NETWORK MANAGEMENT DEVICE, NETWORK MANAGEMENT METHOD, AND VIDEO DISTRIBUTION SYSTEM

Abstract

A network management device (100B) according to the present disclosure comprises: an acquisition unit (111) that acquires required quality indicating video quality required for a video distribution device for video distribution in order to recognize an object from video; a first calculation unit (112) that calculates analysis performance of a video in which an object can be recognized, on the basis of the required quality acquired by the acquisition unit (111); and a second calculation unit (113) that calculates a parameter relating to video distribution on the basis of the analysis performance calculated by the first calculation unit (112).

Description

[Technical field]

[0001] The present disclosure relates to a network management device, a network management method, and a video distribution system. [Background technology]

[0002] In recent years, with the advent of local 5G (Fifth Generation), local governments and private operators (users) have been able to build mobile networks and operate them themselves. For example, a civilian may build a mobile network to collect video from a surveillance camera and detect the presence or absence of an object (e.g., a person, a face, etc.).

[0003] Attention is also being paid to technologies for efficiently operating mobile networks. For example, Patent Document 1 discloses a technology for achieving both improvement of system QoE (Quality of Experience) and improvement of service priority and system throughput. According to the technology disclosed in Patent Document 1, a resource amount (throughput) that satisfies a required QoE is determined using a QoE function that indicates the relationship between bit rate and QoE. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 5011408 Summary of the Invention [Problem to be solved by the invention]

[0005] However, depending on the network conditions, it may not be possible to secure a sufficient communication bandwidth for video distribution. For example, in the case of video used to detect objects, such as video captured by a surveillance camera, the video bit rate tends to be high when the video is distributed in order to maintain high recognition accuracy. However, when the video bit rate becomes high, it becomes difficult to secure a sufficient communication bandwidth.

[0006] In view of the above-mentioned problems, an object of the present disclosure is to provide a network management device, a network management method, and a video distribution system that can achieve efficient video distribution while maintaining recognition accuracy. [Means for solving the problem]

[0007] According to one aspect, a network management device includes: an acquisition unit that acquires a required quality indicating a quality of the video that is required of a video distribution device that distributes the video in order to recognize an object from the video; a first calculation unit that calculates an analysis performance of the image capable of recognizing the object based on the required quality acquired by the acquisition unit; and a second calculation unit that calculates a parameter related to the distribution of the video based on the analysis performance calculated by the first calculation unit.

[0008] A network management method according to one aspect includes the steps of: an acquisition step of acquiring a required quality indicating a quality of the video required for a video distribution device that distributes the video in order to recognize an object from the video; a first calculation step of calculating an analytical performance of the image capable of recognizing the object based on the required quality acquired by the acquisition step; and a second calculation step of calculating parameters related to the distribution of the video based on the analysis performance calculated in the first calculation step.

[0009] A network system according to one aspect includes: an acquisition unit that acquires a required quality indicating a quality of the video that is required of a video distribution device that distributes the video in order to recognize an object from the video; a first calculation unit that calculates an analysis performance of the image capable of recognizing the object based on the required quality acquired by the acquisition unit; and a second calculation unit that calculates a parameter related to the distribution of the video based on the analysis performance calculated by the first calculation unit. Effect of the Invention

[0010] According to the above-described aspects, it is possible to provide a network management device, a network management method, and a network system that can realize efficient video distribution while maintaining recognition accuracy. [Brief description of the drawings]

[0011] [Figure 1] 1 is a block diagram showing a configuration example of a video distribution system according to a first embodiment. [Diagram 2] 4 is a flow diagram illustrating an example of a schematic operation flow of the video distribution system according to the first embodiment. FIG. [Diagram 3] FIG. 11 is a block diagram showing a configuration example of a network management device according to a second embodiment. [Figure 4] FIG. 1 is a diagram showing an example of the configuration of a video distribution system assumed in each of the third to fifth embodiments. [Diagram 5] FIG. 13 is a diagram illustrating an example of a method for setting a video bit rate in each of the third to fifth embodiments and the related art. [Figure 6] FIG. 13 is a diagram for explaining Family-wise recognition performed in each of the third to fifth embodiments from the viewpoint of a solution. [Figure 7] FIG. 1 is a diagram for explaining, from a technical viewpoint, Family-wise recognition performed in each of the third to fifth embodiments. [Figure 8] FIG. 11 is a block diagram showing a configuration example of a network management device according to a third embodiment. [Figure 9] FIG. 13 is a diagram for explaining an example of a definition of a recognition rate according to the third embodiment. [Figure 10]FIG. 13 is a diagram illustrating an example of a Recall versus frame function according to the third embodiment. [Figure 11] FIG. 13 is a diagram illustrating an example of an approximation curve obtained by approximating a function of Recall versus video bit rate in the third embodiment. [Figure 12] FIG. 13 is a diagram illustrating an example of a method for collecting data on video bit rate versus Recall. [Figure 13] 12 is a diagram for explaining an example of a method for calculating the approximation curve shown in FIG. 11. FIG. [Figure 14] FIG. 13 is a diagram illustrating an example of a result of a video bit rate versus a recognition rate according to the third embodiment. [Figure 15] FIG. 11 is a flow diagram illustrating an example of a schematic operation flow of a network management device according to a third embodiment. [Figure 16] FIG. 13 is a block diagram showing a configuration example of a video distribution system according to a fourth embodiment. [Figure 17] FIG. 13 is a block diagram showing a configuration example of a video distribution system according to a fifth embodiment. [Figure 18] FIG. 11 is a block diagram showing an example of a hardware configuration of a computer that realizes the network management device according to the second to fifth embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment of the present disclosure will be described with reference to the drawings. Note that the following description and drawings are omitted and simplified as appropriate for clarity of explanation. In addition, in each of the following drawings, the same elements are given the same reference numerals, and duplicated explanations are omitted as necessary. In addition, the specific numerical values ​​shown below are merely examples for facilitating understanding of the present disclosure, and are not limited thereto.

