CN102902971A - Method and system for conducting statistics on elevator visitor flow based on intelligent visual perception - Google Patents
Method and system for conducting statistics on elevator visitor flow based on intelligent visual perception Download PDFInfo
- Publication number
- CN102902971A CN102902971A CN2012103168626A CN201210316862A CN102902971A CN 102902971 A CN102902971 A CN 102902971A CN 2012103168626 A CN2012103168626 A CN 2012103168626A CN 201210316862 A CN201210316862 A CN 201210316862A CN 102902971 A CN102902971 A CN 102902971A
- Authority
- CN
- China
- Prior art keywords
- elevator
- image
- unit
- moving target
- target
- 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 provides a method for conducting statistics on elevator visitor flow based on intelligent visual perception. The method comprises the steps of S1. establishing a head-and-shoulder model sample database; S2. conducting feature extraction and model training, wherein samples in S1 are subjected to principle component analysis (PCA) feature extraction, and a support vector machine (SVM) trainer is used for training model generating; S3. detecting targets, wherein matching calculation is conducted on images collected in real time according to human body head-and-shoulder model data obtained in S2, and targets in current images are obtained through detection; S4. tracking targets, wherein targets detected in S3 are tracked; and S5. conducing statistics on the visitor flow, wherein corresponding counters are operated according to incoming and outgoing conditions of targets when targets tracked in S4 go cross a crossing line. The method has the advantages that real-time intelligent analyses are conducted on images collected in real time, so that elevator visitor flow data can be obtained, accurate evidences are provided for establishing of effective and energy-saving elevator dispatching strategies, the problem that prediction models of the elevator visitor flow in traditional methods are complex and difficult to establish is solved, and the imprecise prediction caused by special events are prevented.
Description
Technical field
The invention belongs to pattern-recognition and computer vision field, relate in particular to a kind of elevator people flow rate statistical method and system based on the intelligent vision perception.
Background technology
Along with the high rise multi-rise commercial building in city and the increase of residential quarters, obtained using more and more widely as the elevator of important vertical transport instrument.Guaranteeing reliable, efficient, the comfortable and quickly operation of elevator, is the key point of elevator development and technical progress.Will carry out reasonable efficient scheduling to elevator, just must grasp the data of accurate elevator flow of the people, therefore, it is particularly important that stream of people's quantitative statistics just seems.Because the elevator flow of the people is subject to the impact of enchancement factor, has nonlinearity and unpredictability, therefore be difficult to search out the volume of the flow of passengers that specific function is predicted elevator.
Current elevator people flow rate statistical method all is by predicting by the magnitude of traffic flow.Traditional passenger flow volume statistical method is that flow is regarded as a kind of time series, extracts autoregressive model, moving average model, and autoregressive moving-average model.These class methods are easy to realization, but after model parameter is determined, just can not change, and be linear model, therefore, will greatly reduce the degree of accuracy of its prediction to the non-linear elevator flow of the people of complexity.After nineteen ninety, because the self-adaptation of neural network and the advantage of self study, scholars have proposed the elevator people flow rate statistical model based on neural network.But it is slow that neural net method exists again speed of convergence, easily is absorbed in the shortcomings such as local extremum, makes it can not become desirable method.
Therefore, the problem that is difficult to set up for the forecast model that solves complicated elevator flow of the people in the classic method, the coarse situation of prediction of having prevented to bring owing to special event occurs, the present invention proposes a kind of target detection based on PCA-SVM and based on the method for the target following of Kalman filtering, carry out intellectual analysis by real-time image acquisition, draw accurate volume of the flow of passengers data.
Summary of the invention
The present invention is for a kind of volume of the flow of passengers that can effectively know the every day operation of elevator being provided, with according to these accurate data, formulating more effective, energy-conservation elevator dispatching strategy.
For achieving the above object, the technical scheme that the present invention adopts provides a kind of elevator people flow rate statistical method based on the intelligent vision perception, may further comprise the steps:
S100: gather human head and shoulder model image sample;
S200: the human head and shoulder model image that gathers is carried out the PCA feature extraction, and send into training generation Sample Storehouse in the SVM training aids;
S300: differentiate elevator door and whether open, if, gather current frame image, search the moving target in the current frame image, described moving target is carried out the PCA feature extraction, utilize the svm classifier device that moving target and Sample Storehouse are mated classification, if described moving target is the human body head shoulder images, then turn S400; If not, the S300 that then circulates;
S400: judge that described moving target still goes out elevator for advancing elevator, if described moving target is for advancing elevator, then add 1 to advancing the elevator number, if described moving target is for going out elevator, then add 1 to going out the elevator number.
Step S200, concrete comprises the steps:
S201: human head and shoulder model image sample is stretched as one-dimensional vector by row, and setting the image pattern database is high-dimensional data space point set X={x
i, i=1,2,3N;
S202: the mean vector m that asks for X;
Each vector among the S203:X deducts average m, then asks its covariance matrix, and carries out Eigenvalues Decomposition;
S204: several eigenwert characteristic of correspondence vectors before extracting, namely required that major component is vectorial, be designated as Γ
i, i=1,2 ... p(p is the low-dimensional data dimension that dimensionality reduction will obtain);
S205: the major component vector of trying to achieve is carried out dimensionality reduction, and send in the SVM training aids and classify, generate Sample Storehouse.
