric position accordingly, the map image gray scale is corresponding to the height of foreground target;
(5) in map image, carry out Region Segmentation, mark, coupling and the counting of foreground target;
(6) repetitive cycling execution in step (1) is finished real-time passenger flow counting to step (5).
The spacing of above-mentioned biserial counting line is 15% of a video image picture width.
The concrete grammar that parallax is calculated in pointwise on the biserial counting line in the above-mentioned steps (2) is:
Adopt the window matching algorithm of normalized crosscorrelation to carry out parallax and calculate, for the image of 352 * 288 sizes of CIF form, the height of home window, width are got 3 row and 15 row respectively;
Adopt the dynamic window method to solve the parallax computational problem in uniform gray level zone, the gray variance of statistical pixel in the parallax match window, if gray variance is less than certain thresholding, then the size of match window is amplified 50%, continue to calculate its gray variance and judge, so repeatedly until 1/20 of image area;
At last, need carry out cubic spline interpolation, to obtain the parallax result of calculation of sub-pixel to normalized crosscorrelation NCC value.
Extract foreground target according to the parallax of current counting line and the difference of background parallax in the above-mentioned steps (3), the concrete grammar that generates the dynamical parallax spacetime diagram is as follows:
The counting line parallax of present frame and the counting line parallax of background frames are compared, if the difference of the parallax of respective pixel surpasses certain thresholding then judges that this pixel is a foreground pixel, otherwise are background pixel; The parallax value that keeps foreground pixel obtains the foreground target anaglyph, and thresholding is taken as 1 pixel interval, and target is extracted and need be handled respectively on two counting lines,
The column count line foreground target parallax of each frame is launched on the axial time shaft of x, keep 100 nearest column data and constitute row parallax time-space image, whenever there are row of a new frame to enter, row that then will be the oldest are removed, other row move to right, the dynamic foreground target parallax of two width of cloth spacetime diagram that up-to-date row place Far Left, corresponding two column count lines to constitute, follow-up passenger flow counting algorithm is fully based on the parallax time-space image of this two width of cloth foreground target.
In the above-mentioned steps (4) above-mentioned foreground target is carried out the ground level mapping, obtain the map image of planimetric position accordingly, gradation of image is as follows with the concrete grammar that the height spacetime diagram generates corresponding to the height foreground target mapping of foreground target:
According to the parallax of the corresponding depth information of foreground image target is mapped to two-dimensionally on the plane, promptly calculates the corresponding position on plane two-dimensionally of each foreground pixel point in the prospect anaglyph; This gray scale is taken as the height of target in the map image, claims that the image after this mapping is a ground level mapping height time-space image.
Region Segmentation, mark, coupling and the counting of above-mentioned steps (5) foreground target, concrete grammar is as follows:
(1) moving target in above-mentioned two width of cloth row mapping spacetime diagram is extracted and mark, deposit in the target label structural array, and abandon little moving target;
(2) cutting apart and merging of above-mentioned moving target, cut apart accordingly and merge according to its gray scale (representing the passenger flow height), area, width etc.;
(3) coupling of target counting, when target occurring on counting line 2, the target of this target optimum matching on search and the counting line 1 if obtain the optimum matching target, according to barycenter positions relation on time shaft of two coupling targets, is judged the target travel direction; If on the counting line 1 barycenter of target on the time shaft prior to counting line 2 on the barycenter of target, then this target is a positive dirction, otherwise, then be in the other direction, and upgrade counter separately; The information of two coupling targets are deleted from the target label structural array of counting line 1 and counting line 2 respectively, when having a plurality of target to be matched on counting line 2, obtain the optimum matching target according to cost function, cost function is defined as follows:
F(x，y
n)＝aS(x，y
n)+bD
x(x，y
n)+cD
y(x，y
n)
Target y to be matched on target x and the counting line 2 on this function calculation counting line 1
nSimilarity, wherein, S (x, y
n) be x and y
nThe absolute value of difference of height average gray, D
x(x, y
n) be the horizontal direction distance between moving target, D
y(x, y
n) be the vertical direction distance between moving target, a, b, c are weighted values.F (x, y
n) value is more little, and more coupling of two targets be described, selection makes F (x, y
n) the minimum target y of value
nBe the optimum matching target.
