Intelligent basketball live broadcast system based on multi-view camera

Abstract

The application discloses a kind of intelligent basketball live broadcast systems based on multi-view camera, belong to automatic live broadcast field.The system includes: multi-view camera, local server, cloud live broadcast server and mobile terminal App;Wherein, multi-view camera, for collecting court panoramic picture, and in LAN environment court panoramic picture is transmitted to local server in real time;Local server, for identifying focus coordinates from court panoramic picture, to focus coordinates as center on court panoramic picture and carry out area cutting, obtain live broadcast picture and push stream to cloud live broadcast server;Mobile terminal App, for sending preview request to cloud live broadcast server, realize match viewing.The system can realize efficient, accurate, comprehensive intelligent event live broadcast, with wide application prospect and market value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic live broadcast, and more particularly to an intelligent basketball live broadcast system based on a multi-view camera. BACKGROUND

[0002] With the vigorous development of the Internet and network live broadcast technology, sports event live broadcast has become an indispensable part of people's entertainment life. The traditional event live broadcast mode sets up a camera on the court, transmits the picture to a live broadcast vehicle or a live broadcast room through wired or wireless mode, and then pushes the real-time game picture to a cloud live broadcast server by the live broadcast vehicle or the live broadcast room, so that users can watch the picture by pulling the stream through the Internet. The traditional mode has the disadvantages of requiring a large number of equipment, manpower, material resources and time cost, and there are still many problems in manually operating the camera to complete the event live broadcast, such as technical errors caused by long-time live broadcast shooting of the camera operator and slow camera turning caused by manual operation, which leads to poor audience experience.

[0003] In order to solve the inherent problems of the traditional live broadcast mode, the event live broadcast gradually changes from manual live broadcast to digital and intelligent live broadcast. In recent years, under the support of computer vision technology, new devices and new technologies used in automatic live broadcast have emerged. For example, a mobile device is equipped with a holder, the mobile device obtains real-time pictures and analyzes picture information to control the holder to turn, so as to realize automatic live broadcast. However, in the actual application process, due to the limitation of the field angle of the mobile device, the holder is difficult to control, and many problems such as target loss and slow turning are inevitable, and the power of the mobile device itself cannot support long-time and large-scale event live broadcast.

[0004] Therefore, how to realize efficient, accurate and comprehensive intelligent event live broadcast is a problem to be solved by those skilled in the art. SUMMARY

[0005] In view of the above problems, the present application provides an intelligent basketball live broadcast system based on a multi-view camera to at least solve some of the technical problems mentioned in the background.

[0006] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0007] The present application provides an intelligent basketball live broadcast system based on a multi-view camera, which comprises a multi-view camera, a local server, a cloud live broadcast server and a mobile terminal App.

[0008] The multi-view camera is used for collecting a panoramic picture of the court and transmitting the panoramic picture of the court to the local server in real time in a local area network environment.

[0009] The local server is configured for identifying a focus point coordinate from the panoramic picture of the court, performing regional cutting on the panoramic picture of the court with the focus point coordinate as the center, obtaining a live picture and pushing the live picture to the cloud live server.

[0010] The mobile terminal App is configured for sending a preview request to the cloud live server to realize game watching.

[0011] Further, the multi-lens camera is arranged at the center line of the court.

[0012] Further, the local server communicates through an IP address in a local area network of the multi-lens camera.

[0013] Further, the local server is internally arranged with an intelligent detection module, a region determination module, a focus point generation module, a live camera operation module and a zoom control module.

[0014] The intelligent detection module is configured for receiving the panoramic picture of the court from the multi-lens camera, and performing target recognition detection on players and balls in the panoramic picture of the court to obtain a plurality of target coordinates.

[0015] The region determination module is configured for screening the plurality of target coordinates according to a preset rule to obtain in-court target coordinates in a court region.

[0016] The focus point generation module is configured for generating a live focus point coordinate based on the in-court target coordinates.