[0013] <Embodiment 1> First, with reference to FIG. 1, a configuration example of a video distribution system 3 according to the first embodiment will be described.

[0014] As shown in FIG. 1, the video distribution system 3 according to the first embodiment includes an acquisition unit 111, a first calculation unit 112, and a second calculation unit 113. The acquisition unit 111 acquires a required quality indicating the quality of a video that is required of a video distribution device that distributes the video in order to recognize an object from the video. The acquisition unit 111 corresponds to the quality storage unit 101 described later.

[0015] The first calculation unit 112 calculates the analysis performance of a video at which an object can be recognized, based on the required quality acquired by the acquisition unit 111. The first calculation unit 112 corresponds to a configuration combining a required number of frames within delay calculation unit 102, a recall to frame function calculation unit 103, and a required recall calculation unit 104, which will be described later.

[0016] The second calculation unit 113 calculates parameters related to video distribution based on the analysis performance calculated by the first calculation unit 112. The second calculation unit 113 corresponds to a configuration combining a recall vs. video bitrate function storage unit 105 and a required video bitrate calculation unit 106, which will be described later.

[0017] Next, an example of a schematic operation flow of the video distribution system 3 according to the first embodiment will be described with reference to FIG. As shown in FIG. 2, first, the acquisition unit 111 acquires a required quality indicating a quality required of a video in order to recognize an object from the video (step S301).

[0018] Next, the first calculation unit 112 calculates the analysis performance of the video at which the object can be recognized, based on the required quality acquired by the acquisition unit 111 (step S302). Thereafter, the second calculation unit 113 calculates parameters related to video distribution based on the analysis performance calculated by the first calculation unit 112 (step S303).

[0019] As described above, according to the first embodiment, the analysis performance of the video for recognizing the object is calculated based on the required quality, and the parameters related to the video distribution are calculated based on the calculated analysis performance. This makes it possible to realize efficient video distribution while maintaining the required quality (recognition accuracy).

[0020] The acquisition unit 111 may further acquire the video quality of the video. The first calculation unit 112 may calculate the analysis performance based on the required quality and the video quality acquired by the acquisition unit 111.

[0021] The acquisition unit 111 may acquire a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating a delay amount required for the video as the required quality, and may acquire a frame rate of the video as the video quality. The recognition rate may be a rate at which an object is recognized in one or more video frames among a plurality of video frames.

[0022] Furthermore, the first calculation unit 112 may calculate a number of frames within a requested delay, which indicates the number of video frames to be generated within a requested delay, based on the requested delay and frame rate acquired by the acquisition unit 111. Furthermore, the first calculation unit 112 may calculate a first function indicating a relationship between a recall and the number of video frames when the requested recognition rate is the requested recognition rate acquired by the acquisition unit 111. The first function corresponds to a recall vs. frame function, which will be described later. Furthermore, the first calculation unit 112 may calculate a required recall, which indicates a recall required per video frame, as an analysis performance, based on the number of frames within a requested delay calculated above and the first function. Furthermore, the second calculation unit 113 may calculate a required video bitrate, which indicates a video bitrate required for video, as a parameter, based on the required recall calculated by the first calculation unit 112 and a second function indicating a relationship between the recall and video bitrate. The second function corresponds to a recall vs. video bitrate function, which will be described later.

[0023] The second function may be calculated taking into consideration variations in the recognition rate. The acquisition unit 111 may also acquire a required quality according to the object. Furthermore, the video distribution system 3 may further include an encoder that encodes video based on the required video bit rate calculated by the second calculation unit 113, and a video distribution unit that distributes the video encoded by the encoder via a network. The encoder and video distribution unit correspond to the encoder 202 and the video distribution unit 203, which will be described later.

[0024] Furthermore, the video distribution system 3 may further include a guaranteed bandwidth setting unit that sets a guaranteed bandwidth for the network used for video distribution, based on the required video bit rate calculated by the second calculation unit 113. This guaranteed bandwidth setting unit corresponds to guaranteed bandwidth setting unit 402, which will be described later.

[0025] <Embodiment 2> Next, a configuration example of a network management device 100B according to the second embodiment will be described with reference to Fig. 3. The second embodiment corresponds to an embodiment in which the components that are provided separately in the first embodiment are integrated into one device (network management device 100B).

[0026] That is, as shown in FIG. 3, the network management device 100B according to the second embodiment has the same components as those of the first embodiment described above, that is, an acquisition unit 111, a first calculation unit 112, and a second calculation unit 113.

[0027] The second embodiment differs from the first embodiment only in that the same components as those in the first embodiment are integrated into the network management device 100B. Therefore, the operation and effects of the second embodiment are similar to those of the first embodiment, and therefore a description thereof will be omitted.

[0028] <Outline of each of the third to fifth embodiments> The following third to fifth embodiments are more specific examples of the first and second embodiments described above. Before describing each of the third to fifth embodiments, an overview of each of the third to fifth embodiments will be provided. First, with reference to FIG. 4, a configuration example of a video distribution system 9 assumed in each of the third to fifth embodiments will be described.

[0029] 4, the video distribution system 9 includes a network management device 10, a video distribution device 20, and an analysis device 30. The network connecting the network management device 10 and the video distribution device 20 may be a wired network or a wireless network. The network connecting the video distribution device 20 and the analysis device 30 is a wireless network. The wireless network may be Wifi (Wireless Fidelity), LTE (Long Term Evolution), 4G, 5G, local 5G, or the like.

[0030] The video distribution device 20 is a camera such as a surveillance camera, and distributes video captured by the camera. The analysis device 30 analyzes the video distributed by the video distribution device 20. For example, the analysis device 30 detects an object by recognizing an object (e.g., a person, a face, a car, etc.) reflected in the video. Specifically, the analysis device 30 performs a binary classification as to whether or not an object is reflected in the video. In this manner, the analysis device 30 is used for solution applications such as object detection.