The present invention compares based on the passenger flow counting technology of video with existing, have following feature and advantage: the present invention adopts binocular vision technology to carry out the extraction of the foreground target on the biserial counting line, be subjected to the influence of illumination variation and shade hardly, improved the anti-interference of system; Utilize the ground level mapping method of target to extract the ground level position and the height thereof of foreground target, form biserial height time-space image, solved perspective preferably and blocked overlap problem; Carry out extraction, mark and the coupling counting of target at biserial height time-space image.The method of counting that the present invention proposes is applicable to the counting occasion that video camera can only the side be installed, and as the indoor gate of outdoor walkway, roof excessive height and corridor etc., can satisfy the real-time and the accuracy requirement of system well.
Description of drawings
Fig. 1 installs the application scenarios synoptic diagram of the passenger flow counting method of video camera based on binocular vision and side among the present invention.
Fig. 2 is the passenger flow counting method of video camera is installed by system of the present invention based on binocular vision and side a FB(flow block).
Fig. 3 is the mapping graph of foreground target on the ground level background of algorithmic formula derivation usefulness among the present invention.
Embodiment
Below with reference to accompanying drawing passenger flow counting method and the system thereof based on binocular vision and side installation video camera of the present invention is described in further detail.
As shown in Figure 1, based on the passenger flow counting method of binocular vision and side installation video camera, it comprises on-the-spot binocular solid video camera and mounting means thereof.
Left and right video camera A1, A2 take the passenger flow scene, and scene can be occasions such as the indoor gate of outdoor walkway, roof excessive height and corridor.Two biserial counting lines vertical with direction of travel are set in the scene image are used for algorithm counts, its position and spacing can be adjusted according to actual conditions, usually place the both sides in image 3 centre positions, spacing is 15% reliability with the raising counting of picture traverse.Actual passenger flow in image about walking, 2 direction is positive dirction S from counting line 1 to counting line in definition, otherwise be opposite direction.
Shown in Fig. 1～3, above-mentioned passenger flow counting method of video camera being installed, its following steps: (1), obtain left and right two synchronization video image sequences with parallel left and right video camera synchronously based on binocular vision and side; (2), parallax is calculated in pointwise on the biserial counting line, two row parallaxes is launched on time shaft respectively again, forms biserial parallax spacetime diagram; (3), extract foreground target according to the difference of parallax of working as last counting line and background parallax, generation dynamical parallax spacetime diagram; (4), above-mentioned foreground target is carried out ground level mapping, obtain the map image of planimetric position accordingly, the gray scale of this map image is corresponding to the height of preceding bright target; (5), in map image, carry out Region Segmentation, mark, coupling and the counting of foreground target.
The acquisition method of synchronization video image sequence is as follows in the above-mentioned steps (1): parallel left and right video camera is installed in the side on walkway, and height is overlooked the same passenger flow scene of shooting more than 2.5 meters from 2 positions, adopt 2/3 of sidewalk width highly usually.The kinetic characteristic that be to adapt to passenger flow, left and right cameras must be gone field synchronization.By artificial setting, store in advance by the locational parallax of counting line parallax value as a setting in picture for the biserial counting line.Video camera depression angle after the installation is determined by conventional camera marking method, takes for adapting to night, and video camera has the infrared photography function, and infrared LED lamp light filling is set.
