CN110009650A - A kind of escalator handrail borderline region crosses the border detection method and system - Google Patents

A kind of escalator handrail borderline region crosses the border detection method and system Download PDF

Info

Publication number
CN110009650A
CN110009650A CN201811568311.2A CN201811568311A CN110009650A CN 110009650 A CN110009650 A CN 110009650A CN 201811568311 A CN201811568311 A CN 201811568311A CN 110009650 A CN110009650 A CN 110009650A
Authority
CN
China
Prior art keywords
roi
image
border
quadrangle
region
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.)
Granted
Application number
CN201811568311.2A
Other languages
Chinese (zh)
Other versions
CN110009650B (en
Inventor
毕举
施行
王超
吴磊磊
朱鲲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang New Zailing Technology Co Ltd
Original Assignee
Zhejiang New Zailing Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang New Zailing Technology Co Ltd filed Critical Zhejiang New Zailing Technology Co Ltd
Priority to CN201811568311.2A priority Critical patent/CN110009650B/en
Publication of CN110009650A publication Critical patent/CN110009650A/en
Application granted granted Critical
Publication of CN110009650B publication Critical patent/CN110009650B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20092—Interactive image processing based on input by user
    • G06T2207/20104—Interactive definition of region of interest [ROI]

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

It crosses the border detection method the invention discloses a kind of escalator handrail borderline region, comprising the following steps: by video image acquisition unit, acquisition staircase runs video image immediately ahead of staircase, and acquired image data are used for subsequent cell analysis；By cross the border test and analyze unit according to video image acquisition unit be passed to video image information analyzed, obtain geofence signal value；By multi-media cues unit according to the alarm signal received, voice reminder information is exported.The present invention is only handled ROI region, improves processing accuracy and speed.

