Video monitoring device, video monitoring method, and program

Abstract

To provide a video monitoring device capable of appropriately searching for a desired object. [Solution] The feature data included in the tracking data TD includes first through nth feature value groups, and the search condition data SQC includes all or part of the first through nth search feature value groups. A search unit 36, where 1≦k≦n, when the search condition data SQC includes the kth search feature value group, determines the similarity between the kth search feature value group and the kth feature value group included in the feature data belonging to the tracking data TD that is most similar to the kth search feature value group, and searches for tracking data TD that matches the search condition data SQC based on all or part of the first through nth similarities.

Description

[Technical Field]

[0001] The disclosed technology relates to a video monitoring device, a video monitoring method, and a program. [Background technology]

[0002] The following Patent Documents 1 to 6, for example, are known as techniques for searching whether an object (for example, a person) captured in an image obtained from a camera or the like corresponds to a search object (for example, a search target person). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2024-034415 [Patent Document 2] Patent Publication No. 2021-101384 [Patent Document 3] Japanese Patent Publication No. 2022-030846 [Patent Document 4] Patent Publication No. 2021-086573 [Patent Document 5] Japanese Patent Application Publication No. 2019-109709 [Patent Document 6] Japanese Patent Application Laid-Open No. 2013-003964 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above-mentioned technology does not particularly mention the effective use of the history of analysis results for images acquired from a camera or the like. The disclosed technology has been developed in consideration of the above-mentioned circumstances, and aims to provide a video monitoring device, a video monitoring method, and a program that can appropriately search for the object of interest by effectively utilizing the history of analysis results for the video. [Means for solving the problem]

[0005] In order to solve the above problems, the video monitoring device of the present invention comprises a video analysis unit that analyzes video data acquired from a camera, extracts feature data defining multiple features from each of multiple objects in the video, and generates tracking data including the feature data in one or multiple frames for each object identified from the extracted feature data based on predetermined tracking conditions; a search unit that searches for tracking data that matches the search condition data by comparing each of the tracking data with search condition data that defines multiple features of the object to be searched; and if tracking data that matches any of the search condition data is found, The object being searched for was found and a notification unit that notifies the user that the feature data includes first through n-th feature value groups, where n is a plural number, and the search condition data includes all or part of first through n-th search feature value groups corresponding to the first through n-th feature value groups, respectively. When the search condition data includes the k-th search feature value group, the search unit, where 1≦k≦n, determines the similarity between the k-th search feature value group and the k-th feature value group included in the feature data belonging to the tracking data that is most similar to the k-th search feature value group, and searches for the tracking data that matches the search condition data based on all or part of the first through n-th similarities. [Effects of the Invention]

[0006] According to the disclosed technology, by effectively utilizing the history of analysis results for the video, it is possible to appropriately search for the object of interest. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram of a monitoring system according to a first embodiment. [Figure 2] FIG. 2 is a diagram illustrating the operation of a video analysis unit. [Figure 3] FIG. 4 is a diagram illustrating the operation of a search unit. [Figure 4] FIG. 10 is another explanatory diagram of the operation of the search unit. [Figure 5]FIG. 1 is a block diagram of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0008] [First embodiment] <Configuration of the first embodiment> The configuration of a monitoring system 100 according to a first embodiment of the disclosed technology will be described below. FIG. 1 is a block diagram of a monitoring system 100 according to a first embodiment. The monitoring system 100 includes multiple cameras 12, an analysis result storage DB 14, a monitoring device 16 (video monitoring device, computer), and a condition DB 18. Note that "DB" stands for "database." The monitoring device 16 includes a video analysis unit 22 (video analysis means, video analysis process), an analysis result registration unit 24, a search condition setting unit 32, a condition registration unit 34, a search unit 36 ​​(search means, search process), and an alarm unit 38 (alert means, alarm process).

[0009] The video analysis unit 22 samples the video data acquired from the multiple cameras 12 at a predetermined sampling period (for example, every 10 seconds). The sampled results are called sample frame images. Furthermore, the video analysis unit 22 detects one or more person areas, which are video areas in which people are captured, from each sample frame image.