Correct parallaxometer is the basis of entire method in the above-mentioned steps (2), need only the each point on the biserial counting line be calculated, and its concrete grammar is:
(this method is the known algorithms of those skilled in the art to the window matching algorithm of employing normalized crosscorrelation, do not repeat them here) carry out parallax and calculate, for the image of 352 * 288 sizes of CIF form, the height of home window, width are got 3 row and 15 row respectively.The match window interior pixel number that can keep appropriateness so both had been unlikely to cause the fuzzy of parallax, significantly parallax noise also can not occur.
Adopt dynamic window method (this method is the known algorithms of those skilled in the art, does not repeat them here) to solve the parallax computational problem in uniform gray level zone.The gray variance of statistical pixel in the parallax match window, if gray variance is less than certain thresholding, then the size of match window is amplified 50%, continue to calculate its gray variance and judge, so repeatedly until 1/20 of image area, this method can be eliminated the parallax mistake matching problem of homogeneous area effectively.
Calculate background parallax during installation in advance,, for uniform scene ground some grey scale change are set artificially, as paste some lines and pattern for improving the coupling degree of accuracy of static background parallax.
At last, need carry out cubic spline interpolation, to obtain the parallax result of calculation of sub-pixel to normalized crosscorrelation NCC value.
The extraction and the dynamical parallax space-time drawing generating method thereof of foreground target are as follows in the above-mentioned steps (3): the counting line parallax of present frame and the counting line parallax of background frames are compared, surpass certain thresholding as the difference of the parallax of respective pixel and judge that then this pixel is a foreground pixel, otherwise be background pixel.The parallax value that keeps foreground pixel obtains the foreground target anaglyph.Thresholding is taken as 1 pixel interval.Target is extracted and need be handled respectively on two counting lines.
The column count line foreground target parallax of each frame is launched on the axial time shaft of x, keep 100 nearest column data and constitute row parallax time-space image, whenever have row of a new frame to enter, row that then will be the oldest are removed, other row move to right, and up-to-date row are placed Far Left.Like this, the dynamic foreground target parallax of two width of cloth spacetime diagram that corresponding two column count lines have constituted, follow-up passenger flow counting algorithm are fully based on the parallax time-space image of this two width of cloth foreground target.
The ground level map image method that generates foreground target in the above-mentioned steps (4) is as follows: each individual is separated from each other on the ground level, but because perspective effect be exactly overlapped on image, and size also has evident difference.In order to eliminate perspective effect and overlapping influence, according to the parallax of the corresponding depth information of foreground image target is mapped to two-dimensionally on the plane, promptly calculate the corresponding position on plane two-dimensionally of each foreground pixel point in the prospect anaglyph.If mapping point is arranged in ground level, then showing on this aspect has foreground target, and this gray scale is taken as the height of target in the map image, otherwise value is zero.Image after this mapping is the map image of ground level position, and synoptic diagram as shown in Figure 3.
Suppose that the background scene is a plane, OC is a Y direction, and the video camera installation site is P, and depression angle is an alpha+beta, and wherein β is a video camera in half of the visual angle of vertical direction, is decided by lens focus size and video camera target surface size.α is decided by setting angle, can determine by the camera calibration mode.The AC line is the imageable areas on the plane among the figure.When a realistic objective EF stands in the plane, the point of E point on image planes determined angle d β, the position of the vertical direction of E point in the image pixel coordinate of shot by camera is y, if the pixel coordinate initial point is taken at central authorities, picture altitude corresponding to A to C interval is iH, and then the pass of y and d β is
D β value can just can be born.After d β has been arranged, according to the triangle principle, OF/OD=PE/PD, and PE/PD just is decided by the ratio p of E, D point parallax
D/ p
E, can obtain
OF=OD*p
D/ p
E=htan (alpha+beta+d β) p
D/ p
E(2) h is the video camera setting height(from bottom) in the formula, p
DThe parallax p that can order by A
ATry to achieve
So just can obtain
Thereby, as long as known the parallax p at the next line place of background plane scene
AAnd two angle [alpha], β just can obtain upright position on its corresponding plane according to the parallax of foreground target.
For the height of target EF, we have
Represent the height of target for convenience with gray scale, EF is carried out normalization with camera height h.