Description

A kind of escalator handrail borderline region crosses the border detection method and system
Technical field
The invention belongs to escalator security technology areas, and in particular to a kind of escalator handrail borderline region crosses the border the side of detection Method and system.
Background technique
Escalator is answered in each emporium, the place of subway station, railway station, airport et al. traffic-intensive at present With escalator is while bringing greatly convenient, as some improper use cause people's lives and properties Loss.Such as when hand is stretched out handrail boundary or when by leaning out boundary above the waist by someone, if there is shelter in front, if not playing It reminds voice or slows down the staircase speed of service, then passenger is likely to collide with it, endangers passenger's personal safety, even results in Subsequent passenger falls, and drops, and in turn results in Domino effect, causes more pernicious injures and deaths event.Therefore, escalator handrail is detected Borderline region crosses the border event, takes corresponding voice and security implementations at the first time, can be to avoid tragedy occurs.
101695983 A of Chinese invention patent application CN discloses a kind of helping automatically based on omnidirectional computer vision Section of ladder energy and safety monitoring system, including microprocessor, omnibearing vision sensor and escalator PLC controller.It is wherein micro- Processor includes: video image read module, region-of-interest customized module (which includes the determinations on handrail boundary), inspection of crossing the border Survey module and energy-saving control module, wherein cross the border detection module mainly for detection of escalator passenger and belongings It whether has been more than handrail boundary, it is high using edge detection algorithm and mixing to the full-view video image being stored in dynamic memory This background modeling algorithm carries out comprehensive judgement: when detecting non-handrail edge around handrail boundary, then passing through Gauss Foreground object modeling algorithm is confirmed whether to belong to foreground object, if detecting that the outer ledge of foreground object crosses escalator 5 pixels of handrail outer ledge more than, it is determined that crossing the border.The technical solution disadvantage is as follows: being carried out using entire image Detection and modeling, waste CPU and memory source, improve cost；The feelings at non-handrail edge are detected around handrail boundary Under condition, modeling judgement is just carried out, if inspection does not measure non-handrail edge, will appear the case where failing to report；The outer side edges of foreground object Edge crosses 5 pixels of handrail outer ledge of escalator or more it is determined that crossing the border, and decision condition is too simple, only considers empty Between factor do not consider the time, it may appear that more wrong report.
Summary of the invention
Cross the border detection method and system the technical problem to be solved in the present invention is to provide a kind of staircase, only to ROI region into Row processing, improves processing accuracy and speed.
In order to solve the above technical problems, the present invention adopts the following technical scheme that:
A kind of escalator handrail borderline region crosses the border detection method, comprising the following steps:
By video image acquisition unit, acquisition staircase runs video image immediately ahead of staircase, by acquired image number It is analyzed according to for subsequent cell；The unit video image information incoming according to video image acquisition unit is tested and analyzed by crossing the border It is analyzed, obtains geofence signal value；By multi-media cues unit according to the alarm signal received, exports voice and mention Awake information；It is described cross the border test and analyze unit analyze according to the video image information that video image acquisition unit is passed to it is specific The following steps are included:
Obtain region of interest ROI area image, including ROI rectangular region image and ROI quadrilateral area template；Using Gaussian modeling GMM method establishes gray level image background to the rectangular area ROI；Background difference and binaryzation；Binary map form Processing；Count the duration that prospect ratio value and area accounting in ROI quadrangle template are greater than threshold value；
If prospect ratio is greater than given threshold in present frame ROI quadrangle template, while being sustained for longer than setting threshold Value then sends warning of crossing the border, the region of security risk, the region is easy to appear on the outside of the rectangular area the ROI selection handrail Indicated with the quadrangle that four points form, the line of two of them point is parallel and close to handrail boundary, in addition the line of two o'clock to Outer separate handrail boundary.
Preferably, ROI rectangular region image is obtained specifically: according to the vertex of the quadrangle of input, find wherein minimum X coordinate value min_x and y-coordinate value min_y, maximum x coordinate value max_x and y-coordinate value max_y, to obtain ROI rectangle Region obtains subsequent ROI quadrilateral area template for convenience, and a pixel is expanded in the rectangular area ROI outward, then from source figure The subgraph of the rectangular area ROI is obtained as in.
Preferably, ROI quadrilateral area template is obtained specifically:
Quadrangle vertex will be inputted, draws straight line two-by-two in order, to surround closed ROI quadrangle, will be adopted when drawing straight line With Bresenham algorithm；To quadrangle operation fast line filling algorithm.
Preferably, using Gaussian modeling GMM method, gray level image background is established to the rectangular area ROI specifically:
An initial mean value, standard deviation and weight are specified for each pixel of gray level image；
It collects N frame image and obtains the mean value, standard deviation and weight of each pixel using Online EM algorithm；
It is detected since N+1 frame, the method for detection are as follows:
To each pixel: by all Gaussian kernels according to ω/σ descending sort；Selection meets the preceding M Gauss of following formula Core: M=argmin (ω/σ > T)；If the pixel value of current pixel point has a satisfaction in: if (| x- μ _ i |)/σ _ i < K It is considered that it is background dot；
Background image is updated, Online EM algorithm is used.
Preferably, background difference and binaryzation specifically:
Background image is subtracted with present frame, obtains difference image Pd；
Selected threshold th1 carries out binaryzation to difference image Pd, obtains binary map Bd.
A kind of escalator handrail borderline region crosses the border detection system, comprising:
Video image acquisition unit, for the acquisition staircase operation video image immediately ahead of staircase, by acquired image Data are analyzed for subsequent cell；It crosses the border and tests and analyzes unit, the video image for being passed to according to video image acquisition unit Information is analyzed, and geofence signal value is obtained；Multi-media cues unit, for according to the alarm signal received, output Voice reminder information；It is described cross the border test and analyze unit according to video image acquisition unit be passed to video image information divided Analysis specifically includes the following steps:
Obtain area-of-interest (Region Of Interest, ROI) area image, including ROI rectangular region image and ROI quadrilateral area template；Using Gaussian modeling GMM method, gray level image background is established to the rectangular area ROI；Background Difference and binaryzation；Binary map Morphological scale-space；Prospect ratio value and area accounting in statistics ROI quadrangle template are greater than The duration of threshold value；
If prospect ratio is greater than given threshold in present frame ROI quadrangle template, while being sustained for longer than setting threshold Value then sends warning of crossing the border, the region of security risk, the region is easy to appear on the outside of the rectangular area the ROI selection handrail Indicated with the quadrangle that four points form, the line of two of them point is parallel and close to handrail boundary, in addition the line of two o'clock to Outer separate handrail boundary.
Preferably, ROI rectangular region image is obtained specifically: according to the vertex of the quadrangle of input, find wherein minimum X coordinate value min_x and y-coordinate value min_y, maximum x coordinate value max_x and y-coordinate value max_y, to obtain ROI rectangle Region obtains subsequent ROI quadrilateral area template for convenience, and a pixel is expanded in the rectangular area ROI outward, then from source figure The subgraph of the rectangular area ROI is obtained as in.
Preferably, ROI quadrilateral area template is obtained specifically:
Quadrangle vertex will be inputted, draws straight line two-by-two in order, to surround closed ROI quadrangle, will be adopted when drawing straight line With Bresenham algorithm；To quadrangle operation fast line filling algorithm.
Preferably, using Gaussian modeling GMM method, gray level image background is established to the rectangular area ROI specifically:
An initial mean value, standard deviation and weight are specified for each pixel of gray level image；
It collects N frame image and obtains the mean value, standard deviation and weight of each pixel using Online EM algorithm；
It is detected since N+1 frame, the method for detection are as follows:
To each pixel: by all Gaussian kernels according to ω/σ descending sort；Selection meets the preceding M Gauss of following formula Core: M=argmin (ω/σ > T)；If the pixel value of current pixel point has a satisfaction in: if (| x- μ _ i |)/σ _ i < K It is considered that it is background dot；
Background image is updated, Online EM algorithm is used.
Preferably, background difference and binaryzation specifically:
Background image is subtracted with present frame, obtains difference image Pd；
Selected threshold th1 carries out binaryzation to difference image Pd, obtains binary map Bd.
Using the present invention with following the utility model has the advantages that only handling ROI region, the main ROI by input (feels Interest region) quadrangle four vertex, propose that a kind of fast line fill method for quadrangle obtains ROI quadrangle Then region template detects sport foreground to ROI rectangular region image mixed Gauss model, then calculates ROI quadrangle template Interior sport foreground area accounting and area accounting is greater than the duration of a certain given threshold, finally according to the duration Whether a certain given threshold is greater than to determine whether in the presence of crossing the border, and the alarm that will cross the border is sent to multi-media cues unit, multimedia Prompt unit carries out behavior induction according to the warning value received, to passenger on the scene；Meanwhile ladder control unit is helped according to warning value adjusting The terraced speed of service slows down or gradually stops, and avoids causing adverse consequences.
Detailed description of the invention
Fig. 1 is that the staircase of the embodiment of the present invention crosses the border the functional block diagram of detection system；
Fig. 2 is that the staircase of the embodiment of the present invention crosses the border ROI quadrangle photo schematic diagram in detection system；
Fig. 3 is that the staircase of the embodiment of the present invention rectangular area ROI in detection system of crossing the border determines photo schematic diagram；
Fig. 4 is that the staircase implemented of the present invention crosses the border ROI rectangular region image schematic diagram in detection system；
Fig. 5 is that the staircase of the embodiment of the present invention crosses the border the ROI quadrilateral area template schematic diagram of detection system；
Fig. 6 is that the staircase of the embodiment of the present invention crosses the border the fast line filling algorithm for quadrangle of detection system Flow chart；
Fig. 7 is that the staircase of the embodiment of the present invention crosses the border the corrosion principle schematic diagram of detection system；
Fig. 8 is that the staircase of the embodiment of the present invention crosses the border the dilating principle schematic diagram of detection system.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts Example, shall fall within the protection scope of the present invention.
Referring to figs. 1 to Fig. 8, a kind of staircase for showing the embodiment of the present invention crosses the border detection system, comprising:
Video image acquisition unit, for the acquisition staircase operation video image immediately ahead of staircase, by acquired image Data are analyzed for subsequent cell, and wherein video image acquisition unit includes but is not limited to CCD camera, web camera etc., It is installed on immediately ahead of staircase entrance, shoots the video image of staircase operation horizontally forward.
It crossing the border and tests and analyzes unit, the video image information for being passed to according to video image acquisition unit is analyzed, Geofence signal value is obtained, to test and analyze unit include but is not limited to CPU, ARM, DSP, GPU, FPGA, ASIC wherein crossing the border Equal general purpose processing devices.
Multi-media cues unit, for exporting voice reminder information according to the alarm signal received.Multi-media cues list Member including but not limited to liquid crystal display, the equipment that there is loudspeaker etc. video and audio to demonstrate one's ability, after it receives alarm signal, It plays on a display screen and reminds advice video information, advice voice reminder information is played in loudspeaker.
Wherein cross the border test and analyze unit according to video image acquisition unit be passed to video image information carry out analysis tool Body the following steps are included:
Obtain region of interest ROI area image, including ROI rectangular region image and ROI quadrilateral area template；Using Gaussian modeling GMM method establishes gray level image background to the rectangular area ROI；Background difference and binaryzation；Binary map form Processing；Count the duration that prospect ratio value and area accounting in ROI quadrangle template are greater than threshold value；