[0010] Furthermore, the video analysis unit 22 detects feature values ​​corresponding to a plurality of feature items, such as physique and body type, for each person region. A collection of detected feature values ​​is called "feature data." The video analysis unit 22 also creates time-series information of the feature data for the same person as tracking data TD. The analysis result registration unit 24 registers the created tracking data TD in the analysis result accumulation DB 14.

[0011] Furthermore, the search condition setting unit 32 creates search condition data SQC that defines search characteristic values ​​that are characteristic values ​​related to the person to be searched for, based on a user operation. The condition registration unit 34 registers the search condition data SQC in the condition DB 18. The search unit 36 ​​compares each piece of tracking data TD with each piece of search condition data SQC.

[0012] That is, if the number of pieces of tracking data TD is N and the number of pieces of search condition data SQC is M, the search unit 36 ​​performs "N x M" matching. When the search unit 36 ​​finds tracking data TD that matches any of the search condition data SQC, it notifies the user via the notification unit 38 that the person being searched for has been found. The user may determine the search condition data SQC in advance, or may add new search condition data SQC while the monitoring system 100 is in operation.

[0013] As described above, the video analysis unit 22 detects feature values ​​corresponding to a plurality of feature items, such as physique and body type, for each person region. Each feature item belongs to one of a plurality of feature item groups. In this embodiment, although not shown, the feature item groups are assumed to be three: a whole-body feature item group CIB, a facial feature item group CIF, and an attribute feature item group CIA. The feature items included in each feature item group are, for example, as listed below.

[0014] · Whole-body characteristic item group CIB: Includes characteristic items such as physique, body type, height, shoulder width, body proportions (head to torso, limb length ratio), etc. Facial feature item group CIF: Includes feature items such as face shape, forehead width, jaw shape, facial contours (round, oval, square, etc.), eye size, position, spacing, shape, eyelid shape, nose height, nose width, nose shape, etc. Attribute feature item group CIA: Includes feature items such as gender, estimated age, hairstyle, hair color, accessories (masks, hats, headphones, etc.), upper body clothing, lower body clothing, belongings, and footwear.

[0015] <Operation of the video analysis unit 22> FIG. 2 is a diagram illustrating the operation of the video analysis unit 22. Images VA1 to VA4 shown in Fig. 2 are sample frame images, and are assumed to be acquired sequentially at times t1 to t4 in the above-mentioned sampling period (for example, 10 seconds). The contents of the data acquired at each time based on images VA1 to VA4 will be described below. In the following description, symbols for various data are primarily expressed using capital letters, such as "tracking data TD." However, when there are multiple pieces of this data, they may be expressed by adding numbers after capital letters, such as "tracking data TD1, TD2," in order to distinguish between the multiple pieces of data.

[0016] (time t1) Assume that at time t1, image VA1 is supplied to video analysis unit 22 (see FIG. 1), and video analysis unit 22 detects two person areas H1 and H2 in image VA1. Video analysis unit 22 analyzes each of person areas H1 and H2 to generate person frame data HFD1 and HFD2.

[0017] The person frame data HFD (e.g., HFD1) includes a person identification number HID and feature data CD. The feature data CD includes a whole-body feature value group CPB (first feature value group), a face feature value group CPF (second feature value group), and an attribute feature value group CPA (third feature value group, nth feature value group).

[0018] Here, the whole body feature value group CPB, face feature value group CPF, and attribute feature value group CPA are sets of feature values ​​obtained for the above-mentioned whole body feature item group CIB, face feature item group CIF, and attribute feature item group CIA.

[0019] 2, the feature values ​​included in each of the feature value groups CPB, CPF, and CPA are represented by English letters such as "X," "Y," and "Z." Each feature value group, i.e., a set of feature values, is represented by a string of English letters such as "XXXX." In the illustrated example of feature data CD1, the whole-body feature value group CPB1 is "XXXX," the face feature value group CPF1 is "YYYY," and the attribute feature value group CPA1 is "ZZZZ."

[0020] Furthermore, the person frame data HFD2 includes person identification information HID2 and feature data CD2, and the feature data CD2 includes feature value groups CPB2, CPF2, and CPA2.

