Real-time shooting control method and device based on human shape tracking

Abstract

The invention relates to the field of computer vision, and provides a real-time shooting control method and device based on human shape tracking, and the method comprises the steps: obtaining at least one video stream and a position identifier corresponding to each video stream in real time through a camera, and determining the video stream as a first video stream or a second video stream according to the position identifier; analyzing the first video stream through a first human shape tracking algorithm, determining human shape detection frames in the first video stream and confidence degrees corresponding to the human shape detection frames, and determining the human shape detection frames of which the confidence degrees are greater than a confidence degree threshold as target detection frames; performing feature extraction on target objects in the target detection frame to obtain object features, and tracking the target objects in the second video stream based on a second human shape tracking algorithm according to the object features to obtain route information corresponding to each target object; and determining behavior data corresponding to the target object according to the route information, and visually displaying the behavior data.

Description

Technical Field

[0001] The present application relates to the field of computer vision, and in particular to a real-time shooting control method and device based on human figure tracking. Background Art

[0002] Through cameras at various locations in the shopping center, it is very convenient to monitor customer flow in real time, and even track individual objects in the surveillance video, so as to analyze customer shopping preferences and adjust the store layout to maximize space utilization efficiency. However, shopping centers usually have a large flow of people, and the existing human tracking technology also requires a large amount of calculation to track and analyze a single object. There is an urgent need for a human tracking technology that can be applied to scenes such as shopping centers. Summary of the invention

[0003] The main purpose of this application is to provide a real-time shooting control method and device based on human figure tracking, aiming to reduce the computational complexity of human figure tracking in complex environments.

[0004] In a first aspect, the present application provides a real-time shooting control method based on human figure tracking, and the real-time shooting control method based on human figure tracking comprises the following steps:

[0005] Acquire at least one video stream and a position identifier corresponding to each of the video streams in real time through a camera, and determine the video stream as a first video stream or a second video stream according to the position identifier;

[0006] Analyzing the first video stream by using a first human figure tracking algorithm, determining a human figure detection frame in the first video stream and a confidence level corresponding to each of the human figure detection frames, and determining the human figure detection frame whose confidence level is greater than a confidence level threshold as a target detection frame;

[0007] Extracting features of the target objects in the target detection frame to obtain object features, and tracking the target objects in the second video stream based on the object features and a second human figure tracking algorithm to obtain route information corresponding to each of the target objects;

[0008] Behavior data corresponding to the target object is determined according to the route information, and the behavior data is visualized.

[0009] In the second aspect, the present application also provides a real-time shooting control device based on human figure tracking, the real-time shooting control device based on human figure tracking comprises: a first camera arranged in the entrance and exit area, a second camera arranged in the non-entrance and exit area and an information processing module, the information processing module is used to execute the real-time shooting control method based on human figure tracking described in any one of the embodiments of the present application.

[0010] The present application provides a real-time shooting control method and device based on human figure tracking. The present application obtains at least one video stream and a position identifier corresponding to each of the video streams in real time through a camera, and determines the video stream as a first video stream or a second video stream according to the position identifier; analyzes the first video stream through a first human figure tracking algorithm, determines a human figure detection frame in the first video stream, and a confidence level corresponding to each of the human figure detection frames, and determines the human figure detection frame whose confidence level is greater than a confidence level threshold as a target detection frame; extracts features of a target object in the target detection frame to obtain object features, and based on the object features, tracks the target object in the second video stream based on a second human figure tracking algorithm to obtain route information corresponding to each of the target objects; determines behavior data corresponding to the target object according to the route information, and visualizes the behavior data. Since the first human figure tracking algorithm is used to filter out a target detection frame with a higher confidence level from the human figure detection frame, and the second human figure tracking algorithm is used to track the target object corresponding to the target detection frame according to the object features, the computational complexity and complexity of human figure tracking are reduced, so that the real-time shooting control method based on human figure tracking provided in this application can be used in more complex and changeable environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 A flowchart of a real-time shooting control method based on human figure tracking provided by an embodiment of the present application;

[0013] Figure 2 A schematic block diagram of a real-time shooting control device based on human figure tracking provided in one embodiment of the present application. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0015] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0016] The embodiments of the present application provide a real-time shooting control method and device based on human figure tracking.

