Statistical method and device for cross-domain events

Abstract

The application discloses a cross-domain event statistical method, which comprises the following steps: obtaining a target video, and counting cross-domain events according to the target video. In a specific example, the target video can be input into a cross-domain event determination model to obtain the cross-domain event occurrence probability of each video frame in the target video, and the number of cross-domain events included in the target video can be further determined according to the cross-domain event occurrence probability of each video frame. In the embodiment of the application, the cross-domain event model can accurately determine the cross-domain event occurrence probability of each video frame in the target video frame, and the cross-domain event statistical result obtained based on the cross-domain event occurrence probability of each video frame is also relatively accurate. Therefore, the cross-domain event can be accurately counted by using the scheme of the embodiment of the application.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a statistical method and apparatus for cross-domain events. Background Technology

[0002] In some scenarios, it's necessary to count the number of people entering a place. For example, this could involve counting the number of users entering a public place. Currently, for a particular location, a camera can be installed at the entrance to capture video footage. Furthermore, the video frames captured by the camera can be processed to determine the number of people entering the location.

[0003] However, current methods for processing video frames to determine the number of people entering a location are inaccurate. Therefore, a solution is urgently needed to address this problem. Summary of the Invention

[0004] The technical problem to be solved by this application is how to accurately count the number of people, and provides a method and apparatus for counting cross-domain events.

[0005] In a first aspect, embodiments of this application provide a method for statistical analysis of cross-domain events, the method comprising:

[0006] Acquire the target video;

[0007] The target video is input into the cross-domain event determination model to obtain the probability of cross-domain event occurrence for each video frame in the multiple video frames included in the target video.

[0008] The number of cross-domain events included in the target video is determined based on the occurrence probability of cross-domain events corresponding to each video frame.

[0009] Optionally, the number of cross-domain events included in the target video is determined based on the occurrence probability of cross-domain events corresponding to each video frame, including:

[0010] The number of cross-domain events included in the target video is determined by summing the occurrence probabilities of cross-domain events corresponding to each video frame.

[0011] Optionally, the cross-domain event determination model is trained in the following manner:

[0012] Obtain training videos and corresponding tags for the training videos. The training videos include multiple training video frames. The tags for the training videos include the tags of each training video frame among the multiple training video frames. The tags of each training video frame are used to indicate the probability of occurrence of cross-domain events corresponding to the training video frame.

[0013] The cross-domain event determination model is trained based on the training video and the corresponding tags of the training video.

[0014] Optionally, the labels corresponding to the training videos are determined in the following way:

[0015] Obtain at least one of the cross-domain events included in the training video;

[0016] Based on the training video frames involved in each of the at least one cross-domain events, the cross-domain events associated with each training video frame are obtained.

[0017] For each training video frame, the corresponding label for each training video frame is obtained based on the cross-domain event associated with that training video frame.

[0018] Optionally, the training video includes a first training video frame, and the label corresponding to the first training video frame is determined in the following way:

[0019] Determine the occurrence probability of at least one cross-domain event corresponding to the first training video frame, wherein each cross-domain event occurrence probability in the at least one cross-domain event occurrence probability corresponds to a cross-domain event associated with the first training video frame;

[0020] The label corresponding to the first training video frame is obtained based on the probability of occurrence of the at least one cross-domain event.

[0021] Optionally, the probability of at least one cross-domain event occurrence includes the probability of cross-domain event occurrence corresponding to the first cross-domain event, and the probability of cross-domain event occurrence is determined in the following manner:

[0022] Based on the information of the training video frames associated with the first cross-domain event and the preset probability distribution algorithm, the probability of occurrence of the cross-domain event corresponding to the first cross-domain event is obtained.

[0023] Optionally, the information of the training video frames associated with the first cross-domain event includes:

[0024] The information includes the first training video frame among the plurality of training video frames associated with the first cross-domain event, the information of the last training video frame among the plurality of training video frames associated with the first cross-domain event, and the information of the target video frame among the plurality of training video frames associated with the first cross-domain event, wherein the target video frame is the training video frame in which the first cross-domain event was determined to have occurred.

[0025] Optionally, the target video is a video captured by a camera device corresponding to the target location, and the cross-domain event includes: entering the target location, or leaving the target location.