[0021] If the video analysis unit 22 has not acquired other person frame data HFD before acquiring the person frame data HFD1, HFD2 at time t1, the video analysis unit 22 creates new tracking data TD1, TD2 corresponding to the person frame data HFD1, HFD2, respectively. The tracking data TD (for example, TD1) is data including identification information TID that uniquely defines the tracking data TD and one or more person frame data HFD.

[0022] Then, at time t1, the tracking data TD1 becomes data including one person frame data HFD1 and identification information TID1. Similarly, at time t1, the tracking data TD2 becomes data including one person frame data HFD2 and identification information TID2. When the tracking data TD1 and TD2 are created in this manner, the analysis result registration unit 24 (see FIG. 1) accumulates the created tracking data TD1 and TD2 in the analysis result accumulation DB 14.

[0023] (time t2) Thereafter, at time t2, image VA2 is supplied to video analysis unit 22, which detects person area H3 from image VA2. Video analysis unit 22 generates person frame data HFD3 by analyzing person area H3. Then, in feature data CD3 of person frame data HFD3, whole-body feature value group CPB3 is "XXXY," face feature value group CPF3 is "YYYZ," and attribute feature value group CPA3 is "ZZZX."

[0024] The video analysis unit 22 calculates the similarity between the person frame data HFD1, HFD2 acquired at the previous time t1 and the person frame data HFD3 acquired this time. Then, the video analysis unit 22 determines that the person frame data HFD related to the same person if the tracking condition is met, with the tracking condition being that the similarity is equal to or greater than a predetermined tracking determination threshold Tht (not shown). In the example of FIG. 2, the whole-body feature value group CPB1(XXXX) and the whole-body feature value group CPB3(XXXY) have the same three parameter values ​​and differ in one parameter value. Therefore, the similarity between the whole-body feature value groups CPB1, CPB3 is 3 / 4=75%.

[0025] Similarly, the similarity between the facial feature value group CPF1(YYYY) and the facial feature value group CPF3(YYYZ) is also "75%," and the similarity between the attribute feature value group CPA1(ZZZZ) and the attribute feature value group CPA3(ZZZX) is also "75%." The average value of the similarities between these feature value groups CPB, CPF, and CPA is set to the similarity of the person frame data. Therefore, the similarity between the person frame data HFD1 and HFD3 is "75%." If the tracking determination threshold Tht is, for example, "70%,," the video analysis unit 22 determines that the person frame data HFD1 and HFD3 are of the same person.

[0026] As a result, the video analysis unit 22 adds the person frame data HFD3 to the tracking data TD1. That is, the video analysis unit 22 updates the tracking data TD1 to include both of the person frame data HFD1 and HFD3. On the other hand, since there is no new person frame data HFD to be added to the tracking data TD2, the tracking data TD2 is maintained as before. Then, the analysis result registration unit 24 writes the updated tracking data TD1 to the analysis result accumulation DB 14.

[0027] (time t3) Thereafter, at time t3, image VA3 is supplied to video analysis unit 22, which detects person area H4 from image VA3. Video analysis unit 22 analyzes person area H4 to generate person frame data HFD4. Then, in feature data CD4 of person frame data HFD4, whole-body feature value group CPB4 is "XXYY", face feature value group CPF4 is "YYZZ", and attribute feature value group CPA4 is "ZZXX".

[0028] The video analysis unit 22 calculates the similarity between the person frame data HFD3 acquired previously at time t2 and the person frame data HFD4 acquired this time. In the example of Fig. 2, the similarity between the person frame data HFD3 and HFD4 is "75%." As a result, the video analysis unit 22 determines that the person frame data HFD3 and HFD4 are of the same person.

[0029] As a result, the video analysis unit 22 adds the person frame data HFD4 to the tracking data TD1. That is, the video analysis unit 22 updates the tracking data TD1 to include the person frame data HFD1, HFD3, and HFD4. Then, the analysis result registration unit 24 writes the updated tracking data TD1 to the analysis result accumulation DB 14.