By top mapping relations, we just can be mapped to foreground target parallax time-space image on the ground level, are called the map image of ground level position.
Region Segmentation, mark, coupling and the method for counting of foreground target are as follows in the above-mentioned steps (5): through the mapping of ground level position, in the map image of ground level position, originally blocking each overlapping individuality will make a distinction, and because the difference of height, coupling between the target after cutting apart also just not only has positional information, also has elevation information.This step comprises following a few step:
(1) moving target in above-mentioned two width of cloth row mapping spacetime diagram is extracted and mark, deposit in the target label structural array, and abandon little moving target.This algorithm can adopt the method for serial mark to carry out, the serial mark is meant the mark that only carries out moving target at current line according to 8 syntoples of target, every frame is only handled the up-to-date delegation of the map image on counting line 1 and the counting line 2, and the serial computing by the every frame on the time shaft marks the moving target in the mapping spacetime diagram.
(2) above-mentioned moving target cuts apart and merges.Because the crowded and illumination variation of passenger flow, some moving target merges easily, and some moving target then divides, and need cut apart accordingly and merge according to its gray scale (representing the passenger flow height), area, width etc.
(3) coupling of target counting.When target occurring on counting line 2, the target of this target optimum matching on search and the counting line 1 if obtain the optimum matching target, according to barycenter positions relation on time shaft of two coupling targets, is judged the target travel direction.If on the counting line 1 barycenter of target on the time shaft prior to counting line 2 on the barycenter of target, then this target is a positive dirction, otherwise, then be in the other direction, and upgrade counter separately.The information of two coupling targets is deleted from the target label structural array of counting line 1 and counting line 2 respectively.When on counting line 2, having a plurality of target to be matched, obtain the optimum matching target according to cost function, cost function is defined as follows:
F(x，y
n)＝aS(x，y
n)+bD
x(x，y
n)+cD
y(x，y
n) (6)
Target y to be matched on target x and the counting line 2 on this function calculation counting line 1
nSimilarity, wherein, S (x, y
n) be x and y
nThe absolute value of difference of height average gray, D
x(x, y
n) be the horizontal direction distance between moving target, D
y(x, y
n) be the vertical direction distance between moving target, a, b, c are weighted values.F (x, y
n) value is more little, and more coupling of two targets be described, selection makes F (x, y
n) the minimum target y of value
nBe the optimum matching target.
Each frame repetitive cycling execution in step 1 to step 5, is finished the automatic counting of passenger flow.
The present invention adopts that binocular vision technology carries out that biserial counting line parallax calculates, the extraction of passenger flow foreground target and the parallax spacetime diagram generates, the ground level mapping and the object height time-space image generates, based on treatment steps such as the object matching of biserial counting line and counting, final passenger flow quantity and the direction that obtains by counting line.This method is applicable to the counting occasion that video camera can only the side be installed, as the indoor gate of outdoor walkway, roof excessive height and corridor etc.
The present invention is directed to several technological difficulties that passenger flow counting faced that video camera is installed based on the side of video: the cutting apart of foreground target, the correction of perspective effect, target occlusion, overlapping processing, algorithm computation amount was excessive etc., adopted binocular stereo vision and biserial counting line technology to handle.Adopt technique of binocular stereoscopic vision can determine and cut apart the position and the depth information of foreground target more exactly, utilize the target three-dimensional information that foreground target is mapped to two-dimensionally on the plane again, cut apart to carry out perspective correction and target occlusion.Cutting apart and mate counting based on the target of biserial counting line then will reduce calculated amount effectively, reduce disturbing factor.
The foregoing description just lists expressivity principle of the present invention and effect is described, but not is used to limit the present invention.Any personnel that are familiar with this technology all can make amendment to the foregoing description under spirit of the present invention and scope.Therefore, the scope of the present invention should be listed as claims.