[0026] Secondly, embodiments of this application provide a statistical apparatus for cross-domain events, the apparatus comprising:

[0027] The first acquisition unit is used to acquire the target video;

[0028] The first determining unit is used to input the target video into the cross-domain event determining model to obtain the probability of occurrence of cross-domain events corresponding to each video frame in the multiple video frames included in the target video.

[0029] The second determining unit is used to determine the number of cross-domain events included in the target video based on the occurrence probability of cross-domain events corresponding to each video frame.

[0030] Optionally, the second determining unit is used for:

[0031] The number of cross-domain events included in the target video is determined by summing the occurrence probabilities of cross-domain events corresponding to each video frame.

[0032] Optionally, the cross-domain event determination model is trained in the following manner:

[0033] Obtain training videos and corresponding tags for the training videos. The training videos include multiple training video frames. The tags for the training videos include the tags of each training video frame among the multiple training video frames. The tags of each training video frame are used to indicate the probability of occurrence of cross-domain events corresponding to the training video frame.

[0034] The cross-domain event determination model is trained based on the training video and the corresponding tags of the training video.

[0035] Optionally, the labels corresponding to the training videos are determined in the following way:

[0036] Obtain at least one of the cross-domain events included in the training video;

[0037] Based on the training video frames involved in each of the at least one cross-domain events, the cross-domain events associated with each training video frame are obtained.

[0038] For each training video frame, the corresponding label for each training video frame is obtained based on the cross-domain event associated with that training video frame.

[0039] Optionally, the training video includes a first training video frame, and the label corresponding to the first training video frame is determined in the following way:

[0040] Determine the occurrence probability of at least one cross-domain event corresponding to the first training video frame, wherein each cross-domain event occurrence probability in the at least one cross-domain event occurrence probability corresponds to a cross-domain event associated with the first training video frame;

[0041] The label corresponding to the first training video frame is obtained based on the probability of occurrence of the at least one cross-domain event.

[0042] Optionally, the probability of at least one cross-domain event occurrence includes the probability of cross-domain event occurrence corresponding to the first cross-domain event, and the probability of cross-domain event occurrence is determined in the following manner:

[0043] Based on the information of the training video frames associated with the first cross-domain event and the preset probability distribution algorithm, the probability of occurrence of the cross-domain event corresponding to the first cross-domain event is obtained.

[0044] Optionally, the information of the training video frames associated with the first cross-domain event includes:

[0045] The information includes the first training video frame among the plurality of training video frames associated with the first cross-domain event, the information of the last training video frame among the plurality of training video frames associated with the first cross-domain event, and the information of the target video frame among the plurality of training video frames associated with the first cross-domain event, wherein the target video frame is the training video frame in which the first cross-domain event was determined to have occurred.

[0046] Optionally, the target video is a video captured by a camera device corresponding to the target location, and the cross-domain event includes: entering the target location, or leaving the target location.

[0047] Thirdly, embodiments of this application provide a device, the device including: a processor, a memory, and a system bus; the device and the memory are connected via the system bus; the memory is used to store one or more programs, the one or more programs including instructions, the instructions causing the processor to perform the method described in any of the first aspects above when executed by the processor.

[0048] Fourthly, embodiments of this application provide a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform the method described in any of the first aspects above.

[0049] Compared with the prior art, the embodiments of this application have the following advantages:

[0050] This application provides a method for statistical analysis of cross-domain events. The method includes: acquiring a target video, which may include multiple video frames. After acquiring the target video, cross-domain events can be statistically analyzed based on the target video. In a specific example, the target video can be input into a cross-domain event determination model to obtain the probability of cross-domain events occurring for each video frame among the multiple video frames. Furthermore, based on the probability of cross-domain events occurring for each video frame, the number of cross-domain events included in the target video can be determined. Using the solution of this application, a cross-domain event determination model can be used to statistically analyze cross-domain events. The cross-domain event model can accurately determine the probability of cross-domain events occurring for each video frame of the target video frame. Consequently, the statistical results of cross-domain events obtained based on the probability of cross-domain events occurring for each video frame are also relatively accurate. Therefore, the solution of this application can accurately analyze cross-domain events. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of an exemplary application scenario in the embodiments of this application;

[0053] Figure 2 A flowchart illustrating a method for determining cross-domain events provided in an embodiment of this application;