[0017] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0018] Please refer to Figure 1 , Figure 1 A flow chart of a real-time shooting control method based on human tracking provided in an embodiment of the present application. The real-time shooting control method based on human tracking can be used in a terminal or a server to track the target object in the video stream. The terminal can be an electronic device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device; the server can be an independent server, a server cluster, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0019] like Figure 1 As shown, the real-time shooting control method based on human figure tracking includes steps S101 to S104.

[0020] Step S101: acquiring at least one video stream and a position identifier corresponding to each of the video streams in real time through a camera, and determining the video stream as a first video stream or a second video stream according to the position identifier.

[0021] For example, a large number of cameras are usually installed in commercial spaces for real-time monitoring of the situation in the shopping mall. For example, cameras may be installed in stores, entrances and exits, aisles, and other locations. The embodiments of the present application directly obtain the video streams captured by these cameras in real time, avoiding the installation of additional cameras and reducing the hardware cost of the real-time shooting control method based on human tracking provided by the embodiments of the present application.

[0022] Exemplarily, cameras are numbered according to their locations as location identifiers corresponding to the cameras, and the location identifiers of the cameras are obtained while acquiring the video streams. The location identifiers can be used to indicate the building, floor, area, etc. where the cameras are located, which is not limited here.

[0023] In some implementations, determining the video stream as the first video stream or the second video stream according to the location identifier includes:

[0024] Determining, according to the position identifier, whether the position corresponding to the video stream belongs to an entrance and exit area;

[0025] The video stream belonging to the entrance and exit area is determined as the first video stream, and the video stream not belonging to the entrance and exit area is determined as the second video stream.

[0026] For example, shopping malls usually have a large flow of people and high mobility. To track a specific target object from a huge number of customers, a large amount of calculation is required. Therefore, in order to reduce the complexity and calculation amount of human tracking, the real-time shooting control method based on human tracking provided in the embodiment of the present application classifies the acquired video streams, and determines the video stream captured in the entrance and exit area as the first video stream, and the remaining video streams are determined as the second video stream. In this way, the customers entering the mall are analyzed preferentially through the video streams captured in the entrance and exit areas, so that these customers entering the mall can be tracked later, which improves the rationality and tracking efficiency of human tracking and reduces the calculation amount.

[0027] Exemplarily, whether a video stream belongs to an entrance and exit area can be determined according to a location identifier corresponding to the video stream. A shopping mall usually has multiple entrances and exits, such as entrances and exits on different floors and in different directions. Preset fields are set in the location identifiers of these entrances and exits so as to identify the video stream of the entrance and exit area according to the location identifier.

[0028] Step S102: Analyze the first video stream by using a first human figure tracking algorithm to determine the human figure detection frame in the first video stream and the confidence level corresponding to each of the human figure detection frames, and determine the human figure detection frame whose confidence level is greater than a confidence level threshold as a target detection frame.

[0029] Exemplarily, a human detection frame with strong recognition is identified in the first video stream as a target detection frame, so as to track the target detection frame in the second video stream. For example, based on the first video stream of the entrance and exit area, a human detection frame of a customer with distinct features is determined from the customers entering the entrance and exit as a target detection frame, so that in subsequent tracking, only the target object with distinct features needs to be tracked, which reduces the complexity and calculation amount of human tracking.

[0030] The first human figure tracking algorithm may be a SSD (Single Shot Multibox Detector) algorithm with lower resolution and complexity, such as a YOLO algorithm or a YOLO-Lite algorithm, to reduce the amount of calculation when screening a human figure detection frame.