[0031] The network management device 10 determines parameters related to video distribution by the video distribution device 20, and sets the determined parameters in the video distribution device 20. The parameters related to video distribution are, for example, a video bit rate and a frame rate. When the parameter related to video distribution is a video bit rate, the video distribution device 20 encodes and distributes the video based on the video bit rate set by the network management device 10. In the following description, the parameter related to video distribution is described as being a video bit rate.

[0032] Next, with reference to FIG. 5, an example of a method for setting a video bit rate in each of the third to fifth embodiments and the related art will be described. In the video analysis by analysis device 30, a video frame-by-frame analysis technique is generally used, and therefore it is the industry standard to evaluate recognition accuracy on a video frame-by-frame basis.

[0033] Therefore, assuming that the related technology is a combination of the technology described in the above-mentioned Patent Document 1 and the above-mentioned industry standard, the related technology is considered to set the video bit rate so as to satisfy the required recognition accuracy using a recognition accuracy vs. video bit rate function for each video frame. That is, in the related technology, as shown in the upper part of FIG. 5, the analysis device 30 is considered to set the video bit rate so as to guarantee the recognition accuracy of the object for each video frame. Note that in the upper part of FIG. 5, the bar-shaped diagonal line elements indicate that the recognition accuracy is guaranteed for each video frame. That is, the network management device 10 is considered to set the video bit rate assuming that the analysis device 30 performs frame-wise recognition.

[0034] However, in order to guarantee the recognition accuracy for each video frame, it is necessary to distribute the video at a high video bit rate, which makes it difficult to secure a sufficient communication bandwidth for video distribution.

[0035] Therefore, in each of the third to fifth embodiments, as shown in the lower part of FIG. 5, the analysis device 30 sets the video bit rate so that the recognition accuracy of the object is guaranteed in one or more of N (N is an integer equal to or greater than 2) video frames, rather than in every video frame. That is, the network management device 10 sets the video bit rate on the assumption that the analysis device 30 executes Family-wise recognition. Note that FIG. 5 shows an example of N=4, that is, an example of guaranteeing that the object can be recognized in one or more of four video frames. Therefore, in the lower part of FIG. 5, the bar-shaped diagonal line elements indicate that the recognition accuracy is guaranteed in one or more of four video frames.

[0036] In this way, in each of the third to fifth embodiments, although the probability of recognizing an object in one video frame is low, the video bit rate is set so as to ensure that an object can be recognized in at least one video frame out of N video frames. This allows the video bit rate to be lower than in related techniques, making it easier to secure the communication bandwidth required for video distribution.

[0037] Next, with reference to FIG. 6, the Family-wise recognition performed in each of the third to fifth embodiments will be described from the viewpoint of a solution. Fig. 6 shows an example of video frames used for a solution application that generates an alert when a suspicious person is detected in the analysis device 30. Six video frames are shown in Fig. 6. In Fig. 6, the video frames are newer as they move downward, and when a person is detected to be captured, the person is surrounded by a frame.

[0038] In the example of Figure 6, a person is detected in the third video frame, but the person cannot be detected in the following fourth and fifth video frames, resulting in missed detections. However, the person is detected again in the sixth video frame.

[0039] In this way, even if a detection is missed, if a similar context (a person caught in the image in the example of Figure 6) continues for a certain period of time, an alert can be issued if a similar context is detected in even just one of the following several video frames.

[0040] However, contexts that appear in only one video frame are not assumed because they do not appear continuously for a certain period of time. For example, if the video frame rate is 24 fps, contexts that appear for only a few tens of ms are not assumed.

[0041] Next, with reference to FIG. 7, the Family-wise recognition performed in each of the third to fifth embodiments will be described from a technical viewpoint. FIG. 7 is a graph showing the relationship between the video bit rate and the accuracy score when the analysis device 30 detects an object. In FIG. 7, the horizontal axis shows the video bit rate, and the vertical axis shows the accuracy scores of Precision, Accuracy, and Recall. Precision indicates the rate at which the detected object is correctly detected, and is also called the matching rate. Accuracy indicates how accurately the object is detected, and is also called the accuracy rate. Recall indicates the rate at which the object is not missed in detection, and is also called the recall rate. Precision, Accuracy, and Recall are examples of the analysis performance of a video that can recognize an object.

[0042] As shown in Figure 7, the accuracy score of Precision is independent of the video bitrate. On the other hand, the recall accuracy score depends on the video bitrate, but if the video bitrate is about 150 kbps, an accuracy score of about 0.9 or higher can be maintained. This shows that even if the video bitrate is low and the target detection is missed, as long as the target is detected in another video frame, the probability that the detection is correct is about 90% or higher.

[0043] In each of the following embodiments 3 to 5, a required video bit rate for video distribution is calculated as a parameter related to video distribution. Here, in calculating the required video bit rate, it is important to reduce the amount of missed detection. Therefore, in each of the following embodiments 3 to 5, Recall is used to calculate the required video bit rate.

[0044] <Embodiment 3> The third embodiment will now be described. First, a configuration example of the network management device 100 according to the third embodiment will be described with reference to Fig. 8. The network management device 100 according to the third embodiment is, for example, a cloud server, a Multi-access Edge Computing (MEC) server, an edge device, or the like, and the installation location is not limited.

[0045] As shown in FIG. 8, network management device 100 according to the third embodiment includes quality storage unit 101, number of frames within requested delay calculation unit 102, Recall vs. frame function calculation unit 103, required Recall calculation unit 104, Recall vs. video bitrate function storage unit 105, and required video bitrate calculation unit 106. Each component included in network management device 100 may be mounted on a different device. For example, quality storage unit 101 and Recall vs. video bitrate function storage unit 105 may be in the cloud. Each function calculation unit (number of frames within requested delay calculation unit 102, Recall vs. frame function calculation unit 103, and required Recall calculation unit 104) and required video bitrate calculation unit 106 may be mounted on a different device and communicate with each other via wired or wireless communication to acquire information.