[0054] Figure 3 A flowchart illustrating a training method for a cross-domain event determination model provided in an embodiment of this application;

[0055] Figure 4 A schematic diagram illustrating a video frame-associated cross-domain event provided in an embodiment of this application;

[0056] Figure 5 A schematic diagram illustrating a video frame-associated cross-domain event provided in an embodiment of this application;

[0057] Figure 6 This is a schematic diagram of a device for determining cross-domain events provided in an embodiment of this application. Detailed Implementation

[0058] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0059] The inventors of this application have discovered through research that current methods for processing video frames to determine the number of people entering a target location are inaccurate. Similarly, current methods for processing video frames to determine the number of people leaving a target location are also inaccurate. The methods for counting the number of people entering and leaving a target location are similar. The following section will use the counting of people entering a target location as an example to introduce a current method for counting the number of people entering a target location.

[0060] To make it easier to understand, let's first combine Figure 1 Explain the current population statistics scheme.

[0061] See Figure 1 , Figure 1 This is a schematic diagram illustrating an exemplary application scenario in an embodiment of this application. For example... Figure 1 As shown, when counting the number of people entering the target location, it is possible to combine... Figure 1 The regions 102 and 103 shown are located outside door 101 and inside door 101, respectively. Regions 102 and 103 do not overlap. Specifically, a Person Re-identification (REID) model can be used to process video footage captured by cameras installed near door 101 at the target location to obtain pedestrian trajectories. These trajectories are determined by the position of the pedestrian's feet in the video. When the trajectory moves from region 102 to region 103, an entry into the target location is considered an event. Correspondingly, when counting people, this event can be used as the basis for counting people. For example, if N events occur, the number of people entering the target location is determined to be N. The REID model will not be described in detail here.

[0062] However, in some scenarios, such as when a large number of people enter the target location at the same time, these users may block each other's view, meaning the camera cannot capture their feet. This results in discontinuous trajectories determined from the captured video. In such cases, inaccurate person counts may occur. (See reference...) Figure 1To understand this, suppose a pedestrian's trajectory is identified as two discontinuous trajectories, trajectory 110 and trajectory 120. Neither trajectory 110 nor trajectory 120 enters region 103 from region 102. Therefore, based on trajectories 110 and 120, it is determined that no event of entering the target location has occurred. In other words, the current solution's determination of the event of entering the target location is inaccurate. Consequently, when counting the number of people entering the target location based on the event of entering the target location, there will be omissions in the count, resulting in inaccurate statistical results.

[0063] To address the aforementioned issues, embodiments of this application provide a method for determining cross-domain events, a method for counting the number of people involved, and an apparatus.

[0064] The various non-limiting embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0065] Exemplary methods

[0066] See Figure 2 , Figure 2 This is a flowchart illustrating a method for determining cross-domain events provided in an embodiment of this application. In this embodiment, the method can be executed by a server or a terminal device; this application does not impose any specific limitations.

[0067] In one example, the method may include the following steps: S101-S103.

[0068] S101: Acquire the target video.

[0069] In one example, the target video may be a video captured by a camera device corresponding to the target location.

[0070] This application does not specifically limit the target location. The target location can be any location with the need for counting the number of people. In one example, the target location can be a shopping mall, park, cinema or other public place.

[0071] In one example, the capturing device may be, for example, a camera.

[0072] S102: Input the target video into the cross-domain event determination model to obtain the probability of cross-domain event occurrence for each video frame in the multiple video frames included in the target video.

[0073] In this embodiment, after obtaining the target video, a cross-domain event determination model can be used to process the target video, thereby statistically analyzing the cross-domain events corresponding to the target video. The cross-domain event determination model is pre-trained and can determine the probability of cross-domain events occurring for each video frame. Therefore, the target video can be input into the cross-domain event determination model, which can output the probability of cross-domain events occurring for each of the multiple video frames included in the target video. Regarding the multiple video frames, it should be noted that they can be some or all of the video frames included in the target video; this embodiment does not specifically limit this.

[0074] The cross-domain event mentioned in this application embodiment can be either entering or leaving the target location. When the cross-domain event is entering the target location, the solution in this application embodiment can be used to count the number of people entering the target location; when the cross-domain event is leaving the target location, the solution in this application embodiment can be used to count the number of people leaving the target location.