[0031] Specifically, the target detection frame can be determined based on the confidence, and the low-confidence human detection frame is filtered out by Confidence>T, where Confidence represents the confidence of the human detection frame and T is the confidence threshold. If the confidence of the human detection frame is greater than T, it is considered that the human detection frame has the characteristic of being easy to identify, and the human detection frame is determined as the target detection frame.

[0032] Exemplarily, the confidence of the human detection frame may be, for example, an average confidence in multiple video frames, but this is certainly not limited thereto and is not limited here.

[0033] Step S103: extract features of the target objects in the target detection frame to obtain object features, and track the target objects in the second video stream based on the object features and a second human tracking algorithm to obtain route information corresponding to each target object.

[0034] Exemplarily, after the target detection frame is screened out, it can be understood that the tracking of the target object starts with not only determining the target detection frame from the first video stream in S102, but also ends with detecting from the first video stream that the target object leaves the commercial space through the entrance and exit. During this period, the target object can be tracked in the second video stream based on the object features of the target object in the target detection frame.

[0035] The second human tracking algorithm may be a multi-target tracking algorithm with higher accuracy, such as the DeepSORT algorithm.

[0036] Multi-target detection in related technologies usually requires assigning ReID to each detection frame in the picture, and tracking the detection frame according to ReID. However, such tracking requires a lot of computing power when there is a huge flow of people. Therefore, the real-time shooting control method based on human figure tracking provided in the embodiment of the present application selects the target detection frame with obvious features from the human figure detection frame. In the process of human figure tracking, only the target detection frame needs to be tracked. Since the determined detection frame is a human figure detection frame with distinct features, on the one hand, the number of detection frames is reduced, and on the other hand, the difficulty of tracking is reduced, thereby reducing the amount of calculation and complexity.

[0037] In some implementations, extracting features of the target objects in the target detection frame to obtain object features, and tracking the target objects in the second video stream based on the object features and a second human tracking algorithm to obtain route information corresponding to each target object includes:

[0038] Extracting features of the target object in the target detection frame to obtain visual features and movement features of the target object, wherein the movement features include movement direction and movement speed;

[0039] Determining a predicted position of the target object according to the moving direction and the moving speed;

[0040] The target object is selected from the predicted position according to the visual feature, and the target object is tracked.

[0041] For example, during the process of position change, the target object may enter the shooting range of another camera, and thus appear in different video streams. In order to ensure the accuracy of tracking, the position of the target object is first predicted based on the moving direction and speed of the target object, and the predicted position where the target object may appear in the next frame is obtained. The range of the predicted position can be relatively large. After determining the range corresponding to the predicted position, the target object is identified from the predicted position based on visual features, where the visual features can be vectors used to reflect the characteristics of the target object's clothing, body shape, etc., of course, it is not limited to this and is not limited here.

[0042] Exemplarily, the position of the target object is predicted by moving features, thereby avoiding searching for the target object in the entire picture, thereby reducing the amount of calculation for human tracking.

[0043] Exemplarily, if the target object is not found in the predicted position according to the visual features, the range of the predicted position is expanded based on a preset step size, and matching is performed again according to the visual features until the target object is determined in the predicted position.

[0044] Step S104: determining the behavior data corresponding to the target object according to the route information, and visually displaying the behavior data.

[0045] Exemplarily, by tracking the target object, the route information of the target object in the mall is obtained, and then the behavior data of the target object is determined according to the route information, and the behavior data is visualized.

[0046] Specifically, the route information can be determined based on the tracking results of the target object, for example, by obtaining the route of the target object in the second video stream. As for the places not captured by the camera, the positions where the target object appears twice adjacently are connected as the inferred route to obtain the route information. Of course, it is not limited to this and is not limited here.

[0047] In some embodiments, the behavior data includes: a place of stay, a length of stay, and consumption information; determining the behavior data corresponding to the target object according to the route information, and visually displaying the behavior data, includes:

[0048] Determine the target object's stay location and the stay duration corresponding to each of the stay locations according to the route information;

[0049] Acquiring consumption information of the target object at each of the staying places through a consumption management system;

[0050] The stay places and the stay duration and consumption information corresponding to each of the stay places are visually displayed.