If prospect ratio is greater than given threshold in present frame ROI quadrangle template, while being sustained for longer than setting threshold Value then sends warning of crossing the border, the region of security risk, the region is easy to appear on the outside of the rectangular area the ROI selection handrail Indicated with the quadrangle that four points form, the line of two of them point is parallel and close to handrail boundary, in addition the line of two o'clock to Outer separate handrail boundary.
Obtain ROI rectangular region image specifically: according to the vertex of the quadrangle of input, find wherein the smallest x coordinate Value min_x and y-coordinate value min_y, maximum x coordinate value max_x and y-coordinate value max_y are to obtain the rectangular area ROI Facilitate and obtain subsequent ROI quadrilateral area template, a pixel is expanded into the rectangular area ROI outward, then obtain from source images Take the subgraph of the rectangular area ROI.
Obtain ROI quadrilateral area template specifically:
Quadrangle vertex will be inputted, draws straight line two-by-two in order, to surround closed ROI quadrangle, will be adopted when drawing straight line With Bresenham algorithm；To quadrangle operation fast line filling algorithm.
Using Gaussian modeling GMM method, gray level image background is established to the rectangular area ROI specifically:
An initial mean value, standard deviation and weight are specified for each pixel of gray level image；
It collects N frame image and obtains the mean value, standard deviation and weight of each pixel using Online EM algorithm；
It is detected since N+1 frame, the method for detection are as follows:
To each pixel: by all Gaussian kernels according to ω/σ descending sort；Selection meets the preceding M Gauss of following formula Core: M=argmin (ω/σ > T)；If the pixel value of current pixel point has a satisfaction in: if (| x- μ _ i |)/σ _ i < K It is considered that it is background dot；
Background image is updated, Online EM algorithm is used.
Background difference and binaryzation specifically:
Background image is subtracted with present frame, obtains difference image Pd；
Selected threshold th1 carries out binaryzation to difference image Pd, obtains binary map Bd.
In a specific application example, the staircase of the embodiment of the present invention detection system course of work of crossing the border is as follows:
A. video image acquisition unit runs video image, the ROI quadrangle top of input from staircase front acquisition staircase Point is as shown in Fig. 2, ROI quadrangle, the rectangular area ROI are illustrated in fig. 3 shown below, wherein 1 marked in Fig. 2,2,3,4 these tops Putting is the ROI quadrangle vertex inputted, and in Fig. 3, quadrangle is the ROI quadrilateral area drawn according to the vertex of input, square Shape is according to the calculated rectangular area ROI of input vertex, and specific calculation method, which is shown in cross the border to test and analyze, obtains ROI in unit Rectangular image module.
B. the image feeding of video image acquisition unit acquisition, which is crossed the border, tests and analyzes unit, and analytic process is as follows:
1. obtaining ROI region image, including ROI rectangular region image and ROI quadrilateral area template two parts.
1.1. ROI rectangular region image is obtained
According to the vertex of the quadrangle of input, wherein the smallest x coordinate value min_x and y-coordinate value min_y is found, it is maximum X coordinate value max_x and y-coordinate value max_y obtain subsequent ROI quadrangle area for convenience to obtain the rectangular area ROI A pixel is expanded in the rectangular area ROI by domain template outward, then the subgraph of the rectangular area ROI is obtained from source images, such as schemes In 4 shown in image.
1.2. ROI quadrilateral area template is obtained
In order to obtain ROI quadrilateral area template, propose a kind of fast line fill method for quadrangle come ROI quadrilateral area is filled, the memory and cpu resource of this method consumption are few, and space complexity is low, do not need the complicated number of building According to structure, time complexity is small, and algorithm, which only needs to be traversed for a ROI region, can be completed filling.And traditional orderly side table is swept The space complexity for retouching line completion method is higher, needs to establish side table (ET), the data structures such as Sorted Edge table (AET), while the time Complexity is also higher, needs to ask the intersection point of scan line and polygon, is related to floating point arithmetic, also right than relatively time-consuming Intersection point is ranked up, intersection point matches and section filling and etc., increase operation time.
1.2.1. quadrangle vertex will be inputted, straight line is drawn two-by-two in order, to surround closed ROI quadrangle.It draws straight Bresenham algorithm is used when line, the algorithm speed of service is fast, does not need to carry out floating point arithmetic.Assuming that the slope k > of straight line 0, for straight line in first quartile, Bresenham algorithm principle is as follows:
A. starting point (x is drawn1, y1).
B. prepare the next point of picture, x coordinate adds 1, if judgement reaches terminal, completes.Otherwise look for next point or For current point right abutment points or be current point upper right abutment points.By the distance phase that is put on this two points to straight line Subtract, judges that its is positive and negative, if following point is remote to straight line actual point distance, >=0 Δ d=d1-d2, then taking the point of top Y1+1, therefore can directly be judged to choose next point according to the symbol of Δ d, the computation rule of Δ d is as follows:
(1) initial value of Δ d is Δ d=2*dy-dx
(2) the Δ d=Δ d+2*dy as Δ d < 0
(3) the Δ d=Δ d+2*dy-2*dx as Δ d >=0
C. it draws a little
D. step b is jumped back to
E. terminate.
1.2.2. a kind of fast line filling algorithm is proposed for quadrangle, filled ROI quadrangle template is such as Shown in Fig. 5, flow chart is as shown in Figure 6, the specific steps are as follows:
A. by the row scanning rectangular area ROI；
B. using two variable record current pixel value cur_val and previous pixel value pre_val and a state Variable change_state fills to determine when beginning and end；
If it is not 0 that c. cur_val, which is 0, pre_val, then change_state increases by 1.It, will if changestate is 1 Current pixel value is set as 255, while saving x coordinate value xs when starting filling.If change_state is 2, stop The row scanning filling；
If d. change_state is 1 at this time, illustrates scan line and only one intersection point of quadrangle, need to fill out before The pixel filled dates back xs, sets 0 for pixel value；
E. scanning to entire ROI region terminates.
2. background modeling
Using GMM (Gaussian modeling) method, gray level image background is established to the rectangular area ROI.Mixed Gaussian background Modeling procedure is as follows
2.1. an initial mean value, standard deviation and weight are specified for each pixel of gray level image.
2.2. N (generally taking 200 or more, otherwise hardly result in decent result) frame image is collected to obtain using Online EM algorithm To the mean value, standard deviation and weight of each pixel.
2.3. it is detected since N+1 frame, the method for detection:
To each pixel:
2.3.1. by all Gaussian kernels according toDescending sort
2.3.2. selection meets the preceding M Gaussian kernel of following formula:
2.3.3. if the pixel value of current pixel point has a satisfaction in:It can be thought for back Sight spot.
2.4. background image is updated, Online EM algorithm is used.
3. background difference and binaryzation
3.1. background image is subtracted with present frame, obtains difference image Pd；
3.2. selected threshold th1, to difference image PdBinaryzation is carried out, binary map B is obtainedd。
4. binary map Morphological scale-space
4.1. to binary map BdIt carries out corrosion treatment and impurity point is gone using 3x3 template, obtain binary map Be；
Corrosion principle is as follows:
A convolution kernel B is defined first, and core (also known as template or exposure mask) can be any shapes and sizes, and possess One individually defines the reference point-anchor point (anchorpoint) come；
Then core B and image A is subjected to convolution, that is, calculates the pixel minimum value of the overlay area core B；
The minimum value is finally assigned to the specified pixel of reference point.
4.2. to binary map BeExpansion process is carried out, using 3x3 template, compensation expands original foreground part, obtains two-value Scheme Bf。
Dilating principle is as follows:
A convolution kernel B is defined first, and core (also known as template or exposure mask) can be any shapes and sizes, and possess One individually defines the reference point-anchor point (anchorpoint) come；
Then core B and image A is subjected to convolution, calculates the pixel maximum value of the overlay area core B；
This maximum value is finally assigned to the specified pixel of reference point.
5. counting the duration that prospect ratio value and area accounting in ROI quadrangle template are greater than threshold value.
The bianry image B that traversal is obtained by ROI rectangular region image by operations such as Gaussian modelingsfIf current picture Element is 255, while ROI quadrangle template corresponding position pixel value is 255, then the prospect number mask_fg_count in template increases Add 1, in order to count the pixel number mask_count in ROI quadrangle template, as long as ROI quadrangle template image corresponding position picture Element value is that 255, mask_count increases by 1.Then mask_fg_count/mask_count, as ROI quadrangle template are calculated Interior prospect ratio fg_ratio records its duration continue_time when fg_ratio is greater than given threshold.
6. if prospect ratio fg_ratio is greater than given threshold, while duration in present frame ROI quadrangle template Continue_time is greater than given threshold, then the warning that will cross the border is sent.
C, multi-media cues unit receives alarm signal, and multi-media cues unit plays induction voice, and reminding passengers pay attention to It not cross the border, careful front shelter slows down the staircase speed of service after ladder control unit receives alarm signal.If all not receiving Alarm signal, then staircase is in normal speed operating status, continues to keep.
Crossing the border with the staircase of the embodiment of the present invention, detection system is corresponding, and the embodiment of the invention also provides a kind of staircases to get over Boundary's detection method, comprising the following steps:
By video image acquisition unit, acquisition staircase runs video image immediately ahead of staircase, by acquired image number It is analyzed according to for subsequent cell；The unit video image information incoming according to video image acquisition unit is tested and analyzed by crossing the border It is analyzed, obtains geofence signal value；By multi-media cues unit according to the alarm signal received, exports voice and mention Awake information；It is described cross the border test and analyze unit analyze according to the video image information that video image acquisition unit is passed to it is specific The following steps are included:
Obtain region of interest ROI area image, including ROI rectangular region image and ROI quadrilateral area template；Using Gaussian modeling GMM method establishes gray level image background to the rectangular area ROI；Background difference and binaryzation；Binary map form Processing；Count the duration that prospect ratio value and area accounting in ROI quadrangle template are greater than threshold value；
If prospect ratio is greater than given threshold in present frame ROI quadrangle template, while being sustained for longer than setting threshold Value then sends warning of crossing the border, the region of security risk, the region is easy to appear on the outside of the rectangular area the ROI selection handrail Indicated with the quadrangle that four points form, the line of two of them point is parallel and close to handrail boundary, in addition the line of two o'clock to Outer separate handrail boundary.
Further, ROI rectangular region image is obtained specifically: according to the vertex of the quadrangle of input, find wherein most Small x coordinate value min_x and y-coordinate value min_y, maximum x coordinate value max_x and y-coordinate value max_y, to obtain ROI square Shape region obtains subsequent ROI quadrilateral area template for convenience, and a pixel is expanded in the rectangular area ROI outward, then from source The subgraph of the rectangular area ROI is obtained in image.
Further, ROI quadrilateral area template is obtained specifically:
Quadrangle vertex will be inputted, draws straight line two-by-two in order, to surround closed ROI quadrangle, will be adopted when drawing straight line With Bresenham algorithm；To quadrangle operation fast line filling algorithm.
Further, using Gaussian modeling GMM method, gray level image background is established to the rectangular area ROI specifically:
An initial mean value, standard deviation and weight are specified for each pixel of gray level image；
It collects N frame image and obtains the mean value, standard deviation and weight of each pixel using Online EM algorithm；
It is detected since N+1 frame, the method for detection are as follows:
To each pixel: by all Gaussian kernels according to ω/σ descending sort；Selection meets the preceding M Gauss of following formula Core: M=argmin (ω/σ > T)；If the pixel value of current pixel point has a satisfaction in: if (| x- μ _ i |)/σ _ i < K It is considered that it is background dot；
Background image is updated, Online EM algorithm is used.
Further, background difference and binaryzation specifically:
Background image is subtracted with present frame, obtains difference image Pd；
Selected threshold th1 carries out binaryzation to difference image Pd, obtains binary map Bd.
The above staircase corresponding specific implementation process of detection method of crossing the border is crossed the border detection system with above-mentioned staircase, different herein One repeats.
It should be appreciated that exemplary embodiment as described herein is illustrative and be not restrictive.Although being retouched in conjunction with attached drawing One or more embodiments of the invention is stated, it should be understood by one skilled in the art that not departing from through appended right In the case where the spirit and scope of the present invention defined by it is required that, the change of various forms and details can be made.