[0075] S103: Determine the number of cross-domain events included in the target video based on the occurrence probability of cross-domain events corresponding to each video frame.

[0076] After obtaining the occurrence probability of cross-domain events corresponding to each video frame, the number of cross-domain events included in the target video can be determined based on the occurrence probability of cross-domain events corresponding to each video frame among the multiple video frames included in the target video. The number of cross-domain events included in the target video mentioned here can be considered as the result of statistical analysis of the cross-domain events.

[0077] In one example, the probability of cross-domain events occurring for each video frame can be calculated using a specific algorithm to obtain the number of cross-domain events included in the target video. In a specific example, the probabilities of cross-domain events occurring for each video frame can be summed, and the number of cross-domain events included in the target video can be determined based on the summation result. For example, if the summation result is an integer, it can be determined as the number of cross-domain events included in the target video; conversely, if the summation result is not an integer, it can be rounded down or up to obtain the number of cross-domain events included in the target video.

[0078] As described above, in this embodiment, it is unnecessary to use user trajectories to statistically analyze cross-domain events, thus avoiding inaccurate statistical results due to discontinuous trajectories. Using the solution of this embodiment, a cross-domain event determination model can be used to statistically analyze cross-domain events. This model can accurately determine the probability of cross-domain events occurring in each video frame of the target video frame. Consequently, the statistical results of cross-domain events obtained based on the probability of cross-domain events occurring in each video frame are also relatively accurate. Therefore, the solution of this embodiment can accurately statistically analyze cross-domain events.

[0079] Next, the training method of the cross-domain event determination model will be introduced.

[0080] See Figure 3 The figure is a flowchart illustrating a training method for a cross-domain event determination model provided in an embodiment of this application. Figure 3 The training method shown may include the following S201-S202.

[0081] S201: Obtain the training video and the corresponding tags of the training video. The training video includes multiple training video frames. The tags of the training video include the tags of each training video frame in the multiple training video frames. The tags of each training video frame are used to indicate the probability of occurrence of the cross-domain event corresponding to the training video frame.

[0082] In one example, the training video may be a video captured by a camera device corresponding to the aforementioned target location. The training video may include multiple training video frames, and the tags for the training video include tags corresponding to each training video frame. For a given training video frame, the tag corresponding to that training video frame can be used to indicate the probability of the cross-domain event occurring for that training video frame.

[0083] In one example, the labels of the training videos can be manually labeled.

[0084] In yet another example, the labels of the training video can be obtained through the following steps A1-A3.

[0085] Step A1: Obtain at least one of the cross-domain events included in the training video.

[0086] In one example, at least one cross-domain event included in the training video can be identified through manual annotation. For instance, the training video can be played, and an annotator can watch the training video and annotate the cross-domain events in the training video. For a certain cross-domain event, such as a first cross-domain event, when annotating the first cross-domain event, information of the training video frames associated with the first cross-domain event can be annotated.

[0087] In one example, the information of the training video frames associated with the first cross-domain event may include information of the first training video frame among the plurality of training video frames associated with the first cross-domain event, and information of the last training video frame among the plurality of training video frames associated with the first cross-domain event. Wherein:

[0088] As an example, the training video frames in the training video can be numbered in chronological order from earliest to latest. The information of the first training video frame can be its number. Similarly, the information of the last training video frame can be its number.

[0089] As another example, the information of the first training video frame could also be the time corresponding to the first training video frame, where the time corresponding to the training video frame could be the time when the training video frame was captured. Similarly, the information of the last training video frame could also be the time corresponding to the last video frame.

[0090] In another example, the information of the training video frames associated with the first cross-domain event may further include information about a target video frame among the multiple training video frames associated with the first cross-domain event, wherein the target video frame is the training video frame that determines the occurrence of the first cross-domain event. In a specific example, if the cross-domain event is an event of entering a target location, then the target video frame may be the training video frame that determines the user's entry into the target location. The information of the target video frame may be the target video frame number or the time corresponding to the target video frame.

[0091] Step A2: Based on the training video frames involved in each cross-domain event in the at least one cross-domain event, obtain the cross-domain events associated with each training video frame.