Step S300 specifically comprises:
S311: gather current frame image, use inter-frame difference to obtain bianry image, judge whether described bianry image exists connected region, if exist, then turn S312; If do not exist, then continue to gather the next frame image.
S312: set up the kalman filter models that described bianry image is followed the tracks of, search the moving target in the image.
Among the step S300, whether described detection elevator door is opened by the image that collects is carried out rim detection, comprising:
S321: whether the left-hand component of judging the defined area detects straight line; If, then turn next step, if not, judge that then elevator is as closing the door;
S322: whether the right-hand component of judging the defined area detects straight line; If, then turn next step, if not, judge that then elevator is as closing the door;
S323: whether the straight line of judging above the right and left is symmetrical about intermediate point, if, then judge door-opened elevator, if not, judge that then elevator closes the door.
Step S400 specifically comprises:
The coordinate relation of the coordinate line of judging eve, the current quarter of the tracking target in the testing image and presetting;
If eve tracking target place coordinate is below ruler, namely in the elevator, current time place coordinate is the ruler top, namely outside the elevator, judges that then this target is as going out elevator;
If eve tracking target place coordinate is the ruler top, namely in the elevator, current time place coordinate is the ruler below, namely in the elevator, judges that then this target is as advancing elevator.
Concrete, the Hough transformation method is adopted in described detection to straight line.
The present invention also provides a kind of elevator people flow rate statistical system based on the intelligent vision perception, comprising:
Information acquisition unit is used for gathering human head and shoulder model image sample; The Database unit is used for the human head and shoulder model image that gathers is carried out the PCA feature extraction, and sends into training generation Sample Storehouse in the SVM training aids; Whether the first judging unit is used for differentiating elevator door and opens; Data analysis unit is used for gathering current frame image, searches the moving target in the current frame image, described moving target is carried out the PCA feature extraction, utilize the svm classifier device that moving target and Sample Storehouse are mated classification, if described moving target is the human body head shoulder images, then turn statistic unit; Statistic unit is used for judging that described moving target still goes out elevator for advancing elevator, if described moving target is for advancing elevator, then adds 1 to advancing the elevator number, if described moving target is for going out elevator, then adds 1 to going out the elevator number.
Described data are set up concrete the comprising in unit: the first numerical analysis unit, be used for human head and shoulder model image sample is stretched as one-dimensional vector by row, and setting the image pattern database is high-dimensional data space point set X={x
i, i=1,2,3N; Average is asked for the unit, is used for asking for the mean vector m of X; The second value analytic unit is used for each vector of X is deducted average m, then asks its covariance matrix, and carries out Eigenvalues Decomposition; Major component vector acquiring unit, namely required that major component is vectorial for several eigenwert characteristics of correspondence vectors before extracting, be designated as Γ
i, i=1,2 ... p(p is the low-dimensional data dimension that dimensionality reduction will obtain); The Sample Storehouse generation unit is used for the major component vector of trying to achieve is carried out dimensionality reduction, and sends in the SVM training aids and classify, and generates Sample Storehouse.
Described data analysis unit comprises: image acquisition units, be used for gathering current frame image, and use inter-frame difference to obtain bianry image, judge whether described bianry image exists connected region, if exist, then transport the moving-target acquiring unit; If do not exist, then continue to gather the next frame image.The moving target acquiring unit is used for setting up the kalman filter models that described bianry image is followed the tracks of, and searches the moving target in the image.
Described data analysis unit also comprises: the first detecting unit is used for judging whether the left-hand component of defined area detects straight line; If, then turn the second detecting unit, if not, judge that then elevator is as closing the door; The second detecting unit is used for judging whether the right-hand component of defined area detects straight line; If, then turn the 3rd detection, if not, judge that then elevator is as closing the door; The 3rd detecting unit be used for to judge whether the straight line of above the right and left is symmetrical about intermediate point, if, then judge door-opened elevator, if not, judge that then elevator closes the door.
What described statistic unit was concrete comprises: coordinate analysis unit and counting unit; The coordinate analysis unit concerns with the coordinate of the coordinate line that presets eve, current quarter that be used for to judge the tracking target in the testing image; If eve tracking target place coordinate is below ruler, namely in the elevator, current time place coordinate is the ruler top, namely outside the elevator, then judges this target as going out elevator, and then described counting unit adds 1 to going out the elevator number; If eve tracking target place coordinate is the ruler top, namely in the elevator, current time place coordinate is the ruler below, namely in the elevator, then judges this target as advancing elevator, and then described counting unit adds 1 to advancing the elevator number.
The invention has the beneficial effects as follows: the sorter that has adopted PCA-SVM to mix carries out human head and shoulder identification, the human head and shoulder image pattern that gathers is carried out the PCA feature extraction, and send in the SVM training aids and train, to generate Sample Storehouse, for the current frame image that obtains, by obtaining the moving target in the current frame image, and it is carried out the classification of PCA feature extraction and svm classifier device, and carry out matching ratio with Sample Storehouse, the moving target that identifies in the current frame image is the human body head shoulder images, and by analyzing the motion state of described moving target, the volume of the flow of passengers of the every day operation of statistics elevator.Adopt this technical scheme higher than the accuracy rate of single sorter identification, the volume of the flow of passengers by the every day operation of statistics elevator, the degree of wear of estimating device for the elevator maintenance personnel provides foundation, and then formulates more effective elevator maintenance plan, reduces the generation of fault.