Claims (10)

  1. The detection method 1. a kind of escalator handrail borderline region crosses the border, which comprises the following steps:
    By video image acquisition unit, acquisition staircase runs video image immediately ahead of staircase, and acquired image data are used It is analyzed in subsequent cell；The unit video image information progress incoming according to video image acquisition unit is tested and analyzed by crossing the border Analysis obtains geofence signal value；By multi-media cues unit according to the alarm signal received, voice reminder letter is exported Breath；It is described cross the border test and analyze unit according to video image acquisition unit be passed to video image information carry out analysis specifically include Following steps:
    Obtain region of interest ROI area image, including ROI rectangular region image and ROI quadrilateral area template；Using mixing Gauss models GMM method, establishes gray level image background to the rectangular area ROI；Background difference and binaryzation；At binary map morphology Reason；Count the duration that prospect ratio value and area accounting in ROI quadrangle template are greater than threshold value；
    If prospect ratio is greater than given threshold in present frame ROI quadrangle template, while being sustained for longer than given threshold, then Transmission is crossed the border warning, the region of security risk is easy to appear on the outside of the rectangular area the ROI selection handrail, the region is with four The quadrangle of point composition indicates that the line of two of them point is parallel and close to handrail boundary, and in addition the line of two o'clock is outwardly away from Handrail boundary.
  2. The detection method 2. staircase hand borderline region according to claim 1 crosses the border, which is characterized in that obtain ROI rectangle region Area image specifically: according to the vertex of the quadrangle of input, wherein the smallest x coordinate value min_x and y-coordinate value min_y is found, Maximum x coordinate value max_x and y-coordinate value max_y obtains subsequent tetra- side ROI to obtain the rectangular area ROI for convenience A pixel is expanded in the rectangular area ROI by shape region template outward, then the subgraph of the rectangular area ROI is obtained from source images.
  3. The detection method 3. staircase hand borderline region according to claim 1 crosses the border, which is characterized in that obtain ROI quadrangle Region template specifically:
    Quadrangle vertex will be inputted, draws straight line two-by-two in order, to surround closed ROI quadrangle, will be used when drawing straight line Bresenham algorithm；To quadrangle operation fast line filling algorithm.
  4. The detection method 4. staircase hand borderline region according to claim 1 crosses the border, which is characterized in that built using mixed Gaussian Mould GMM method establishes gray level image background to the rectangular area ROI specifically:
    An initial mean value, standard deviation and weight are specified for each pixel of gray level image；
    It collects N frame image and obtains the mean value, standard deviation and weight of each pixel using Online EM algorithm；
    It is detected since N+1 frame, the method for detection are as follows:
    To each pixel: by all Gaussian kernels according to ω/σ descending sort；Selection meets the preceding M Gaussian kernel of following formula: M= argmin(ω/σ>T)；If the pixel value of current pixel point has a satisfaction in: (| x- μ _ i |)/σ _ i < K can think It is background dot；
    Background image is updated, Online EM algorithm is used.
  5. The detection method 5. staircase hand borderline region according to claim 1 crosses the border, which is characterized in that background difference and two-value Change specifically:
    Background image is subtracted with present frame, obtains difference image Pd；
    Selected threshold th1 carries out binaryzation to difference image Pd, obtains binary map Bd.
  6. The detection system 6. a kind of escalator handrail borderline region crosses the border characterized by comprising
    Video image acquisition unit, for the acquisition staircase operation video image immediately ahead of staircase, by acquired image data It is analyzed for subsequent cell；It crosses the border and tests and analyzes unit, the video image information for being passed to according to video image acquisition unit It is analyzed, obtains geofence signal value；Multi-media cues unit, for exporting voice according to the alarm signal received Prompting message；It is described cross the border test and analyze unit according to video image acquisition unit be passed to video image information carry out analysis tool Body the following steps are included:
    Obtain region of interest ROI area image, including ROI rectangular region image and ROI quadrilateral area template；Using mixing Gauss models GMM method, establishes gray level image background to the rectangular area ROI；Background difference and binaryzation；At binary map morphology Reason；Count the duration that prospect ratio value and area accounting in ROI quadrangle template are greater than threshold value；
    If prospect ratio is greater than given threshold in present frame ROI quadrangle template, while being sustained for longer than given threshold, then Transmission is crossed the border warning, the region of security risk is easy to appear on the outside of the rectangular area the ROI selection handrail, the region is with four The quadrangle of point composition indicates that the line of two of them point is parallel and close to handrail boundary, and in addition the line of two o'clock is outwardly away from Handrail boundary.
  7. The detection system 7. staircase hand borderline region according to claim 6 crosses the border, which is characterized in that obtain ROI rectangle region Area image specifically: according to the vertex of the quadrangle of input, wherein the smallest x coordinate value min_x and y-coordinate value min_y is found, Maximum x coordinate value max_x and y-coordinate value max_y obtains subsequent tetra- side ROI to obtain the rectangular area ROI for convenience A pixel is expanded in the rectangular area ROI by shape region template outward, then the subgraph of the rectangular area ROI is obtained from source images.
  8. The detection system 8. staircase hand borderline region according to claim 6 crosses the border, which is characterized in that obtain ROI quadrangle Region template specifically:
    Quadrangle vertex will be inputted, draws straight line two-by-two in order, to surround closed ROI quadrangle, will be used when drawing straight line Bresenham algorithm；To quadrangle operation fast line filling algorithm.
  9. The detection system 9. staircase hand borderline region according to claim 6 crosses the border, which is characterized in that built using mixed Gaussian Mould GMM method establishes gray level image background to the rectangular area ROI specifically:
    An initial mean value, standard deviation and weight are specified for each pixel of gray level image；
    It collects N frame image and obtains the mean value, standard deviation and weight of each pixel using Online EM algorithm；
    It is detected since N+1 frame, the method for detection are as follows:
    To each pixel: by all Gaussian kernels according to ω/σ descending sort；Selection meets the preceding M Gaussian kernel of following formula: M= argmin(ω/σ>T)；If the pixel value of current pixel point has a satisfaction in: (| x- μ _ i |)/σ _ i < K can think It is background dot；
    Background image is updated, Online EM algorithm is used.
  10. The detection system 10. staircase hand borderline region according to claim 6 crosses the border, which is characterized in that background difference and two Value specifically:
    Background image is subtracted with present frame, obtains difference image Pd；
    Selected threshold th1 carries out binaryzation to difference image Pd, obtains binary map Bd.
CN201811568311.2A 2018-12-20 2018-12-20 Escalator handrail boundary area border crossing detection method and system Active CN110009650B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811568311.2A CN110009650B (en) 2018-12-20 2018-12-20 Escalator handrail boundary area border crossing detection method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811568311.2A CN110009650B (en) 2018-12-20 2018-12-20 Escalator handrail boundary area border crossing detection method and system