[0051] For example, the location of the target object can be determined based on the location identifier of the video stream and the specific location of the target object in the video stream, thereby determining the target object's stay location, and determining the corresponding stay duration based on the length of time the target object stays at the stay location. Specifically, the stay location can be a specific store or functional area (such as a seating area, restroom, etc.), which is not limited here.

[0052] For example, since the consumption behavior of customers is important data for analyzing customer behavior, consumption records can be obtained in real time through the cash register system of the commercial center to determine the consumption information of the target object.

[0053] Exemplarily, the various places where the target object stayed and the corresponding stay duration and consumption information are displayed in chronological order. Specifically, the user's stays at different locations in the same place can be combined into the same stay place, such as different areas of the same store. Of course, it is not limited to this and is not limited here.

[0054] It is understandable that there is a certain time threshold for the length of stay at a stay location. If the target object's stay time at the stay location is less than the time threshold, it is considered that the target object is just passing by or appears at the stay location by chance, and has no actual intention to stay.

[0055] In some implementations, determining the behavior data corresponding to the target object according to the route information and visually displaying the behavior data includes:

[0056] Acquire attribute information corresponding to each of the stay locations of the target object;

[0057] Clustering is performed on each stay location according to the attribute information and the behavior data to obtain a sub-attribute corresponding to each attribute information.

[0058] For example, various locations in a shopping mall are usually divided in advance and their attribute information is determined. In addition to the division of rest areas and commercial areas, or the commercial areas are usually further divided into clothing attributes, catering attributes, daily necessities attributes, health care attributes, etc. according to the business scope of the store. However, even for different stores with the same attribute information, customers usually have consumption preferences. For example, customers who want to buy clothes will not traverse all stores with clothing attributes, but will choose a few stores with clothing attributes for comparison. In order to determine the competitive relationship between each store, the sub-attributes of the stores under the same attribute information can be further determined through clustering.

[0059] In some implementations, clustering each stay location according to the attribute information and the behavior data to obtain a sub-attribute corresponding to each attribute information includes:

[0060] Determine the stay locations with corresponding consumption information and the stay locations with a number of occurrences greater than a preset number of times as target stay locations;

[0061] In the behavior information of the target object, a stay location that has the same attribute information as the target stay location is determined as a related stay location;

[0062] According to the target stay location of at least one target object and the related stay locations respectively corresponding to the target stay location, each of the stay locations is clustered to obtain a sub-attribute corresponding to each attribute information.

[0063] For example, when analyzing the behavior of the target object, the most noteworthy thing is the target object's consumption behavior and the behavior of repeatedly appearing in the same store, which can usually show that the customer has a real willingness to consume. Therefore, the stay place with corresponding consumption information and the stay place with a frequency greater than a preset number of times are determined as the target stay place. It can be understood that the target stay place is a stay place with commercial attributes.

[0064] For example, after the target stay location is determined, the stay locations that have the same attribute information as the target stay location may be related to the target location. For example, if a customer makes a purchase in a clothing store, then other clothing stores corresponding to the customer may be competitors of the clothing store. Therefore, the stay locations that have the same attribute information as the target stay location are determined as related stay locations.

[0065] For example, at least one target stay location and one or more related stay locations corresponding to each target location are determined for different target objects, and then the stay locations are clustered based on a clustering algorithm to determine the strength of the correlation between the related stay locations, and the target stay locations and related stay locations with strong correlation are determined as the same sub-attribute. The clustering algorithm may be, for example, a K-means clustering algorithm, but is of course not limited thereto and is not limited here.

[0066] In some embodiments, the method further comprises:

[0067] Based on a preset density analysis algorithm, hotspot areas and non-hotspot areas in the current environment are determined according to the route information.

[0068] For example, through the route information of multiple target objects, hot spots and non-hot spots in the business center can also be determined, so as to adjust the services in the business center according to the popularity of each area and improve the service quality.