[0046] The quality storage unit 101 prestores a required quality (requested recognition quality) indicating the quality of video (recognition quality) that an analysis device (e.g., analysis device 30 in FIG. 4) requests of a video delivery device (e.g., video delivery device 20 in FIG. 4) that delivers the video in order to recognize an object from the video. Specifically, the quality storage unit 101 prestores a required delay amount d[s] and a required recognition rate p as the required quality. The required delay amount d[s] indicates the upper limit of the delay amount of video that the analysis device requests of the video delivery device, and the requested recognition rate p indicates the recognition rate of video that the analysis device requests of the video delivery device.

[0047] Here, the definition of the recognition rate will be explained with reference to Fig. 9. Fig. 9 shows an example in which the object to be detected is a face. In a video with a changed video bit rate, starting from a video frame in which a face was detected in the original video, a face is detected in the following N video frames, and if a face can be detected at least once in the N video frames, the detection is deemed successful. Then, statistics on successful and unsuccessful detections are collected, and the success rate is defined as the recognition rate.

[0048] Furthermore, the quality storage unit 101 stores in advance video quality indicating the actual quality of a video delivered by a video delivery device (for example, video delivery device 20 in FIG. 4) and received by the analysis device. Specifically, the quality storage unit 101 stores in advance a frame rate m [fps] as the video quality. Other examples of video quality include bit rate and resolution (quantization parameter (QP value)).

[0049] The quality storage unit 101 may obtain the requested delay amount d[s] and the requested recognition rate p from the analysis device, and may obtain the frame rate m[fps] from the video distribution device or the analysis device. Alternatively, the quality storage unit 101 may obtain the requested delay amount d[s], the requested recognition rate p, and the frame rate m[fps] from input by the user.

[0050] The number of frames within the required delay calculation unit 102 calculates the number of frames within the required delay n , which indicates the number of video frames to be generated within the required delay d [s], based on the required delay d [s] and the frame rate m [fps] stored in the quality storage unit 101. max Calculate n max is calculated by the following Equation 1, where the required delay amount is d [s] and the frame rate is m [fps].

number

[0051] The recall versus frame function calculation unit 103 calculates a recall versus frame function that indicates the relationship between Recall and the number of video frames when the required recognition rate is p, based on the required recognition rate p stored in the quality storage unit 101. The recall versus frame function is a function that indicates the number of video frames required to recognize an object when the required recognition rate is p and Recall is q. When the required recognition rate is p and the recall is q, the probability of recognizing the object within the Nth image frame is expressed by the following Equation 2.

number

number

[0052] When the required recognition rate is p and Recall is q, Recall versus frame function calculation unit 103 calculates the above-mentioned formula 3 as the Recall versus frame function. An example of a graph of the above-mentioned formula 3 when the required recognition rate p is 0.95 is shown in Fig. 10. In Fig. 10, the horizontal axis indicates Recall and the vertical axis indicates the number of video frames.

[0053] The required recall calculation unit 104 calculates the required number of frames within the delay n calculated by the required number of frames within the delay calculation unit 102. max Based on the frame-to-frame function calculated by the frame-to-frame function calculation unit 103, a required frame recall q is calculated, which indicates the required frame recall per image frame in order to recognize an object by the Nth image frame. min Calculate.

[0054] For example, when the required delay amount d is 1 [s] and the frame rate m is 24 [fps], the number of frames within the required delay calculated by the required delay frame number calculation unit 102 is n max is 24 according to the formula 1. When the required recognition rate is 0.95, the graph of the recall vs. frame function calculated by the recall vs. frame function calculation unit 103 is as shown in FIG. 10. Therefore, the required recall calculation unit 104 calculates n max The Recall value of 0.12 corresponds to =24, and the required Recall q min It is calculated as follows.

[0055] In addition, when frame-wise recognition is assumed as in the related technology, it is necessary to guarantee the recognition accuracy of the object for each video frame, so N is the minimum n min The value is 1. In the graph of Figure 10, n min The recall value corresponding to =1 is 0.95, which is the required recall.

[0056] Recall vs. video bitrate function storage unit 105 stores a Recall vs. video bitrate function that indicates the relationship between Recall and video bitrate. FIG. 11 shows an example of an approximation curve that approximates the Recall vs. video bitrate function. In FIG. 11, the horizontal axis indicates video bitrate, and the vertical axis indicates Recall. FIG. 11 also shows two approximation curves, with the dotted approximation curve indicating Recall (μ) for the entire video, and the solid approximation curve indicating μ-σ (standard deviation). The μ-σ approximation curve was derived in consideration of variations in recognition rate, and as described below, required video bitrate calculation unit 106 calculates the required video bitrate using the μ-σ approximation curve.

[0057] Here, a method for creating the two approximation curves shown in Fig. 11 will be described with reference to Fig. 12 and Fig. 13. Fig. 12 and Fig. 13 show an example in which the object to be detected is a face. Note that here, for the sake of convenience, an operating entity that creates the approximation curves is referred to as a creation unit (not shown). The creation unit may be provided in the network management device 10, or may be provided in a separate device.

[0058] (1) Collecting video bitrate vs. recall data First, as shown in FIG. 12, the creation unit receives delivery of an original video (step S101), performs face detection (step S102), stores the detection result (step S103), and calculates Recall (step S107). Next, the creation unit receives delivery of a video with a reduced video bit rate (step S104), performs face detection (step S105), stores the detection result (step S106), and calculates Recall (step S107). As long as the video bit rate does not fall below the threshold, the creation unit repeats the loop of steps S104 to S107. Thereafter, the creation unit accumulates video bit rate vs. Recall data indicating the relationship between the video bit rate and Recall (step S108).