Publications (2)

Publication Number Publication Date
CN110009650A true CN110009650A (en) 2019-07-12
CN110009650B CN110009650B (en) 2021-06-29

Family

ID=67165186

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811568311.2A Active CN110009650B (en) 2018-12-20 2018-12-20 Escalator handrail boundary area border crossing detection method and system

Country Status (1)

Country Link
CN (1) CN110009650B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110472550A (en) * 2019-08-02 2019-11-19 南通使爱智能科技有限公司 A kind of text image shooting integrity degree judgment method and system
CN111275910A (en) * 2019-11-05 2020-06-12 杭州安脉盛智能技术有限公司 Method and system for detecting border crossing behavior of escalator based on Gaussian mixture model
WO2021077878A1 (en) * 2019-10-22 2021-04-29 华为技术有限公司 Image processing method and apparatus, and electronic device
CN113344874A (en) * 2021-06-04 2021-09-03 温州大学 Pedestrian boundary crossing detection method based on Gaussian mixture modeling
WO2024125192A1 (en) * 2022-12-16 2024-06-20 中兴通讯股份有限公司 Safety alert method, terminal, and computer readable storage medium

Citations (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008094553A2 (en) * 2007-01-31 2008-08-07 Objectvideo, Inc. Target detection and tracking from video streams
CN101408987A (en) * 2007-10-12 2009-04-15 中兴通讯股份有限公司 Method for generating deuce-symmetrical straight line
CN101665210A (en) * 2008-09-03 2010-03-10 株式会社日立制作所 Image recognition device
CN101670973A (en) * 2008-09-09 2010-03-17 东芝电梯株式会社 Escalator monitoring system
CN101695983A (en) * 2009-10-23 2010-04-21 浙江工业大学 Omnibearing computer vision based energy-saving and safety monitoring system of escalator
CN101920603A (en) * 2010-08-06 2010-12-22 东莞市升力智能科技有限公司 NURBS figure laser broom method
CN102087707A (en) * 2009-12-03 2011-06-08 索尼株式会社 Image processing equipment and image processing method
CN105701760A (en) * 2016-01-12 2016-06-22 中国人民解放军国防科学技术大学 Histogram real-time generation method of geographic raster data optional polygon area
CN106503632A (en) * 2016-10-10 2017-03-15 南京理工大学 A kind of escalator intelligent and safe monitoring method based on video analysis
CN107273852A (en) * 2017-06-16 2017-10-20 华南理工大学 Escalator floor plates object and passenger behavior detection algorithm based on machine vision
CN107368786A (en) * 2017-06-16 2017-11-21 华南理工大学 A kind of passenger based on machine vision crosses handrail detection algorithm
US20180029835A1 (en) * 2016-07-29 2018-02-01 Otis Elevator Company Monitoring system of a passenger conveyor, a passenger conveyor, and a monitoring method thereof
CN107662872A (en) * 2016-07-29 2018-02-06 奥的斯电梯公司 The monitoring system and its monitoring method of passenger conveyor
CN108182381A (en) * 2017-12-05 2018-06-19 华南理工大学 Escalator occupant detection algorithm based on quick Adaboost training algorithms
CN108230305A (en) * 2017-12-27 2018-06-29 浙江新再灵科技股份有限公司 Method based on the detection of video analysis staircase abnormal operating condition
CN108639921A (en) * 2018-07-05 2018-10-12 江苏瑞奇海力科技有限公司 A kind of staircase passenger safety prior-warning device and method
CN108805093A (en) * 2018-06-19 2018-11-13 华南理工大学 Escalator passenger based on deep learning falls down detection algorithm