[0092] After determining the at least one cross-domain event, the cross-domain events associated with each video frame can be obtained based on the training video frames involved in each of the at least one cross-domain event. For example, if the training video includes 23 video frames, and these 23 video frames correspond to two cross-domain events, the first cross-domain event involves video frames 1 to 10 of these 23 video frames, and the second cross-domain event involves video frames 4 to 23 of these 23 video frames. Then: (Refer to...) Figure 4 To understand, Figure 4 This is a schematic diagram illustrating a video frame-associated cross-domain event provided in an embodiment of this application. For example... Figure 4 As shown, the first video frame to the fourth video frame are associated with the first cross-domain event, the fourth video frame to the tenth video frame are associated with the two cross-domain events, and the eleventh video frame to the twenty-third video frame are associated with the second cross-domain event.

[0093] Step A3: For each training video frame, obtain the label corresponding to each training video frame based on the cross-domain event associated with that training video frame.

[0094] After determining the cross-domain events associated with each training video frame, for each training video frame, the label corresponding to that training video frame can be determined based on the cross-domain events associated with that training video frame.

[0095] Next, taking the first training video frame as an example, we will introduce the specific implementation method of step A3.

[0096] In one example, step A3 can be implemented by determining the label corresponding to the first training video frame based on the number of cross-domain events associated with the first training video frame.

[0097] In yet another example, step A3 can be achieved through steps A31-A32 as follows.

[0098] Step A31: Determine the probability of occurrence of at least one cross-domain event corresponding to the first training video frame, wherein each cross-domain event probability corresponds to a cross-domain event associated with the first training video frame.

[0099] In one example, if the first training video frame corresponds to M cross-domain events, then the occurrence probabilities of the M cross-domain events corresponding to the first training video frame can be determined, and these M occurrence probabilities correspond one-to-one with the M cross-domain events. Combined with... Figure 4To understand this, for the fourth training video frame, which is associated with two cross-domain events, we can determine the occurrence probabilities of two cross-domain events for the fourth training video frame, namely the occurrence probability of the cross-domain event corresponding to the first cross-domain event and the occurrence probability of the cross-domain event corresponding to the second cross-domain event.

[0100] In one example, the M cross-domain events may include a first cross-domain event. The probability of occurrence of the cross-domain event corresponding to the first cross-domain event can be determined based on the number of training video frames associated with the first cross-domain event and the position information of the first training video frames. The position information of the first training video frames may be, for example, the position of the first training video frame among the multiple training video frames associated with the first cross-domain event.

[0101] In another example, the probability of the occurrence of the cross-domain event corresponding to the first cross-domain event can be determined based on the information of the training video frames associated with the first cross-domain event and a preset probability distribution algorithm.

[0102] Information regarding the training video frames associated with the first cross-domain event can be found in the relevant description above, and will not be repeated here.

[0103] Regarding the preset probability distribution algorithm, it can be a uniform distribution algorithm or a Gaussian distribution algorithm, and this application embodiment does not make specific limitations.

[0104] When the preset probability distribution algorithm is a uniform distribution algorithm, the information of the training video frame associated with the first cross-domain event can be the information of the first training video frame and the information of the last training video frame. In this case, assuming the time corresponding to the first training video frame is t1 and the time corresponding to the last training video frame is t3, then for the first training video frame, the probability of the cross-domain event occurring with the first cross-domain event is...

[0105] When the preset probability distribution algorithm is a Gaussian distribution algorithm, the information of the training video frame associated with the first cross-domain event can be the information of the first training video frame, the information of the last training video frame, and the information of the target video frame. In this case, assuming the time corresponding to the first training video frame is t1, the time corresponding to the last training video frame is t3, and the time corresponding to the target video frame is t2, then the mean μ of the Gaussian distribution is μ = t2, and the variance σ = min(t2-t1, t3-t2) / 1.96. For the first training video frame, the probability of the cross-domain event corresponding to the first cross-domain event occurring is...

[0106] Step A32: Obtain the label corresponding to the first training video frame based on the occurrence probability of the at least one cross-domain event.

[0107] In one example, the probability of occurrence of the at least one cross-domain event can be summed to obtain the label corresponding to the first training video frame.