[0017] The live camera operation module is configured for cutting a rectangular frame of a preset size from the panoramic picture of the court as a live picture according to the live focus point coordinate, and continuously updating the rectangular frame with the live focus point coordinate to simulate a camera operation effect.

[0018] The zoom control module is configured for gradually reducing the current rectangular frame to simulate a zoom-in effect when the live focus point coordinate is at a left basket position or a right basket position in the court for a preset time length, and gradually increasing the current rectangular frame to simulate a zoom control when the live focus point coordinate is away from the basket position.

[0019] Further, the intelligent detection module comprises:

[0020] A detection frame generation sub-module is configured for calling a self-trained yolov5 model to perform target recognition detection on players and balls in the panoramic picture of the court to generate a detection frame corresponding to each target, which is expressed as:

[0021] boundingBox = [x, y, w, h]

[0022] boundingBox={x,y,w,h} (1)

[0023] A target coordinate generation sub-module is configured to generate a target coordinate by taking the sum of the horizontal coordinate of the top-left position of the bounding box and the midpoint of the width as the horizontal coordinate of the target coordinate, and taking the sum of the vertical coordinate of the top-left position of the bounding box and the height as the vertical coordinate of the target coordinate, and the target coordinate is represented as:

[0024] targetPos=((x+0.5w),(y+h))

[0025] wherein targetPos represents the target coordinate.

[0026] Further, the region determination module comprises:

[0027] A calibration sub-module is configured to calibrate the corner coordinates of the court by means of marking points to obtain a court rectangle.

[0028] An initialization sub-module is configured to initialize a count variable count, wherein the count variable count is used to record the number of intersection points of each target coordinate and the court rectangle.

[0029] An edge traversal processing sub-module is configured to traverse each edge of the court rectangle and obtain two end points p1 and p2 of the current edge.

[0030] An intersection point judgment sub-module is configured to judge whether the vertical coordinate of each target coordinate is within the vertical coordinate range of p1 to p2; if it is within the range, it means that the target coordinate produces an intersection point with the current edge.

[0031] An intersection point coordinate calculation sub-module is configured to calculate the intersection x coordinate of the intersection point based on the slope of the current edge; if the intersection x coordinate is greater than the horizontal coordinate of the corresponding target coordinate, it means that the target coordinate is on the right side of the current edge, and at this time, the value of the count variable count is increased.

[0032] A range judgment sub-module is configured to judge whether the count variable count is odd for each target coordinate after all edge traversals are completed; if it is odd, it means that the target coordinate is inside the court rectangle; if it is even, it means that the target coordinate is outside the court rectangle.

[0033] Further, the focal point generation module comprises:

[0034] The basketball positioning submodule is configured to determine whether the number of basketball targets corresponding to the target coordinates in the field is 1; if not, calculate the distance between each player target and each basketball target, and take the minimum distance corresponding to the basketball target coordinate as the basketball coordinate in the field;

[0035] The basketball movement analysis submodule is configured to record the basketball coordinate in the field at the current time, and determine the basketball movement direction by calculating the difference between the basketball coordinate in the field at the current time and the basketball coordinate in the field at the previous time;

[0036] The player matching submodule is configured to find the player target coordinate farthest in the basketball movement direction.

[0037] The focal point coordinate generation submodule is configured to generate a circumscribed circle with the player target coordinate farthest in the same movement direction and the basketball coordinate in the field as the center of the circumscribed circle as the live focal point coordinate.

[0038] Further, the live camera module comprises:

[0039] The center distance calculation submodule is configured to calculate the first Euclidean distance between the current live focal point coordinate and the target live focal point coordinate.

[0040] The maximum movement distance calculation submodule is configured to calculate the maximum movement distance at a preset speed and a preset time interval.

[0041] The movement judgment submodule is configured to determine whether the first Euclidean distance is greater than the maximum movement distance; if less than or equal to, the rectangular frame corresponding to the current live focal point coordinate is taken as the updated rectangular frame; if greater than, the movement process is entered, and the movement ratio is calculated.