[0030] (time t4) Thereafter, at time t4, when image VA4 is supplied to video analysis unit 22, video analysis unit 22 detects person area H5 from image VA4. Video analysis unit 22 generates person frame data HFD5 by analyzing person area H5. Then, in feature data CD5 of person frame data HFD5, whole-body feature value group CPB5 is "XYYY", face feature value group CPF5 is "YZZZ", and attribute feature value group CPA5 is "ZXXX".

[0031] The video analysis unit 22 calculates the similarity between the person frame data HFD4 acquired at the previous time t3 and the person frame data HFD5 acquired this time. In the example of Fig. 2, the similarity between the person frame data HFD4 and HFD5 is "75%." As a result, the video analysis unit 22 determines that the person frame data HFD4 and HFD5 are of the same person.

[0032] As a result, the video analysis unit 22 adds the person frame data HFD5 to the tracking data TD1. That is, the video analysis unit 22 updates the tracking data TD1 to include the person frame data HFD1, HFD3, HFD4, and HFD5. Then, the analysis result registration unit 24 writes the updated tracking data TD1 to the analysis result accumulation DB 14. As a result, at time t4, the tracking data TD1 and TD2 have the content as shown in FIG. 2.

[0033] <Operation of Search Unit 36> FIG. 3 is a diagram illustrating the operation of the search unit 36. 3, tracking data TD1 and TD2 are data stored in the analysis result accumulation DB 14. Tracking data TD1 includes person frame data HFD1, HFD3, HFD4, and HFD5, and tracking data TD2 includes person frame data HFD2. Search condition data SQC11, SQC12, and SQC13 in the figure are data stored in the condition DB 18.

[0034] The search condition data SQC includes similarity conditions LD and search feature data SD, where the search feature data SD includes all or part of a whole-body search feature value group SPB (first search feature value group), a face search feature value group SPF (second search feature value group), and an attribute search feature value group SPA (third search feature value group, nth search feature value group).

[0035] These search feature value groups SPB, SPF, and SPA are sets of feature values ​​that serve as search conditions for the above-mentioned feature item groups CIB, CIF, and CIA, respectively. Also, the similarity conditions LD included in the search condition data SQC define the conditions for determining that the search condition data SQC and the tracking data TD are "similar," i.e., "presumed to be the same person."

[0036] 3, the search condition data SQC11 includes a similarity condition LD11 and search feature data SD11. The search feature data SD11 includes a whole-body search feature value group SPB11, a face search feature value group SPF11, and an attribute search feature value group SPA11.

[0037] The search condition data SQC12 also includes similarity conditions LD12 and search feature data SD12. Here, the search feature data SD12 includes search feature value groups SPB12 and SPF12, but does not include the attribute search feature value group SPA.

[0038] The search condition data SQC13 includes a similarity condition LD13 and search feature data SD13. Here, the search feature data SD13 includes only the whole-body search feature value group SPB13, and does not include the face search feature value group SPF or the attribute search feature value group SPA.

[0039] FIG. 4 is another explanatory diagram of the operation of the search unit 36. In FIG. 4 shows the detailed operation when comparing the trace data TD1 with the similarity condition LD11. The operation content in this case will be explained for each of the times t1 to t4 when the trace data TD1 is updated.

[0040] (time t1) At time t1, the tracking data TD1 contains only person frame data HFD1, so the search unit 36 ​​compares the feature data CD1 contained in the person frame data HFD1 with the search feature data SD11 contained in the search condition data SQC11.

[0041] The similarities between the feature value group CPB, CPF, CPA of the latest person frame data HFD (HFD1) at the current time (here, time t1) and the search feature value group SPB, SPF, SPA in the applied search condition data SQC are the current whole-body similarity CB, the current face similarity CF, and the current attribute similarity CA, respectively.

[0042] As shown in the figure, at time t1, the current whole-body similarity CB is 80%, the current face similarity CF is 20%, and the current attribute similarity CA is 50% detected by the search unit 36. Furthermore, the search unit 36 ​​identifies the highest values ​​of the similarities CB, CF, and CA up to this point as the highest whole-body similarity MB (first similarity), the highest face similarity MF (second similarity), and the highest attribute similarity MA (third similarity, nth similarity), respectively.