[0108] Regarding the tags corresponding to each video frame, we will now take a uniform distribution algorithm as an example, based on the preset probability distribution algorithm. Figure 5 Let's illustrate with examples. Figure 5 This is a schematic diagram illustrating a video frame-associated cross-domain event provided in an embodiment of this application. For example... Figure 5 As shown, the first to fourth video frames are associated with the first cross-domain event, the fourth to tenth video frames are associated with these two cross-domain events, and the eleventh to twenty-third video frames are associated with the second cross-domain event. Therefore:

[0109] For the first to the tenth video frames, the probability of the cross-domain event corresponding to the first cross-domain event is 0.1.

[0110] For video frames 4 through 23, the probability of the cross-domain event corresponding to the second cross-domain event is 0.05.

[0111] Therefore, for the first to third video frames, the probability of a cross-domain event is 0.1; for the fourth to tenth video frames, the probability is 0.15; and for the eleventh to twenty-third video frames, the probability is 0.05.

[0112] S202: Train the cross-domain event determination model based on the training video and the corresponding tags of the training video.

[0113] The cross-domain event determination model is trained by obtaining the training video and the labels corresponding to the training video frames. This application does not specifically limit the structure of the cross-domain event determination model. In one example, the cross-domain event determination model may include a feature extraction module and a probability prediction module. The feature extraction module is used to extract features from the training video frames and process the features of the training video frames to obtain the probability of occurrence of the cross-domain event corresponding to the training video frames. In a specific example, the feature extraction module may employ a Convolutional Neural Network (CNN), and the probability prediction module may employ a Recurrent Neural Network (RNN).

[0114] Exemplary device

[0115] Based on the methods provided in the above embodiments, this application also provides an apparatus, which will be described below with reference to the accompanying drawings.

[0116] See Figure 6 The figure is a schematic diagram of the structure of a cross-domain event statistics device provided in an embodiment of this application. The device 600 includes: a first acquisition unit 601, a first determination unit 602, and a second determination unit 603.

[0117] The first acquisition unit 601 is used to acquire the target video;

[0118] The first determining unit 602 is used to input the target video into the cross-domain event determining model to obtain the probability of occurrence of cross-domain events corresponding to each video frame in the multiple video frames included in the target video.

[0119] The second determining unit 603 is used to determine the number of cross-domain events included in the target video based on the occurrence probability of cross-domain events corresponding to each video frame.

[0120] Optionally, the second determining unit 603 is used for:

[0121] The number of cross-domain events included in the target video is determined by summing the occurrence probabilities of cross-domain events corresponding to each video frame.

[0122] Optionally, the cross-domain event determination model is trained in the following manner:

[0123] Obtain training videos and corresponding tags for the training videos. The training videos include multiple training video frames. The tags for the training videos include the tags of each training video frame among the multiple training video frames. The tags of each training video frame are used to indicate the probability of occurrence of cross-domain events corresponding to the training video frame.

[0124] The cross-domain event determination model is trained based on the training video and the corresponding tags of the training video.

[0125] Optionally, the labels corresponding to the training videos are determined in the following way:

[0126] Obtain at least one of the cross-domain events included in the training video;

[0127] Based on the training video frames involved in each of the at least one cross-domain events, the cross-domain events associated with each training video frame are obtained.

[0128] For each training video frame, the corresponding label for each training video frame is obtained based on the cross-domain event associated with that training video frame.

[0129] Optionally, the training video includes a first training video frame, and the label corresponding to the first training video frame is determined in the following way:

[0130] Determine the occurrence probability of at least one cross-domain event corresponding to the first training video frame, wherein each cross-domain event occurrence probability in the at least one cross-domain event occurrence probability corresponds to a cross-domain event associated with the first training video frame;

[0131] The label corresponding to the first training video frame is obtained based on the probability of occurrence of the at least one cross-domain event.

[0132] Optionally, the probability of at least one cross-domain event occurrence includes the probability of cross-domain event occurrence corresponding to the first cross-domain event, and the probability of cross-domain event occurrence is determined in the following manner:

[0133] Based on the information of the training video frames associated with the first cross-domain event and the preset probability distribution algorithm, the probability of occurrence of the cross-domain event corresponding to the first cross-domain event is obtained.