[0042] The updated focal point coordinate submodule is configured to calculate the intermediate live focal point coordinate according to the movement ratio.

[0043] The updated rectangular frame submodule is configured to calculate the intermediate rectangular frame corresponding to the intermediate live focal point coordinate, and if the second Euclidean distance between the intermediate live focal point coordinate and the target live focal point coordinate is less than or equal to the maximum movement distance, the intermediate rectangular frame is taken as the updated rectangular frame.

[0044] Further, the cloud live server is internally deployed with a live push stream module and a live pull stream module.

[0045] The live push stream module is configured to provide a push stream address for the local server to push the live picture to the cloud live server.

[0046] The live pull stream module is configured to provide a pull stream address for the mobile terminal App to pull stream and watch.

[0047] Further, the mobile terminal App is internally deployed with a live preview module and a live management module.

[0048] The live preview module is configured to send a preview request to the live pull module, and the live pull module is configured to transmit a live picture to the mobile terminal App.

[0049] The live management module is configured to select a live on state or a live off state.

[0050] Compared with the prior art, the intelligent basketball live broadcast system based on a multi-view camera has the following beneficial effects:

[0051] The application based on a multi-view camera splices the collected picture into a panoramic picture of a court, and identifies a live focus coordinate based on a local server to realize automatic live broadcast, effectively avoiding the problems of target loss and slow turning in traditional live broadcast, and realizing efficient, accurate and comprehensive intelligent live broadcast of a competition.

[0052] The technical solutions of the application will be further described in detail below with reference to the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the accompanying drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description are only embodiments of the application, and other accompanying drawings can be obtained by those skilled in the art without creative effort on the basis of the provided accompanying drawings.

[0054] Figure 1 The intelligent basketball live broadcast system based on a multi-view camera provided in the embodiments of the application is shown in the schematic diagram.

[0055] Figure 2 The multi-view camera provided in the embodiments of the application is shown in the schematic diagram.

[0056] Figure 3 The panoramic picture of a court taken by the multi-view camera provided in the embodiments of the application is shown in the schematic diagram.

[0057] Figure 4 The court calibration and detection example diagram of the intelligent basketball live broadcast system based on a multi-view camera provided in the embodiments of the application is shown in the schematic diagram.

[0058] Figure 5 The live broadcast content example diagram of the intelligent basketball live broadcast system based on a multi-view camera provided in the embodiments of the application is shown in the schematic diagram.

[0059] Figure 6The live picture example provided by the mobile terminal App is provided for the embodiment of the application.

[0060] Figure 7 The mobile terminal App interface schematic diagram is provided for the embodiment of the application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the application will be clearly and completely described with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, but not all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the application.

[0062] The embodiment of the application discloses a smart basketball live broadcast system based on a multi-view camera, referring to Figure 1 As shown in the figure, it comprises a multi-view camera, a local server, a cloud live broadcast server and a mobile terminal App.

[0063] The multi-view camera is used for collecting a panoramic picture of a court and transmitting the panoramic picture of the court to the local server in real time in a local area network environment.

[0064] The local server is used for identifying a focal point coordinate from the panoramic picture of the court, performing regional cutting on the panoramic picture of the court with the focal point coordinate as the center, obtaining a live picture and pushing the live picture to the cloud live broadcast server.

[0065] The mobile terminal App is used for sending a preview request to the cloud live broadcast server to realize match watching.

[0066] Next, each part is described in detail.

[0067] 1. Multi-view camera:

[0068] In the embodiment of the application, the multi-view camera is erected at the court center line position to ensure that more pictures can be obtained. Figure 2 For example, the multi-view camera has four independent lenses, which are spliced into a panoramic picture of the court as shown in Figure 3 by image splicing operation.

