WO2020001302A1 - 一种基于视觉传感器的人流量统计方法、装置及系统 - Google Patents

一种基于视觉传感器的人流量统计方法、装置及系统 Download PDF

Info

Publication number
WO2020001302A1
WO2020001302A1 PCT/CN2019/091456 CN2019091456W WO2020001302A1 WO 2020001302 A1 WO2020001302 A1 WO 2020001302A1 CN 2019091456 W CN2019091456 W CN 2019091456W WO 2020001302 A1 WO2020001302 A1 WO 2020001302A1
Authority
WO
WIPO (PCT)
Prior art keywords
identification number
statistics
list
portrait
identification
Prior art date
Application number
PCT/CN2019/091456
Other languages
English (en)
French (fr)
Inventor
郑天航
林彬
颜王辉
Original Assignee
苏州欧普照明有限公司
欧普照明股份有限公司
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 苏州欧普照明有限公司, 欧普照明股份有限公司 filed Critical 苏州欧普照明有限公司
Publication of WO2020001302A1 publication Critical patent/WO2020001302A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/50Context or environment of the image
    • G06V20/52Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/53Recognition of crowd images, e.g. recognition of crowd congestion

Definitions

  • the present invention relates to the technical field of data statistics, and in particular, to a method, a device, and a system for counting human traffic based on a vision sensor.
  • the present invention provides a method and device for counting human traffic based on a vision sensor to overcome the above problems or at least partially solve the above problems.
  • a method for counting human traffic based on a vision sensor including:
  • Use vision sensors to collect image data in a specified area at a specific frequency
  • the identification number determined to be valid is counted so as to achieve statistics on the flow of people in the designated area.
  • the detecting each frame image in the image data, and when identifying that any frame image includes a portrait, assigning an identification number to the identified portrait includes:
  • an identification number is assigned to the identified portrait, and the identification number is recorded to a preset number statistics list.
  • the method further includes:
  • the coordinates and / or corresponding timestamps of the central pixel points of the respective portraits are read, and recorded with the identification number of each portrait to the numbered statistical list at the same time.
  • the analyzing the validity of the identification number based on the multi-frame image continuous with the any frame image includes:
  • identification number matches any of the numbers in the number statistics list, tracking data of the identification number in consecutive M frames of images;
  • the validity of the identification number is analyzed based on the data in the consecutive M frames of images.
  • the method further includes:
  • the analyzing the validity of the identification number based on data in the consecutive M frames of images includes:
  • the data of the identification number in consecutive M frames of images includes: the identification number of consecutive M frames The coordinates and / or timestamp of the central pixel of the portrait in each frame of the image;
  • the identification number is determined to be a valid number, and it is retained in the number statistics list.
  • the method further includes:
  • M is less than the preset minimum Mmin, determine that the corresponding identification number is an invalid number, and add it to the invalid statistics list or discard it;
  • M is greater than the preset maximum value Mmax, the moving distance of the central pixel of the identification number within consecutive M frames of images is obtained; if the moving distance is less than the first specified distance A, it is determined that the identification number is an invalid number , Add it to the invalid statistics list or discard it.
  • the method further includes:
  • the identification number does not match any of the numbers in the number statistics list, recording the coordinates of the central pixel of the portrait corresponding to the identification number and / or the current time stamp;
  • the identification number corresponding to the disappeared portrait is assigned as the identification number.
  • the method further includes:
  • identification number has an assignment record, record related data of the identification number to the number statistics list;
  • the identification number is not assigned a record, the identification number is retained as a valid number in the number statistics list.
  • the counting the identification numbers determined to be valid to implement statistics on the flow of people in the designated area includes:
  • the method further includes:
  • the statistical data is output in a specific format.
  • a vision sensor-based person flow statistics device including:
  • Vision sensor hardware module for collecting image data in a specified area at a specific frequency
  • a recognition component configured to detect each frame image in the image data, and when it is recognized that any frame image includes a portrait, assign an identification number to the recognized portrait;
  • An analysis component configured to analyze the validity of the identification number based on a plurality of frames of images continuous with the any frame of images
  • the statistics component is configured to count the identification numbers determined to be valid, so as to implement statistics on the flow of people in the specified area.
  • the identification component includes:
  • a detection unit configured to obtain image data collected by the vision sensor, and sequentially detect each frame of the image
  • a number assigning unit is configured to assign an identification number to the identified portrait when it is identified that a portrait is included in any frame image, and record the identification number to a preset number statistics list.
  • the identification component further includes:
  • the recording unit is configured to read the coordinates and / or corresponding timestamps of the central pixel points of the identified individual portraits, and record them into the numbered statistical list simultaneously with the identification numbers of the individual portraits.
  • the analysis component includes:
  • a comparison unit configured to compare the identification number with the number statistical list
  • a tracking unit configured to track data of the identification number in consecutive M frames of images when the identification number matches any of the numbers in the number statistical list
  • a validity analysis unit is configured to analyze the validity of the identification number based on data in the consecutive M frames of images.
  • the comparison unit is further configured to compare the identification number with a preset invalid number list before performing comparison with the numbered statistics list; if the identification number is ineffective with the invalid number list If any number in the number list matches, it is determined that the identification number is an invalid identification number;
  • the validity analysis unit is further configured to obtain data of the identification number in consecutive M frames of images, and determine whether M is within a specified numerical range; where the identification number is in consecutive M frames
  • the data in the image includes: the coordinates and / or timestamps of the central pixel points of the portrait in each frame of the consecutive M frames of the identification number;
  • the identification number is determined to be a valid number, and it is retained in the number statistics list.
  • the validity analysis unit is further configured to determine that the corresponding identification number is an invalid number when M is less than the preset minimum value Mmin, and add it to the invalid statistics list or discard it;
  • the moving distance of the central pixel of the identification number within consecutive M frames of images is obtained; when the moving distance is less than the first specified distance A, it is determined that the identification number is an invalid number , Add it to the invalid statistics list or discard it.
  • the analysis component further includes:
  • An assigning unit configured to record, when the identification number does not match any of the numbers in the numbered statistical list, the center pixel point coordinates and / or the current time stamp of the portrait corresponding to the portrait;
  • the identification number corresponding to the disappeared portrait is assigned as the identification number.
  • the validity analysis unit is further configured to, after obtaining a moving distance of the identification number within consecutive M frames, determine that the moving distance is greater than the first specified distance A, perform statistics according to the number A list to determine whether there is an assignment record for the identification number;
  • identification number has an assignment record, record related data of the identification number to the number statistics list;
  • the identification number is not assigned a record, the identification number is retained as a valid number in the number statistics list.
  • the statistics component is further configured to count the identification numbers included in the numbered statistics list, so as to implement statistics on the flow of people in the designated area.
  • the vision sensor module includes:
  • the lens is used for imaging the specified area and collecting light onto the vision sensor.
  • the above device further includes: a main processor connected to the vision sensor, the identification component, the analysis component, and the statistics component, and configured to manage and / or data of the vision sensor, the identification component, the analysis component, and the statistics component analysis.
  • a main processor connected to the vision sensor, the identification component, the analysis component, and the statistics component, and configured to manage and / or data of the vision sensor, the identification component, the analysis component, and the statistics component analysis.
  • the above device further includes: an output component, configured to output the statistical data in a specific format.
  • a vision sensor-based human traffic statistics system for performing human traffic statistics on a region to be detected having a plurality of sub-regions, wherein each of the sub-regions is provided with any one of the foregoing.
  • the visual sensor-based person flow statistics device according to the item.
  • the system further includes: a cloud server, configured to receive and store statistical data transmitted by the sub-areas, and count the human traffic in the area to be detected.
  • a cloud server configured to receive and store statistical data transmitted by the sub-areas, and count the human traffic in the area to be detected.
  • the method further includes: an access terminal, configured to obtain and view the statistics of human traffic in each sub-region stored by the cloud server; wherein the access terminal includes a terminal client program.
  • the present invention provides a method, device, and system for counting human traffic based on a vision sensor. After acquiring image data in a specified area by using the vision sensor, each frame of image can be identified. Assign an identification number to each portrait.
  • the human traffic statistics method provided by the present invention adds an analysis step after identifying a portrait, to further effectively analyze each identification number, and to count the human traffic by counting the effective identification numbers, thereby extremely Greatly improved the accuracy of people flow statistics.
  • FIG. 1 is a schematic flowchart of a method for counting human traffic based on a vision sensor according to an embodiment of the present invention
  • FIG. 2 is a schematic diagram of defining a center point coordinate of a portrait according to an embodiment of the present invention
  • FIG. 3 is a schematic flowchart of a method for analyzing validity of an identification number according to an embodiment of the present invention
  • FIG. 4 is a schematic flowchart of a method for counting human flow based on a vision sensor according to a preferred embodiment of the present invention
  • FIG. 5 is a schematic structural diagram of a device for counting human traffic based on a vision sensor according to an embodiment of the present invention
  • FIG. 6 is a schematic structural diagram of a human traffic statistics device based on a vision sensor according to a preferred embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram of a human traffic statistics system based on a vision sensor according to an embodiment of the present invention.
  • FIG. 1 is a schematic flowchart of a vision sensor-based human flow statistics method according to an embodiment of the present invention.
  • a vision sensor-based human flow statistics method according to an embodiment of the present invention may include:
  • Step S102 Use a vision sensor to collect image data in a specified area at a specific frequency
  • Step S104 Detect each frame image in the image data, and when it is recognized that any frame image includes a portrait, assign an identification number to the identified portrait;
  • Step S106 Analyze the validity of the identification number based on the multi-frame image continuous with any frame image
  • step S108 the identification numbers determined to be valid are counted, so as to implement statistics on the flow of people in the designated area.
  • An embodiment of the present invention provides a method for counting human traffic based on a vision sensor. After acquiring image data in a specified area by using the vision sensor, each frame of the image can be identified. Portraits are assigned identification numbers.
  • the method for counting human traffic provided in this embodiment also adds an analysis step after identifying a portrait to further effectively analyze each identification number, and to count the number of people through counting the effective identification numbers, thereby Greatly improved the accuracy of people flow statistics.
  • the vision sensor when it collects image data, it can acquire at a specific frequency, such as continuously or uninterruptedly acquiring image data or video data of a specified area, or performing periodic acquisition every 1 second.
  • a specific frequency such as continuously or uninterruptedly acquiring image data or video data of a specified area, or performing periodic acquisition every 1 second.
  • the present invention No restrictions.
  • the above step S104 may further include: first, acquiring the image data collected by the vision sensor, sequentially detecting each frame image in the image data, and identifying the portrait included in each frame image; and The identified portrait is assigned an identification number, and the identification number is recorded to a preset number statistics list.
  • assigning the identification number it may be sequentially assigned as a natural number starting from 1, or may be assigned according to other rules, which is not limited in the present invention.
  • the detection may be started from the first frame of the collected image data, or may be detected from any one of the frames according to user requirements. Because the vision sensor collects image data at a specific frequency, after determining the initial frame image that needs to be detected, it can always detect and identify subsequent image frames, or the user determines the time point at which the recognition detection ends or Image frame.
  • each identification number in the numbered statistics list in this embodiment corresponds to the identified portrait.
  • the corresponding identification number can be added to the numbered statistics list. That is to say, in this embodiment, all the assigned identification numbers are defaulted as valid numbers. Since the validity is further judged later, when counting the flow of people in the specified area, the number is directly listed in the statistics list The identification number can be counted without causing double counting of the number of persons.
  • the coordinates of the central pixel point may be the coordinates of the diagonal focus of the box marked on the identified portrait in the entire image after recognizing the presence of the portrait in the image using the portrait recognition algorithm.
  • the coordinates of the center pixel point of the portrait, and the corresponding time stamp can be the time stamp of the image frame where the portrait is located.
  • the coordinates of the center pixel can be expressed in the form of (X, Y), where X represents the coordinates of the pixels on the abscissa and Y represents the coordinates of the pixels on the vertical direction. Assuming that the image pixels are 640 ⁇ 480, the value range of X is 0. ⁇ 640, Y ranges from 0 to 480.
  • the timestamp can be represented by T. The above is only a schematic list of the recording format of the coordinates and timestamp of the central pixel point. In actual applications, the recording may be performed in other ways, which is not limited in the present invention.
  • FIG. 3 is a schematic flowchart of a method for analyzing the validity of an identification number according to an embodiment of the present invention. As shown in FIG. 3, in this embodiment, identifying the validity of a number may include:
  • Step S302 comparing the identification number with the number statistics list
  • Step S304 If the identification number matches any number in the numbered statistics list, track data of the identification number in consecutive M frames of images;
  • step S306 the validity of the identification number is analyzed based on the data in the consecutive M frames of images.
  • the identification number After the identification number is assigned to the identified portrait, the identification number will be recorded in the number statistics list first. Therefore, when judging the validity of any identification number, you can first compare it with the number in the number statistics list. Make a match. It should be noted that, because the identification number is analyzed based on continuous images in this embodiment, if a portrait is recognized in a certain frame and the identification number is assigned, the same portrait is detected and recognized in consecutive multiple frames of images. At this time, it is equivalent to tracking the portrait with the same identification number, without repeatedly assigning its identification number. For example, the identification number 1 is assigned to the identified portrait for the first time.
  • the identification number needs to be combined with the identification number to distinguish the portrait at different times, and then the time when the subsequent portrait appears Use 1-1,1-2 ... 1-n numbering. Therefore, if an identification number matches any of the numbers in the number statistics list, it means that the identification number has been tracked. At this time, the data of the identification number in subsequent consecutive M-frame images can continue to be tracked, and then based on consecutive M-frame images. The data in the analysis of the validity of the identification number.
  • misrecognition may often occur.
  • the misrecognized object is usually a static object.
  • the forward and backward motion characteristics of the object are analyzed and identified (such as within a specified time period). (No displacement), so as to determine whether it is misidentification.
  • identification numbers may be assigned during the detection and identification process, and these identification numbers are not displaced in the subsequent processes. At this time, the above identification numbers can be unifiedly managed and an invalid number list can be created and managed uniformly.
  • the method may further include:
  • Step S308 comparing the identification number with a preset invalid number list
  • Step S310 if the identification number matches any number in the invalid number list, determine that the identification number is an invalid identification number;
  • step S312 if the identification number does not match any number in the invalid number list, step S302 is performed to compare with the number statistics list.
  • the number in the invalid number list may be a number that is stationary within a certain period of time without displacement. That is, if an identification number is assigned to a still portrait (such as a model model of a store, etc.), it is determined that the portrait is always in a static state through subsequent consecutive frames of images, indicating that the identification of the portrait is incorrectly identified, and The assigned identification number is added to the invalid statistics list. When the still portrait is identified again and assigned an identification number, it only needs to be compared with the invalid number list first. If it matches any of the identification numbers in the invalid number list, , It will be discarded directly to improve the accuracy of analysis and identification.
  • a still portrait such as a model model of a store, etc.
  • step S306 the validity of the identification number is analyzed based on the data in consecutive M frames of images.
  • it may include:
  • the data of identification numbers in consecutive M frames of images include: identification numbers in each frame of consecutive M frames of images
  • the coordinates and / or timestamp of the central pixel of the portrait preferably, the coordinates of the pixels may include the coordinates of the pixels in the X and Y directions; wherein the data of the identification number in consecutive M frames of images can be understood as the identification number If the number of M frames that appear consecutively is too small, the identification number may not correspond to a portrait. If the number of M frames appears too many, it indicates that the identification number corresponds to a still image. Need further judgment. Therefore, the obtained range of M can be judged to further determine the validity of the identification number.
  • the identification number is judged to be a valid number, and it is retained in the number statistics list.
  • an identification number X when analyzing the validity of an identification number X, if a portrait corresponding to the identification number X is tracked within the range of Mmin to Mmax, it means that the identification number X is a valid number and can be retained in Numbered statistics list.
  • the corresponding identification number may be judged to be an invalid number, and it may be added to the invalid statistics list or discarded, indicating that the identification number X does not belong to the statistical range, and invalid processing is performed at this time.
  • M is greater than a preset maximum value Mmax, the moving distance of the central pixel of the identification number in consecutive M frames of images is obtained, and the moving distance is determined.
  • Mmax a preset maximum value
  • the moving distance is less than the first specified distance A, indicating that it may be a still portrait, then it is determined that the identification number is an invalid number, and it is added to the invalid statistics list or discarded.
  • the process of portrait recognition generally, after taking a frame of image, it will be identified according to the characteristics of the face. For a feature, the same ID number (that is, the identification number) will be assigned, and it will be recognized again in subsequent frames of the frame. When the same feature is reached, this ID will be matched again. However, when a person is in motion, for the same person, actions such as face rotation may cause different detected feature values. At this time, the system may assign a new ID to this person. At this time, the ID of the portrait needs to be re-assigned, that is, when an ID disappears and a new ID appears, both of them are less than a fixed value in time and / or position, then it can be determined that the two IDs correspond to the same ID. Personal portrait.
  • an identification number ID1 is assigned first, but the same feature value is not tracked in the next frame image, and after the next frame or multiple frames, for The same portrait feature value may be assigned as the new portrait with the identification number ID2 and recorded.
  • ID1 and ID2 correspond to the same portrait.
  • the time and / or position of ID1 and ID2 can be judged.
  • ID1 and ID2 can be considered.
  • the ID2 is recorded as the identification number of the corresponding portrait of ID1
  • ID2 is replaced with ID1 to update the number statistics list.
  • the method may further include:
  • Step S314 if the identification number does not match any of the numbers in the numbered statistics list, record the coordinates and / or the current time stamp of the central pixel of the portrait corresponding to the identification number;
  • step S316 it is determined whether the distance between the center pixel point of the disappeared portrait in the last frame image and the current center pixel point of the portrait corresponding to the identification number is smaller than the second specified distance B within the specified time interval;
  • Step S320 if it does not exist, continue to analyze the next frame image.
  • the above process details the conditions and processes for reassigning numbers.
  • a certain time interval if there is a change in the position of the disappeared portrait and the newly identified portrait within a certain range, you can determine that the change in the position is the same person portrait The generated distance changes. At this time, it can be considered that the newly identified portrait is the same person as the disappeared portrait, and then the number can be re-assigned.
  • the numbers recorded in the numbered statistics list are all valid numbers. Therefore, when counting the flow of people in a specified area, the identification numbers included in the numbered statistics list can be calculated. Further, after the statistics of the flow of people in the designated area, the statistical data can also be output in a specific format for users to view at any time.
  • the method for counting human traffic based on a vision sensor in this preferred embodiment may include:
  • Step S401 After the work is started, the visual sensor collected by the visual sensor captures image or video data, and then performs human detection and tracking on the above data;
  • Step S402 identifying whether there is any person in the initial frame image of the image data
  • step S404 it is compared with the pre-created invalid ID list to determine whether it matches any one of the IDs; if they match, step S405 is performed; if they do not match, step S406 is performed;
  • Step S406 compare with the ID statistics list; if it matches IDn, start to execute step S407, track the movement of subsequent IDn, and record the time point and position of each frame; if not, go to step S419;
  • Step S407 tracking the movement of the subsequent IDn and assigning IDn_0;
  • Step S408 record the position of IDn_0, the coordinates of the pixel points in the X and Y directions: (Xn0, Yn0) and the time stamp Tn0;
  • Step S409 Continue to track the movement of IDn in the next frame of image and assign IDn_1;
  • Step S410 Record the position of IDn_1, the coordinates of the pixel points in the X and Y directions: (Xn1, Yn1) and the time stamp Tn1;
  • step S411 the M-th frame data continuously appearing in IDn is tracked, and IDn_m is assigned; the M can be adjusted according to different application scenarios, which is not limited by the present invention.
  • Step S412 Record the position of IDn_m, the coordinates of the pixel points in the X and Y directions: (Xnm, Ynm) and the time stamp Tnm;
  • step S413 the validity of IDn is determined after the data of M frames in which IDn appears consecutively. In this embodiment, it is first determined whether M is less than 10; if yes, step S414 is performed; if not, step S416 is performed;
  • step S414 it is determined whether M is greater than 3. If yes, the IDn is considered valid, and step S415 is performed; if not, it indicates that the IDn is an invalid ID, added to the invalid ID list or discarded;
  • Step S415 the output IDn is a valid ID, that is, it is retained in the ID statistics list;
  • step S417 If yes, go to step S417; if not, it means that the IDn is an invalid ID, add it to the invalid ID list or discard it;
  • Step S417 it is determined whether there is an ID assignment record; if so, step S418 is performed; if not, step S415 is performed;
  • Step S418, record the latest data of the identification number in the ID statistics list
  • Step S419 the above is the data processing required for IDn after the recognition program recognizes the ID and continues to track IDn.
  • the recognition program does not track it in the next frame of image IDn, it will judge that the portrait corresponding to IDn disappears, and record the last position (Xn, Yn) of IDn in the image and the time stamp Tn.
  • the recognition program does not track it in the next frame of image IDn, it will judge that the portrait corresponding to IDn disappears, and record the last position (Xn, Yn) of IDn in the image and the time stamp Tn.
  • Xn, Yn last position of IDn in the image and the time stamp
  • Step S420 recording the position (Xm, Ym) of the IDm in the image and the time stamp Tm;
  • Step S421 using Tm-Tn ⁇ 2 seconds as an example, to determine whether the position change satisfies the assignment condition; if so, execute step S422 to determine whether the position change satisfies the assignment condition; if not, it means that no assignment is required, and the tracking is continued just
  • step S424 if the assignment conditions are not satisfied, the process proceeds to reading the next frame of image for judgment and analysis.
  • an ID is assigned to the identified portrait. Further, the validity of the ID is judged according to the subsequent images to determine the final effective ID for statistics. In this preferred embodiment, not only invalid identification of the IDs that are incorrectly identified, but also the situation of multiple IDs that exist for an identified person can be eliminated to further improve the accuracy of traffic statistics.
  • an embodiment of the present invention further provides a visual sensor-based person flow statistics device 100.
  • the visual sensor-based person flow statistics device may include:
  • the vision sensor 10 is configured to collect image data in a specified area at a specific frequency
  • the identification component 20 is used for detecting each frame image in the image data. When it is recognized that any frame image includes a portrait, an identification number is assigned to the identified portrait; the identification component 20 may use an image recognition algorithm for the image data Each frame of the image is detected to identify whether a portrait exists in the image;
  • An analysis component 30 configured to analyze the validity of the identification number based on multiple frames of images that are continuous with any one frame of images
  • the statistics component 40 is configured to count the identification numbers determined to be valid, so as to implement statistics on the flow of people in the designated area.
  • the analysis component 30 may directly perform statistics after analyzing the validity of the numbers, without setting the statistics component 40 separately.
  • the identification component 20 may include:
  • the detecting unit 21 is configured to acquire image data collected by a vision sensor, and sequentially detect each frame of the image;
  • the number assigning unit 22 is configured to assign an identification number to the identified portrait when it is identified that a portrait is included in any frame image, and record the identification number to a preset number statistics list.
  • the recording unit 23 is configured to read the coordinates and / or corresponding timestamps of the central pixel points of the identified individual portraits, and record them into the number statistics list together with the identification numbers of the individual portraits.
  • the analysis component 30 may include:
  • a comparison unit 31 configured to compare the identification number with the number statistics list
  • a tracking unit 32 configured to track data of the identification number in consecutive M frames of images when the identification number matches any number in the number statistics list;
  • the validity analysis unit 33 is configured to analyze the validity of the identification number based on data in consecutive M frames of images.
  • the comparison unit 31 is further configured to compare the identification number with a preset invalid number list before performing comparison with the number statistics list; if the identification number matches any number in the invalid number list If it matches, it is determined that the identification number is an invalid identification number; if the identification number does not match any of the numbers in the invalid number list, it is compared with the number statistics list.
  • the validity analysis unit 33 is further configured to obtain data of the identification number in consecutive M frames of images, and determine whether M is within a specified numerical range.
  • the data of the identification numbers in consecutive M frames of images includes: identification numbers Coordinates and / or timestamps of the central pixel of the portrait in each frame of consecutive M frames of images; when M is greater than a preset minimum value Mmin and less than a preset maximum value Mmax, the identification number is determined to be a valid number, and It remains in the numbered statistics list.
  • the validity analysis unit 33 is further configured to judge that the corresponding identification number is an invalid number when M is less than a preset minimum value Mmin, and add it to the invalid statistics list or discard it; when M is greater than the preset maximum value Mmax, obtain The moving distance of the central pixel of the identification number in consecutive M frames of images; when the moving distance is less than the first specified distance A, the identification number is determined to be an invalid number, and it is added to the invalid statistics list or discarded.
  • the analysis component 30 may further include: an assigning unit 34 configured to record the coordinates of the central pixel of the portrait corresponding to the portrait and / or the current number when the identification number does not match any number in the number statistics list. Timestamp; determines whether the distance between the center pixel point of the disappeared portrait in the last frame of the image and the current center pixel point of the portrait corresponding to the identification number is less than the second specified distance B within the specified time interval; if it exists, then Assign the identification number corresponding to the disappeared portrait as the identification number.
  • an assigning unit 34 configured to record the coordinates of the central pixel of the portrait corresponding to the portrait and / or the current number when the identification number does not match any number in the number statistics list. Timestamp; determines whether the distance between the center pixel point of the disappeared portrait in the last frame of the image and the current center pixel point of the portrait corresponding to the identification number is less than the second specified distance B within the specified time interval; if it exists, then Assign the identification number corresponding to the disappeared portrait as
  • the validity analysis unit 33 is further configured to, after acquiring the moving distance of the identification number within consecutive M frames, determine whether the identification number has an assignment record according to the number statistics list when determining that the moving distance is greater than the first specified distance A; If the identification number has an assignment record, the relevant data related to the identification number is recorded to the number statistics list; if the identification number has no assignment record, the identification number is retained as a valid number in the number statistics list.
  • the statistics component 40 is further configured to count the identification numbers included in the numbered statistics list, so as to implement statistics on the flow of people in the designated area.
  • the device for counting traffic may further include:
  • the lens 50 is configured to image a specified area and collect light onto a vision sensor.
  • the main processor 60 is connected to the vision sensor 10, the identification component 20, the analysis component 30, and the statistics component 40, and is used for management and / or data analysis of the vision sensor 10, the identification component 20, the analysis component 30, and the statistics component 40.
  • the device for counting human flow may further include an output component 70 for outputting the statistical data in a specific format for users to view at any time.
  • An embodiment of the present invention further provides a vision sensor-based person flow statistics system for performing flow statistics on an area to be detected with multiple sub-areas, wherein each sub-area is provided with the vision-based sensor provided in the foregoing embodiment.
  • Traffic statistics device the lens 50, the vision sensor 10, the main processor 60, and the output component 70 may constitute a vision sensor hardware module in the traffic statistics device of this embodiment.
  • the above-mentioned statistical system may further include a cloud server for receiving and storing statistical data transmitted in each sub-region, and counting the human flow in the region to be detected.
  • the vision sensor-based person flow statistics system may further include: an access terminal for acquiring and viewing the flow statistics data of each sub-region stored by the cloud server; wherein the access terminal includes: a terminal Client program.
  • a terminal Client program For example, a computer client program, a mobile phone application program, or other programs in an access terminal, the present invention is not limited.
  • FIG. 7 illustrates a vision sensor-based human traffic statistics system according to an embodiment of the present invention, as shown in FIG. 7.
  • the area to be detected can be divided into node 1, node 2, node 3 ... node n, each of the above n analysis nodes, and each child node represents a sub-region, so a network can be organized to obtain the data of each node in the entire area, and then Accurately calculate the flow of people in each area in the area to be detected.
  • each node After each node obtains the human flow of the corresponding sub-area, it can send its own human flow to the cloud server through the router, and then the access terminal can read the statistical human flow data at any time through the external network.
  • An embodiment of the present invention provides a more effective method for counting people traffic, which can analyze the recognition characteristics of an object before and after the recognition that often occurs during the detection and recognition process, so as to determine whether it is a mistaken recognition.
  • the assigned identification number is removed, thereby improving the accuracy of the output data.
  • the identification number can be re-assigned based on the judgment of specific conditions, thereby eliminating the need for identification.
  • each functional unit in each embodiment of the present invention may be physically independent of each other, or two or more functional units may be integrated together, or all functional units may be integrated into one processing unit.
  • the above integrated functional units may be implemented in the form of hardware, or in the form of software or firmware.
  • the integrated functional unit is implemented in the form of software and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the present invention is essentially or all or part of the technical solution may be embodied in the form of a software product.
  • the computer software product is stored in a storage medium and includes several instructions for making one
  • a computing device for example, a personal computer, a server, or a network device
  • the foregoing storage media include: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, and other media that can store program codes.
  • all or part of the steps of implementing the foregoing method embodiments may be completed by a program instructing related hardware (such as a computing device such as a personal computer, a server, or a network device), and the program instructions may be stored in a computer-readable storage
  • a program instructing related hardware such as a computing device such as a personal computer, a server, or a network device
  • the program instructions may be stored in a computer-readable storage
  • the computing device executes all or part of the steps of the method according to the embodiments of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

本发明提供了一种基于视觉传感器的人流量统计方法、装置及系统,该方法包括:利用视觉传感器按特定频率采集指定区域内的图像数据;对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号;基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性;对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计。基于本发明提供的人流量统计方法,在识别出人像之后还加入了分析步骤,以对各识别编号进行进一步有效分析,通过对有效识别编号进行计数实现对人流量的统计,从而极大的提高了人流量统计的准确性。

Description

一种基于视觉传感器的人流量统计方法、装置及系统 技术领域
本发明涉及数据统计技术领域,特别是涉及一种基于视觉传感器的人流量统计方法、装置及系统。
背景技术
随着大数据时代的到来,通常会对人们的各种信息进行采集以分析人们的喜好,尤其是通过识别分析技术来分析图像中包含的人像数据已经越来越广泛的被应用,目前已在安防、智能家居、以及智能商业上已经有各种各样的方案被实施。人流量作为一个统计数据对商业领域有重要价值,比如店铺在不同时段的进店人数,以及进店人数在店铺内的分布情况,同时结合店铺的销售数据,可以分析出很多有价值的数据,可以为店铺的有效运营和业务成长提供指导。
当前对于人数或人流量的分析普遍存在两个很大的问题:第一,分析识别程序经常会发生错误识别,即,把不是一个人的物体识别成一个人;第二,识别到了一个人,会启动跟踪,但经常会导致某些时刻,跟踪丢失,以至于对同一个人计数多次,进而导致识别结果的不准确。
发明内容
本发明提供了一种基于视觉传感器的人流量统计方法及装置以克服上述问题或者至少部分地解决上述问题。
根据本发明的一个方面,提供了一种基于视觉传感器的人流量统计方法,包括:
利用视觉传感器按特定频率采集指定区域内的图像数据;
对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号;
基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性;
对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计。
可选地,所述对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出人像分配识别编号,包括:
获取所述视觉传感器采集的图像数据,依次对各帧图像进行检测;
当识别到任一帧图像中包括的人像时,为识别出的人像分配识别编号, 并将所述识别编号记录至预设的编号统计列表。
可选地,所述为识别出的人像分配识别编号,并将所述识别编号记录至预设的编号统计列表之后,还包括:
读取识别出的各人像的中心像素点坐标和/或对应时间戳,与各人像的识别编号同时记录至所述编号统计列表。
可选地,所述基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性,包括:
将所述识别编号与所述编号统计列表进行比对;
若所述识别编号与所述编号统计列表中的任一编号匹配,则追踪该识别编号在连续M帧图像中的数据;
基于所述连续M帧图像中的数据分析该识别编号的有效性。
可选地,所述将所述识别编号与所述编号统计列表进行比对之前,还包括:
将所述识别编号与预设的无效编号列表进行比对;
若所述识别编号与所述无效编号列表中的任一编号匹配,则确定该识别编号为无效识别编号;
若所述识别编号与所述无效编号列表中的任一编号不匹配,则与所述编号统计列表进行比对。
可选地,所述基于所述连续M帧图像中的数据分析该识别编号的有效性,包括:
获取所述识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内;其中,所述识别编号在连续M帧图像中的数据包括:所述识别编号在连续M帧图像中每一帧图像中人像的中心像素点坐标和/或时间戳;
若M大于预设最小值Mmin,且小于预设最大值Mmax,则判断所述识别编号为有效编号,将其保留至编号统计列表中。
可选地,所述获取所述识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内之后,还包括:
若M小于所述预设最小值Mmin,则判断其对应的识别编号为无效编号,将其添加至所述无效统计列表或丢弃;
若M大于所述预设最大值Mmax,则获取所述识别编号在连续M帧图像内的中心像素点的移动距离;若该移动距离小于第一指定距离A,则确定该识别编号为无效编号,将其添加至所述无效统计列表或丢弃。
可选地,所述将所述识别编号与所述编号统计列表进行比对之后,还包括:
若所述识别编号与所述编号统计列表中的任一编号不匹配,则记录所述识别编号对应人像的中心像素点坐标和/或当前时间戳;
判断在指定时间间隔内,是否存在已消失人像在出现的最后一帧图像中的中心像素点与所述识别编号对应人像当前中心像素点的距离小于第二指定距离B;
若存在,则将所述已消失人像对应的识别编号赋值为该识别编号。
可选地,所述若M大于所述预设最大值Mmax,则获取所述识别编号在连续M帧内的移动距离之后,还包括:
若判断该移动距离大于所述第一指定距离A,则根据所述编号统计列表判断所述识别编号是否存在赋值记录;
若所述识别编号有赋值记录,则将与该识别编号的相关数据记录至所述编号统计列表;
若所述识别编号没有赋值记录,则将该识别编号作为有效编号保留至所述编号统计列表。
可选地,所述对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计,包括;
对所述编号统计列表中包括的识别编号进行计数,以实现对所述指定区域内的人流量统计。
可选地,对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计之后,还包括:
将统计后的数据按照特定格式输出。
根据本发明的另一个方面,还提供了一种基于视觉传感器的人流量统计装置,包括:
视觉传感器硬件模组,用于按特定频率采集指定区域内的图像数据;
识别组件,用于对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号;
分析组件,用于基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性;
统计组件,用于对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计。
可选地,所述识别组件包括:
检测单元,用于获取所述视觉传感器采集的图像数据,依次对各帧图像进行检测;
编号分配单元,用于当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号,并将所述识别编号记录至预设的编号统计列表。
可选地,所述识别组件还包括:
记录单元,用于读取识别出的各人像的中心像素点坐标和/或对应时间戳,与各人像的识别编号同时记录至所述编号统计列表。
可选地,所述分析组件包括:
比对单元,用于将所述识别编号与所述编号统计列表进行比对;
追踪单元,用于当所述识别编号与所述编号统计列表中的任一编号匹配时,追踪该识别编号在连续M帧图像中的数据;
有效性分析单元,用于基于所述连续M帧图像中的数据分析该识别编号的有效性。
可选地,所述比对单元,还用于在与所述编号统计列表进行比对之前,将所述识别编号与预设的无效编号列表进行比对;若所述识别编号与所述无效编号列表中的任一编号匹配,则确定该识别编号为无效识别编号;
若所述识别编号与所述无效编号列表中的任一编号不匹配,则与所述编号统计列表进行比对。
可选地,所述有效性分析单元,还用于获取所述识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内;其中,所述识别编号在连续M帧图像中的数据包括:所述识别编号在连续M帧图像中每一帧图像中人像的中心像素点坐标和/或时间戳;
当M大于预设最小值Mmin,且小于预设最大值Mmax时,判断所述识别编号为有效编号,将其保留至编号统计列表中。
可选地,所述有效性分析单元,还用于当M小于所述预设最小值Mmin时,判断其对应的识别编号为无效编号,将其添加至所述无效统计列表或丢弃;
当M大于所述预设最大值Mmax时,获取所述识别编号在连续M帧图像内的中心像素点的移动距离;当该移动距离小于第一指定距离A时,确定该识别编号为无效编号,将其添加至所述无效统计列表或丢弃。
可选地,所述分析组件还包括:
赋值单元,用于当所述识别编号与所述编号统计列表中的任一编号不匹配时,记录所述识别编号对应人像的中心像素点坐标和/或当前时间戳;
判断在指定时间间隔内,是否存在已消失人像在出现的最后一帧图像中的中心像素点与所述识别编号对应人像当前中心像素点的距离小于第二指定距离B;
若存在,则将所述已消失人像对应的识别编号赋值为该识别编号。
可选地,所述有效性分析单元,还用于在获取所述识别编号在连续M帧内的移动距离之后,当判断该移动距离大于所述第一指定距离A时,根据 所述编号统计列表判断所述识别编号是否存在赋值记录;
若所述识别编号有赋值记录,则将与该识别编号的相关数据记录至所述编号统计列表;
若所述识别编号没有赋值记录,则将该识别编号作为有效编号保留至所述编号统计列表。
可选地,所述统计组件,还用于对所述编号统计列表中包括的识别编号进行计数,以实现对所述指定区域内的人流量统计。
可选地,所述视觉传感器模组包括:
镜头,用于对所述指定区域进行成像,并收集光线至所述视觉传感器上。
可选地,上述装置还包括:主处理器,与所述视觉传感器、识别组件、分析组件和统计组件连接,用于对视觉传感器、识别组件、分析组件和统计组件的管理和/或数据的分析。
可选地,上述装置还包括:输出组件,用于将统计后的数据按照特定格式输出。
根据本发明的另一个方面,还提供了一种基于视觉传感器的人流量统计系统,用于对具备多个子区域的待检测区域进行人流量统计,其中,各所述子区域设置有上述任一项所述的基于视觉传感器的人流量统计装置。
可选地,上述系统还包括:云服务器,用于接收和存储所述各子区域传输的统计数据,统计所述待检测区域的人流量。
可选地,还包括:访问终端,用于获取和查看所述云服务器存储的各子区域的人流量统计数据;其中,所述访问终端包括:终端客户端程序。
本发明提供了一种基于视觉传感器的人流量统计方法、装置及系统,在利用视觉传感器获取到指定区域内的图像数据后,可对各帧图像进行识别,当识别到其中包括人像时,可对各人像分配识别编号。除此之外,本发明提供的人流量统计方法在识别出人像之后还加入了分析步骤,以对各识别编号进行进一步有效分析,通过对有效识别编号进行计数实现对人流量的统计,从而极大的提高了人流量统计的准确性。
上述说明仅是本发明技术方案的概述,为了能够更清楚了解本发明的技术手段,而可依照说明书的内容予以实施,并且为了让本发明的上述和其它目的、特征和优点能够更明显易懂,以下特举本发明的具体实施方式。
根据下文结合附图对本发明具体实施例的详细描述,本领域技术人员将会更加明了本发明的上述以及其他目的、优点和特征。
附图说明
通过阅读下文优选实施方式的详细描述,各种其他的优点和益处对于本 领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的,而并不认为是对本发明的限制。而且在整个附图中,用相同的参考符号表示相同的部件。在附图中:
图1是根据本发明实施例的基于视觉传感器的人流量统计方法的流程示意图;
图2是根据本发明实施例的人像的中心点坐标的定义示意图;
图3是根据本发明实施例的分析识别编号有效性的方法的流程示意图;
图4是根据本发明优选实施例的基于视觉传感器的人流量统计方法的流程示意图;
图5是根据本发明实施例的基于视觉传感器的人流量统计装置结构示意图;
图6是根据本发明优选实施例的基于视觉传感器的人流量统计装置结构示意图;
图7是根据本发明实施例的基于视觉传感器的人流量统计系统结构示意图。
具体实施方式
下面将参照附图更详细地描述本公开的示例性实施例。虽然附图中显示了本公开的示例性实施例,然而应当理解,可以以各种形式实现本公开而不应被这里阐述的实施例所限制。相反,提供这些实施例是为了能够更透彻地理解本公开,并且能够将本公开的范围完整的传达给本领域的技术人员。
图1是根据本发明实施例的基于视觉传感器的人流量统计方法流程示意图,如图1所示,根据本发明实施例的基于视觉传感器的人流量统计方法可以包括:
步骤S102,利用视觉传感器按特定频率采集指定区域内的图像数据;
步骤S104,对上述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出人像分配识别编号;
步骤S106,基于与任一帧图像连续的多帧图像分析识别编号的有效性;
步骤S108,对确定为有效的识别编号进行计数,以实现对指定区域内的人流量统计。
本发明实施例提供了一种基于视觉传感器的人流量统计方法,在利用视觉传感器获取到指定区域内的图像数据后,可对各帧图像进行识别,当识别到其中包括人像时,可对各人像分配识别编号。除此之外,本实施例提供的人流量统计方法在识别出人像之后还加入了分析步骤,以对各识别编号进行进一步有效分析,通过对有效识别编号进行计数实现对人流量的统计,从而 极大的提高了人流量统计的准确性。
在本实施例中,视觉传感器采集图像数据时,可按照特定频率采集,如连续不间断地采集指定区域的图像数据或是视频数据,或每隔1秒进行周期性采集,对此,本发明不做限定。
优选地,采集到图像数据之后,上述步骤S104可进一步包括:首先,获取视觉传感器所采集的图像数据,依次对图像数据中的各帧图像进行检测,识别各帧图像中包括的人像;再为识别出的人像分配识别编号,并将识别编号记录至预设的编号统计列表。分配识别编号时,可以从1开始,以自然数的方式进行依次分配,也可以按照其他规则进行分配,本发明不做限定。而依次对图像数据中的各帧数据进行检测时,可以从所采集的图像数据的第一帧开始检测,也可以是根据用户需求从其中任意一帧开始检测。由于视觉传感器采集图像数据时是按特定频率进行采集的,因此,在确定需要检测的初始帧图像后,即可一直对后续的图像帧进行检测识别,或由用户确定识别检测结束的时间点或图像帧。
上文提及,为识别出的人像分配识别编号后,可将其记录至编号统计列表中。本实施例中的编号统计列表中各识别编号是对所识别出的人像对应的,当初始识别出一个人像,即可将其对应的识别编号添加至编号统计列表中。也就是说,本实施例中对所有分配的识别编号默认为有效编号,由于后续才对其有效性进行进一步判断,因此,当对该指定区域人流量进行统计时,直接对该编号统计列表中的识别编号计数即可,不会造成人数的重复统计。
编号统计列表中除了记录各识别编号之外,还可以读取识别出的各人像的中心像素点坐标和/或对应时间戳,与各人像的识别编号同时记录至编号统计列表中,以便在后面有效性分析过程中提供参考条件。其中,如图2所示,中心像素点坐标可以是利用人像识别算法识别到图像中存在人像后,将所识别到的人像上标定的方框的对角线焦点在整个图像中的坐标作为该人像的中心像素点坐标,对应的时间戳则可以是人像所在图像帧的时间戳。中心像素点坐标可用(X,Y)的形式表示,其中,X代表在横坐标的像素点坐标,Y代表纵向的像素点坐标,假设图像像素为640×480,则X的取值范围是0~640,Y的取值范围是0~480。时间戳可以用T表示。上述只是示意性地列举了对中心像素点坐标和时间戳的记录形式,实际应用中还可以采用其他方式进行记录,本发明不做限定。
为了保证对人数的有效统计,本发明实施例提供的方案在为识别出的人像分配识别编号后,还会基于与各帧图像连续的多张图像分析各识别编号的有效性。图3是根据本发明实施例的分析识别编号有效性的方法的流程示意图,如图3所示,在本实施例中对编号的有效性进行识别时可以包括:
步骤S302,将识别编号与编号统计列表进行比对;
步骤S304,若该识别编号与编号统计列表中的任一编号匹配,则追踪该识别编号在连续M帧图像中的数据;
步骤S306,基于上述连续M帧图像中的数据分析该识别编号的有效性。
由于在对识别出的人像分配识别编号后,会先将该识别编号记录至编号统计列表,因此,在对任一识别编号的有效性进行判断时,可以先将其与编号统计列表中的编号进行匹配。需要说明的是,因为本实施例是基于连续的图像对识别编号进行分析,所以,如果在某一帧识别到一个人像而分配识别编号后,在连续的多帧图像中检测识别到同一个人像时,相当于是对该人像以同一识别编号进行追踪,而不会对其的识别编号进行重复分配。例如,初次为识别出的人像分配了识别编号1,此时,在后续对该识别编号1进行追踪过程中需要结合识别编号对该人像在不同时刻进行区分的话,可以对后续该人像出现的时刻采用1-1,1-2...1-n的编号方式。因此,若一识别编号与编号统计列表中的任一编号匹配,则说明已经追踪到该识别编号,此时可继续追踪后续连续的M帧图像中该识别编号的数据,进而基于连续M帧图像中的数据分析该识别编号的有效性。
对于图像的识别过程中可能经常会发生错误识别,通常情况下,错误识别的对象通常为静态的物体,这样,识别到一个物体后,通过分析识别物体的前后动作特征(如在指定时间段内无位移),从而判断是否是错误识别。对于此类识别物体可能在检测识别过程中可能会被分配识别编号,而这些识别编号在后续过程中是没有位移的,此时便可将上述识别编号统一管理,创建无效编号列表统一管理。
优选地,如图3所示,在上述步骤S302之前,还可以包括:
步骤S308,将识别编号与预设的无效编号列表进行比对;
步骤S310,若该识别编号与无效编号列表中的任一编号匹配,则确定该识别编号为无效识别编号;
步骤S312,若该识别编号与无效编号列表中的任一编号不匹配,则执行步骤S302与编号统计列表进行比对。
其中,无效编号列表中的编号可以是在一定时间段内静止且没有位移的编号。即,如果分配了识别编号给一静止人像(如商店的模特模型等),则通过后续连续的多帧图像判断该人像一直处于静止状态,则说明对于该人像的识别为错误识别,将其所赋予的识别编号添加至无效统计列表中,当后续再次识别到该静止人像并分配识别编号时,只需先将其与无效编号列表进行比对,若与无效编号列表中的任一识别编号匹配,则将其直接丢弃,从而提高分析识别的准确性。
进一步地,上述步骤S306提及,会根据连续M帧图像中的数据分析识别编号的有效性。优选地,可以包括:
获取识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内;其中,识别编号在连续M帧图像中的数据包括:识别编号在连续M帧图像中每一帧中人像的中心像素点坐标和/或时间戳;优选地,像素点坐标可以包括X、Y两个方向上的像素点坐标;其中,识别编号在连续M帧图像中的数据可以理解为该识别编号连续出现在的M帧图像,如果识别编号出现M帧数太少,则说明该识别编号对应的可能不是人像,如果识别编号出现M帧数太多,则说明所该识别编号对应为静止物像,需要进一步判断。因此,可以对所获取的M的范围可进行判断,以进一步确定识别编号的有效性。
对于M与指定数值范围的判断可以分为以下三种情况:
一、Mmin>M>Mmax
若M大于预设最小值Mmin,且小于预设最大值Mmax,则判断该识别编号为有效编号,将其保留至编号统计列表中。
也就是说,在对一个识别编号X的有效性进行分析时,如果在Mmin~Mmax这一范围内追踪到该识别编号X对应的人像,则说明该识别编号X为有效编号,则可保留在编号统计列表中。
二、M<Mmin
如果M小于预设最小值Mmin,则可以判断其对应的识别编号为无效编号,将其添加至无效统计列表或丢弃,说明该识别编号X不属于统计范围内,此时做无效处理。
三、M>Mmax
如果M大于预设最大值Mmax,则获取该识别编号在连续M帧图像内的中心像素点的移动距离,对其移动距离进行判断。上文介绍,记录识别编号在连续M帧图像中的数据时,可以记录该识别编号在X、Y方向上的像素点坐标,因此,在计算移动距离时,可以通过计算在X、Y上移动的距离计算识别编号的移动距离。在X、Y上移动的距离可分别采用以下公式进行计算:Xnm-Xn0的绝对值,或Ynm-Yn0的绝对值。
1.若判断该移动距离小于第一指定距离A,说明其可能是静止人像,则判断该识别编号为无效编号,将其添加至无效统计列表或丢弃。
2.若判断该移动距离大于第一指定距离A,则根据编号统计列表判断该识别编号是否存在赋值记录;若该识别编号有赋值记录,则将该识别编号的最新相关数据记录在编号统计列表中;若该识别编号没有赋值记录,则将该识别编号作为有效编号保留至编号统计列表。
在人像识别过程中,一般是拍到一帧图像后,会按人脸特征进行识别,对于一个特征会分配同一个ID号(即识别编号),当在该帧图片的后续帧图片中再次识别到相同的特征时,会再次匹配这个ID。但是,但是人在运动中,对于同一个人来讲,可能会进行脸部转动等动作而导致所检测到的特征值不同,此时,系统对这个人可能会分配新的ID。这时,就需要将该人像的ID重新赋值,即当一个ID消失、新的ID出现时,两者时间上和/或位置上都小于一个定值,那么可以判断为是两个ID对应同一个人像。
举例来说,在对识别出的人像进行识别编号分配时,先分配到一个识别编号ID1,但是在下一帧图像中并未跟踪到相同的特征值,而在下一帧或多帧图像之后,对于同一人像特征值可能就会作为新的人像而分配识别编号ID2并进行记录。但实际上,ID1和ID2对应的是同一个人像,此时,则可以对ID1和ID2的时间和/或位置进行判断,当时间差和/或位置差在一定数值范围内,可认为ID1和ID2对应的是同一个人像,这样,则将该人像的ID号赋值为ID2,即ID2=ID1,以该ID2记录为ID1对应人像的识别编号,将ID2替换ID1以更新编号统计列表。
因此,如图3所示,在上述步骤S302,将各识别编号与编号统计列表进行比对之后,还可以包括:
步骤S314,若该识别编号与编号统计列表中的任一编号不匹配,则记录该识别编号对应人像的中心像素点坐标和/或当前时间戳;
步骤S316,判断在判断在指定时间间隔内,是否存在已消失人像在出现的最后一帧图像中的中心像素点与该识别编号对应人像当前中心像素点的距离小于第二指定距离B;
步骤S318,若存在,则将上述已消失人像对应的识别编号赋值为该识别编号;
步骤S320,若不存在,则继续分析下一帧图像。
上述过程详细介绍了编号重新赋值的条件及过程,在一定的时间间隔内,如果有已消失人像与新识别人像的位置的变化处于一定范围内,则可以判断该位置的变化为同一个人像所产生的距离变化,此时,可以认为新识别人像与该已消失人像为同一个人,接下来可进行编号重新赋值。
经过上述介绍可知,编号统计列表中所记录的编号均为有效编号,因此,对指定区域内的人流量统计时,对编号统计列表中包括的识别编号进行计算即可。进一步地,对指定区域的人流量统计之后,还可以将统计后的数据按照特定格式输出,以供用户随时进行查看。
下面通过一个优选实施例对上述实施例进行详细说明。如图4所示,本优选实施例的基于视觉传感器的人流量统计方法可以包括:
步骤S401,开始工作后,视觉传感器采集的视觉传感器捕捉图像或视频数据,然后对上述数据进行人体检测和跟踪;