[0059] (2) Approximation curve calculation Next, the creation unit performs fitting to the following Equation 4 using the video bit rate vs. Recall data, and creates f1(x) to f4(x) as shown in FIG.

number

[0060] Specifically, the creation unit first performs fitting to the above-mentioned Equation 4 to create f1(x) and f2(x). Next, the creation unit finds the standard deviation σ of f1(x) and creates f3(x) by f1(x)-σ. Next, the creation unit connects a point in f2(x) that corresponds to the bit rate of the original video and a point in f3(x) that corresponds to the x-coordinate of the intersection of f1(x) and f2(x), and sets this as f4(x). Therefore, a1=a3. Also, the x-coordinate of the intersection of f1(x) and f2(x) is the same as the intersection of f3(x) and f4(x).

[0061] In this manner, f1(x) to f4(x) are created as shown in Fig. 13. Of these, the approximation curve consisting of f1(x) and f2(x) is the approximation curve corresponding to μ shown in Fig. 11. Moreover, the approximation curve consisting of f3(x) and f4(x) is the approximation curve corresponding to μ-σ shown in Fig. 11.

[0062] The required video bit rate calculation unit 106 calculates the required Recall q min Based on the Recall vs. video bit rate function stored in the Recall vs. video bit rate function storage unit 105, a required video bit rate β indicating a video bit rate required for recognizing an object by the Nth video frame is calculated. min Calculate.

[0063] For example, Recall q min= 0.12, the approximate curve of the recall vs. video bit rate function is as shown in Figure 11. In the μ-σ approximate curve of Figure 11, q min The video bit rate corresponding to β = 0.12 is 73 [kbps]. Therefore, the required video bit rate calculation unit 106 calculates 73 [kbps] as the required video bit rate β min It is calculated as follows.

[0064] In addition, when frame-wise recognition is assumed as in the related art, N=1, so the required recall is 0.95 as described above. In the μ approximation curve in Figure 11, the video bit rate corresponding to q=0.95 is 425 [kbps]. Therefore, 425 [kbps] is the required video bit rate.

[0065] Next, an example of the result of the video bit rate versus the recognition rate (the recognition rate of Family-wise recognition) will be described with reference to Fig. 14. In Fig. 14, the horizontal axis indicates the video bit rate, and the vertical axis indicates the recognition rate.

[0066] When frame-wise recognition is assumed, as in the related technology, N = 1 and the required video bit rate is 425 [kbps]. In this case, the required recognition rate (0.95) or higher can be maintained, but the required video bit rate becomes higher.

[0067] In contrast, when family-wise recognition is assumed as in the third embodiment, N=24 and the required video bit rate is 73 kbps. In this case, although the required video bit rate is low, the required recognition rate (0.95) or higher can be maintained. From this, it is understood that according to the third embodiment, it is possible to maintain the required recognition rate (0.95) or higher while reducing the video bit rate (from 425 [kbps] to 73 [kbps]).

[0068] Next, an example of a schematic operation flow of the network management device 100 according to the third embodiment will be described with reference to FIG. As shown in FIG. 15, first, the quality storage unit 101 stores a requested delay amount, a requested recognition rate, and a frame rate (step S201).

[0069] Next, the number of frames within requested delay calculation unit 102 calculates the number of frames within requested delay based on the requested delay amount and the frame rate stored in the quality storage unit 101 (step S202).

[0070] Next, the Recall vs. frame function calculation unit 103 calculates a Recall vs. frame function for the requested recognition rate based on the requested recognition rate stored in the quality storage unit 101 (step S203). The processes of steps S202 and S203 may be performed in the reverse order, or may be performed simultaneously in parallel.

[0071] Next, the required Recall calculation unit 104 calculates the required Recall per video frame based on the number of frames within requested delay calculated by the number of frames within requested delay calculation unit 102 and the Recall versus frame function calculated by the Recall versus frame function calculation unit 103 (step S204).

[0072] Thereafter, required video bitrate calculation unit 106 calculates the required video bitrate based on the required Recall calculated by required Recall calculation unit 104 and the Recall vs. video bitrate function stored in Recall vs. video bitrate function storage unit 105 (step S205).

[0073] As described above, according to the third embodiment, the required video bit rate is calculated assuming family-wise recognition. That is, the required recall per video frame is calculated so that the recognition rate, which indicates the proportion of objects recognized in one or more video frames out of N video frames, satisfies the required recognition rate, and the required video bit rate is calculated based on the calculated required recall.

[0074] This makes it possible to maintain the required recognition rate (recognition accuracy) while keeping the video bit rate low compared to the case of frame-wise recognition, thereby enabling efficient video distribution. Also, since the video bit rate can be kept low, the communication bandwidth required for video distribution can be reduced.

[0075] <Fourth embodiment> Next, a configuration example of the video distribution system 1 according to the fourth embodiment will be described with reference to FIG. 16, the video distribution system 1 according to the fourth embodiment includes a network management device 100A, a video distribution device 200, and an analysis device 300. The wireless network connecting the video distribution device 200 and the analysis device 300 may be Wifi, LTE, 4G, 5G, local 5G, or the like (the same applies to the following FIG. 17).

[0076] Network management device 100A differs from network management device 100 according to the above-described first embodiment in that required video bit rate transmission unit 107 is added. Required video bit rate transmission unit 107 transmits information on the required video bit rate calculated by required video bit rate calculation unit 106 to video distribution device 200 .

[0077] The video distribution device 200 is a device that distributes video, such as a camera such as a security camera. The video distribution device 200 includes a required video bit rate receiving unit 201, an encoder 202, a video distribution unit 203, and a shooting unit 204. Required video bit rate receiver 201 receives information on the required video bit rate from network management device 100A. The image capturing unit 204 captures video. The encoder 202 encodes the video captured by the image capture unit 204 based on the required video bit rate received by the required video bit rate receiving unit 201 . The video distributor 203 transmits the video encoded by the encoder 202 to the analysis device 300 via the network.