[0069] The multi-view camera is provided with a data acquisition module, and the local server communicates through the IP address of the multi-view camera in the local area network. In specific use, the SDK interface provided by the official camera can be used to read the real-time picture of the camera and perform transcoding to realize real-time picture transmission.

[0070] 2. Local server:

[0071] In this embodiment of the invention, the local server is equipped with an intelligent detection module, a region determination module, a focus generation module, a live streaming camera movement module, and a zoom control module; wherein:

[0072] (1) Intelligent detection module:

[0073] This intelligent detection module receives panoramic images of the stadium from multiple cameras and performs target recognition and detection on players and the ball within the panoramic images to obtain the coordinates of multiple targets. Specifically, the intelligent detection module includes a detection box generation submodule and a target coordinate generation submodule; wherein:

[0074] 1) The detection box generation submodule is used to call the self-trained YOLOv5 model to perform target detection and recognition of players and balls in the panoramic view of the stadium, such as... Figure 4 As shown, generate a detection box corresponding to each target (player target or ball target); represented as:

[0075] boundingBox = [x, y, w, h]

[0076] Where boundingBox represents the detection box; x represents the horizontal coordinate of the upper left position of the detection box; y represents the vertical coordinate of the upper left position of the detection box; w represents the width of the detection box; h represents the height of the detection box; and the units of x, y, w, and h are all pixels.

[0077] 2) The target coordinate generation submodule is used to generate target coordinates by using the sum of the x-coordinate of the upper left position of the detection box and the midpoint of its width as the target x-coordinate, and the sum of the y-coordinate of the upper left position of the detection box and its height as the target y-coordinate; expressed as:

[0078] targetPos = ((x + 0.5w), (y + h))

[0079] Here, targetPos represents the target coordinates.

[0080] (2) Region determination module:

[0081] In actual live streaming, because multi-camera panoramic shots inevitably include various targets outside the court, such as basketballs and spectators, the court area needs to be demarcated before the live stream begins to eliminate the impact of off-court factors on the live stream effect. Specifically, this area determination module is used to filter multiple target coordinates according to preset rules to obtain the coordinates of targets within the court area. Specifically, this area determination module includes a calibration submodule, an initialization submodule, an edge traversal processing submodule, an intersection point determination submodule, an intersection point coordinate calculation submodule, and a range determination submodule; among which:

[0082] 1) calibration sub-module, used for calibrating the corner coordinates of the court by the way of punctuation, to obtain the court rectangle;

[0083] 2) initialization sub-module, used for initializing the count variable count; the count variable count is used for recording the number of intersection points of each target coordinate targetPos and the court rectangle;

[0084] 3) edge traversal processing sub-module, used for traversing each edge of the court rectangle, and obtaining two end points p1 and p2 of the current edge;

[0085] 4) intersection point judgment sub-module, used for judging whether the vertical coordinate of each target coordinate is within the vertical coordinate range of p1 to p2; if it is within the range, it means that the target coordinate produces intersection points with the current edge; if it is not within the range, it means that the point is above or below the current edge, and does not produce intersection points with the point;

[0086] 5) intersection point coordinate calculation sub-module, used for calculating the intersection x coordinate of the intersection point based on the slope of the current edge; if the intersection x coordinate is greater than the corresponding target coordinate horizontal coordinate, it means that the target coordinate is on the right side of the current edge, and the value of the count variable count is increased at this time;

[0087] 6) range judgment sub-module, used for judging whether the count variable count is odd after all edge traversals for each target coordinate; if it is odd, it means that the target coordinate is inside the court rectangle; if it is even, it means that the target coordinate is outside the court rectangle.

[0088] After screening the target coordinates, live focus points are generated for the in-court coordinates and the camera is controlled. The above operations are completed by the focus point generation module and the live camera control module deployed in the local server. For details, see the subsequent description.