[0043] At time t1, the highest similarities MB, MF, and MA are equal to the current similarities CB, CF, and CA, that is, 80%, 20%, and 50%. Here, suppose that the similarity condition LD11 is "the average value of the highest similarities MB, MF, and MA is 80% or more." At time t1, the average value of the highest similarities MB, MF, and MA is 50%, so this similarity condition LD11 is not satisfied.

[0044] (time t2) At time t2, person frame data HFD3 is added to the tracking data TD1. The search unit 36 ​​then compares the feature data CD3 in the person frame data HFD3 with the search feature data SD11 in the search condition data SQC11.

[0045] At time t2, the search unit 36 ​​compares the feature data CD3 with the search feature data SD11, and as a result, the current whole-body similarity CB is 50%, the current facial similarity CF is 20%, and the current attribute similarity CA is 80%, as shown in the figure.

[0046] Of these, the current attribute similarity CA (80%) exceeds the highest attribute similarity MA (50%) before time t2. Therefore, the search unit 36 ​​updates the highest attribute similarity MA to the currently detected current attribute similarity CA (80%). As a result, the highest similarities MB, MF, and MA become 80%, 20%, and 80%. Because the average of these highest similarities MB, MF, and MA is 60%, the similarity condition LD11, which states that "the average of the highest similarities MB, MF, and MA is 80% or more," has not yet been satisfied.

[0047] (time t3) Next, at time t3, person frame data HFD4 is added to the tracking data TD1. The search unit 36 ​​then compares the feature data CD4 in the person frame data HFD4 with the search feature data SD11 in the search condition data SQC11.

[0048] Assume that the search unit 36 ​​compares the feature data CD4 with the searched feature data SD11 at time t3 and finds that the current whole-body similarity CB is 50%, the current facial similarity CF is 80%, and the current attribute similarity CA is 30%, as shown in the figure. Of these, the current facial similarity CF (80%) exceeds the highest facial similarity MF (20%) before time t3. Therefore, the search unit 36 ​​updates the highest facial similarity MF to the currently detected current attribute similarity CA (80%).

[0049] As a result, the highest similarities MB, MF, and MA are 80%, 80%, and 80%. Since the average value of these highest similarities MB, MF, and MA is 80%, the similarity condition LD11, which states that "the average value of the highest similarities MB, MF, and MA is 80% or more," is satisfied.

[0050] Therefore, the search unit 36 ​​notifies the user via the notification unit 38 that "it is presumed that the person associated with the tracking data TD1 and the person associated with the search condition data SQC11 are the same person." At this point, the comparison between the tracking data TD1 and the search condition data SQC11 is completed. Therefore, even if new person frame data HFD5 is added to the tracking data TD1 after time t4, the comparison between the tracking data TD1 and the search condition data SQC11 is omitted.

[0051] The similarity condition LD can be variously varied in addition to the similarity condition LD11 described above. For example, the search feature data SD12 of the search condition data SQC12 shown in FIG. 3 does not include the attribute search feature value group SPA. In such a case, the similarity condition LD12 may be a condition that does not involve the attribute maximum similarity MA, such as "the average value of the maximum similarities MB and MF is 80% or more." Similarly, the search feature data SD13 of the search condition data SQC13 does not include the face search feature value group SPF and the attribute search feature value group SPA. Therefore, the similarity condition LD13 may be a condition based only on the whole-body maximum similarity MB, such as "the whole-body maximum similarity MB is 80% or more."

[0052] Furthermore, the similarity condition LD can assign "weighting" or "priority settings" to the highest similarities MB, MF, and MA. For example, the similarity condition LD can be set as "double weighting is given to the highest similarity MB, and the average value of the highest similarities MB, MF, and MA is 80% or more." In this case, the similarity condition LD is satisfied if "(MB x 2 + MF + MA) / 4" is 80% or more.