Patent Citations (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008094553A2 (en) * 2007-01-31 2008-08-07 Objectvideo, Inc. Target detection and tracking from video streams
CN101408987A (en) * 2007-10-12 2009-04-15 中兴通讯股份有限公司 Method for generating deuce-symmetrical straight line
CN101665210A (en) * 2008-09-03 2010-03-10 株式会社日立制作所 Image recognition device
CN101670973A (en) * 2008-09-09 2010-03-17 东芝电梯株式会社 Escalator monitoring system
CN101695983A (en) * 2009-10-23 2010-04-21 浙江工业大学 Omnibearing computer vision based energy-saving and safety monitoring system of escalator
CN102087707A (en) * 2009-12-03 2011-06-08 索尼株式会社 Image processing equipment and image processing method
CN101920603A (en) * 2010-08-06 2010-12-22 东莞市升力智能科技有限公司 NURBS figure laser broom method
CN105701760A (en) * 2016-01-12 2016-06-22 中国人民解放军国防科学技术大学 Histogram real-time generation method of geographic raster data optional polygon area
CN107662872A (en) * 2016-07-29 2018-02-06 奥的斯电梯公司 The monitoring system and its monitoring method of passenger conveyor
US20180029835A1 (en) * 2016-07-29 2018-02-01 Otis Elevator Company Monitoring system of a passenger conveyor, a passenger conveyor, and a monitoring method thereof
CN106503632A (en) * 2016-10-10 2017-03-15 南京理工大学 A kind of escalator intelligent and safe monitoring method based on video analysis
CN107273852A (en) * 2017-06-16 2017-10-20 华南理工大学 Escalator floor plates object and passenger behavior detection algorithm based on machine vision
CN107368786A (en) * 2017-06-16 2017-11-21 华南理工大学 A kind of passenger based on machine vision crosses handrail detection algorithm
CN108182381A (en) * 2017-12-05 2018-06-19 华南理工大学 Escalator occupant detection algorithm based on quick Adaboost training algorithms
CN108230305A (en) * 2017-12-27 2018-06-29 浙江新再灵科技股份有限公司 Method based on the detection of video analysis staircase abnormal operating condition
CN108805093A (en) * 2018-06-19 2018-11-13 华南理工大学 Escalator passenger based on deep learning falls down detection algorithm
CN108639921A (en) * 2018-07-05 2018-10-12 江苏瑞奇海力科技有限公司 A kind of staircase passenger safety prior-warning device and method

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
VOLKER A. PLIEM 等: "Crack Detectionon an Escalator Handrail", 《IEEE INSTRUMENTATIONAND MEASUREMENT TECHNOLOGY CONFERENCE》 *
卫洪春 等: "扫描线多边形区域填充算法研究", 《四川文理学院学报》 *
杜启亮 等: "基于Adaboost和码本模型的手扶电梯出入口视频监控方法", 《计算机应用》 *
汤一平 等: "基于计算机视觉的自动扶梯节能***", 《计算机测量与控制》 *
程滔 等: "Bresenham算法及其在遥感影像图形绘制中的应用", 《合肥工业大学学报(自然科学版)》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110472550A (en) * 2019-08-02 2019-11-19 南通使爱智能科技有限公司 A kind of text image shooting integrity degree judgment method and system
WO2021077878A1 (en) * 2019-10-22 2021-04-29 华为技术有限公司 Image processing method and apparatus, and electronic device
CN111275910A (en) * 2019-11-05 2020-06-12 杭州安脉盛智能技术有限公司 Method and system for detecting border crossing behavior of escalator based on Gaussian mixture model
CN113344874A (en) * 2021-06-04 2021-09-03 温州大学 Pedestrian boundary crossing detection method based on Gaussian mixture modeling
CN113344874B (en) * 2021-06-04 2024-02-09 温州大学 Pedestrian boundary crossing detection method based on Gaussian mixture modeling
WO2024125192A1 (en) * 2022-12-16 2024-06-20 中兴通讯股份有限公司 Safety alert method, terminal, and computer readable storage medium

Also Published As

Publication number Publication date
CN110009650B (en) 2021-06-29

Similar Documents

Publication Publication Date Title
CN110009650A (en) A kind of escalator handrail borderline region crosses the border detection method and system
EP2706507A1 (en) Method and apparatus for generating morphing animation
CN103049084A (en) Electronic device and method for adjusting display direction according to face direction
CN103235938A (en) Method and system for detecting and identifying license plate
US8131013B2 (en) Method and detecting system for positions of facial parts
CN103310422B (en) Obtain the method and device of image
EP3043343A1 (en) Information processing device, information processing method, and program
CN103413288A (en) LCD general defect detecting method
CN102799893A (en) Method for processing monitoring video in examination room
CN113554008B (en) Method and device for detecting static object in area, electronic equipment and storage medium
US9111346B2 (en) Image processing apparatus, image processing method, and recording medium
CN103065163B (en) A kind of fast target based on static images detects recognition system and method
CN105991885A (en) Image processing apparatus and image processing method
CN108416304B (en) Three-classification face detection method using context information
CN111695615A (en) Vehicle damage assessment method and device based on artificial intelligence, electronic equipment and medium
CN108229421A (en) A kind of falling from bed behavior real-time detection method based on deep video information
CN103049748A (en) Behavior-monitoring method and behavior-monitoring system
CN104282013B (en) A kind of image processing method and device for foreground target detection
CN117041484B (en) People stream dense area monitoring method and system based on Internet of things
CN113553922A (en) Mask wearing state detection method based on improved convolutional neural network
CN104156689B (en) Method and device for positioning feature information of target object
CN107316011A (en) Data processing method, device and storage medium
CN106943116A (en) A kind of infant eyesight automatic testing method
CN116206155A (en) Waste steel classification and identification method based on YOLOv5 network
CN112070859B (en) Photo image automatic filling method and system for photo book

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant