WO2014097530A1 - Système de commande d'un processus d'analyse - Google Patents

Système de commande d'un processus d'analyse Download PDF

Info

Publication number
WO2014097530A1
WO2014097530A1 PCT/JP2013/006535 JP2013006535W WO2014097530A1 WO 2014097530 A1 WO2014097530 A1 WO 2014097530A1 JP 2013006535 W JP2013006535 W JP 2013006535W WO 2014097530 A1 WO2014097530 A1 WO 2014097530A1
Authority
WO
WIPO (PCT)
Prior art keywords
analysis processing
analysis
influence
target data
degree
Prior art date
Application number
PCT/JP2013/006535
Other languages
English (en)
Japanese (ja)
Inventor
有熊　威
小山　和也
Original Assignee
日本電気株式会社
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 日本電気株式会社 filed Critical 日本電気株式会社
Publication of WO2014097530A1 publication Critical patent/WO2014097530A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects

Definitions

  • the present invention relates to an analysis processing control system, and more particularly to an analysis processing control system for controlling analysis processing of continuously input data.
  • an analysis system that can obtain various analysis processing results from input data by combining multiple analysis engines of the same or different types has been developed.
  • analysis processing such as processing moving image data input from a camera in parallel or in series using a flow line extraction engine, a face extraction engine, an age discrimination engine, etc., and determining a person with a predetermined behavior.
  • an analysis system such as determining a person from moving image data captured by such a camera, it is required to obtain an analysis processing result quickly and preferably in real time without delay from the input of analysis target data. .
  • the analysis target data is continuously input data such as moving image data or audio data and has a relatively large capacity
  • the load of analysis processing performed by the analysis engine increases, and the analysis processing result can be quickly obtained. It becomes difficult to obtain.
  • the analysis target data volume is reduced in advance, such as by setting the frame rate of moving image data low, there arises a problem that an analysis result with desired analysis accuracy cannot be obtained.
  • Patent Document 1 discloses a technique of deleting data with an old recording time from among still image data captured by an imaging device.
  • still image data is simply deleted for the purpose of ensuring the number of images that can be captured by the imaging apparatus.
  • such a technique is applied to an analysis apparatus for the purpose of obtaining an analysis result of a desired analysis accuracy quickly when processing data to be continuously input. I can't do it.
  • an object of the present invention is to provide an analysis processing apparatus that can solve the problem that a desired analysis result cannot be obtained quickly from analysis target data.
  • An analysis processing control system is Data accepting means for accepting continuously inputted analysis target data; An influence degree calculating means for calculating an influence degree representing the degree of influence on the accuracy of the analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; According to the calculated degree of influence, data management means for specifying the analysis target data of the analysis processing unit to be discarded without being subject to analysis processing by the analysis processing means, The influence calculation means calculates a plurality of different influences of different types set in advance, The data management means identifies the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences.
  • the configuration is as follows.
  • An analysis processing system Data accepting means for accepting continuously inputted analysis target data; Analysis processing means for performing analysis processing on the received analysis target data for each analysis processing unit set in advance of the analysis target data; An influence degree calculating means for calculating an influence degree representing the degree of influence on the accuracy of the analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; According to the calculated degree of influence, data management means for specifying the analysis target data of the analysis processing unit to be discarded without being subject to analysis processing by the analysis processing means, The influence calculation means calculates a plurality of different influences of different types set in advance, The data management means identifies the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences.
  • the configuration is as follows.
  • the program which is the other form of this invention is: In the information processing device, Data accepting means for accepting continuously inputted analysis target data; An influence degree calculating means for calculating an influence degree representing the degree of influence on the accuracy of the analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; According to the calculated degree of influence, data management means for specifying the analysis target data of the analysis processing unit to be discarded without being subjected to analysis processing by the analysis processing means; And realize The influence calculation means calculates a plurality of different influences of different types set in advance, The data management means identifies the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences. It is a program for realizing this.
  • an analysis processing control method includes: Accept analysis target data input continuously, Calculating the degree of influence representing the degree of influence on the accuracy of analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; According to the calculated degree of influence, specify the analysis target data of the analysis processing unit to be discarded without being subject to analysis processing by the analysis processing means, When calculating the degree of influence, each of a plurality of different types of influence set in advance is calculated, Identifying the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences;
  • the configuration is as follows.
  • the present invention can provide an analysis processing control system, an analysis processing system, a program, and an analysis processing control method that can quickly obtain a desired analysis result from analysis target data. .
  • FIG. 4 It is a block diagram which shows the structure of the analysis system in Embodiment 4 of this invention. It is a figure which shows an example of the data memorize
  • FIGS. 1 to 4 are diagrams for explaining the configuration of the analysis system
  • FIG. 5 is a diagram for explaining the operation of the analysis system.
  • the analysis system in the present embodiment is configured by one or a plurality of information processing apparatuses, and performs a predetermined analysis process on the acquired analysis target data. For example, as shown in FIG. 1, video data acquired by the camera (1) is acquired and accumulated as a camera data source (2), and moving object detection / extraction (3) or feature amount extraction is performed on the video data. In addition, an analysis process such as verification (5) and (6) is performed, and alert notifications (4) and (7) are performed when an analysis result satisfying a predetermined condition is obtained.
  • the system analysis processing control system, analysis processing system
  • the system is not necessarily limited to a system that performs the analysis processing on the analysis target data described above. That is, the system according to the present invention may analyze data different from video data, and may perform other analysis processing without being limited to analysis processing such as moving object detection.
  • the analysis system 10 is connected to one or more data acquisition units 20 for acquiring analysis target data such as video data.
  • the data acquisition unit 20 is a device that acquires analysis target data that is continuously input, such as video data. Specifically, the data acquisition unit 20 acquires the analysis target data in order in units of frame images, which are analysis processing units to be analyzed by the analysis processing unit 16 to be described later, at a preset frame rate interval.
  • the analysis system 10 also includes a data reception unit 11, a data management unit 13, a freshness calculation unit 14, an effectiveness calculation unit 15, and an analysis processing unit 16, which are constructed by incorporating a program into the equipped arithmetic device. ing.
  • the analysis system 10 includes a data storage unit 12 in the storage device equipped. Note that the analysis system 10 may include other configurations in order to obtain a target analysis processing result, but a description of the other configurations is omitted.
  • the analysis system 10 is illustrated as being configured by one information processing device, but the analysis system 10 may be configured by one information processing device, You may be comprised with several information processing apparatus. Furthermore, the analysis system 10 is not necessarily limited to including the analysis processing unit 16, and another information processing apparatus including the analysis processing device 16 may be connected to the analysis system 10.
  • the data reception unit 11 receives the analysis target data acquired by the data acquisition unit 20 and stores it in the data storage unit 12. At this time, the data receiving unit 11 receives the analysis target data for each analysis processing unit, and in the analysis target data of the analysis processing unit, an ID (identification information) for identifying the analysis target data of the analysis processing unit, The attribute information is associated and stored in the data storage unit 12.
  • the attribute information is, for example, a date and time when a frame image that is analysis target data in an analysis processing unit is captured and acquired, and a frame ID that represents the order of the frame image.
  • FIG. 12 an example of information stored in the data storage unit 12 is shown in FIG.
  • a frame image “data body 1” is received as analysis processing data in an analysis processing unit
  • when “1” is associated as the ID and a frame image is acquired and acquired as attribute information thereof “20120731T01: 11: 11.111” representing the date and time (year / month / day and time) and “11234” representing the frame ID are stored in association with each other.
  • the “date and time” as the attribute information of the frame image includes the time when the frame image is input to the analysis system, the time stored in the data storage unit 12, and the analysis processing unit stored in the data storage unit 12. 16 may be the time when the processing wait state is reached.
  • the “date and time” is a value that represents an older time as the acquired time is older, and a value that represents a new value as the acquired time is newer.
  • the attribute information including the “date and time” and the “frame ID” represents the reception status when the analysis target data (frame image) is received for each analysis processing unit (for each frame image).
  • the data management unit 13 controls the calculation process and the analysis process of the influence on the analysis target data stored in the data storage unit 12. Specifically, the data management unit 13 first starts from the data reception unit 11 at a preset timing such as a timing when the remaining amount of data capacity that can be stored in the data storage unit 12 is less than a predetermined threshold. When the issued trigger is received, the analysis target data is acquired from the data storage unit 12. Then, the data management unit 13 controls the freshness calculation unit 14 (the influence degree calculation unit) and the validity level calculation unit 15 (the influence degree calculation unit) to calculate the freshness and the validity level, respectively.
  • a preset timing such as a timing when the remaining amount of data capacity that can be stored in the data storage unit 12 is less than a predetermined threshold.
  • the data management unit 13 controls the freshness calculation unit 14 (the influence degree calculation unit) and the validity level calculation unit 15 (the influence degree calculation unit) to calculate the freshness and the validity level, respectively.
  • the data management unit 13 acquires the freshness and effectiveness calculated by the freshness calculation unit 14 and the effectiveness calculation unit 15. Then, the data management unit 13 determines analysis target data to be discarded from the data storage unit 12 based on the acquired freshness and validity, and discards the analysis target data determined to be discarded from the data storage unit 12. Control as follows.
  • the data management unit 13 takes out the analysis target data stored in the data storage unit 12 and remains without being discarded from the data storage unit 12 and causes the analysis processing unit 16 to execute the analysis processing. Control.
  • One or a plurality of analysis processing units 16 are provided, and are analysis engines that perform analysis processes such as the above-described moving object detection / extraction, feature amount extraction, and collation.
  • the data management unit 13 acquires analysis target data from the data storage unit 12 for each analysis processing unit, that is, each frame image of video data that is analysis target data. It outputs to both the calculation part 14 and the effectiveness calculation part 15. At this time, the data management unit 13 also acquires attribute information associated with each frame image that is analysis target data of the analysis processing unit, that is, the date and time when the frame image was captured and the frame ID, and the frame image At the same time, it is output to both the freshness calculator 14 and the validity calculator 15. The data management unit 13 does not acquire each frame image itself from the data storage unit 12, acquires only the ID and attribute information of the frame image, and outputs them to both the freshness calculation unit 14 and the validity calculation unit 15. May be.
  • the freshness calculation unit 14 calculates “freshness” for each frame image using “date and time” in the attribute information.
  • the “freshness” in the present embodiment is defined as [0, 1 as a function of “elapsed time from the reference time to the current time: t” with the time based on “date and time” in the attribute information as “reference time”. ] (Value between 0 and 1), expressed as “fi (t)”.
  • freshness: fi (t) takes a value closer to “1” as the date and time when the frame image is acquired is closer to the current time, and takes a value closer to “0” as the distance from the current time is further away. That is, the freshness of the new frame image is close to “1” with respect to the current time, and the freshness of the old frame image is close to “0” with respect to the current time.
  • the “freshness” is represented by “fi (T)” (T: time from the system input time to the present).
  • the “reference time” is set to “the time when the frame image is in a state waiting for processing by the analysis processing unit”
  • the “freshness” is “fi (t buf )” (t buf : The time from when the frame image is in a process waiting state to the present time).
  • “reference time” is “the time when the analysis processing unit last processed data”
  • “freshness” is “fi ( ⁇ T)” ( ⁇ T: the analysis processing unit last The difference between the time when the data was processed and the above T).
  • the horizontal axis represents a new frame image as the value is small, and the vertical axis represents “freshness”.
  • “freshness” takes a value closer to “1” as the frame image is newer, and takes a value closer to “0” as the frame image is older.
  • the “freshness” having such a value represents the degree of influence (influence degree) on the accuracy of analysis processing by the analysis processing unit 16. This is because it is considered that the older the frame image, the lower the degree of influence on the accuracy of the analysis processing by the analysis processing unit 16 and the higher the influence on the accuracy of the analysis processing.
  • the validity calculating unit 15 uses the “frame ID” in the attribute information to generate the “effectiveness” for each frame image. Is calculated.
  • “effectiveness” is represented by “ei (d; D)” which takes [0, 1] (a value between 0 and 1), and is a frame image (data (data ( It is a value that represents the degree of influence on the accuracy of the analysis processing in d)).
  • “D” represents a set of data (d), and an attribute of “d” is represented by “.attribute name”.
  • the odd-numbered frame images are treated as having high effectiveness. That is, in this example, as shown in the graph (symbol *) of the symbol g2 in FIG. 4, the odd-numbered frame image has an effectiveness of “1” and the even-numbered frame image has an effectiveness of “0.5”. take. This is because if the input frame rate is at least 1 ⁇ 2, a decrease in the accuracy of the analysis process can be suppressed.
  • the freshness calculation unit 14 and the effectiveness calculation unit 15 affect the analysis processing by the analysis processing unit 16 of the analysis target data for each frame image that is the analysis target data of the analysis processing unit.
  • two different values such as “freshness” and “effectiveness” are calculated.
  • the “freshness” and “validity” for each frame image calculated by the freshness calculation unit 14 and the validity calculation unit 15 are passed to the data management unit 13.
  • the data management unit 13 Based on the “freshness” and “validity” for each frame image, the data management unit 13 specifies a frame image to be discarded from the data storage unit 12 without performing an analysis process by the analysis processing unit 16. At this time, the data management unit 13 first determines a value for determining whether or not to discard the frame image (data (d)) based on “freshness” and “validity” for each frame image. A certain “data rejection index: di (d)” is calculated.
  • data rejection index is an example, and may be calculated by other methods. For example, either “fi” or “ei” may be selected as a “data rejection index”, or each value may be weighted and added to calculate a “data rejection index”. .
  • the data management unit 13 identifies a frame image having a small value as a deletion candidate, and deletes the deletion candidate frame image from the data storage unit 12. To do.
  • the frame images indicated by reference signs D1, D2, and D3 are specified as deletion candidates in the order of D1, D2, and D3, and the free space in the data storage unit 12 exceeds a predetermined value or ratio. Until it becomes, the frame images as deletion candidates are deleted in the specified order.
  • the data management unit 13 performs control so that the frame image remaining in the data storage unit 12 without being deleted is taken out and the analysis processing unit 16 executes the analysis processing. Then, if necessary, output according to the contents of the analysis processing result is performed.
  • the analysis system 10 acquires a frame image which is data to be analyzed by the data acquisition unit 20 which is a camera (step S1), receives the frame image by the data reception unit 11 (step S2), and stores data. Store in the unit 12 (step S3).
  • the data receiving unit 11 stores the date information and attribute information including the frame ID in the data storage unit 12 in association with each frame image. Note that the analysis system 10 always acquires the analysis target data described above.
  • the data reception unit 11 monitors the free space in the data storage unit 12, and the data reception unit 11 triggers the data management unit 13 when the free space becomes smaller than a predetermined threshold. Is issued (step S4). Upon receiving this trigger, the data management unit 13 acquires the analysis target data from the data storage unit 12 (step S5), and the frame that is the acquired analysis target data is sent to the freshness calculation unit 14 and the validity calculation unit 15. Instructs to calculate the freshness and effectiveness for each image.
  • the freshness calculation unit 14 and the validity calculation unit 15 calculate the freshness and the validity for each frame image, and notify the data management unit 13 (steps S6 and S7).
  • the data management unit 13 Based on the freshness and validity notified from the freshness calculation unit 14 and the validity calculation unit 15, the data management unit 13 identifies a frame image that is a deletion candidate from the data storage unit 12 as described above. . Then, the data management unit 13 discards the frame image that is a deletion candidate from the data storage unit 12 according to the free space in the data storage unit 12 (steps S8 and S9).
  • the data management unit 13 takes out the analysis target data stored in the data storage unit 12 and remains without being discarded from the data storage unit 12, and executes the analysis process in the analysis processing unit 16. Control.
  • the analysis system 10 As described above, in the analysis system 10 according to the present embodiment, two types of influences (freshness and effectiveness described above) representing the degree of influence on the analysis processing accuracy of the analysis target data are calculated and based on these values.
  • the data to be analyzed is discarded from the data storage unit 12. For this reason, since it is possible to reduce the analysis target data to be analyzed, the analysis processing load can be reduced, and the analysis result can be obtained quickly.
  • the analysis target data having a low influence on the analysis processing accuracy is discarded based on the two kinds of influence degrees, a decrease in the accuracy of the analysis result can be suppressed and a desired analysis result can be obtained.
  • the “influence degree” used when specifying the analysis target data to be discarded the above-described “freshness” (first influence degree) and “effectiveness” (second influence degree) are two.
  • the type of information is used, it is not limited to using these two types of information.
  • the analysis target data to be discarded may be specified using information representing another situation at the time of reception of the analysis target data or information representing the status of the analysis processing of the analysis target data, as will be described later. Further, the analysis target data to be discarded may be specified using not only two different types of information but also three or more types of information.
  • FIG. 6 is a diagram for explaining the configuration of the analysis system
  • FIG. 7 is a diagram for explaining the operation of the analysis system.
  • the analysis system in the present embodiment has substantially the same configuration as that of the first embodiment described above, but information used as an “influence degree” used when specifying analysis target data to be discarded is different from that in the first embodiment. .
  • information used as an “influence degree” used when specifying analysis target data to be discarded is different from that in the first embodiment. .
  • differences from the first embodiment will be mainly described.
  • FIG. 10 An example of the configuration of the analysis system 10 in the present embodiment is shown in FIG.
  • the analysis system 10 includes a processing status management unit 17 in addition to the configuration described in the first embodiment. Accordingly, the “effectiveness” calculated by the effectiveness calculation unit 15 (influence calculation means) in the analysis system 10 in the present embodiment is different from that in the first embodiment.
  • the processing status management unit 17 manages the processing status of the analysis target data by the analysis processing unit 16 in cooperation with the data management unit 13. For example, the processing status management unit 17 acquires an analysis result for each analysis processing unit of the analysis target data by the analysis processing unit 16 that performs a specific analysis process set in advance, that is, for each frame image, and notifies the data management unit 13 of the result. . As another example, the processing status management unit 17 measures the processing load of the analysis processing unit 16 that performs a specific analysis process set in advance and notifies the data management unit 13 of the processing load.
  • the effectiveness calculation part 15 in this embodiment is based on the analysis process situation, such as the analysis result by the analysis process part 16 managed by the said process condition management part 17, and the processing load of the said analysis process part 16. “Effectiveness” representing the degree of influence (influence degree) on the accuracy of analysis processing for each analysis processing unit (each frame image) of the analysis target data is calculated.
  • the effectiveness calculation unit 15 detects “subject size” as an analysis result by the analysis processing unit 16 that detects a specific subject appearing in the frame image, and the larger the “subject size”, the larger the frame image.
  • the “efficiency” of is calculated so as to increase.
  • the effectiveness calculation unit 15 calculates that the “effectiveness” is lower as the “subject size” shown in the frame image is smaller. If the subject cannot be detected, the “efficiency” is more effective than the case where the subject is detected.
  • the degree is calculated to be low.
  • a plurality of analysis processing units 16 are provided, and a predetermined analysis process is performed on the subject detected by the analysis processing unit 16 located on the front stage side in the other analysis processing unit 16 on the rear stage side. Is assumed. In such a configuration, as the subject size is smaller, the analysis processing unit 16 on the downstream side cannot expect a high-precision analysis result, so the degree of influence on the analysis processing unit 16 on the subsequent stage is calculated to be low.
  • the effectiveness calculation unit 15 detects a preset suspicious behavior object as an analysis result by the specific analysis processing unit 16 for the frame image
  • the “effectiveness” of the frame image is determined. Calculate to be higher.
  • the frame image in which the suspicious action object is detected is assumed to be used later, and it is desirable to store the frame image. Therefore, the effectiveness is calculated as described above.
  • the effectiveness calculation unit 15 includes a load at the time of analysis processing by the specific analysis processing unit 16 on the frame image, that is, a CPU and a memory constituting the analysis system 10 used by the specific analysis processing unit 16.
  • the load of hardware resources such as the above is detected, and the higher the load, the lower the “effectiveness” for the frame image to be processed later by the specific analysis processing unit 16.
  • the frame image to be processed later is highly likely to be thinned out, and thus the effectiveness is calculated as described above.
  • the data management unit 13 calculates the “freshness” for each frame image calculated by the freshness calculation unit 14 as described in the first embodiment, and the effectiveness calculation unit 15 according to the present embodiment as described above. Based on the “effectiveness”, a “data rejection index” is calculated, and the analysis processing unit 16 specifies a frame image to be discarded from the data storage unit 12 without performing analysis processing. For example, a new frame image showing a large specific subject, a series of frame images in which a specific suspicious action object is detected even if it is old, or a frame image that is processed after an old and high load is detected , Etc. are left in the data storage unit 12, and the other frame images are specified as deletion candidates in the order of "data rejection index".
  • the analysis system 10 acquires a frame image which is data to be analyzed by the data acquisition unit 20 which is a camera (step S1), receives the frame image by the data reception unit 11 (step S2), and stores data. Store in the unit 12 (step S3). At this time, the data receiving unit 11 stores the date information and attribute information including the frame ID in the data storage unit 12 in association with each frame image. Note that the analysis system 10 always acquires the analysis target data described above.
  • the data reception unit 11 monitors the free space in the data storage unit 12, and the data reception unit 11 triggers the data management unit 13 when the free space becomes smaller than a predetermined threshold. Is issued (step S4).
  • the data management unit 13 acquires the analysis target data from the data storage unit 12 (step S5), and further acquires the analysis processing status from the analysis processing unit 16 (step S5 ').
  • the data management unit 13 instructs the freshness calculation unit 14 and the validity calculation unit 15 to calculate the freshness and validity for each frame image that is the acquired analysis target data.
  • the data management unit 13 notifies the effectiveness calculation unit 15 of the acquired analysis processing status, and instructs to calculate the effectiveness by referring to the analysis processing status.
  • the freshness calculation unit 14 and the validity calculation unit 15 calculate the freshness and the validity for each frame image, and notify the data management unit 13 (steps S6 and S7).
  • the data management unit 13 Based on the freshness and validity notified from the freshness calculation unit 14 and the validity calculation unit 15, the data management unit 13 identifies a frame image that is a deletion candidate from the data storage unit 12 as described above. . Then, the data management unit 13 discards the frame image that is a deletion candidate from the data storage unit 12 according to the free space in the data storage unit 12 (steps S8 and S9).
  • the data management unit 13 takes out the analysis target data stored in the data storage unit 12 and remains without being discarded from the data storage unit 12, and executes the analysis process in the analysis processing unit 16. Control.
  • the analysis system 10 As described above, in the analysis system 10 according to the present embodiment, two types of influences (freshness and effectiveness described above) representing the degree of influence on the analysis processing accuracy of the analysis target data are calculated and based on these values.
  • the data to be analyzed is discarded from the data storage unit 12. For this reason, since it is possible to reduce the analysis target data to be analyzed, the analysis processing load can be reduced, and the analysis result can be obtained quickly. In addition, since the analysis target data that has a low influence on the analysis processing accuracy is discarded, a decrease in the accuracy of the analysis result can be suppressed, and a desired analysis result can be obtained.
  • the “influence degree” used when specifying the analysis target data to be discarded the above-described “freshness” (first influence degree) and “effectiveness” (second influence degree) are two.
  • the type of information is used, it is not limited to using these two types of information.
  • the analysis target data to be discarded may be specified using information representing another situation at the time of receiving the analysis target data or information representing another situation of the analysis processing of the analysis target data. Further, the analysis target data to be discarded may be specified using not only two different types of information but also three or more types of information.
  • FIG. 8 is a diagram for explaining the configuration of the analysis system.
  • the analysis system in the present embodiment has substantially the same configuration as in the first and second embodiments described above, but information used as an “influence degree” used when specifying analysis target data to be discarded is the first and second embodiments. Different from 2.
  • the difference between the analysis system in the present embodiment and the first and second embodiments will be mainly described.
  • FIG. 8 shows an example of the configuration of the analysis system 10 in the present embodiment.
  • the analysis system 10 includes a first effectiveness calculation unit 15a and a second effectiveness calculation unit 15b in place of the freshness calculation unit 14 and the effectiveness calculation unit 15 having the configuration described in the second embodiment. It has.
  • the first effectiveness calculation unit 15a and the second effectiveness calculation unit 15b have an influence on the accuracy of the analysis processing for each analysis processing unit (for each frame image) of the analysis target data based on the analysis processing status by the analysis processing unit 16.
  • “Effectiveness” first influence degree, second influence degree) representing the degree (influence degree) of each is calculated.
  • the first effectiveness calculation unit 15a and the second effectiveness calculation unit 15b have different “efficiency” of the “efficiency” calculated by the effectiveness calculation unit 15 described in the second embodiment. calculate.
  • the first effectiveness calculation unit 15a detects “subject size” as an analysis result by the analysis processing unit 16 that detects a specific subject appearing in the frame image, and the larger the “subject size”, the more Calculation is performed so that the “first effectiveness” of the frame image is increased. Further, when the second susceptibility calculating unit 15b detects a suspicious action object set in advance as an analysis result by the specific analysis processing unit 16 for the frame image, the “second effectiveness” of the frame image is determined. Calculate to be higher.
  • the data management unit 13 uses the two types of “effectiveness” (first effectiveness and second effectiveness calculated by the first effectiveness calculation unit 15a and the second effectiveness calculation unit 15b. ) Is calculated, and the frame image to be discarded from the data storage unit 12 without specifying the analysis processing by the analysis processing unit 16 is specified. For example, a series of frame images in which a specific subject appears large and a specific suspicious action object is continuously detected are left in the data storage unit 12, and other frame images are displayed with a “data rejection index”. Specify deletion candidates in ascending order.
  • the analysis system 10 calculates two types of influences (the first validity and the second validity described above) that indicate the degree of influence on the analysis processing accuracy of the analysis target data.
  • the analysis target data is discarded from the data storage unit 12 based on the value. For this reason, since it is possible to reduce the analysis target data to be analyzed, the analysis processing load can be reduced, and the analysis result can be obtained quickly.
  • the analysis target data that has a low influence on the analysis processing accuracy is discarded, a decrease in the accuracy of the analysis result can be suppressed, and a desired analysis result can be obtained.
  • the above two types of “effectiveness” information are used as the “influence” used when specifying the analysis target data to be discarded.
  • the present invention is limited to using these two types of information.
  • the analysis target data to be discarded may be specified by using information representing another situation of the analysis processing of the analysis target data, and is not limited to two different types of information, but may be discarded by using three or more types of information.
  • Analysis target data to be identified may be specified.
  • FIGS. 9 to 10 are diagrams for explaining the configuration of the analysis system
  • FIG. 11 is a diagram for explaining the operation of the analysis system.
  • the analysis system according to the present embodiment has a configuration that is substantially the same as that of the first embodiment described above, but a method for calculating the “influence” used when specifying the analysis target data to be discarded is different from that of the first embodiment.
  • a method for calculating the “influence” used when specifying the analysis target data to be discarded is different from that of the first embodiment.
  • FIG. 9 shows an example of the configuration of the analysis system 10 in the present embodiment.
  • the analysis system 10 includes an index update determination unit 18 and an index storage unit 19 (impact management unit) in addition to the configuration described in the first embodiment.
  • the index update determination unit 18 uses the freshness calculation unit 14 and the validity calculation unit 15 to calculate the freshness and validity calculated for each frame image that is an analysis processing unit of analysis target data. Are stored in the index storage unit 19 together with identification information (ID) for specifying the frame image.
  • ID identification information
  • the “freshness time limit” and the “effectiveness time limit” may be set and stored with any time limit, for example, a time obtained by adding a preset time from the time when the freshness or validity is accepted.
  • the freshness calculation unit 14 and the effectiveness calculation unit 15 cooperate with the index update determination unit 18 to calculate the “freshness” and “effectiveness” of the frame image. Then, it is checked whether or not the same “ID” as the ID of the frame image exists in the index storage unit 19. Then, whether the current time is within the “freshness due date” and the “effectiveness due date” associated with the same “ID” is checked. If the current time is within the due date, the current time is stored in the index storage unit 19.
  • the freshness calculation unit 14 and the validity calculation unit 15 When the freshness calculation unit 14 and the validity calculation unit 15 read “freshness” and “effectiveness” from the index storage unit 19 as described above, the freshness calculation unit 14 and the validity calculation unit 15 obtain the read “freshness” and “effectiveness”.
  • the data management unit 13 is notified.
  • the data management unit 13 calculates a “data rejection index” based on the notified “freshness” and “effectiveness”, and discards the data from the data storage unit 12 without performing analysis processing by the analysis processing unit 16. Is identified.
  • the freshness calculation unit 14 or the validity calculation unit 15 newly calculates “freshness” or “validity”
  • the data management unit 13 sets the “freshness” or “validity” based on the calculated “freshness” or “validity”.
  • a “data rejection index” is calculated, and a frame image to be discarded from the data storage unit 12 is specified. Then, the data management unit 12 cooperates with the index update determination unit 18 to store and update the newly calculated “freshness” and “validity” in the index storage unit 19 together with the expiration date.
  • the analysis system 10 acquires a frame image which is data to be analyzed by the data acquisition unit 20 which is a camera (step S1), receives the frame image by the data reception unit 11 (step S2), and stores data. Store in the unit 12 (step S3). At this time, the data receiving unit 11 stores the date information and attribute information including the frame ID in the data storage unit 12 in association with each frame image. Note that the analysis system 10 always acquires the analysis target data described above.
  • the data reception unit 11 monitors the free space in the data storage unit 12, and the data reception unit 11 triggers the data management unit 13 when the free space becomes smaller than a predetermined threshold. Is issued (step S4). Upon receiving this trigger, the data management unit 13 acquires analysis target data from the data storage unit 12 (step S5). Then, the data management unit 13 instructs the freshness calculation unit 14 and the validity calculation unit 15 to calculate the freshness and validity for each frame image that is the acquired analysis target data. In particular, the data management unit 13 notifies the effectiveness calculation unit 15 of the acquired analysis processing status, and instructs to calculate the effectiveness by referring to the analysis processing status.
  • the freshness calculation unit 14 and the validity calculation unit 15 calculate the freshness and the validity for each frame image, and notify the data management unit 13 (steps S6 and S7).
  • the data management unit 13 cooperates with the index update determination unit 18 to display the freshness and validity notified from the freshness calculation unit 14 and the validity calculation unit 15 together with the respective expiration dates. 19 (step S7 ′′).
  • the freshness calculation unit 14 and the validity calculation unit 15 have already calculated the freshness and the validity by referring to the ID of the target frame image.
  • the freshness and validity stored in the index storage unit 19 are read (step S7 ′). If the current date is not within the expiration date, freshness and effectiveness are newly calculated (steps S6 and S7), and stored in the index update unit 19 together with the respective expiration dates (step S7 ").
  • the data management unit 13 identifies a frame image as a deletion candidate from the data storage unit 12 as described above based on the calculated freshness and validity. Then, the data management unit 13 discards the frame image that is a deletion candidate from the data storage unit 12 according to the free space in the data storage unit 12 (steps S8 and S9).
  • the data management unit 13 takes out the analysis target data stored in the data storage unit 12 and remains without being discarded from the data storage unit 12, and executes the analysis process in the analysis processing unit 16. Control.
  • the analysis system 10 As described above, in the analysis system 10 according to the present embodiment, it is set when calculating the two types of influences (freshness and effectiveness described above) used to determine whether to discard the analysis target data. If it is within the deadline, it is calculated in the past without newly calculating. This is effective for the analysis target data that can assume a situation in which the freshness and the validity do not change with the passage of time. In such a case, the processing load of the analysis system 10 can be reduced.
  • the “influence degree” used when specifying the analysis target data to be discarded the above-described “freshness” (first influence degree) and “effectiveness” (second influence degree) are two.
  • the type of information is used, it is not limited to using these two types of information.
  • the analysis target data to be discarded may be specified using information representing another situation at the time of receiving the analysis target data or information representing another situation of the analysis processing of the analysis target data. Further, the analysis target data to be discarded may be specified using not only two different types of information but also three or more types of information.
  • (Appendix 1) Data accepting means 101 for accepting continuously input analysis target data; An influence degree calculating means 102 for calculating an influence degree indicating the degree of influence on the accuracy of the analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; , According to the calculated degree of influence, the data management means 103 for specifying the analysis target data of the analysis processing unit to be discarded without being subject to analysis processing by the analysis processing means, The influence degree calculation means 102 calculates a plurality of different kinds of influence degrees set in advance, The data management unit 103 identifies the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences. Analysis processing control system 100.
  • the influence degree calculation means is configured to change the influence degree for each analysis processing unit of the analysis target data based on the reception status of the analysis target data received by the data reception means for each analysis processing unit. Calculating at least one kind of the degree of influence among the plurality of kinds of influences; Analysis processing control system.