[0078] Analysis device 300 is a device that analyzes video, and is a cloud server, etc. Analysis device 300 includes a video receiving unit 301, a decoder 302, and a video analysis unit 303. Video receiving unit 301 receives video from video distribution device 200 via a network. The decoder 302 decodes the video received by the video receiving unit 301 . The video analysis unit 303 analyzes the video decoded by the decoder 302. For example, the video analysis unit 303 detects an object by recognizing an object (e.g., a person, a face, a car, etc.) reflected in the video. Specifically, the video analysis unit 303 performs a binary classification as to whether or not an object is reflected in the video.

[0079] As described above, according to the fourth embodiment, video distribution device 200 can encode the video captured by imaging unit 204 based on the required video bit rate calculated by network management device 100A, and distribute the encoded video to analysis device 300. Furthermore, the required video bit rate calculated by network management device 100A is calculated assuming Family-wise recognition and so as to satisfy a required recognition rate, so that the detection accuracy of the object in analysis device 300 can be maintained.

[0080] <Fifth Preferred Embodiment> Next, a configuration example of the video distribution system 2 according to the fifth embodiment will be described with reference to FIG. As shown in FIG. 17, a video distribution system 2 according to the fifth embodiment includes a network management device 100A and a bandwidth control device 400.

[0081] The network management device 100A has the same configuration as the network management device 100A according to the above-mentioned embodiment 4. However, the destination of the information on the required video bit rate is the bandwidth control device 400.

[0082] Bandwidth control device 400 is a device that controls the bandwidth of a network, and is, for example, a router disposed between video distribution device 200 and analysis device 300. However, bandwidth control device 400 may also be, for example, a router disposed between video distribution device 200 and a video receiving device (not shown). Bandwidth control device 400 includes required video bitrate receiving unit 401 and guaranteed bandwidth setting unit 402. Required video bit rate receiver 401 receives information on the required video bit rate from network management device 100A. Based on the required video bit rate received by required video bit rate receiver 401 , guaranteed bandwidth setting unit 402 sets a guaranteed bandwidth for the network used for video distribution by video distribution device 200 .

[0083] As described above, according to the third embodiment, the bandwidth control device 400 can set a guaranteed bandwidth for the network used for video distribution by the video distribution device 200, based on the required video bit rate calculated by the network management device 100A.

[0084] <Hardware configuration of the network management device according to the embodiment> Next, with reference to FIG. 18, an example of a hardware configuration of a computer 900 that realizes the network management devices 100, 100A, and 100B according to the above-mentioned second to fifth embodiments will be described.

[0085] 18, a computer 900 includes a processor 901, a memory 902, a storage 903, an input / output interface (input / output I / F) 904, and a communication interface (communication I / F) 905. The processor 901, the memory 902, the storage 903, the input / output interface 904, and the communication interface 905 are connected by a data transmission path for transmitting and receiving data to and from each other.

[0086] The processor 901 is, for example, an arithmetic processing device such as a central processing unit (CPU) or a graphics processing unit (GPU). The memory 902 is, for example, a random access memory (RAM) or a read only memory (ROM). The storage 903 is, for example, a storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a memory card. The storage 903 may also be a memory such as a RAM or a ROM.

[0087] The storage 903 stores programs that realize the functions of the components of the network management devices 100, 100A, and 100B. The processor 901 executes each of these programs to realize the functions of the components of the network management devices 100, 100A, and 100B. When executing each of the above programs, the processor 901 may read these programs onto the memory 902 before executing them, or may execute them without reading them onto the memory 902. The memory 902 and storage 903 also fulfill the role of realizing the storage functions of the network management devices 100, 100A, and 100B.

[0088] Moreover, the above-mentioned program includes a set of instructions (or software code) for making the computer perform one or more functions of the network management devices 100, 100A, and 100B described in the above-mentioned embodiments when the program is loaded into the computer. The program may be stored in a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, the computer-readable medium or the tangible storage medium includes RAM, ROM, flash memory, SSD or other memory technology, compact disc (CD)-ROM, digital versatile disk (DVD), Blu-ray (registered trademark) disk or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may be transmitted on a transitory computer-readable medium or a communication medium. By way of example and not limitation, the transitory computer-readable medium or the communication medium includes electrical, optical, acoustic, or other forms of propagating signals.

[0089] The input / output interface 904 is connected to a display device 9041, an input device 9042, a sound output device 9043, and the like. The display device 9041 is a device that displays a screen corresponding to drawing data processed by the processor 901, such as an LCD (Liquid Crystal Display), a CRT (Cathode Ray Tube) display, or a monitor. The input device 9042 is a device that accepts an operation input by an operator, and is, for example, a keyboard, a mouse, or a touch sensor. The display device 9041 and the input device 9042 may be integrated and realized as a touch panel. The sound output device 9043 is a device that acoustically outputs a sound corresponding to the sound data processed by the processor 901, such as a speaker.

[0090] The communication interface 905 transmits and receives data to and from an external device. For example, the communication interface 905 communicates with the external device via a wired communication path or a wireless communication path.

[0091] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure.