[0053] Furthermore, for example, the similarity condition LD can be set to prioritize a certain highest similarity (the highest whole-body similarity MB in the above example), such as "the highest whole-body similarity MB is 80% or more, or the average of the highest similarities MB, MF, and MA is 80% or more." In this case, even if the average of the highest similarities MB, MF, and MA is less than 80%, the similarity condition LD is satisfied as long as the highest whole-body similarity MB is 80% or more.

[0054] [Computer Configuration] 5 is a block diagram of the computer 980. The monitoring device 16 shown in FIG. 1 includes one or more computers 980 shown in FIG. 5, a computer 980 includes a CPU 981, a storage unit 982, a communication port 983, an input / output port 984, and a media port 985. The storage unit 982 includes a RAM 982a, a ROM 982b, and an SSD (Solid State Drive) 982c. The communication port 983 is connected to a communication circuit 986. The input / output port 984 is connected to an input / output device 987. The media port 985 reads and writes data from a recording medium 988.

[0055] The ROM 982b stores an IPL (Initial Program Loader) executed by the CPU, etc. The SSD 982c stores application programs, various data, etc. The CPU 981 executes application programs, etc. loaded from the SSD 982c to the RAM 982a, thereby realizing various functions. The interior of the monitoring device 16 shown in FIG. 1 is primarily a block diagram of functions realized by application programs, etc.

[0056] [Variations] The disclosed technology is not limited to the above-described embodiments, and various modifications are possible. The above-described embodiments are provided as examples to facilitate understanding of the disclosed technology, and are not necessarily limited to those including all of the described configurations. Furthermore, other configurations may be added to the configurations of the above-described embodiments, and some of the configurations may be replaced with other configurations. Furthermore, the control lines and information lines shown in the figures are those considered necessary for explanation, and do not necessarily represent all control lines and information lines necessary in the product. In reality, it can be assumed that almost all configurations are interconnected. Possible modifications of the above-described embodiments include, for example, the following:

[0057] (1) In the above embodiment, the object to be searched for is a "person." However, in the present disclosure, the object is not limited to a person, and may be an animal, a vehicle, or other object.

[0058] (2) In the above embodiment, n, which is the number of feature value groups CPB, CPF, CPA, search feature value groups SPB, SPF, SPA, similarities CB, CF, CA, and highest similarities MB, MF, MA, was all 3, but n may be 2 or a natural number equal to or greater than 4. For example, the face feature value group CPF may be subdivided into "eyes," "nose," "ears," "lips," etc., and each of these may be an independent feature value group.

[0059] (3) Since the hardware of the monitoring device 16 in the above embodiment can be realized by a general computer, the programs for executing the various processes described above may be stored on a storage medium (a computer-readable storage medium on which the programs are recorded) or distributed via a transmission path.

[0060] (4) In the above embodiment, the various processes described above are described as software processes using programs, but some or all of them may be replaced with hardware processes using ASICs (Application Specific Integrated Circuits) or FPGAs (Field Programmable Gate Arrays), etc.

[0061] (5) The various processes executed in the above-described embodiments may be executed by a server computer via a network (not shown), and the various data stored in the above-described embodiments may also be stored in the server computer.

[0062] [Effects of the embodiment] As described above, according to the embodiment, the feature data CD includes first through n-th feature value groups (CPB, CPF, CPA), where n is a plural number, and the search condition data SQC includes all or part of first through n-th search feature value groups (SPB, SPF, SPA) corresponding to the first through n-th feature value groups (CPB, CPF, CPA), respectively. When the search condition data SQC includes the k-th search feature value group, the search unit 36, where 1≦k≦n, determines the similarity between the k-th search feature value group and the feature value group most similar to the k-th search feature value group included in the feature data CD belonging to the tracing data TD as the k-th similarity, and searches for tracing data TD that matches the search condition data SQC based on all or part of the first through n-th similarities (MB, MF, MA).

[0063] This allows the tracking data TD that matches the search condition data SQC to be searched for based on the first to nth similarities (MB, MF, MA), making it possible to effectively utilize the history of analysis results for the video and appropriately search for the object of interest.