  • the influence degree calculating means is configured to change the influence degree for each analysis processing unit of the analysis target data based on the reception time for each analysis processing unit of the analysis target data received by the data reception means. Calculating at least one kind of the degree of influence among the plurality of kinds of influences; Analysis processing control system.
  • the influence degree calculating means is configured to change the influence degree for each analysis processing unit of the analysis target data based on the reception order of the analysis target data received by the data reception means for each analysis processing unit. Calculating at least one kind of the degree of influence among the plurality of kinds of influences; Analysis processing control system.
  • the influence degree calculating means is configured to determine the influence of each analysis processing unit of the analysis target data based on the analysis processing status by the analysis processing means for each analysis processing unit of the analysis target data received by the data receiving means. Calculating the degree as the degree of influence of at least one of the plurality of degrees of influence of the different types; Analysis processing control system.
  • the influence degree calculating means is based on the analysis processing result by the specific analysis processing means for each analysis processing unit of the analysis target data received by the data receiving means, for each analysis processing unit of the analysis target data. Calculating the degree of influence as the degree of influence of at least one of the plurality of different kinds of influences; Analysis processing control system.
  • the influence degree calculating unit is configured to analyze the analysis processing unit of the analysis target data based on a load at the time of analysis processing by a specific analysis processing unit for each analysis processing unit of the analysis target data received by the data receiving unit. Calculating each degree of influence as the degree of influence of at least one of the plurality of different degrees of influence; Analysis processing control system.
  • the influence calculating means is Based on the reception time for each analysis processing unit of the analysis target data received by the data reception means, the influence degree for each analysis processing unit of the analysis target data is calculated as a first influence degree; Based on the receiving order for each analysis processing unit of the analysis target data received by the data receiving means, the influence degree for each analysis processing unit of the analysis target data is calculated as a second influence degree, The data management means specifies the analysis target data of the analysis processing unit to be discarded based on the first influence degree and the second influence degree. Analysis processing control system.
  • the influence calculating means is Based on the reception time for each analysis processing unit of the analysis target data received by the data reception means, the influence degree for each analysis processing unit of the analysis target data is calculated as a first influence degree; Based on the analysis processing status by the analysis processing unit for each analysis processing unit of the analysis target data received by the data reception unit, the influence degree for each analysis processing unit of the analysis target data is determined as a second influence degree. As The data management means specifies the analysis target data of the analysis processing unit to be discarded based on the first influence degree and the second influence degree. Analysis processing control system.
  • the influence calculating means is Based on the first analysis processing status by the analysis processing unit for each analysis processing unit of the analysis target data received by the data receiving unit, the degree of influence of the analysis target data for each analysis processing unit is first determined. And calculating the influence degree for each analysis processing unit of the analysis target data as the second influence degree based on the second analysis processing state different from the first analysis processing state. And The data management means specifies the analysis target data of the analysis processing unit to be discarded based on the first influence degree and the second influence degree. Analysis processing control system.
  • the analysis processing control system according to any one of appendices 1 to 10,
  • the influence degree for each analysis processing unit of the analysis target data calculated by the influence degree calculating means is stored together with information for specifying the analysis target data of the analysis processing unit and information indicating the effective period of the influence degree.
  • the influence degree calculating means calculates the influence degree that the influence degree of the analysis target data of the analysis processing unit is within an effective period based on the information stored in the influence degree management means. Use as the degree of influence calculated for the analysis target data, Analysis processing control system.
  • (Appendix 12) Data accepting means for accepting continuously inputted analysis target data; Analysis processing means for performing analysis processing on the received analysis target data for each analysis processing unit set in advance of the analysis target data; An influence degree calculating means for calculating an influence degree representing the degree of influence on the accuracy of the analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; According to the calculated degree of influence, data management means for specifying the analysis target data of the analysis processing unit to be discarded without being subject to analysis processing by the analysis processing means, The influence calculation means calculates a plurality of different influences of different types set in advance, The data management means identifies the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences. Analysis processing system.
  • Data accepting means for accepting continuously inputted analysis target data;
  • An influence degree calculating means for calculating an influence degree representing the degree of influence on the accuracy of the analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means;
  • data management means for specifying the analysis target data of the analysis processing unit to be discarded without being subjected to analysis processing by the analysis processing means;
  • the influence calculation means calculates a plurality of different influences of different types set in advance,
  • the data management means identifies the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences. A program to make things happen.
  • (Appendix 14) Accept analysis target data input continuously, Calculating the degree of influence representing the degree of influence on the accuracy of analysis processing by the predetermined analysis processing means for the received analysis target data for each analysis processing unit of the analysis target data by the analysis processing means; According to the calculated degree of influence, specify the analysis target data of the analysis processing unit to be discarded without being subject to analysis processing by the analysis processing means, When calculating the degree of influence, each of a plurality of different types of influence set in advance is calculated, Identifying the analysis target data of the analysis processing unit to be discarded based on a plurality of different types of influences; Analysis processing control method.
  • Appendix 15 An analysis processing control method according to appendix 14, When calculating the influence degree, the influence degree for each analysis processing unit of the analysis target data is calculated based on the reception status of the received analysis target data for each analysis processing unit. An analysis processing control method as at least one of the influence degrees.
  • Appendix 16 The analysis processing control method according to appendix 14 or 15, When calculating the influence degree, the influence degree for each analysis processing unit of the analysis target data is calculated based on the analysis processing status by the analysis processing unit for each analysis processing unit of the received analysis target data. Calculating at least one kind of the degree of influence among a plurality of different kinds of influences, Analysis processing control method.
  • the above-described program is stored in a storage device or recorded on a computer-readable recording medium.
  • the recording medium is a portable medium such as a flexible disk, an optical disk, a magneto-optical disk, and a semiconductor memory.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