Description of drawings
Fig. 1 is the process flow diagram of the elevator people flow rate statistical method based on the intelligent vision perception of the present invention;
Fig. 2 is the process flow diagram whether described elevator of detection of the present invention opens the door;
Fig. 3 is the structured flowchart of the elevator people flow rate statistical system based on the intelligent vision perception of the present invention.
Embodiment
By describing technology contents of the present invention, structural attitude in detail, realized purpose and effect, below in conjunction with embodiment and cooperate that accompanying drawing is detailed to give explanation.
In the technical program, described SVM is support vector machine, its basic thought is for the given learning tasks with limited quantity training sample, how between the complexity of the model study precision of specific training sample (namely to) and learning ability (namely identifying error-free the ability of arbitrary sample), seek best compromise, in the hope of obtaining best Generalization Ability.
SVM is as a kind of method of the detection based on machine learning, it is extremely important that the choosing of training sample database set up, head shoulder based on PCA-SVM needs two class samples as the detection method model, and a class is based on the head of PCA feature extraction and takes on decent, decent of class right and wrong head shoulder.In the technical program, adopt for the collection of decent of head shoulder to artificially collect.When selecting head shoulder images, to choose the representative image of human head and shoulder partial geometry clear in structure.For decent of non-head shoulder, in fact any other image can be as decent of non-head shoulder.After training, utilize great amount of samples to test, to guarantee to obtain best sorter.
PCA is principal component analytical method, according to the inner link of research object variable to be carried out comprehensively, take out some regular things of being with, consist of the mathematical model simplified in a way, and then in order to one of multivariate statistical method of studying complicated spontaneous phenomenon, its basic thought be exactly inside dependency structure from variance-covariance matrix be starting point, manage to find out less comprehensive characteristics and represent original more feature, and these preferably comprehensive characteristics can reflect as much as possible again the information of more feature, be should be separate between these comprehensive characteristics, representative best again.Principal component analysis (PCA) is a kind of effective method of analyzing data in the statistics, and its purpose mainly is dimensionality reduction, original P dimension space dimensionality reduction is projected to the N dimension space, and preserved the main information in the former data behind dimensionality reduction, thereby make data be easier to process.
Consult Fig. 1, in order to solve technical matters of the present invention, the technical scheme that the present invention adopts provides a kind of elevator people flow rate statistical method based on the intelligent vision perception, may further comprise the steps:
S100: gather human head and shoulder model image sample; In this technical scheme, choose the image that comprises in a large number human body, and manually select the representative head shoulder images of a large amount of human head and shoulder partial geometry of cutting clear in structure, and and the non-head shoulder images sample storage of equal number, be used for step S200 sample storage and training.
S200: the human head and shoulder model image that gathers is carried out the PCA feature extraction, and send into training generation Sample Storehouse in the SVM training aids;
In this step, establish that the head shoulder images number of samples is N in the Sample Storehouse, each schedule of samples is shown S
i, i=1,2 ..., N.At first each sample image in the Sample Storehouse is drawn into a row vector, regards it as in high-dimensional data space data point, then adopt the PCA algorithm to ask for the major component vector of all data points in the Sample Storehouse, and send in the svm classifier device and train.Testing image is carried out Data Dimensionality Reduction, and classify by the svm classifier device, concrete, its step is as follows:
S201: human head and shoulder model image sample is stretched as one-dimensional vector by row, and for example, the image of 20 x, 20 sizes will be drawn as the row vector of 1 x, 400 sizes, and setting the image pattern database is high-dimensional data space point set X={x
i, i=1,2,3N;
S202: the mean vector m that asks for X;
Each vector among the S203:X deducts average m, then asks its covariance matrix, and carries out Eigenvalues Decomposition;
S204: several eigenwert characteristic of correspondence vectors before extracting, namely required that major component is vectorial, be designated as Γ
i, i=1,2 ... p(p is the low-dimensional data dimension that dimensionality reduction will obtain);
S205: the major component vector of trying to achieve is carried out dimensionality reduction, and send in the SVM training aids and classify, generate Sample Storehouse.
Testing image is repeated process among the step S200, try to achieve the major component vector, and classify with the svm classifier device, generate Sample Storehouse.After Sample Storehouse is set up and to be finished, carry out the collection that on-the-spot people on the spot takes on image, and the people who gathers is takeed on after image and Sample Storehouse compare, count into elevator and the concrete number that goes out elevator.It is specifically realized by following step:
S300: differentiate elevator door and whether open, if, gather current frame image by video capture devices such as video cameras, search the moving target in the current frame image, described moving target is carried out the PCA feature extraction, utilize the svm classifier device that moving target and Sample Storehouse are mated classification, if described moving target is the human body head shoulder images, then turn S400; If not, the S300 that then circulates; Consult Fig. 2, whether described detection elevator door is opened by the image that collects is carried out rim detection, comprising:
S321: whether the left-hand component of judging the defined area detects straight line; If, then turn S322, if not, judge that then elevator is as closing the door;
S322: whether the right-hand component of judging the defined area detects straight line; If, then turn S323, if not, judge that then elevator is as closing the door;
S323: whether the straight line of judging above the right and left is symmetrical about intermediate point, if, then judge door-opened elevator, if not, judge that then elevator closes the door.
In step S300, search the moving target in the current frame image, concrete is achieved by the following technical solution:
S311: gather current frame image, use inter-frame difference to obtain bianry image, judge whether described bianry image exists connected region, if exist, then turn S312; If do not exist, then continue to gather the next frame image.