[0134] Optionally, the information of the training video frames associated with the first cross-domain event includes:

[0135] The information includes the first training video frame among the plurality of training video frames associated with the first cross-domain event, the information of the last training video frame among the plurality of training video frames associated with the first cross-domain event, and the information of the target video frame among the plurality of training video frames associated with the first cross-domain event, wherein the target video frame is the training video frame in which the first cross-domain event was determined to have occurred.

[0136] Optionally, the target video is a video captured by a camera device corresponding to the target location, and the cross-domain event includes: entering the target location, or leaving the target location.

[0137] Since the device 600 is a device corresponding to the method provided in the above method embodiments, the specific implementation of each unit of the device 600 is based on the same concept as the above method embodiments. Therefore, the specific implementation of each unit of the device 600 can be referred to the description section of the above method embodiments, and will not be repeated here.

[0138] This application provides a device comprising: a processor, a memory, and a system bus; the device and the memory are connected via the system bus; the memory is used to store one or more programs, the one or more programs including instructions, which, when executed by the processor, cause the processor to perform the cross-domain event determination method described in the above method embodiments.

[0139] This application provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform the cross-domain event determination method described in the above method embodiments.

[0140] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0141] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

[0142] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A statistical method for cross-domain events, characterized in that, The method includes: Acquire the target video; The target video is input into the cross-domain event determination model to obtain the probability of cross-domain event occurrence for each video frame in the multiple video frames included in the target video. The number of cross-domain events included in the target video is determined by summing the occurrence probabilities of cross-domain events corresponding to each video frame.

2. The method according to claim 1, characterized in that, The cross-domain event determination model is trained in the following manner: Obtain training videos and corresponding tags for the training videos. The training videos include multiple training video frames. The tags for the training videos include the tags of each training video frame among the multiple training video frames. The tags of each training video frame are used to indicate the probability of occurrence of cross-domain events corresponding to the training video frame. The cross-domain event determination model is trained based on the training video and the corresponding tags of the training video.

3. The method according to claim 2, characterized in that, The labels corresponding to the training videos are determined in the following way: Obtain at least one of the cross-domain events included in the training video; Based on the training video frames involved in each of the at least one cross-domain events, the cross-domain events associated with each training video frame are obtained. For each training video frame, the corresponding label for each training video frame is obtained based on the cross-domain event associated with that training video frame.

4. The method according to claim 3, characterized in that, The training video includes a first training video frame, and the label corresponding to the first training video frame is determined in the following way: Determine the occurrence probability of at least one cross-domain event corresponding to the first training video frame, wherein each cross-domain event occurrence probability in the at least one cross-domain event occurrence probability corresponds to a cross-domain event associated with the first training video frame; The label corresponding to the first training video frame is obtained based on the probability of occurrence of the at least one cross-domain event.

5. The method according to claim 4, characterized in that, The probability of at least one cross-domain event occurrence includes the probability of cross-domain event occurrence corresponding to the first cross-domain event, and the probability of cross-domain event occurrence is determined in the following manner: Based on the information of the training video frames associated with the first cross-domain event and the preset probability distribution algorithm, the probability of occurrence of the cross-domain event corresponding to the first cross-domain event is obtained.

6. The method according to claim 5, characterized in that, The information of the training video frames associated with the first cross-domain event includes: The information includes the first training video frame among the plurality of training video frames associated with the first cross-domain event, the information of the last training video frame among the plurality of training video frames associated with the first cross-domain event, and the information of the target video frame among the plurality of training video frames associated with the first cross-domain event, wherein the target video frame is the training video frame in which the first cross-domain event was determined to have occurred.

7. The method according to claim 1, characterized in that, The target video is a video captured by a camera device corresponding to the target location, and the cross-domain event includes: entering the target location, or leaving the target location.

8. A statistical device for cross-domain events, characterized in that, The device includes: The first acquisition unit is used to acquire the target video; The first determining unit is used to input the target video into the cross-domain event determining model to obtain the probability of occurrence of cross-domain events corresponding to each video frame in the multiple video frames included in the target video. The second determining unit is used to determine the number of cross-domain events included in the target video based on the sum of the occurrence probabilities of cross-domain events corresponding to each of the video frames.

9. A device, characterized in that, The device includes: a processor, a memory, and a system bus; the device and the memory are connected via the system bus; the memory is used to store one or more programs, the one or more programs including instructions, which, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 7.

Images