La présente invention concerne le système (100) de commande d'un processus d'analyse qui est équipé des éléments suivants : un moyen de calcul (102) de degré d'influence qui calcule, pour chaque unité d'un processus d'analyse effectué sur des données de sujet d'analyse par un moyen de processus d'analyse prescrit, un degré d'influence représentant le degré d'influence sur le processus d'analyse effectué par le moyen de processus d'analyse par rapport aux données de sujet d'analyse reçues ; et un moyen de gestion (103) de données qui, conformément au degré d'influence calculé, spécifie, pour chaque unité du processus d'analyse, les données de sujet d'analyse qui sont à écarter sans être soumises au processus d'analyse par le moyen de processus d'analyse. La configuration est telle que le moyen de calcul (102) de degré d'influence calcule individuellement de multiples types différents et prédéfinis de degrés d'influence et le moyen de gestion (103) de données spécifie les données de sujet d'analyse à écarter pour chaque unité du processus d'analyse, en fonction des multiples types différents de degrés d'influence.
PCT/JP2013/006535 2012-12-19 2013-11-06 Système de commande d'un processus d'analyse WO2014097530A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2012276653 2012-12-19
JP2012-276653 2012-12-19

Publications (1)

Publication Number Publication Date
WO2014097530A1 true WO2014097530A1 (fr) 2014-06-26