[0064] Furthermore, it is more preferable that the search condition data SQC includes weightings to be assigned to all or some of the first through n-th similarities (MB, MF, MA), or priority settings for any of the first through n-th similarities (MB, MF, MA). By weighting or prioritizing in this way, it is possible to more appropriately search for the object to be searched for. [Explanation of symbols]

[0065] 12 Camera 16 Surveillance equipment (video surveillance equipment, computers) 22 Video analysis unit (video analysis means, video analysis process) 36 Search section (search means, search process) 38 Notification section (notification means, notification process) CD feature data MB Highest whole-body similarity (first similarity) MF highest face similarity (second similarity) MA Attribute Highest Similarity (3rd Similarity, nth Similarity) TD Tracking Data CPB whole body feature value group (first feature value group) CPF facial feature value group (second feature value group) CPA attribute feature value group (3rd feature value group, nth feature value group) SPB Whole-body search feature value set (first search feature value set) SPF face search feature value set (second search feature value set) SPA attribute search feature value group (third search feature value group, nth search feature value group) SQC search condition data

Claims

1. a video analysis unit that analyzes video data acquired from the camera, extracts feature data defining a plurality of features from each of a plurality of objects in the video, and generates tracking data including the feature data in one or a plurality of frames for each object identified from the extracted feature data based on predetermined tracking conditions; a search unit that searches for the tracking data that matches the search condition data by comparing each of the tracking data with search condition data that defines a plurality of characteristics of an object to be searched; a notification unit that notifies the user that an object to be searched for has been found when the tracking data matches any of the search condition data, the feature data includes first to n-th feature value groups, where n is a plural number; the search condition data includes all or part of first to nth search feature value groups corresponding to the first to nth feature value groups, respectively; When the search condition data includes a k-th search feature value group, the search unit determines the similarity between the k-th search feature value group and the feature value that is most similar to the k-th search feature value group among the k-th feature value group included in the feature data belonging to the tracing data as the k-th similarity, and searches for the tracing data that matches the search condition data based on all or part of the first to n-th similarities. A video surveillance device characterized by:

2. The search condition data is Weighting of all or part of the first to n-th similarities, or priority setting of any of the first to n-th similarities, is included.

2. The video monitoring device according to claim 1.

3. a video analysis process of analyzing video data acquired from a camera, extracting feature data defining a plurality of features from each of a plurality of objects in the video, and generating tracking data including the feature data in one or a plurality of frames for each object identified from the extracted feature data based on predetermined tracking conditions; a search process for searching for tracking data that matches the search condition data by comparing each of the tracking data with search condition data that defines a plurality of characteristics of an object to be searched; a notification step of notifying the user that an object to be searched for has been found when the tracking data matches any of the search condition data; the feature data includes first to n-th feature value groups, where n is a plural number; the search condition data includes all or part of first to nth search feature value groups corresponding to the first to nth feature value groups, respectively; The search step is a step of searching for the tracing data that matches the search condition data based on all or part of the first to n-th similarities, where 1≦k≦n, when the search condition data includes a k-th search feature value group, and the similarity between the k-th search feature value group and the feature value that is most similar to the k-th search feature value group among the k-th feature value group included in the feature data belonging to the tracing data is set as the k-th similarity. A video monitoring method comprising:

4. Computer, a video analysis means for analyzing video data acquired from a camera, extracting feature data defining a plurality of features from each of a plurality of objects in the video, and generating tracking data including the feature data in one or a plurality of frames for each object identified from the extracted feature data based on predetermined tracking conditions; a search means for comparing each of the tracking data with search condition data defining a plurality of characteristics of an object to be searched, and searching for the tracking data that matches the search condition data; a notification means for notifying that an object to be searched for has been found when the tracking data that matches any of the search condition data is found; A program for functioning as the feature data includes first to n-th feature value groups, where n is a plural number; the search condition data includes all or part of first to nth search feature value groups corresponding to the first to nth feature value groups, respectively; The search means is a means for searching for the tracing data that matches the search condition data based on all or part of the first to nth similarities, when the search condition data includes a kth search feature value group, and the similarity between the kth search feature value group and the feature value that is most similar to the kth search feature value group among the kth feature value group included in the feature data belonging to the tracing data, where 1≦k≦n. A program characterized by:

Images