Claims (6)
1. based on the passenger flow counting method of stereoscopic vision and side installation video camera, it is characterized in that, comprise the steps:
(1) obtain left and right two synchronization video image sequences with parallel synchronous left and right video camera, and at default two parallel to each other and vertical with the direction of travel biserial counting lines of video image picture, the locational parallax of the biserial that prestores counting line is parallax as a setting;
(2) parallax is calculated in pointwise on the biserial counting line, two row parallaxes is launched on time shaft respectively again, forms biserial parallax spacetime diagram;
(3) extract foreground target according to the parallax of a counting line and the difference of background parallax, generate the dynamical parallax spacetime diagram;
(4) above-mentioned foreground target is carried out the ground level mapping, obtain the map image of planimetric position accordingly, the map image gray scale is corresponding to the height of foreground target;
(5) in map image, carry out Region Segmentation, mark, coupling and the counting of foreground target;
(6) repetitive cycling execution in step (1) is finished real-time passenger flow counting to step (5).
2. the passenger flow counting method based on stereoscopic vision and side installation video camera according to claim 1 is characterized in that the spacing of above-mentioned biserial counting line is 15% of a video image picture width.
3. the passenger flow counting method based on stereoscopic vision and side installation video camera according to claim 1 is characterized in that the concrete grammar that parallax is calculated in the pointwise on the biserial counting line in the above-mentioned steps (2) is:
Adopt the window matching algorithm of normalized crosscorrelation to carry out parallax and calculate, for the image of 352 * 288 sizes of CIF form, the height of home window, width are got 3 row and 15 row respectively;
The gray variance of statistical pixel in the parallax match window if gray variance less than certain thresholding, then amplifies 50% with the size of match window, continues to calculate its gray variance and judge, so repeatedly until 1/20 of image area;
At last, need carry out cubic spline interpolation, to obtain the parallax result of calculation of sub-pixel to normalized crosscorrelation NCC value.
4. the passenger flow counting method based on stereoscopic vision and side installation video camera according to claim 3, it is characterized in that, extract foreground target according to the parallax of current counting line and the difference of background parallax in the above-mentioned steps (3), the concrete grammar that generates the dynamical parallax spacetime diagram is as follows:
The counting line parallax of present frame and the counting line parallax of background frames are compared, if the difference of the parallax of respective pixel surpasses certain thresholding then judges that this pixel is a foreground pixel, otherwise are background pixel; The parallax value that keeps foreground pixel obtains the foreground target anaglyph, and thresholding is taken as 1 pixel interval, and target is extracted and need be handled respectively on two counting lines,
The column count line foreground target parallax of each frame is launched on the axial time shaft of x, keep 100 nearest column data and constitute row parallax time-space image, whenever there are row of a new frame to enter, row that then will be the oldest are removed, other row move to right, the dynamic foreground target parallax of two width of cloth spacetime diagram that up-to-date row place Far Left, corresponding two column count lines to constitute, follow-up passenger flow counting algorithm is fully based on the parallax time-space image of this two width of cloth foreground target.
5. the passenger flow counting method based on stereoscopic vision and side installation video camera according to claim 4, it is characterized in that, in the above-mentioned steps (4) above-mentioned foreground target is carried out the ground level mapping, obtain the map image of planimetric position accordingly, gradation of image is as follows with the concrete grammar that the height spacetime diagram generates corresponding to the height foreground target mapping of foreground target:
According to the parallax of the corresponding depth information of foreground image target is mapped to two-dimensionally on the plane, promptly calculates the corresponding position on plane two-dimensionally of each foreground pixel point in the prospect anaglyph; This gray scale is taken as the height of target in the map image, obtains the map image of planimetric position accordingly.