S312: set up the kalman filter models that described bianry image is followed the tracks of, search the moving target in the image.The kalman filter models of setting up head shoulder model following is to use Kalman Prediction in image sequence, establishes motion state parameters (certain is position and the speed at the human body target place constantly) X of human body target
kFor:
State transition equation is:
Because the time interval between the consecutive frame image is very short, so the target state variation is less, can suppose that human body target is linear uniform motion in the unit interval, then system state transition matrix A in the tracing process
kCan be set as:
T is the time interval between adjacent two two field pictures.
Because the position of human body target in image can directly obtain, and can not observe directly his speed and acceleration, the state equation of system is described be the tracking characteristics point that detects in the position of the plane of delineation, its expression formula is:
Observation matrix H (k) can be made as:
W in the above formula
k, v
kBe the process noise that adds, can see the zero-mean white noise sequence as.
State vector systematic error covariance matrix Q is:
Measuring error covariance matrix R is:
According to above equation, can obtain the motion state parameters X of described human body target
kThereby, find the moving target in the image.
In step S300, the high Hough transformation method of degree of accuracy is adopted in described detection to straight line, considers a point (x
i, y
i), the slope-intercept form of an equation of straight line: y=ax+b.Write equation the form of b=-xa+y as and with reference to the ab plane, with obtain for fixed point (x
i, y
i) unique straight-line equation.If describe straight line with polar coordinates, that is: ρ=xcos θ+ysin θ, ρ's is the former distance of point to line of rectangular coordinate system in the formula, θ is the angle of straight line and x axle.
The Hough mapping algorithm is that (ρ, θ) plane is quantized, and each point in the parameter space distributes a counter, before the conversion each counter is initialized as 0, with every bit (x
k, y
k) quantized value of substitution θ, calculate each ρ, income value (after quantizing) must drop in certain little lattice, counter corresponding to little lattice adds 1 simultaneously, after the calculating of whole (x, y) point is complete, all little lattice are tested, namely corresponding to the collinear point of parameter space, its (ρ, θ) can be used as Straight Line Fitting Parameters (ρ to the little lattice of count value maximum
0, θ
0).The point of other subtotal numerical value abandons corresponding to non-colinear point.
S400: according to the coordinate relation of the eve of the tracking target in the testing image, current quarter and the coordinate line that presets, judge that target still goes out elevator for advancing elevator, the preferred switch of judging elevator door, can effectively reduce the calculated amount of algorithm, when elevator when closing the door state, can not produce the event of target turnover elevator this moment, Resurvey next frame image, when elevator door when opening state, then carry out the judgement of subsequent step.
If described moving target is for advancing elevator, then add 1 to advancing the elevator number, if described moving target is for going out elevator, then add 1 to going out the elevator number.
Concrete, if eve tracking target place coordinate below ruler, namely in the elevator, current time place coordinate is the ruler top, namely outside the elevator, then judges this target as going out elevator, described counting unit adds 1 to going out the elevator number; If eve tracking target place coordinate is the ruler top, namely in the elevator, current time place coordinate is the ruler below, namely in the elevator, then judges this target as advancing elevator, and described counting unit adds 1 to advancing the elevator number.
The on off state of technical scheme of the present invention based on the elevator door in the image after the method judgement rim detection of Hough transformation detection of straight lines, be under the prerequisite of door opening state at elevator, utilize the head shoulder model of human body, use the PCA-SVM combination algorithm to detect the human head and shoulder target; Utilize Kalman filtering to realize the tracking of human body, judge the turnover state of target, and add up the volume of the flow of passengers.The present invention is used for human head and shoulder identification in conjunction with PCA and two kinds of methods of SVM, identifies than single method, and recognition accuracy is higher; By the volume of the flow of passengers of the every day operation of statistics elevator, estimate that for the elevator maintenance personnel degree of wear of device provides foundation, and then formulate more effective elevator maintenance plan, reduce the generation of fault.
Consult Fig. 3, the present invention also provides a kind of elevator people flow rate statistical system based on the intelligent vision perception, comprising:
Information acquisition unit is used for gathering human head and shoulder model image sample, and what described information acquisition unit was concrete can be the video capture devices such as video camera, video recorder.
The Database unit carries out the PCA feature extraction to the human head and shoulder model image that information acquisition unit gathers, and sends into training generation Sample Storehouse in the SVM training aids.Described data are set up concrete the comprising in unit: the first numerical analysis unit, be used for human head and shoulder model image sample is stretched as one-dimensional vector by row, and setting the image pattern database is high-dimensional data space point set X={x
i, i=1,2,3N; Average is asked for the unit, is used for asking for the mean vector m of X; The second value analytic unit is used for each vector of X is deducted average m, then asks its covariance matrix, and carries out Eigenvalues Decomposition; Major component vector acquiring unit, namely required that major component is vectorial for several eigenwert characteristics of correspondence vectors before extracting, be designated as Γ
i, i=1,2 ... p(p is the low-dimensional data dimension that dimensionality reduction will obtain); The Sample Storehouse generation unit is used for the major component vector of trying to achieve is carried out dimensionality reduction, and sends in the SVM training aids and classify, and generates Sample Storehouse.