步骤S402,识别上述图像数据的初始帧图像中是否有人存在;
步骤S403,如果检测到有人存在,则分配一个ID号给检测到的人,例如ID=01,接下来执行步骤S404,对此ID进行分析和跟踪处理,并继续对下一帧图像进行检测;如果没有检测到人,则继续对下一帧图像进行检测;
步骤S404,与预先创建的无效ID列表对比,判断是否与其中任一ID匹配;若匹配,则执行步骤S405;若不匹配,则执行步骤S406;
步骤S405,将该ID丢弃;
步骤S406,与ID统计列表比对;若与IDn匹配上,则开始执行步骤S407,追踪后续IDn的移动,并记录每一帧的时间点和位置;若不匹配,则执行步骤S419;
步骤S407,追踪后续IDn的移动,赋予IDn_0;
步骤S408,记录IDn_0的位置,X、Y方向的像素点坐标:(Xn0,Yn0)和时间戳Tn0;
步骤S409,继续追踪下一帧图像中IDn的移动,赋予IDn_1;
步骤S410,记录IDn_1的位置,X、Y方向的像素点坐标:(Xn1,Yn1)和时间戳Tn1;
步骤S411,追踪到IDn连续出现的第M帧数据,赋予IDn_m;该M可以根据不同的应用场景进行调整,本发明不做限定;
步骤S412,记录IDn_m的位置,X、Y方向的像素点坐标:(Xnm,Ynm)和时间戳Tnm;
步骤S413,获取到IDn连续出现的M帧的数据后,判断IDn的有效性;在本实施例中,首先判断M是否小于10;若是,则执行步骤S414;若否,则执行步骤S416;
步骤S414,判断M是否大于3;若是,则认为IDn有效,执行步骤S415;若否,则说明该IDn为无效ID,添加至无效ID列表或丢弃;
步骤S415,输出IDn为有效ID,即保留在ID统计列表中;
步骤S416,若M>10,则进入有关位置的判断和分析以判断IDn是否有效;判断IDn在M帧范围内的移动距离是否大于10pix;移动距离的判断可以利用第10帧的X或Y的坐标判断其移动位置,如:Xn9-Xn0的绝对值大于10pix,或Yn9-Yn0的绝对值大于10pix;
若是,则执行步骤S417;若否,则说明该IDn为无效ID,添加至无效ID列表或丢弃;
步骤S417,判断是否存在ID赋值记录;若是,则执行步骤S418,若否, 则执行步骤S415;
步骤S418,将该识别编号的最新数据记录在ID统计列表中;
步骤S419,以上是针对IDn在识别程序识别到ID后,并继续跟踪IDn的情况下所需要进行的数据处理,某些情况下,识别到IDn后,识别程序在下一帧图像中并未跟踪到IDn,则会判断IDn对应的人像消失,同时记录IDn最后在图像中的位置(Xn,Yn)以及时间戳Tn。此时当前图像帧中识别到新的人像出现,则为人像分配新的IDm,这样接下来需进入ID赋值判断程序;
步骤S420,记录IDm在图像中的位置(Xm,Ym)以及时间戳Tm;
步骤S421,利用Tm-Tn<2秒作为示例,来判断位置的变化是否满足赋值条件;若是,则执行步骤S422判断位置的变化是否满足赋值条件;若否,则说明无需赋值,此时继续跟踪即可;
步骤S422,判断是否满足条件:Xm-Xn<B或Ym-Yn<B,如B=10pixel;若满足,则执行步骤S423;若不满足,则执行步骤S424;
步骤S423,若满足赋值条件,则将IDn对应人像的识别编号记录为IDm,即,在ID统计列表中记录IDm=IDn,将IDm记录为IDn对应人像的ID;
步骤S424,若不满足赋值条件,则进入读取下一帧图像进行判断分析流程。
基于本发明优选实施例提供的方法,在识别出图像中的人像之后,会为识别出的人像分配ID。进一步地,根据后续的图像对ID的有效性进行判断,以确定最终的有效ID进行统计。在本优选实施例中,不仅可以对错误识别的ID做无效处理,还可以消除针对识别的一个人存在的多个ID的情况,以进一步提升人流量统计的准确性。
基于同一发明构思,本发明实施例还提供了一种基于视觉传感器的人流量统计装置100,如图5所示,本发明实施例的基于视觉传感器的人流量统计装置可以包括:
视觉传感器10,用于按特定频率采集指定区域内的图像数据;
识别组件20,用于对图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号;该识别组件20可采用图像识别算法对图像数据中的各帧图像进行检测,以识别出图像中是否存在人像;
分析组件30,用于基于与任一帧图像连续的多帧图像分析该识别编号的有效性;
统计组件40,用于对确定为有效的识别编号进行计数,以实现对指定区域内的人流量统计。可选地,分析组件30可以在对编号的有效性进行分析 后直接进行统计,而无需单独设置统计组件40。
在本发明一优选实施例中,如图6所示,识别组件20可以包括:
检测单元21,用于获取视觉传感器采集的图像数据,依次对各帧图像进行检测;
编号分配单元22,用于用于当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号,并将该识别编号记录至预设的编号统计列表。
记录单元23,用于读取识别出的各人像的中心像素点坐标和/或对应时间戳,与各人像的识别编号同时记录至编号统计列表。
继续参见图6,本发明一优选实施例中,分析组件30可以包括:
比对单元31,用于将该识别编号与编号统计列表进行比对;
追踪单元32,用于当该识别编号与编号统计列表中的任一编号匹配时,追踪该识别编号在连续M帧图像中的数据;
有效性分析单元33,用于基于连续M帧图像中的数据分析该识别编号的有效性。
可选地,比对单元31,还用于在与编号统计列表进行比对之前,将该识别编号与预设的无效编号列表进行比对;若该识别编号与无效编号列表中的任一编号匹配,则确定该识别编号为无效识别编号;若该识别编号与无效编号列表中的任一编号不匹配,则与编号统计列表进行比对。
有效性分析单元33,还用于获取该识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内;其中,识别编号在连续M帧图像中的数据包括:识别编号在连续M帧图像中每一帧图像中人像的中心像素点坐标和/或时间戳;当M大于预设最小值Mmin,且小于预设最大值Mmax时,判断该识别编号为有效编号,将其保留至编号统计列表中。
有效性分析单元33,还用于当M小于预设最小值Mmin时,判断其对应的识别编号为无效编号,将其添加至无效统计列表或丢弃;当M大于预设最大值Mmax时,获取该识别编号在连续M帧图像内的中心像素点的移动距离;当该移动距离小于第一指定距离A时,确定该识别编号为无效编号,将其添加至无效统计列表或丢弃。
继续参见图6,分析组件30还可以包括:赋值单元34,用于当该识别编号与编号统计列表中的任一编号不匹配时,记录该识别编号对应人像的中心像素点坐标和/或当前时间戳;判断在指定时间间隔内,是否存在已消失人像在出现的最后一帧图像中的中心像素点与该识别编号对应人像当前中心像素点的距离小于第二指定距离B;若存在,则将上述已消失人像对应的识别编号赋值为该识别编号。
有效性分析单元33,还用于在获取该识别编号在连续M帧内的移动距 离之后,当判断该移动距离大于第一指定距离A时,根据编号统计列表判断该识别编号是否存在赋值记录;若该识别编号有赋值记录,则将与该识别编号的相关数据记录至编号统计列表;若该识别编号没有赋值记录,则将该识别编号作为有效编号保留至编号统计列表。
统计组件40,还用于对编号统计列表中包括的识别编号进行计数,以实现对指定区域内的人流量统计。
另外,如图6所示,本发明实施例提供的人流量统计装置还可以包括:
镜头50,用于对指定区域进行成像,并收集光线至视觉传感器上。
主处理器60,与视觉传感器10、识别组件20、分析组件30和统计组件40连接,用于对视觉传感器10、识别组件20、分析组件30和统计组件40的管理和/或数据的分析。
可选地,本实施例提供的人流量统计装置还可以包括输出组件70,用于将统计后的数据按照特定格式输出,以供用户随时进行查看。
本发明实施例还提供了一种基于视觉传感器的人流量统计系统,用于对具备多个子区域的待检测区域进行人流量统计,其中,各子区域设置有上述实施例所提供的基于视觉传感器的人流量统计装置。在本实施例中,镜头50,视觉传感器10,主处理器60以及输出组件70可构成本实施例人流量统计装置中的视觉传感器硬件模组。
上述统计系统还可以包括云服务器,用于接收和存储各子区域传输的统计数据,统计所述待检测区域的人流量。
优选地,本实施例提供的基于视觉传感器的人流量统计系统还可以包括:访问终端,用于获取和查看云服务器存储的各子区域的人流量统计数据;其中,所述访问终端包括:终端客户端程序。如电脑客户端程序、手机应用程序或是其他访问终端中的程序,本发明不做限定。
图7示出了根据本发明实施例的基于视觉传感器的人流量统计系统,如图7所示。待检测区域可分为节点1、节点2、节点3...节点n上述n各分析节点,每个子节点代表一个子区域,由此可组织一个网络来获取整个区域内各节点的数据,进而准确出统计出待检测区域内各分区的人流量。
在各个节点获取相应子区域的人流量之后,可通过路由器将各自的人流量发送至云服务器,进而由访问终端通过外网对所统计的人流量数据随时读取。
本发明实施例提供了一种更加有效的人流量统计方法,可以针对检测识别过程经常发生的错误识别,通过分析识别物体的前后动作特征,从而判断是否是错误识别,如果是错误识别,将其所赋予的识别编号去除,从而提高输出数据的准确性;还可以针对检测识别程序对一个人赋予多个识别编号的 情况,可通过特定条件的判断进行识别编号的重新赋值,从而消除针对识别的一个人存在的多个识别编号的情况,来提高分析准确度。
所属领域的技术人员可以清楚地了解到,上述描述的系统、装置和单元的具体工作过程,可以参考前述方法实施例中的对应过程,为简洁起见,在此不另赘述。
另外,在本发明各个实施例中的各功能单元可以物理上相互独立,也可以两个或两个以上功能单元集成在一起,还可以全部功能单元都集成在一个处理单元中。上述集成的功能单元既可以采用硬件的形式实现,也可以采用软件或者固件的形式实现。
本领域普通技术人员可以理解:所述集成的功能单元如果以软件的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本发明的技术方案本质上或者该技术方案的全部或部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,其包括若干指令,用以使得一台计算设备(例如个人计算机,服务器,或者网络设备等)在运行所述指令时执行本发明各实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(ROM)、随机存取存储器(RAM),磁碟或者光盘等各种可以存储程序代码的介质。
或者,实现前述方法实施例的全部或部分步骤可以通过程序指令相关的硬件(诸如个人计算机,服务器,或者网络设备等的计算设备)来完成,所述程序指令可以存储于一计算机可读取存储介质中,当所述程序指令被计算设备的处理器执行时,所述计算设备执行本发明各实施例所述方法的全部或部分步骤。
最后应说明的是:以上各实施例仅用以说明本发明的技术方案,而非对其限制;尽管参照前述各实施例对本发明进行了详细的说明,本领域的普通技术人员应当理解:在本发明的精神和原则之内,其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分或者全部技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案脱离本发明的保护范围。

Claims (27)

  1. 一种基于视觉传感器的人流量统计方法,包括:
    利用视觉传感器按特定频率采集指定区域内的图像数据;
    对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号;
    基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性;
    对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计。
  2. 根据权利要求1所述的方法,其中,所述对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出人像分配识别编号,包括:
    获取所述视觉传感器采集的图像数据,依次对各帧图像进行检测;
    当识别到任一帧图像中包括的人像时,为识别出的人像分配识别编号,并将所述识别编号记录至预设的编号统计列表。
  3. 根据权利要求2所述的方法,其中,所述为识别出的人像分配识别编号,并将所述识别编号记录至预设的编号统计列表之后,还包括:
    读取识别出的各人像的中心像素点坐标和/或对应时间戳,与各人像的识别编号同时记录至所述编号统计列表。
  4. 根据权利要求2所述的方法,其中,所述基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性,包括:
    将所述识别编号与所述编号统计列表进行比对;
    若所述识别编号与所述编号统计列表中的任一编号匹配,则追踪该识别编号在连续M帧图像中的数据;
    基于所述连续M帧图像中的数据分析该识别编号的有效性。
  5. 根据权利要求4所述的方法,其中,所述将所述识别编号与所述编号统计列表进行比对之前,还包括:
    将所述识别编号与预设的无效编号列表进行比对;
    若所述识别编号与所述无效编号列表中的任一编号匹配,则确定该识别编号为无效识别编号;
    若所述识别编号与所述无效编号列表中的任一编号不匹配,则与所述编号统计列表进行比对。
  6. 根据权利要求4所述的方法,其中,所述基于所述连续M帧图像中的数据分析该识别编号的有效性,包括:
    获取所述识别编号在连续出现的M帧图像中的数据,并判断M是否处 于指定数值范围内;其中,所述识别编号在连续M帧图像中的数据包括:所述识别编号在连续M帧图像中每一帧图像中人像的中心像素点坐标和/或时间戳;
    若M大于预设最小值Mmin,且小于预设最大值Mmax,则判断所述识别编号为有效编号,将其保留至编号统计列表中。
  7. 根据权利要求6所述的方法,其中,所述获取所述识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内之后,还包括:
    若M小于所述预设最小值Mmin,则判断其对应的识别编号为无效编号,将其添加至所述无效统计列表或丢弃;
    若M大于所述预设最大值Mmax,则获取所述识别编号在连续M帧图像内的中心像素点的移动距离;若该移动距离小于第一指定距离A,则确定该识别编号为无效编号,将其添加至所述无效统计列表或丢弃。
  8. 根据权利要求4所述的方法,其中,所述将所述识别编号与所述编号统计列表进行比对之后,还包括:
    若所述识别编号与所述编号统计列表中的任一编号不匹配,则记录所述识别编号对应人像的中心像素点坐标和/或当前时间戳;
    判断在指定时间间隔内,是否存在已消失人像在出现的最后一帧图像中的中心像素点与所述识别编号对应人像当前中心像素点的距离小于第二指定距离B;
    若存在,则将所述已消失人像对应的识别编号赋值为该识别编号。
  9. 根据权利要求7所述的方法,其中,所述若M大于所述预设最大值Mmax,则获取所述识别编号在连续M帧内的移动距离之后,还包括:
    若判断该移动距离大于所述第一指定距离A,则根据所述编号统计列表判断所述识别编号是否存在赋值记录;
    若所述识别编号有赋值记录,则将与该识别编号的相关数据记录至所述编号统计列表;
    若所述识别编号没有赋值记录,则将该识别编号作为有效编号保留至所述编号统计列表。
  10. 根据权利要求2所述的方法,其中,所述对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计,包括;
    对所述编号统计列表中包括的识别编号进行计数,以实现对所述指定区域内的人流量统计。
  11. 根据权利要求1-10任一项所述的方法,其中,对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计之后,还包括:
    将统计后的数据按照特定格式输出。
  12. 一种基于视觉传感器的人流量统计装置,包括:
    视觉传感器,用于按特定频率采集指定区域内的图像数据;
    识别组件,用于对所述图像数据中的各帧图像进行检测,当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号;
    分析组件,用于基于与所述任一帧图像连续的多帧图像分析所述识别编号的有效性;
    统计组件,用于对确定为有效的识别编号进行计数,以实现对所述指定区域内的人流量统计。
  13. 根据权利要求12所述的装置,其中,所述识别组件包括:
    检测单元,用于获取所述视觉传感器采集的图像数据,依次对各帧图像进行检测;
    编号分配单元,用于当识别到任一帧图像中包括人像时,为识别出的人像分配识别编号,并将所述识别编号记录至预设的编号统计列表。
  14. 根据权利要求13所述的装置,其中,所述识别组件还包括:
    记录单元,用于读取识别出的各人像的中心像素点坐标和/或对应时间戳,与各人像的识别编号同时记录至所述编号统计列表。
  15. 根据权利要求13所述的装置,其中,所述分析组件包括:
    比对单元,用于将所述识别编号与所述编号统计列表进行比对;
    追踪单元,用于当所述识别编号与所述编号统计列表中的任一编号匹配时,追踪该识别编号在连续M帧图像中的数据;
    有效性分析单元,用于基于所述连续M帧图像中的数据分析该识别编号的有效性。
  16. 根据权利要求15所述的装置,其中,
    所述比对单元,还用于在与所述编号统计列表进行比对之前,将所述识别编号与预设的无效编号列表进行比对;若所述识别编号与所述无效编号列表中的任一编号匹配,则确定该识别编号为无效识别编号;
    若所述识别编号与所述无效编号列表中的任一编号不匹配,则与所述编号统计列表进行比对。
  17. 根据权利要求15所述的装置,其中,
    所述有效性分析单元,还用于获取所述识别编号在连续出现的M帧图像中的数据,并判断M是否处于指定数值范围内;其中,所述识别编号在连续M帧图像中的数据包括:所述识别编号在连续M帧图像中每一帧图像中人像的中心像素点坐标和/或时间戳;
    当M大于预设最小值Mmin,且小于预设最大值Mmax时,判断所述识别编号为有效编号,将其保留至编号统计列表中。
  18. 根据权利要求17所述的装置,其中,
    所述有效性分析单元,还用于当M小于所述预设最小值Mmin时,判断其对应的识别编号为无效编号,将其添加至所述无效统计列表或丢弃;
    当M大于所述预设最大值Mmax时,获取所述识别编号在连续M帧图像内的中心像素点的移动距离;当该移动距离小于第一指定距离A时,确定该识别编号为无效编号,将其添加至所述无效统计列表或丢弃。
  19. 根据权利要求15所述的装置,其中,所述分析组件还包括:
    赋值单元,用于当所述识别编号与所述编号统计列表中的任一编号不匹配时,记录所述识别编号对应人像的中心像素点坐标和/或当前时间戳;
    判断在指定时间间隔内,是否存在已消失人像在出现的最后一帧图像中的中心像素点与所述识别编号对应人像当前中心像素点的距离小于第二指定距离B;
    若存在,则将所述已消失人像对应的识别编号赋值为该识别编号。
  20. 根据权利要求18所述的装置,其中,
    所述有效性分析单元,还用于在获取所述识别编号在连续M帧内的移动距离之后,当判断该移动距离大于所述第一指定距离A时,根据所述编号统计列表判断所述识别编号是否存在赋值记录;
    若所述识别编号有赋值记录,则将与该识别编号的相关数据记录至所述编号统计列表;
    若所述识别编号没有赋值记录,则将该识别编号作为有效编号保留至所述编号统计列表。
  21. 根据权利要求13所述的装置,其中,
    所述统计组件,还用于对所述编号统计列表中包括的识别编号进行计数,以实现对所述指定区域内的人流量统计。
  22. 根据权利要求12-21任一项所述的装置,其中,还包括:
    镜头,用于对所述指定区域进行成像,并收集光线至所述视觉传感器上。
  23. 根据权利要求12-21任一项所述的装置,其中,还包括:
    主处理器,与所述视觉传感器、识别组件、分析组件和统计组件连接,用于对视觉传感器、识别组件、分析组件和统计组件的管理和/或数据的分析。
  24. 根据权利要求12-21任一项所述的装置,其中,还包括:
    输出组件,用于将统计后的数据按照特定格式输出。
  25. 一种基于视觉传感器的人流量统计系统,用于对具备多个子区域的待检测区域进行人流量统计;其中,各子区域设置有权利要求12-22中任一项所述的基于视觉传感器的人流量统计装置。
  26. 根据权利要求25所述的系统,其中,还包括:
    云服务器,用于接收和存储所述各子区域传输的统计数据,统计所述待检测区域的人流量。
  27. 根据权利要求26所述的系统,其中,还包括:
    访问终端,用于获取和查看所述云服务器存储的各子区域的人流量统计数据;其中,所述访问终端包括:终端客户端程序。
PCT/CN2019/091456 2018-06-25 2019-06-17 一种基于视觉传感器的人流量统计方法、装置及系统 WO2020001302A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810663182.9 2018-06-25
CN201810663182.9A CN108921072B (zh) 2018-06-25 2018-06-25 一种基于视觉传感器的人流量统计方法、装置及系统

Publications (1)

Publication Number Publication Date
WO2020001302A1 true WO2020001302A1 (zh) 2020-01-02

Family

ID=64422427

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/091456 WO2020001302A1 (zh) 2018-06-25 2019-06-17 一种基于视觉传感器的人流量统计方法、装置及系统

Country Status (2)

Country Link
CN (1) CN108921072B (zh)
WO (1) WO2020001302A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111274900A (zh) * 2020-01-15 2020-06-12 北京航空航天大学 一种基于底层特征提取的空基人群计数方法
CN113297888A (zh) * 2020-09-18 2021-08-24 阿里巴巴集团控股有限公司 一种图像内容检测结果核查方法及装置
CN113297888B (zh) * 2020-09-18 2024-06-07 阿里巴巴集团控股有限公司 一种图像内容检测结果核查方法及装置

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108921072B (zh) * 2018-06-25 2021-10-15 苏州欧普照明有限公司 一种基于视觉传感器的人流量统计方法、装置及系统
CN112597879A (zh) * 2020-12-21 2021-04-02 上海商米科技集团股份有限公司 一种基于人头识别的过店客流统计方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101196991A (zh) * 2007-12-14 2008-06-11 同济大学 密集客流计数和行人步行速度自动检测方法及系统
CN104134078A (zh) * 2014-07-22 2014-11-05 华中科技大学 一种人流量统计系统中分类器的自动选择方法
CN105139425A (zh) * 2015-08-28 2015-12-09 浙江宇视科技有限公司 一种人数统计方法及装置
US20170076171A1 (en) * 2015-09-14 2017-03-16 Hanwha Techwin Co., Ltd. Foreground extracting method and apparatus
CN108921072A (zh) * 2018-06-25 2018-11-30 苏州欧普照明有限公司 一种基于视觉传感器的人流量统计方法、装置及系统

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101477641A (zh) * 2009-01-07 2009-07-08 北京中星微电子有限公司 基于视频监控的人数统计方法和系统
US9025823B2 (en) * 2013-03-12 2015-05-05 Qualcomm Incorporated Tracking texture rich objects using rank order filtering
CN103425967B (zh) * 2013-07-21 2016-06-01 浙江大学 一种基于行人检测和跟踪的人流监控方法
CN107368789B (zh) * 2017-06-20 2021-01-19 华南理工大学 一种基于Halcon视觉算法的人流量统计装置及方法
CN108090493B (zh) * 2017-11-15 2019-03-05 南京光普信息技术有限公司 一种基于wifi定位的进店顾客数据统计方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101196991A (zh) * 2007-12-14 2008-06-11 同济大学 密集客流计数和行人步行速度自动检测方法及系统
CN104134078A (zh) * 2014-07-22 2014-11-05 华中科技大学 一种人流量统计系统中分类器的自动选择方法
CN105139425A (zh) * 2015-08-28 2015-12-09 浙江宇视科技有限公司 一种人数统计方法及装置
US20170076171A1 (en) * 2015-09-14 2017-03-16 Hanwha Techwin Co., Ltd. Foreground extracting method and apparatus
CN108921072A (zh) * 2018-06-25 2018-11-30 苏州欧普照明有限公司 一种基于视觉传感器的人流量统计方法、装置及系统

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111274900A (zh) * 2020-01-15 2020-06-12 北京航空航天大学 一种基于底层特征提取的空基人群计数方法
CN113297888A (zh) * 2020-09-18 2021-08-24 阿里巴巴集团控股有限公司 一种图像内容检测结果核查方法及装置
CN113297888B (zh) * 2020-09-18 2024-06-07 阿里巴巴集团控股有限公司 一种图像内容检测结果核查方法及装置

Also Published As

Publication number Publication date
CN108921072B (zh) 2021-10-15
CN108921072A (zh) 2018-11-30

Similar Documents

Publication Publication Date Title
WO2020001302A1 (zh) 一种基于视觉传感器的人流量统计方法、装置及系统
CN107784282B (zh) 对象属性的识别方法、装置及系统
KR101971866B1 (ko) 동화상에 있어서 물체를 검출하는 방법 및 장치, 및 그 프로그램을 기억한 기억 매체
JP5603403B2 (ja) 対象物計数方法、対象物計数装置および対象物計数プログラム
CN111914635B (zh) 人体测温方法、装置、系统和电子设备
CN110428449B (zh) 目标检测跟踪方法、装置、设备及存储介质
JP6589321B2 (ja) システム、検索方法およびプログラム
CN111126122B (zh) 人脸识别算法评估方法及装置
JP6503079B2 (ja) 特定人物検知システム、特定人物検知方法および検知装置
WO2022156234A1 (zh) 一种目标重识别方法、装置及计算机可读存储介质
CN111881740A (zh) 人脸识别方法、装置、电子设备及介质
KR20110103223A (ko) 얼굴 인식을 위한 sift 알고리즘을 이용한 키포인트 매칭 시스템 및 방법
CN109146913B (zh) 一种人脸跟踪方法及装置
JP2015230616A (ja) 画像処理方法および画像処理装置
CN113903066A (zh) 轨迹生成方法、系统、装置及电子设备
CN107231519B (zh) 视频处理装置及控制方法
JP2020095757A (ja) 情報処理装置、情報処理方法及びプログラム
JP2020095651A (ja) 生産性評価システム、生産性評価装置、生産性評価方法、及びプログラム
CN112070094B (zh) 训练数据的筛选方法、装置、电子设备及存储介质
CN114758384A (zh) 人脸检测方法、装置、设备及存储介质
JP2015118583A (ja) 情報処理装置、情報処理方法、並びにプログラム
CN113141433A (zh) 测试屏幕灵敏度的方法、装置和处理器
JP6958020B2 (ja) 画像解析プログラム、画像解析装置、及び画像解析方法
CN111461028A (zh) 复杂场景中的口罩检测模型训练及检测方法、介质及装置
JP2021056671A (ja) 作業状態判別装置、作業状態判別方法およびプログラム

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19824830

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19824830

Country of ref document: EP

Kind code of ref document: A1