6. the passenger flow counting method based on stereoscopic vision and side installation video camera according to claim 5 is characterized in that, Region Segmentation, mark, coupling and the counting of above-mentioned steps (5) foreground target, and concrete grammar is as follows:
(1) moving target in above-mentioned two width of cloth row mapping spacetime diagram is extracted and mark, deposit in the target label structural array, and abandon little moving target;
(2) cutting apart and merging of above-mentioned moving target, cut apart accordingly and merge according to its gray scale (representing the passenger flow height), area, width etc.;
(3) coupling of target counting, when target occurring on counting line 2, the target of this target optimum matching on search and the counting line 1 if obtain the optimum matching target, according to barycenter positions relation on time shaft of two coupling targets, is judged the target travel direction; If on the counting line 1 barycenter of target on the time shaft prior to counting line 2 on the barycenter of target, then this target is a positive dirction, otherwise, then be in the other direction, and upgrade counter separately; The information of two coupling targets are deleted from the target label structural array of counting line 1 and counting line 2 respectively, when having a plurality of target to be matched on counting line 2, obtain the optimum matching target according to cost function, cost function is defined as follows:
F(x，y
n)＝aS(x，y
n)+bD
x(x，y
n)+cD
y(x，y
n)
Target y to be matched on target x and the counting line 2 on this function calculation counting line 1
nSimilarity, wherein, S (x, y
n) be x and y
nThe absolute value of difference of height average gray, D
x(x, y
n) be the horizontal direction distance between moving target, D
y(x, y
n) be the vertical direction distance between moving target, a, b, c are weighted values.F (x, y
n) value is more little, and more coupling of two targets be described, selection makes F (x, y
n) the minimum target y of value
nBe the optimum matching target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010145863XA CN102222265A (en) | 2010-04-13 | 2010-04-13 | Binocular vision and laterally mounted video camera-based passenger flow counting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010145863XA CN102222265A (en) | 2010-04-13 | 2010-04-13 | Binocular vision and laterally mounted video camera-based passenger flow counting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102222265A true CN102222265A (en) | 2011-10-19 |
Family
ID=44778810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010145863XA Pending CN102222265A (en) | 2010-04-13 | 2010-04-13 | Binocular vision and laterally mounted video camera-based passenger flow counting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102222265A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102890791A (en) * | 2012-08-31 | 2013-01-23 | 浙江捷尚视觉科技有限公司 | Depth information clustering-based complex scene people counting method |
| CN103455792A (en) * | 2013-08-20 | 2013-12-18 | 深圳市飞瑞斯科技有限公司 | Guest flow statistics method and system |
| CN104537668A (en) * | 2014-12-29 | 2015-04-22 | 浙江宇视科技有限公司 | Fast anaglyph calculating method and device |
| CN104715471A (en) * | 2014-01-03 | 2015-06-17 | 杭州海康威视数字技术股份有限公司 | Target positioning and tracking method and device |
| CN106127137A (en) * | 2016-06-21 | 2016-11-16 | 长安大学 | A kind of target detection recognizer based on 3D trajectory analysis |
| CN106446788A (en) * | 2016-08-31 | 2017-02-22 | 山东恒宇电子有限公司 | Method for passenger flow statistic by means of high-dynamic range image based on optic nerve mechanism |
| 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 |
| CN106503605A (en) * | 2015-09-01 | 2017-03-15 | 南京理工大学 | Human body target recognition methods based on stereovision technique |
| CN108280402A (en) * | 2017-12-27 | 2018-07-13 | 武汉长江通信智联技术有限公司 | A kind of passenger flow volume statistical method and system based on binocular vision |
| CN113422928A (en) * | 2021-05-28 | 2021-09-21 | 佛山市诚智鑫信息科技有限公司 | Safety monitoring snapshot method and system |