After Sample Storehouse generates, then, in real life, the people passes in and out quantity and adds up in the running process of elevator, when only elevator door is out just the someone pass in and out elevator, therefore to judge first whether described elevator opens the door, concrete pass through the function that judgement is realized in following unit, at first, whether the first judging unit differentiation elevator door is opened; This first judging unit comprises the first detecting unit, the second detecting unit and the 3rd detecting unit, and the first detecting unit judges whether the left-hand component of defined area detects straight line; If, then turn the second detecting unit, if not, judge that then elevator is as closing the door; The second detecting unit judges whether the right-hand component of defined area detects straight line; If, then turn the 3rd detecting unit, if not, judge that then elevator is as closing the door; The 3rd detecting unit judges whether the straight line of above the right and left is symmetrical about intermediate point, if, then judge door-opened elevator, if not, judge that then elevator closes the door.
After judging that elevator is door opening state, carry out the collection of human body image, collection to described human body image realizes by described data analysis unit, data analysis unit, be used for gathering current frame image, search the moving target in the current frame image, this data analysis unit comprises image acquisition units and moving target acquiring unit, described image acquisition units gathers current frame image, use inter-frame difference to obtain bianry image, judge whether described bianry image exists connected region, if exist, then search moving target in the image by the moving target acquiring unit, it is concrete is to realize that by setting up kalman filter models that bianry image follows the tracks of moving target searches; If do not exist, then described image acquisition units continues to gather the next frame image.
Described data analysis unit utilizes the svm classifier device that moving target and Sample Storehouse are mated classification also to described moving target is carried out the PCA feature extraction.After finding target by above-described moving target acquiring unit, classify by the PCA-SVM sorter, mate classification with Sample Storehouse, if the Sample Similarity in the moving target that gathers and the Sample Storehouse is higher, judge that then described moving target is the human body head shoulder images, after definite described moving target is the human body head shoulder images, judge that by statistic unit described moving target still goes out elevator for advancing elevator, what described statistic unit was concrete comprises: coordinate analysis unit and counting unit; The coordinate analysis unit concerns with the coordinate of the coordinate line that presets eve, current quarter that be used for to judge the tracking target in the testing image; If eve tracking target place coordinate is below ruler, namely in the elevator, current time place coordinate is the ruler top, namely outside the elevator, then judges this target as going out elevator, and then described counting unit adds 1 to going out the elevator number; If eve tracking target place coordinate is the ruler top, namely in the elevator, current time place coordinate is the ruler below, namely in the elevator, then judges this target as advancing elevator, and then described counting unit adds 1 to advancing the elevator number.
The above only is embodiments of the invention; be not so limit claim of the present invention; every equivalent structure or equivalent flow process conversion that utilizes instructions of the present invention and accompanying drawing content to do; or directly or indirectly be used in other relevant technical fields, all in like manner be included in the scope of patent protection of the present invention.
Claims (10)
1. the elevator people flow rate statistical method based on the intelligent vision perception is characterized in that, may further comprise the steps:
S100: gather human head and shoulder model image sample;
S200: the human head and shoulder model image sample that gathers is carried out the PCA feature extraction, and send into training generation Sample Storehouse in the SVM training aids;
S300: differentiate elevator door and whether open, if, gather current frame image, search the moving target in the current frame image, described moving target is carried out the PCA feature extraction, utilize the svm classifier device that moving target and Sample Storehouse are mated classification, if described moving target is the human body head shoulder images, then turn S400; If elevator does not open the door, S300 then circulates;
S400: judge that described moving target still goes out elevator for advancing elevator, if described moving target is for advancing elevator, then add 1 to advancing the elevator number, if described moving target is for going out elevator, then add 1 to going out the elevator number.
2. the elevator people flow rate statistical method based on the intelligent vision perception according to claim 1 is characterized in that, step S200, and concrete comprises the steps:
S201: human head and shoulder model image sample is stretched as one-dimensional vector by row, and setting the image pattern database is high-dimensional data space point set X={x
i, i=1,2,3N;
S202: the mean vector m that asks for X;
Each vector among the S203:X deducts average m, then asks its covariance matrix, and carries out Eigenvalues Decomposition;
S204: several eigenwert characteristic of correspondence vectors before extracting, namely required that major component is vectorial, be designated as Γ
i, i=1,2 ... p; Wherein, p is the low-dimensional data dimension that dimensionality reduction will obtain;
S205: the major component vector of trying to achieve is carried out dimensionality reduction, and send in the SVM training aids and classify, generate Sample Storehouse.
3. the elevator people flow rate statistical method based on the intelligent vision perception according to claim 1, it is characterized in that: step S300 specifically comprises:
S311: gather current frame image, use inter-frame difference to obtain bianry image, judge whether described bianry image exists connected region, if exist, then turn S312; If do not exist, then continue to gather the next frame image.
S312: set up the kalman filter models that described bianry image is followed the tracks of, search the moving target in the image.
4. the elevator people flow rate statistical method based on the intelligent vision perception according to claim 1, it is characterized in that: among the step S300, whether described detection elevator door is opened by the image that collects is carried out rim detection, comprising:
S321: whether the left-hand component of judging the defined area detects straight line; If, then turn next step, if not, judge that then elevator is as closing the door;
S322: whether the right-hand component of judging the defined area detects straight line; If, then turn next step, if not, judge that then elevator is as closing the door;
S323: whether the straight line of judging above the right and left is symmetrical about intermediate point, if, then judge door-opened elevator, if not, judge that then elevator closes the door.