[0092] Furthermore, some or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes. (Appendix 1) an acquisition unit that acquires a required quality indicating a quality of the video that is required of a video distribution device that distributes the video in order to recognize an object from the video; a first calculation unit that calculates an analysis performance of the image capable of recognizing the object based on the required quality acquired by the acquisition unit; A second calculation unit that calculates a parameter related to the distribution of the video based on the analysis performance calculated by the first calculation unit. Network management device. (Appendix 2) The acquisition unit further acquires a video quality of the video, The first calculation unit calculates the analysis performance based on the required quality and the video quality acquired by the acquisition unit. 2. A network management device as described in claim 1. (Appendix 3) the acquisition unit acquires, as the required quality, a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating an upper limit of a delay amount required for the video, and acquires, as the video quality, a frame rate of the video; The recognition rate is a ratio of the object being recognized in one or more image frames among a plurality of image frames. 3. A network management device as described in claim 2. (Appendix 4) The first calculation unit is calculating a number of frames within a requested delay, which indicates a number of video frames to be generated within the requested delay, based on the requested delay and the frame rate acquired by the acquisition unit; calculating a first function indicating a relationship between a recall rate and a frame number of video frames when the requested recognition rate is the requested recognition rate acquired by the acquisition unit; calculating a required recall indicating a recall required per one video frame based on the calculated number of frames within the required delay and the first function as the analysis performance; The second calculation unit is calculating, as the parameter, a required video bit rate indicating a video bit rate required for the video based on the required recall calculated by the first calculation unit and a second function indicating a relationship between the recall and a video bit rate; 4. A network management device as described in appendix 3. (Appendix 5) The second function is calculated taking into consideration the variation in the recognition rate. 5. A network management device as described in appendix 4. (Appendix 6) The acquisition unit acquires the required quality according to the object. 6. A network management device according to any one of claims 1 to 5. (Appendix 7) an acquisition step of acquiring a required quality indicating a quality of the video required for a video distribution device that distributes the video in order to recognize an object from the video; a first calculation step of calculating an analytical performance of the image capable of recognizing the object based on the required quality acquired by the acquisition step; A second calculation step of calculating a parameter related to the distribution of the video based on the analysis performance calculated in the first calculation step. Network management methods. (Appendix 8) The acquiring step further includes acquiring an image quality of the image; In the first calculation step, the analysis performance is calculated based on the required quality and the video quality acquired in the acquisition step. 8. The network management method of claim 7. (Appendix 9) In the obtaining step, a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating an upper limit of a delay amount required for the video are obtained as the required quality, and a frame rate of the video is obtained as the video quality; The recognition rate is a ratio of the object being recognized in one or more image frames among a plurality of image frames. 9. A network management method as claimed in claim 8. (Appendix 10) In the first calculation step, calculating a number of frames within a required delay, which indicates a number of video frames to be generated within the required delay, based on the required delay and the frame rate acquired in the acquisition step; calculating a first function indicating a relationship between a recall rate and a number of video frames when the requested recognition rate is the requested recognition rate acquired in the acquiring step; calculating a required recall indicating a recall required per one video frame based on the calculated number of frames within the required delay and the first function as the analysis performance; In the second calculation step, calculating, as the parameter, a required video bit rate indicating a video bit rate required for the video based on the required recall calculated in the first calculation step and a second function indicating a relationship between the recall and a video bit rate; 10. The network management method of claim 9. (Appendix 11) The second function is calculated taking into consideration the variation in the recognition rate. 11. The network management method of claim 10. (Appendix 12) In the obtaining step, the required quality corresponding to the object is obtained. 12. A network management method according to any one of claims 7 to 11. (Appendix 13) an acquisition unit that acquires a required quality indicating a quality of the video that is required of a video distribution device that distributes the video in order to recognize an object from the video; a first calculation unit that calculates an analysis performance of the image capable of recognizing the object based on the required quality acquired by the acquisition unit; A second calculation unit that calculates a parameter related to the distribution of the video based on the analysis performance calculated by the first calculation unit. Video distribution system. (Appendix 14) The acquisition unit further acquires a video quality of the video, The first calculation unit calculates the analysis performance based on the required quality and the video quality acquired by the acquisition unit. 14. The video distribution system according to claim 13. (Appendix 15) the acquisition unit acquires, as the required quality, a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating an upper limit of a delay amount required for the video, and acquires, as the video quality, a frame rate of the video; The recognition rate is a ratio of the object being recognized in one or more image frames among a plurality of image frames. 15. A video distribution system as described in appended claim 14. (Appendix 16) The first calculation unit is calculating a number of frames within a requested delay, which indicates a number of video frames to be generated within the requested delay, based on the requested delay and the frame rate acquired by the acquisition unit; calculating a first function indicating a relationship between a recall rate and a frame number of video frames when the requested recognition rate is the requested recognition rate acquired by the acquisition unit; calculating a required recall indicating a recall required per one video frame as the analysis performance based on the calculated number of frames within the required delay and the first function; The second calculation unit is Calculating, as the parameter, a required video bit rate indicating a video bit rate required for the video based on the required recall calculated by the first calculation unit and a second function indicating a relationship between the recall and a video bit rate. 16. The video distribution system according to claim 15. (Appendix 17) The second function is calculated taking into consideration the variation in the recognition rate. 17. The video distribution system according to claim 16. (Appendix 18) an encoder that encodes the video based on the required video bitrate calculated by the second calculation unit; and a video distribution unit that distributes the video encoded by the encoder via a network. 18. A video distribution system according to claim 16 or 17. (Appendix 19) a guaranteed bandwidth setting unit that sets a guaranteed bandwidth for a network used for distributing the video based on the required video bit rate calculated by the second calculation unit; 18. A video distribution system according to claim 16 or 17. (Appendix 20) The acquisition unit acquires the required quality according to the object. 20. A video distribution system according to any one of appendix 13 to 19. [Explanation of symbols]

[0093] 1,2,3 Video distribution system 100,100A,100B Network Management Device 101 Quality storage section 102 Calculation unit for number of frames within required delay 103 Recall vs. frame function calculation section 104 Required Recall Calculation Unit 105 Recall vs. video bitrate function memory 106 Required video bit rate calculation unit 107 Required video bit rate transmitter 111 Acquisition Department 112 First Calculation Section 113 Second Calculation Section 200 Video distribution device 201 Required video bit rate receiver 202 Encoder 203 Video Distribution Department 204 Photography Department 300 Analyzer 301 Video Receiving Unit 302 Decoder 303 Video Analysis Department 400 Bandwidth Control Device 401 Required video bit rate receiver 402 Guaranteed Bandwidth Setting Unit 900 Computers 901 Processor 902 Memory 903 Storage 904 Input / Output Interface 9041 Display device 9042 Input Device 9043 Sound output device 905 Communication Interface

Claims

1. an acquisition unit that acquires a required quality indicating a quality of the video that is required of a video distribution device that distributes the video in order to recognize an object from the video; a first calculation unit that calculates an analysis performance of the video capable of recognizing the object based on the required quality acquired by the acquisition unit; A second calculation unit that calculates a parameter related to the distribution of the video based on the analysis performance calculated by the first calculation unit. Network management device.