Family

ID=50977898

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2013/006535 WO2014097530A1 (fr) 2012-12-19 2013-11-06 Système de commande d'un processus d'analyse

Country Status (1)

Country Link
WO (1) WO2014097530A1 (fr)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005534205A (ja) * 2001-11-21 2005-11-10 イー・トレピッド・テクノロジーズ・エルエルシー 複数のカメラから与えられたデジタルビデオ内容を記憶する方法および装置
JP2008011457A (ja) * 2006-06-30 2008-01-17 Olympus Imaging Corp カメラ
JP2010204892A (ja) * 2009-03-03 2010-09-16 Nippon Telegr & Teleph Corp <Ntt> 映像解析装置，映像解析方法および映像解析プログラム
WO2012098853A1 (fr) * 2011-01-20 2012-07-26 日本電気株式会社 Système de répartition de données de processus de détection de ligne de production, procédé de répartition de données de processus de détection de ligne de production et programme

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005534205A (ja) * 2001-11-21 2005-11-10 イー・トレピッド・テクノロジーズ・エルエルシー 複数のカメラから与えられたデジタルビデオ内容を記憶する方法および装置
JP2008011457A (ja) * 2006-06-30 2008-01-17 Olympus Imaging Corp カメラ
JP2010204892A (ja) * 2009-03-03 2010-09-16 Nippon Telegr & Teleph Corp <Ntt> 映像解析装置，映像解析方法および映像解析プログラム
WO2012098853A1 (fr) * 2011-01-20 2012-07-26 日本電気株式会社 Système de répartition de données de processus de détection de ligne de production, procédé de répartition de données de processus de détection de ligne de production et programme

Similar Documents

Publication Publication Date Title
CN104050449B (zh) 一种人脸识别方法及装置
US11423054B2 (en) Information processing device, data processing method therefor, and recording medium
KR20140091599A (ko) 소셜 네트워킹 정보를 이용한 얼굴 인식
JP5300585B2 (ja) 情報処理装置、情報処理方法
US20150146006A1 (en) Display control apparatus and display control method
CN106295598A (zh) 一种跨摄像头目标跟踪方法及装置
US20160239712A1 (en) Information processing system
US20130022244A1 (en) Image processing apparatus and image processing method
US20170177423A1 (en) Management system for notifications using contextual metadata
JP2010211485A (ja) 注視度合測定装置、注視度合測定方法、注視度合測定プログラムおよびそのプログラムを記録した記録媒体
JP2018045302A (ja) 情報処理装置、情報処理方法及びプログラム
JP2011070629A (ja) 広告効果測定システム及び広告効果測定装置
US20180139347A1 (en) Timed image deletion from a personal electronic device
CN113674313A (zh) 一种行人跟踪方法、装置、存储介质和电子设备
WO2018179119A1 (fr) Appareil d&#39;analyse d&#39;image, procédé d&#39;analyse d&#39;image et support d&#39;enregistrement
CN110348367B (zh) 视频分类方法、视频处理方法、装置、移动终端及介质
US20180197000A1 (en) Image processing device and image processing system
WO2014097530A1 (fr) Système de commande d&#39;un processus d&#39;analyse
JP2005157134A (ja) 情報出力方法及び装置及びプログラム及び情報出力プログラムを格納したコンピュータ読み取り可能な記憶媒体
CN108334811B (zh) 一种人脸图像处理方法及装置
US10283166B2 (en) Video indexing method and device using the same
JPWO2019163699A1 (ja) 特徴抽出方法、照合システム、およびプログラム
WO2014122879A1 (fr) Système de traitement d&#39;analyse
JP2018137639A (ja) 動画像処理システム、並びに、符号化装置及びプログラム、並びに、復号装置及びプログラム
CN109033924A (zh) 一种视频中人形检测的方法及装置

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: 13866285

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: 13866285

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: JP