5. the elevator people flow rate statistical method based on the intelligent vision perception according to claim 1 is characterized in that step S400 specifically comprises:
The coordinate relation of the coordinate line of judging eve, the current quarter of the tracking target in the testing image and presetting;
If eve tracking target place coordinate is below ruler, namely in the elevator, current time place coordinate is the ruler top, namely outside the elevator, judges that then this target is as going out elevator;
If eve tracking target place coordinate is the ruler top, namely in the elevator, current time place coordinate is the ruler below, namely in the elevator, judges that then this target is as advancing elevator.
6. the elevator people flow rate statistical method based on the intelligent vision perception according to claim 5 is characterized in that: described detection employing Hough transformation method to straight line.
7. the elevator people flow rate statistical system based on the intelligent vision perception is characterized in that, comprising:
Information acquisition unit is used for gathering human head and shoulder model image sample;
The Database unit is used for the human head and shoulder model image that gathers is carried out the PCA feature extraction, and sends into training generation Sample Storehouse in the SVM training aids;
Whether the first judging unit is used for differentiating elevator door and opens;
Data analysis unit is used for gathering current frame image, searches the moving target in the current frame image, described moving target is carried out the PCA feature extraction, utilize the svm classifier device that moving target and Sample Storehouse are mated classification, if described moving target is the human body head shoulder images, then turn statistic unit;
Statistic unit is used for judging that described moving target still goes out elevator for advancing elevator, if described moving target is for advancing elevator, then adds 1 to advancing the elevator number, if described moving target is for going out elevator, then adds 1 to going out the elevator number.
8. the elevator people flow rate statistical system based on the intelligent vision perception according to claim 7 is characterized in that described data are set up concrete the comprising in unit:
The first numerical analysis unit is used for human head and shoulder model image sample is stretched as one-dimensional vector by row, and setting the image pattern database is high-dimensional data space point set X={x
i, i=1,2,3N;
Average is asked for the unit, is used for asking for the mean vector m of X;
The second value analytic unit is used for each vector of X is deducted average m, then asks its covariance matrix, and carries out Eigenvalues Decomposition;
Major component vector acquiring unit, namely required that major component is vectorial for several eigenwert characteristics of correspondence vectors before extracting, be designated as Γ
i, i=1,2 ... p; Wherein, p is the low-dimensional data dimension that dimensionality reduction will obtain;
The Sample Storehouse generation unit is used for the major component vector of trying to achieve is carried out dimensionality reduction, and sends in the SVM training aids and classify, and generates Sample Storehouse.
9. the elevator people flow rate statistical system based on the intelligent vision perception according to claim 7 is characterized in that described data analysis unit comprises:
Image acquisition units is used for gathering current frame image, uses inter-frame difference to obtain bianry image, judges whether described bianry image exists connected region, if exist, then transports the moving-target acquiring unit; If do not exist, then continue to gather the next frame image.
The moving target acquiring unit is used for setting up the kalman filter models that described bianry image is followed the tracks of, and searches the moving target in the image.
Described data analysis unit also comprises:
The first detecting unit is used for judging whether the left-hand component of defined area detects straight line; If, then turn the second detecting unit, if not, judge that then elevator is as closing the door;
The second detecting unit is used for judging whether the right-hand component of defined area detects straight line; If, then turn the 3rd detecting unit, if not, judge that then elevator is as closing the door;
The 3rd detecting unit be used for to judge whether the straight line of above the right and left is symmetrical about intermediate point, if, then judge door-opened elevator, if not, judge that then elevator closes the door.
10. the elevator people flow rate statistical system based on the intelligent vision perception according to claim 7 is characterized in that what described statistic unit was concrete comprises: coordinate analysis unit and counting unit;
The coordinate analysis unit concerns with the coordinate of the coordinate line that presets eve, current quarter that be used for to judge the tracking target in the testing image;
If eve tracking target place coordinate is below ruler, namely in the elevator, current time place coordinate is the ruler top, namely outside the elevator, then judges this target as going out elevator, and then described counting unit adds 1 to going out the elevator number;
If eve tracking target place coordinate is the ruler top, namely in the elevator, current time place coordinate is the ruler below, namely in the elevator, then judges this target as advancing elevator, and then described counting unit adds 1 to advancing the elevator number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012103168626A CN102902971A (en) | 2012-08-31 | 2012-08-31 | Method and system for conducting statistics on elevator visitor flow based on intelligent visual perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012103168626A CN102902971A (en) | 2012-08-31 | 2012-08-31 | Method and system for conducting statistics on elevator visitor flow based on intelligent visual perception |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102902971A true CN102902971A (en) | 2013-01-30 |
Family
ID=47575192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012103168626A Pending CN102902971A (en) | 2012-08-31 | 2012-08-31 | Method and system for conducting statistics on elevator visitor flow based on intelligent visual perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102902971A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646250A (en) * | 2013-09-13 | 2014-03-19 | 魏运 | Pedestrian monitoring method and device based on distance image head and shoulder features |
| CN104063253A (en) * | 2014-07-07 | 2014-09-24 | 无锡智广厦科技有限公司 | Method for automatic statistics of passenger flow, all-in-one machines and distributed system for automatic statistics of passenger flow |
| CN105321187A (en) * | 2014-07-22 | 2016-02-10 | 无锡慧眼电子科技有限公司 | Pedestrian counting method based on head detection |
| CN106155298A (en) * | 2015-04-21 | 2016-11-23 | 阿里巴巴集团控股有限公司 | Man-machine recognition methods and device, the acquisition method of behavior characteristics data and device |
| CN106530324A (en) * | 2016-10-21 | 2017-03-22 | 华中师范大学 | Visual cortex mechanism simulated video object tracking method |
| CN107798409A (en) * | 2016-08-30 | 2018-03-13 | 中兴智能交通股份有限公司 | A kind of crowd massing Forecasting Methodology based on time series models |
| CN107986127A (en) * | 2017-11-20 | 2018-05-04 | 江苏省特种设备安全监督检验研究院 | The stranded detection method of passenger in a kind of elevator |
| CN109164813A (en) * | 2018-10-31 | 2019-01-08 | 长安大学 | A kind of unmanned express delivery robot elevator device and control method up and down |
| CN109977783A (en) * | 2019-02-28 | 2019-07-05 | 浙江新再灵科技股份有限公司 | Method based on the independent boarding detection of vertical ladder scene perambulator |
| CN110012046A (en) * | 2018-09-19 | 2019-07-12 | 浙江新再灵科技股份有限公司 | A kind of real-time supplying system of elevator card based on algorithm of target detection and method |
| CN110334674A (en) * | 2019-07-10 | 2019-10-15 | 哈尔滨理工大学 | A kind of tracking of plane free body track identification and prediction technique |
| CN110865415A (en) * | 2018-08-28 | 2020-03-06 | 浙江大华技术股份有限公司 | Security inspection method and device |
| CN110930432A (en) * | 2019-11-19 | 2020-03-27 | 北京文安智能技术股份有限公司 | Video analysis method, device and system |
| CN111967323A (en) * | 2020-07-16 | 2020-11-20 | 电子科技大学 | Electric power live working safety detection method based on deep learning algorithm |
| WO2021103423A1 (en) * | 2019-11-28 | 2021-06-03 | 深圳市商汤科技有限公司 | Method and apparatus for detecting pedestrian events, electronic device and storage medium |
| CN113516016A (en) * | 2021-04-21 | 2021-10-19 | 新疆爱华盈通信息技术有限公司 | Method and system for identifying opening and closing states of parking garage roller shutter |
| CN114054897A (en) * | 2021-12-14 | 2022-02-18 | 北京新风航天装备有限公司 | Robot intermittent welding method for linear welding seam |
| CN114241015A (en) * | 2022-02-28 | 2022-03-25 | 长沙莫之比智能科技有限公司 | Method and device for counting number of people going in and out based on millimeter wave radar |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040091137A1 (en) * | 2002-11-04 | 2004-05-13 | Samsung Electronics Co., Ltd. | System and method for detecting face |
| CN1931697A (en) * | 2006-09-29 | 2007-03-21 | 浙江工业大学 | Intelligent dispatcher for group controlled lifts based on image recognizing technology |
| KR100714112B1 (en) * | 2005-12-09 | 2007-05-02 | 한국전자통신연구원 | Face recognition and confirmation method and apparatus |
| CN101456501A (en) * | 2008-12-30 | 2009-06-17 | 北京中星微电子有限公司 | Method and apparatus for controlling elevator button |
| CN102609716A (en) * | 2012-01-10 | 2012-07-25 | 银江股份有限公司 | Pedestrian detecting method based on improved HOG feature and PCA (Principal Component Analysis) |
-
2012
- 2012-08-31 CN CN2012103168626A patent/CN102902971A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040091137A1 (en) * | 2002-11-04 | 2004-05-13 | Samsung Electronics Co., Ltd. | System and method for detecting face |
| KR100714112B1 (en) * | 2005-12-09 | 2007-05-02 | 한국전자통신연구원 | Face recognition and confirmation method and apparatus |
| CN1931697A (en) * | 2006-09-29 | 2007-03-21 | 浙江工业大学 | Intelligent dispatcher for group controlled lifts based on image recognizing technology |
| CN101456501A (en) * | 2008-12-30 | 2009-06-17 | 北京中星微电子有限公司 | Method and apparatus for controlling elevator button |
| CN102609716A (en) * | 2012-01-10 | 2012-07-25 | 银江股份有限公司 | Pedestrian detecting method based on improved HOG feature and PCA (Principal Component Analysis) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646250A (en) * | 2013-09-13 | 2014-03-19 | 魏运 | Pedestrian monitoring method and device based on distance image head and shoulder features |
| CN103646250B (en) * | 2013-09-13 | 2015-04-22 | 黄卫 | Pedestrian monitoring method and device based on distance image head and shoulder features |
| CN104063253A (en) * | 2014-07-07 | 2014-09-24 | 无锡智广厦科技有限公司 | Method for automatic statistics of passenger flow, all-in-one machines and distributed system for automatic statistics of passenger flow |
| CN105321187A (en) * | 2014-07-22 | 2016-02-10 | 无锡慧眼电子科技有限公司 | Pedestrian counting method based on head detection |
| CN106155298A (en) * | 2015-04-21 | 2016-11-23 | 阿里巴巴集团控股有限公司 | Man-machine recognition methods and device, the acquisition method of behavior characteristics data and device |
| CN107798409A (en) * | 2016-08-30 | 2018-03-13 | 中兴智能交通股份有限公司 | A kind of crowd massing Forecasting Methodology based on time series models |
| CN106530324A (en) * | 2016-10-21 | 2017-03-22 | 华中师范大学 | Visual cortex mechanism simulated video object tracking method |
| CN107986127A (en) * | 2017-11-20 | 2018-05-04 | 江苏省特种设备安全监督检验研究院 | The stranded detection method of passenger in a kind of elevator |
| CN110865415B (en) * | 2018-08-28 | 2024-03-22 | 浙江大华技术股份有限公司 | Security check method and device |
| CN110865415A (en) * | 2018-08-28 | 2020-03-06 | 浙江大华技术股份有限公司 | Security inspection method and device |
| CN110012046A (en) * | 2018-09-19 | 2019-07-12 | 浙江新再灵科技股份有限公司 | A kind of real-time supplying system of elevator card based on algorithm of target detection and method |
| CN109164813A (en) * | 2018-10-31 | 2019-01-08 | 长安大学 | A kind of unmanned express delivery robot elevator device and control method up and down |
| CN109977783A (en) * | 2019-02-28 | 2019-07-05 | 浙江新再灵科技股份有限公司 | Method based on the independent boarding detection of vertical ladder scene perambulator |
| CN109977783B (en) * | 2019-02-28 | 2021-06-04 | 浙江新再灵科技股份有限公司 | Method for detecting whether baby carriage takes advantage of ladder independently based on vertical ladder scene |
| CN110334674A (en) * | 2019-07-10 | 2019-10-15 | 哈尔滨理工大学 | A kind of tracking of plane free body track identification and prediction technique |
| CN110930432A (en) * | 2019-11-19 | 2020-03-27 | 北京文安智能技术股份有限公司 | Video analysis method, device and system |
| WO2021103423A1 (en) * | 2019-11-28 | 2021-06-03 | 深圳市商汤科技有限公司 | Method and apparatus for detecting pedestrian events, electronic device and storage medium |
| CN111967323A (en) * | 2020-07-16 | 2020-11-20 | 电子科技大学 | Electric power live working safety detection method based on deep learning algorithm |
| CN111967323B (en) * | 2020-07-16 | 2022-07-26 | 电子科技大学 | Electric power live working safety detection method based on deep learning algorithm |
| CN113516016A (en) * | 2021-04-21 | 2021-10-19 | 新疆爱华盈通信息技术有限公司 | Method and system for identifying opening and closing states of parking garage roller shutter |
| CN113516016B (en) * | 2021-04-21 | 2023-08-01 | 新疆爱华盈通信息技术有限公司 | Method and system for identifying opening and closing states of rolling gate of parking garage |
| CN114054897A (en) * | 2021-12-14 | 2022-02-18 | 北京新风航天装备有限公司 | Robot intermittent welding method for linear welding seam |
| CN114241015A (en) * | 2022-02-28 | 2022-03-25 | 长沙莫之比智能科技有限公司 | Method and device for counting number of people going in and out based on millimeter wave radar |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102902971A (en) | Method and system for conducting statistics on elevator visitor flow based on intelligent visual perception | |
| CN102163290B (en) | Method for modeling abnormal events in multi-visual angle video monitoring based on temporal-spatial correlation information | |
| CN103049740B (en) | Fatigue state detection method based on video image and device | |
| CN105488456B (en) | Method for detecting human face based on adaptive threshold adjustment rejection sub-space learning | |
| CN107871126A (en) | Model recognizing method and system based on deep-neural-network | |
| CN103049751A (en) | Improved weighting region matching high-altitude video pedestrian recognizing method | |
| CN111462488A (en) | Intersection safety risk assessment method based on deep convolutional neural network and intersection behavior characteristic model | |
| CN110084165A (en) | The intelligent recognition and method for early warning of anomalous event under the open scene of power domain based on edge calculations | |
| CN105303191A (en) | Method and apparatus for counting pedestrians in foresight monitoring scene | |
| CN107679516B (en) | Lower extremity movement recognition methods based on multiple dimensioned Gauss-Markov random field models | |
| CN103886305B (en) | Specific face searching method for grassroots policing, safeguard stability and counter-terrorism | |
| CN103400391A (en) | Multiple-target tracking method and device based on improved random forest | |
| CN110728252B (en) | Face detection method applied to regional personnel motion trail monitoring | |
| CN101726498B (en) | Intelligent detector and method of copper strip surface quality on basis of vision bionics | |
| CN103971106A (en) | Multi-view human facial image gender identification method and device | |
| CN102999951A (en) | Intelligent personnel attendance checking method based on wireless network received signal strength | |
| CN109948501A (en) | The detection method of personnel and safety cap in a kind of monitor video | |
| CN103902966A (en) | Video interaction event analysis method and device base on sequence space-time cube characteristics | |
| CN106710228B (en) | A kind of implementation method of passenger-cargo shunting traffic parameter monitoring system | |
| CN103150552B (en) | A kind of driving training management method based on number of people counting | |
| CN107590427A (en) | Monitor video accident detection method based on space-time interest points noise reduction | |
| CN105930808A (en) | Moving object tracking method based on vector boosting template updating | |
| CN111666823A (en) | Pedestrian re-identification method based on individual walking motion space-time law collaborative identification | |
| CN113450573A (en) | Traffic monitoring method and traffic monitoring system based on unmanned aerial vehicle image recognition | |
| CN104809455B (en) | Action identification method based on the ballot of discriminability binary tree |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130130 |