2. The acquisition unit further acquires a video quality of the video, The first calculation unit calculates the analysis performance based on the required quality and the video quality acquired by the acquisition unit. The network management device according to claim 1 .

3. the acquisition unit acquires, as the required quality, a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating an upper limit of a delay amount required for the video, and acquires, as the video quality, a frame rate of the video; The recognition rate is a ratio of the object being recognized in one or more image frames among a plurality of image frames. The network management device according to claim 2.

4. The first calculation unit is calculating a number of frames within a requested delay, which indicates a number of video frames to be generated within the requested delay, based on the requested delay and the frame rate acquired by the acquisition unit; calculating a first function indicating a relationship between a recall rate and a frame number of video frames when the requested recognition rate is the requested recognition rate acquired by the acquisition unit; calculating, as the analysis performance, a required recall indicating a recall required per one video frame based on the calculated number of frames within the required delay and the first function; The second calculation unit is calculating, as the parameter, a required video bit rate indicating a video bit rate required for the video based on the required recall calculated by the first calculation unit and a second function indicating a relationship between the recall and a video bit rate; The network management device according to claim 3.

5. The second function is calculated taking into consideration the variation in the recognition rate.

5. The network management device according to claim 4.

6. The acquisition unit acquires the required quality according to the object. A network management device according to any one of claims 1 to 5.

7. an acquisition step of acquiring a required quality indicating a quality of the video required for a video distribution device that distributes the video in order to recognize an object from the video; a first calculation step of calculating an analytical performance of the image capable of recognizing the object based on the required quality acquired by the acquisition step; A second calculation step of calculating a parameter related to the distribution of the video based on the analysis performance calculated in the first calculation step. Network management methods.

8. The acquiring step further includes acquiring a video quality of the video; In the first calculation step, the analysis performance is calculated based on the required quality and the video quality acquired in the acquisition step. The network management method according to claim 7.

9. In the obtaining step, a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating an upper limit of a delay amount required for the video are obtained as the required quality, and a frame rate of the video is obtained as the video quality; The recognition rate is a ratio of the object being recognized in one or more image frames among a plurality of image frames. The network management method according to claim 8.

10. In the first calculation step, calculating a number of frames within a required delay, which indicates a number of video frames to be generated within the required delay, based on the required delay and the frame rate acquired in the acquisition step; calculating a first function indicating a relationship between a recall rate and a number of video frames when the requested recognition rate is the requested recognition rate acquired in the acquiring step; calculating, as the analysis performance, a required recall indicating a recall required per one video frame based on the calculated number of frames within the required delay and the first function; In the second calculation step, calculating, as the parameter, a required video bit rate indicating a video bit rate required for the video based on the required recall calculated in the first calculation step and a second function indicating a relationship between the recall and a video bit rate; The network management method according to claim 9.

11. The second function is calculated taking into consideration the variation in the recognition rate. The network management method according to claim 10.

12. In the obtaining step, the required quality corresponding to the object is obtained. A network management method according to any one of claims 7 to 11.

13. an acquisition unit that acquires a required quality indicating a quality of the video that is required of a video distribution device that distributes the video in order to recognize an object from the video; a first calculation unit that calculates an analysis performance of the video capable of recognizing the object based on the required quality acquired by the acquisition unit; A second calculation unit that calculates a parameter related to the distribution of the video based on the analysis performance calculated by the first calculation unit. Video distribution system.

14. The acquisition unit further acquires a video quality of the video, The first calculation unit calculates the analysis performance based on the required quality and the video quality acquired by the acquisition unit. The video distribution system according to claim 13.

15. the acquisition unit acquires, as the required quality, a required recognition rate indicating a recognition rate required for the video and a required delay amount indicating an upper limit of a delay amount required for the video, and acquires, as the video quality, a frame rate of the video; The recognition rate is a ratio of the object being recognized in one or more image frames among a plurality of image frames. The video distribution system according to claim 14.

16. The first calculation unit is calculating a number of frames within a requested delay, which indicates a number of video frames to be generated within the requested delay, based on the requested delay and the frame rate acquired by the acquisition unit; calculating a first function indicating a relationship between a recall rate and a frame number of video frames when the requested recognition rate is the requested recognition rate acquired by the acquisition unit; calculating, as the analysis performance, a required recall indicating a recall required per one video frame based on the calculated number of frames within the required delay and the first function; The second calculation unit is calculating, as the parameter, a required video bit rate indicating a video bit rate required for the video based on the required recall calculated by the first calculation unit and a second function indicating a relationship between the recall and a video bit rate; The video distribution system according to claim 15.

17. The second function is calculated taking into consideration the variation in the recognition rate. The video distribution system according to claim 16.

18. an encoder that encodes the video based on the required video bit rate calculated by the second calculation unit; and a video distribution unit that distributes the video encoded by the encoder via a network.

18. A video distribution system according to claim 16 or 17.

19. a guaranteed bandwidth setting unit that sets a guaranteed bandwidth for a network used for distributing the video based on the required video bit rate calculated by the second calculation unit; 18. A video distribution system according to claim 16 or 17.

20. The acquisition unit acquires the required quality according to the object.

20. A video distribution system according to any one of claims 13 to 19.

Images