-
2010
- 2010-04-13 CN CN201010145863XA patent/CN102222265A/en active Pending
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102890791A (en) * | 2012-08-31 | 2013-01-23 | 浙江捷尚视觉科技有限公司 | Depth information clustering-based complex scene people counting method |
| CN103455792A (en) * | 2013-08-20 | 2013-12-18 | 深圳市飞瑞斯科技有限公司 | Guest flow statistics method and system |
| CN104715471B (en) * | 2014-01-03 | 2018-01-02 | 杭州海康威视数字技术股份有限公司 | Target locating method and its device |
| CN104715471A (en) * | 2014-01-03 | 2015-06-17 | 杭州海康威视数字技术股份有限公司 | Target positioning and tracking method and device |
| CN104537668B (en) * | 2014-12-29 | 2017-08-15 | 浙江宇视科技有限公司 | A kind of quick parallax image computational methods and device |
| CN104537668A (en) * | 2014-12-29 | 2015-04-22 | 浙江宇视科技有限公司 | Fast anaglyph calculating method and device |
| 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 |
| CN106503605A (en) * | 2015-09-01 | 2017-03-15 | 南京理工大学 | Human body target recognition methods based on stereovision technique |
| CN106127137A (en) * | 2016-06-21 | 2016-11-16 | 长安大学 | A kind of target detection recognizer based on 3D trajectory analysis |
| CN106446788A (en) * | 2016-08-31 | 2017-02-22 | 山东恒宇电子有限公司 | Method for passenger flow statistic by means of high-dynamic range image based on optic nerve mechanism |
| CN108280402A (en) * | 2017-12-27 | 2018-07-13 | 武汉长江通信智联技术有限公司 | A kind of passenger flow volume statistical method and system based on binocular vision |
| CN108280402B (en) * | 2017-12-27 | 2021-09-24 | 武汉长江通信智联技术有限公司 | Binocular vision-based passenger flow volume statistical method and system |
| CN113422928A (en) * | 2021-05-28 | 2021-09-21 | 佛山市诚智鑫信息科技有限公司 | Safety monitoring snapshot method and system |
| CN113422928B (en) * | 2021-05-28 | 2022-02-18 | 佛山市诚智鑫信息科技有限公司 | Safety monitoring snapshot method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102222265A (en) | Binocular vision and laterally mounted video camera-based passenger flow counting method | |
| CN101714293A (en) | Stereoscopic vision based acquisition method of congestion degree of bus passenger flow | |
| AU2020100323A4 (en) | Solar Power Forecasting | |
| AU2006203491B2 (en) | Method for processing stereo vision data using image density | |
| CN105225482A (en) | Based on vehicle detecting system and the method for binocular stereo vision | |
| CN102168954B (en) | Monocular-camera-based method for measuring depth, depth field and sizes of objects | |
| CN102110376B (en) | Roadside parking space detection device based on computer vision | |
| CN105516654A (en) | Scene-structure-analysis-based urban monitoring video fusion method | |
| WO2016199244A1 (en) | Object recognition device and object recognition system | |
| US20140086451A1 (en) | Method and apparatus for detecting continuous road partition | |
| CN101567097B (en) | Bus passenger flow automatic counting method based on two-way parallactic space-time diagram and system thereof | |
| CN104504688A (en) | Method and system based on binocular stereoscopic vision for passenger flow density estimation | |
| KR20190058230A (en) | System and method for modeling surveillance camera layout | |
| CN103473554A (en) | People flow statistical system and people flow statistical method | |
| CN104408746A (en) | Passenger flow statistical system based on depth information | |
| CN103268470A (en) | Method for counting video objects in real time based on any scene | |
| Lin et al. | Detection of buildings from monocular images | |
| CN104966062A (en) | Video monitoring method and device | |
| WO2017193172A1 (en) | "solar power forecasting" | |
| CN116824859B (en) | Intelligent traffic big data analysis system based on Internet of things | |
| KR20190059395A (en) | Gis 3 cctv 3 cctv | |
| JPH09297849A (en) | Vehicle detector | |
| CN103453890A (en) | Nighttime distance measuring method based on taillight detection | |
| CN111783690A (en) | Urban travelable area CIM information processing method based on vehicle density perception | |
| Shaaban et al. | Parking space detection system using video images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20111019 |