[0069] For example, since the route information is obtained by sampling people entering the commercial center according to confidence, and confidence cannot be regarded as a sampling preference, it can be regarded as random sampling. Therefore, the hotspot area determined according to the target object is representative to a certain extent.

[0070] In some implementations, the determining of hotspot areas and non-hotspot areas in the current environment according to the route information based on a preset density analysis algorithm includes:

[0071] Determine density estimates for each dwelling location

[0072]

[0073] and determining the hotspot area and the non-hotspot area according to the density estimation value; wherein, represents the density estimation value of the target area x, n represents the number of human detection boxes in the target area, h is the preset bandwidth parameter, and K n (xX i ) represents the kernel function, X i Indicates the position of the human detection frame.

[0074] Exemplarily, the hotspot area and non-hotspot area in the current environment can be determined by a kernel function. For example, each preset density estimation value is determined according to the kernel function, and a target area whose average density estimation value over a period of time is greater than a density threshold is determined as a hotspot area, and a target area whose average density estimation value is less than or equal to the density threshold is determined as a non-hotspot area.

[0075] For example, the target area is pre-divided, and the target area may include multiple stay locations. The kernel function K is calculated based on the distance between the human detection frame and the target area. n (xX i ), where the kernel function can be a Gaussian kernel, a uniform kernel, a triangular kernel, etc., which are not limited here.

[0076] Exemplarily, the bandwidth parameter h is related to the scale of the target area, and may also be selected according to actual needs. A larger bandwidth parameter may be selected for a larger target area, and conversely, a smaller bandwidth parameter may be selected for a smaller target area.

[0077] The real-time shooting control method based on human figure tracking provided in the embodiment of the present application obtains at least one video stream and the position identifier corresponding to each of the video streams in real time through a camera, and determines the video stream as a first video stream or a second video stream according to the position identifier; analyzes the first video stream through a first human figure tracking algorithm, determines a human figure detection frame in the first video stream, and the confidence corresponding to each of the human figure detection frames, and determines the human figure detection frame with the confidence greater than the confidence threshold as a target detection frame; extracts features of the target object in the target detection frame to obtain object features, and tracks the target object in the second video stream based on the object features and a second human figure tracking algorithm to obtain route information corresponding to each of the target objects; determines behavior data corresponding to the target object according to the route information, and visualizes the behavior data. Since the first human figure tracking algorithm is used to filter out a target detection frame with a higher confidence level from the human figure detection frame, and the second human figure tracking algorithm is used to track the target object corresponding to the target detection frame according to the object features, the computational complexity and complexity of human figure tracking are reduced, so that the real-time shooting control method based on human figure tracking provided in this application can be used in more complex and changeable environments.

[0078] Please refer to Figure 2 , Figure 2 A schematic block diagram of a real-time shooting control device based on human figure tracking provided in one embodiment of the present application.

[0079] like Figure 2 As shown, the real-time shooting control device based on human figure tracking includes: a first camera arranged in the entrance and exit area, a second camera arranged in the non-entrance and exit area, and an information processing module, and the information processing module is used to execute the real-time shooting control method based on human figure tracking described in any one of the embodiments of the present application.

[0080] For example, the embodiment of the present application does not limit the number of the first camera and the second camera, and the number of the first camera and the second camera can be arbitrary.

[0081] Exemplarily, the information processing module is communicatively connected with each of the first camera and the second camera, obtains the video stream of each of the first camera and the second camera in real time, and executes the real-time shooting control method based on human tracking as described in any one of the embodiments of the present application based on the obtained video stream.

[0082] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0083] It should also be understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0084] The serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A real-time shooting control method based on human figure tracking, characterized in that: The method comprises: Acquire at least one video stream and a position identifier corresponding to each of the video streams in real time through a camera, and determine the video stream as a first video stream or a second video stream according to the position identifier; Analyzing the first video stream by using a first human figure tracking algorithm, determining a human figure detection frame in the first video stream and a confidence level corresponding to each of the human figure detection frames, and determining the human figure detection frame whose confidence level is greater than a confidence level threshold as a target detection frame; Extracting features of the target objects in the target detection frame to obtain object features, and tracking the target objects in the second video stream based on the object features and a second human figure tracking algorithm to obtain route information corresponding to each of the target objects; Behavior data corresponding to the target object is determined according to the route information, and the behavior data is visualized.