[0089] (3) focus point generation module:

[0090] The focus point generation module is used for generating live focus point coordinates based on the in-court target coordinates; specifically, the focus point generation module includes an in-court basketball positioning sub-module, a basketball movement analysis sub-module, a player matching sub-module, and a focus point coordinate generation sub-module; wherein:

[0091] 1) the in-court basketball positioning sub-module is used for judging whether the number of corresponding basketball targets in the in-court target coordinates is 1; if it is not 1, the distance between each player target and each basketball target is calculated, and the basketball target coordinate corresponding to the minimum distance is taken as the in-court basketball coordinate;

[0092] 2) the basketball movement analysis sub-module is used for recording the in-court basketball coordinate CurrentBallPos at the current moment, and judging the basketball movement direction by calculating the difference value with the in-court basketball coordinate of the previous stage;

[0093] 3) The player matching submodule is used to find the coordinates of the player target furthest in the direction of basketball movement;

[0094] 4) The focus coordinate generation submodule is used to generate an outer circle using the coordinates of the farthest player target and the basketball coordinates in the same direction of movement, and the center of the outer circle is used as the live focus coordinate LiveFocusPos.

[0095] (4) Live camera movement module:

[0096] This live streaming camera movement module is used to crop a rectangle of a preset size from the panoramic view of the stadium based on the coordinates of the live focus, thus simulating the effect of a gimbal rotation. Figure 5 As shown; and the rectangle is continuously updated with the coordinates of the live focus to simulate the camera movement effect; the operation completed by this camera movement is to move the camera from the current position to a new position; for example, just like shooting a game as a cameraman, the content being shot must be centered on the "focus", but at the same time, the camera must be operated according to the dynamics of the field to predict the running situation below, that is, to track the "new focus";

[0097] Specifically, the live camera movement module includes a center distance calculation submodule, a maximum movement distance calculation submodule, a movement judgment submodule, a focus coordinate update submodule, and a matrix frame update submodule:

[0098] 1) Center distance calculation submodule, used to calculate the first Euclidean distance between the current live focus coordinate currentCenter and the target live focus coordinate targetCenter;

[0099] Since live streaming is always centered on the focus point, if a live stream is already in progress, the current focus point coordinate, currentCenter, is the center position of the current screen; for example, at a resolution of 1920*1080, 960*540 is the current focus point position.

[0100] The target live focus coordinate, targetCenter, is the LiveFocusPos of the next frame or several frames predicted by the focus generation module based on the movement direction and position changes of the player and basketball coordinates in the current frame.

[0101] 2) Maximum movement distance calculation submodule, used to calculate the maximum movement distance maxMoveDistance by setting the speed and time interval deltaTime; so as to control the maximum step size of movement in each frame and avoid the screen jumping too fast;

[0102] 3) a movement judgment sub-module, configured to judge whether the first Euclidean distance is greater than the maximum movement distance; if less than or equal to, the current live focus coordinate corresponding rectangular frame is taken as the updated rectangular frame; if greater than, the movement process is entered, and the movement ratio is calculated; represented as:

[0103]

[0104] wherein, distance represents the first Euclidean distance between the current live focus coordinate currentCenter and the target live focus coordinate targetCenter; t is the movement ratio, i.e. the proportion of the movement amount that the lens can complete within the preset time interval deltaTime to the first Euclidean distance distance;

[0105] 4) an update focus coordinate sub-module, configured to calculate the intermediate live focus coordinate newCenter (i.e. the live focus coordinate in the movement process of the current live focus coordinate to the target live focus coordinate) according to the movement ratio; so as to ensure that the picture gradually approaches the target in proportion;

[0106] wherein, the intermediate live focus coordinate is represented as:

[0107] newCenter = currentCenter + (targetCenter - currentCenter) * t

[0108] 5) an update matrix frame sub-module, configured to calculate the corresponding intermediate rectangular frame according to the intermediate live focus coordinate; if the second Euclidean distance between the intermediate live focus coordinate and the target live focus coordinate is less than or equal to the maximum movement distance, the intermediate rectangular frame is taken as the updated matrix frame

[0109] (5) a zoom control module:

[0110] The zoom control module is configured to, when the live focus coordinate is at the left basket position or the right basket position within the court for a preset time length, gradually reduce the current rectangular frame to simulate the pushing effect (the cropped image, i.e. the live picture size, can be restored to 1920*1080); when the live focus coordinate leaves the basket position, gradually increase the current rectangular frame to simulate the zoom control.

[0111] 3、a cloud live server:

[0112] The cloud live server is internally deployed with a live stream pushing module and a live stream pulling module; wherein:

[0113] 1、the live stream pushing module:

[0114] The live streaming pushing module is configured to provide a pushing address for the local server to push live pictures to the cloud live server.

[0115] 2. A live streaming pulling module

[0116] The live streaming pulling module is configured to provide a pulling address for a mobile terminal App to pull and watch live pictures.

[0117] 4. A mobile terminal App

[0118] The mobile terminal App is internally disposed with a live preview module and a live management module, as shown in Figure 6 and Figure 7 The live preview module is configured to send a preview request to the live streaming pulling module, and the live streaming pulling module is configured to transmit live pictures to the mobile terminal App.

[0119] 1. A live preview module

[0120] The live preview module is configured to send a preview request to the live streaming pulling module, and the live streaming pulling module is configured to transmit live pictures to the mobile terminal App.

[0121] 2. A live management module

[0122] The live preview module is configured to select a live opening state or a live closing state, and manage the live created by itself.

[0123] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be mutually referred to.

[0124] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart basketball live streaming system based on multi-camera setup, characterized in that, include: Multi-view cameras, local servers, cloud live streaming servers, and mobile apps; The multi-view camera is used to capture panoramic images of the stadium and transmit the panoramic images of the stadium to the local server in real time under a local area network environment. The local server is used to identify the focal coordinates from the panoramic view of the stadium, crop the area on the panoramic view of the stadium with the focal coordinates as the center, obtain the live broadcast image, and push it to the cloud live broadcast server. The mobile app is used to send a preview request to the cloud-based live streaming server to enable the viewing of the match; The local server is equipped with an intelligent detection module, a region determination module, a focus generation module, a live camera movement module, and a zoom control module. The intelligent detection module is used to receive the panoramic view of the stadium from the multi-view camera, and to perform target recognition and detection on the players and ball in the panoramic view of the stadium to obtain the coordinates of multiple targets. The area determination module is used to filter the multiple target coordinates according to preset rules to obtain the on-field target coordinates within the field area; The focus generation module is used to generate live focus coordinates based on the target coordinates within the field. The live camera module is used to crop out a rectangular frame of a preset size from the panoramic view of the stadium as the live screen based on the coordinates of the live focus. The rectangle is continuously updated based on the live stream focus coordinates to simulate camera movement effects; The zoom control module is used to gradually reduce the size of the current rectangle after the live focus coordinates are at the left or right basket position in the court for a preset time to simulate the advancing effect. When the live focus coordinates leave the position of the basket, the current rectangle is gradually enlarged to simulate zoom control; The live streaming camera movement module includes: The center distance calculation submodule is used to calculate the first Euclidean distance between the current live focus coordinates and the target live focus coordinates; The maximum travel distance calculation submodule is used to calculate the maximum travel distance based on preset speed and preset time interval. The movement determination submodule is used to determine whether the first Euclidean distance is greater than the maximum movement distance; if it is less than or equal to, the rectangle corresponding to the current live focus coordinates is used as the updated rectangle; if it is greater, the movement process is initiated and the movement ratio is calculated. Update the focus coordinates submodule to calculate the coordinates of the intermediate live stream focus based on the movement ratio; The update matrix frame module is used to calculate the corresponding intermediate rectangle frame based on the intermediate live focus coordinates. If the second Euclidean distance between the intermediate live focus coordinates and the target live focus coordinates is less than or equal to the maximum movement distance, then the intermediate rectangle frame is used as the updated matrix frame.

2. The intelligent basketball live streaming system based on multi-cameras according to claim 1, characterized in that, The multi-camera system was installed at the center line of the court.

3. The intelligent basketball live streaming system based on a multi-camera system according to claim 1, characterized in that, The local server communicates via IP addresses within the local area network of the multi-camera system.

4. The intelligent basketball live streaming system based on a multi-camera system according to claim 1, characterized in that, The intelligent detection module includes: The detection box generation submodule is used to call the self-trained YOLOv5 model to perform target detection of players and the ball in the panoramic view of the stadium, and generate a detection box corresponding to each target; represented as: boundingBox = [x, y, w, h] Where boundingBox represents the detection box; x represents the horizontal coordinate of the upper left position of the detection box; y represents the vertical coordinate of the upper left position of the detection box; w represents the width of the detection box; h represents the height of the detection box; and the units of x, y, w, and h are all pixels. The target coordinate generation submodule generates target coordinates by using the sum of the x-coordinate of the top-left position of the detection box and the midpoint of its width as the target x-coordinate, and the sum of the y-coordinate of the top-left position of the detection box and its height as the target y-coordinate; represented as: targetPos = ((x + 0.5w), (y + h)) Here, targetPos represents the target coordinates.

5. The intelligent basketball live streaming system based on a multi-camera system according to claim 1, characterized in that, The region determination module includes: The calibration submodule is used to mark the coordinates of the corners of the court by punctuation, thus obtaining the court rectangle; An initialization submodule is used to initialize the count variable count; the count variable count is used to record the number of intersections between each target coordinate and the rectangle of the court; The edge traversal processing submodule is used to traverse each edge of the stadium rectangle and obtain the two endpoints p1 and p2 of the current edge; The intersection point determination submodule is used to determine whether the ordinate of each target is within the range of the ordinates from p1 to p2; if it is within the range, it means that the target coordinate intersects with the current edge. The intersection point coordinate calculation submodule is used to calculate the x-coordinate of the intersection point generated based on the slope of the current edge; if the x-coordinate of the intersection point is greater than the corresponding x-coordinate of the target coordinate, it indicates that the target coordinate is to the right of the current edge, and the value of the count variable is increased. The range determination submodule is used to determine whether the count variable count is odd after all edges have been traversed for each target coordinate. If it is odd, it means that the target coordinate is inside the rectangle of the court; if it is even, it means that the target coordinate is outside the rectangle of the court.

6. The intelligent basketball live streaming system based on multi-cameras according to claim 1, characterized in that, The focus generation module includes: The in-court basketball positioning submodule is used to determine whether the number of basketball targets corresponding to the in-court target coordinates is 1; if it is not 1, the distance between each player target and each basketball target is calculated, and the basketball target coordinates corresponding to the minimum distance are taken as the in-court basketball coordinates. The basketball motion analysis submodule is used to record the current basketball coordinates on the court and determine the direction of the basketball's motion by calculating the difference between the basketball coordinates on the court and the previous stage. The player matching submodule is used to find the coordinates of the player target that is furthest away in the direction of basketball movement; The focus coordinate generation submodule is used to generate a circumcircle using the coordinates of the furthest player target and the basketball coordinates in the same direction of movement, and the center of the circumcircle is used as the focus coordinates for the live broadcast.

7. The intelligent basketball live streaming system based on a multi-camera system according to claim 1, characterized in that, The cloud-based live streaming server is equipped with a live streaming push module and a live streaming pull module. The live streaming module is used to provide a streaming address so that the local server can push the live stream to the cloud live streaming server. The live streaming module is used to provide a streaming address for the mobile app to stream and watch.

8. The intelligent basketball live streaming system based on a multi-camera system according to claim 7, characterized in that, The mobile app includes a live streaming preview module and a live streaming management module. The live streaming preview module is used to send a preview request to the live streaming pull module, which then transmits the live stream to the mobile app. The live streaming management module is used to select whether the live streaming is on or off.

Images