2. The real-time shooting control method based on human tracking according to claim 1 is characterized in that: The step of determining the video stream as the first video stream or the second video stream according to the position identifier includes: Determining, according to the position identifier, whether the position corresponding to the video stream belongs to an entrance and exit area; The video stream belonging to the entrance and exit area is determined as the first video stream, and the video stream not belonging to the entrance and exit area is determined as the second video stream.

3. The real-time shooting control method based on human tracking according to claim 1 is characterized in that: The extracting features of the target objects in the target detection frame to obtain object features, and tracking the target objects in the second video stream based on the object features and a second human tracking algorithm to obtain route information corresponding to each target object includes: Extracting features of the target object in the target detection frame to obtain visual features and movement features of the target object, wherein the movement features include movement direction and movement speed; Determining a predicted position of the target object according to the moving direction and the moving speed; The target object is selected from the predicted position according to the visual feature, and the target object is tracked.

4. The real-time shooting control method based on human tracking according to claim 1 is characterized in that: The behavior data includes: the place of stay, the length of stay and consumption information; the determining the behavior data corresponding to the target object according to the route information and visually displaying the behavior data includes: Determine the target object's stay location and the stay duration corresponding to each of the stay locations according to the route information; Acquiring consumption information of the target object at each of the staying places through a consumption management system; The stay places and the stay duration and consumption information corresponding to each of the stay places are visually displayed.

5. The real-time shooting control method based on human figure tracking according to claim 4 is characterized in that: Determining the behavior data corresponding to the target object according to the route information, and visually displaying the behavior data, includes: Acquire attribute information corresponding to each of the stay locations of the target object; Clustering is performed on each stay location according to the attribute information and the behavior data to obtain a sub-attribute corresponding to each attribute information.

6. The real-time shooting control method based on human tracking according to claim 5 is characterized in that: The clustering of each stay location according to the attribute information and the behavior data to obtain sub-attributes corresponding to each attribute information includes: Determine the stay locations with corresponding consumption information and the stay locations with a number of occurrences greater than a preset number of times as target stay locations; In the behavior information of the target object, a stay location that has the same attribute information as the target stay location is determined as a related stay location; According to the target stay location of at least one target object and the related stay locations respectively corresponding to the target stay location, each of the stay locations is clustered to obtain a sub-attribute corresponding to each attribute information.

7. The real-time shooting control method based on human tracking according to claim 1 is characterized in that: The method further comprises: Based on a preset density analysis algorithm, hotspot areas and non-hotspot areas in the current environment are determined according to the route information.

8. The real-time shooting control method based on human tracking according to claim 7 is characterized in that: The method of determining the hotspot area and the non-hotspot area in the current environment according to the route information based on the preset density analysis algorithm includes: Determine density estimates for each dwelling location and determining the hotspot area and the non-hotspot area according to the density estimation value; wherein, represents the density estimation value of the target area x, n represents the number of human detection boxes in the target area, h is the preset bandwidth parameter, and K n (xX i ) represents the kernel function, X i Indicates the position of the human detection frame.

9. The real-time shooting control method based on human tracking according to any one of claims 1 to 8, characterized in that: The first human figure tracking algorithm includes the YOLO algorithm; the second human figure tracking algorithm includes the DeepSORT algorithm.

10. A real-time shooting control device based on human figure tracking, characterized in that: The real-time shooting control device based on human tracking includes: a first camera set in the entrance and exit area, a second camera set in the non-entrance and exit area, and an information processing module, and the information processing module is used to execute the real-time shooting control method based on human tracking as described in any one of claims 1-9.

Images