Intelligent activity and match scheduling method for sports complex

Abstract

The invention discloses a sports complex-oriented intelligent activity and competition scheduling method. And judging the busy degree of the worker and judging whether the worker adapts to the current work or not according to the posture and the position of the worker through the posture key point network, and scoring the worker so as to obtain the state rating of the worker. According to the invention, the staff category score, the staff state rating and the staff position map are adopted, so that the distance from the staff to the scheduling position, the adaptation degree of the staff to the current work, and the busy degree of the staff can be calculated. And the matching degree of the workers and the types of the workers required by the scheduling position is taken as a judgment condition, so that the technical effect of real-time and accurate personnel scheduling is achieved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an intelligent scheduling method for activities and events in a sports complex. Background Art

[0002] At present, large-scale sports events last for a long time, have a tight schedule, involve many types of work and a large number of tasks, and need to recruit a large number of service personnel of various types to complete them through reasonable division of labor and cooperation. Therefore, there must be an effective method to measure the demand scale of various service personnel, and to reasonably organize and dispatch these personnel to provide auxiliary decision-making support for managers. The personnel demand measurement and organization and dispatch of existing large-scale sports events mainly rely on the experience of managers, and there is still a lack of a strict and efficient scientific decision-making method. Empirical methods are difficult to make a quick and effective assessment of the quality of decision-making plans, so it is easy to cause decision-making deviations or errors. When faced with new situations or new scenarios, it is also impossible to make scientific, timely and efficient decision responses based on new changes.

[0003] Because a staff member has the ability to undertake multiple work modules. Therefore, during the activities and events of the sports complex, the staff can be flexibly dispatched to perform multiple tasks. However, how to judge whether the staff is suitable for the current work and whether the staff is busy or idle, so as to dispatch the staff is a problem that needs to be solved. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent scheduling method for activities and events in a sports complex to solve the above-mentioned problems existing in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides a method for intelligently scheduling activities and events for a sports complex, including: Obtaining a staff position map and corresponding staff monitoring images at multiple time points; the staff position map indicates the position of a staff member corresponding to a time point; the staff monitoring image indicates an image containing a staff member taken by multiple cameras at a time point; Through the posture key point network, based on the staff monitoring images at multiple time points, the staff status is judged by the staff posture, and multiple staff key point graphs and posture cycle lengths are obtained; the key points in the staff key point graph represent the positions of the cargo contacting parts and the sinking posture parts; one staff monitoring image at one time point corresponds to multiple staff key point graphs at one time point; Based on the length of the posture cycle, the key point diagram of the staff at multiple time points and the staff position diagram, it is judged whether the staff is busy or idle, and the staff status rating is obtained; Personnel scheduling is performed based on the worker status rating and the worker location map at the current time point.

[0006] Optionally, the posture key point network includes a human body detection network, a key point detection network and a change detection network.

[0007] Optionally, judging the worker status based on the worker monitoring image and the worker posture to obtain a plurality of worker key point graphs and posture cycle lengths includes: Inputting the worker monitoring image into a human body detection network, segmenting the position of the human body, and obtaining a plurality of single human body images; Input the single human body image into a key point detection network, detect the key parts of the human body, and obtain a first key point image to be determined and a key point boundary box; the key point boundary box represents a rectangular frame containing the key parts of the human body; the first key point image to be determined is an image containing key points; the key point is the center point of the key point boundary box; A single human body image at multiple time points corresponds to obtaining a first force key point map to be determined at multiple time points; Based on the first force key point graph to be determined at the plurality of time points, detecting changes in the key points of the staff member to obtain the posture cycle length; Based on the first key point graph to be determined at multiple time points, the key point bounding box and the posture cycle length, the key points of the continuous working state are determined through the change detection network to obtain multiple key point graphs of the staff.

[0008] Optionally, judging whether a worker is busy or idle based on the posture cycle length, the worker key point graph and the worker position graph at multiple time points, and obtaining a worker status rating includes: The absolute value of the difference between the ordinates of the corresponding key points in the staff key point graphs at the multiple time points is used as the key point change value; multiple key point change values ​​are correspondingly obtained for multiple key points in a staff key point graph at a time point; The change values ​​of multiple key points are averaged to determine the burden of the work on the staff and obtain the first staff rating; Based on the worker position graphs at multiple time points, detect the change of worker positions and obtain the position cycle length; obtaining a second staff rating based on the attitude cycle length and the position cycle length; The first staff rating is added to the second staff rating to obtain a staff status rating.

[0009] Optionally, the first force key point map to be determined based on multiple time points, the key point boundary box and the posture cycle length, through a change detection network, determines the key points of the continuous working state to obtain multiple staff key point maps, including: According to the time from morning to night and the length of the posture cycle, the first key point graphs to be determined at multiple time points are sequentially filled into different sets to obtain multiple sets of first key point graphs to be determined; Inputting a plurality of first key point graphs to be determined in the first key point graph set to a work change detection network in order from morning to night, detecting changes of key points in the work process, and obtaining key point change features; A plurality of key point graph sets to be determined correspond to obtaining a plurality of key point change features; Through the change detection network, based on multiple key point change features, multiple first key point map sets to be determined and corresponding key point bounding boxes, the positions of key points are adjusted to obtain multiple staff key point maps.

[0010] Optionally, obtaining a second staff rating based on the posture cycle length and the position cycle length includes: According to the time from morning to night and the length of the posture cycle, the position images of the staff at multiple time points are sequentially filled into different sets to obtain multiple periodic posture image sets; the periodic posture image sets include multiple key point images of the staff for a repeated action; Calculate the distance between the key points corresponding to the key point graphs of the staff at two adjacent time points in the periodic posture image set to obtain the key point change value; obtain w-1 key point change values ​​corresponding to the w key point graphs of the staff in the periodic posture image set; Cluster the w-1 key point change values ​​to obtain the clustered key point change values; Based on the cluster key point change values ​​and location cycle lengths, a second staff rating is obtained.

[0011] Optionally, obtaining a second staff rating based on the cluster key point change value and the position cycle length includes: Get the position change threshold and key point change threshold; If the position cycle length is less than the position change threshold, and the cluster key point change value is less than the key point change threshold, the second staff rating is set to 0; the second staff rating of 0 indicates that the staff is resting; If the position cycle length is less than the position change threshold, and the cluster key point change value is greater than or equal to the key point change threshold, the second worker rating is set to 0.5; the second worker rating of 0.5 indicates that the scheduling can be completed in a short time; If the position cycle length is greater than or equal to the position change threshold, and the cluster key point change value is greater than or equal to the key point change threshold, the second staff rating is set to 0; the second staff rating of 0 indicates that the staff is not suitable for the current position and needs to be dispatched to other positions; If the position cycle length is greater than or equal to the position change threshold, and the cluster key point change value is less than the key point change threshold, the second staff rating is set to 1; the second staff rating of 1 indicates that the staff is suitable for the current position and scheduling requires a long time.

[0012] Optionally, the position of the key point is adjusted through the change detection network based on multiple key point change features, multiple first force key point map sets to be determined and corresponding key point bounding boxes to obtain multiple staff key point maps, including: Inputting the first key point graph to be determined in the first set of key point graphs to be determined into a first convolutional network to obtain a first convolutional feature; Inputting the first convolution feature, the corresponding key point change feature and the corresponding key point bounding box into a first neural network to obtain an adjustment feature; Inputting the adjusted features into a reconstruction network, adjusting the positions of the key points within the key point boundary box, and obtaining a worker key point map; the key points in the worker key point map represent positions where the human body exerts force; A first key point diagram of force to be determined corresponds to a key point diagram of a staff member.

[0013] Optionally, the first force key point map to be determined based on the multiple time points detects changes in key points of the staff member to obtain the posture cycle length, including: Get multiple split lengths; According to the time points from early to late, and according to the segmentation length, the first key point graphs to be determined at the multiple time points are put into different sets to obtain multiple segmentation key point graph sets; the segmentation key point graph sets include the first key point graphs to be determined corresponding to the segmentation length; Calculate the similarity between the first force key point graphs to be determined corresponding to the subscripts of the plurality of segmentation key point graph sets to obtain a plurality of key point similarity values; The average of multiple key point similarity values ​​is calculated to obtain an average key point similarity value; Multiple segmentation lengths correspond to multiple average key point similarity values; Among multiple average key point similarity values, an average key point similarity value that is greater than other average key point similarity values ​​is used as a posture cycle similarity value; The segmentation length corresponding to the posture cycle similarity value is taken as the posture cycle length.

[0014] Optionally, the performing of personnel scheduling based on the personnel status rating and the personnel location map at the current time point includes: Obtaining a worker category score, a required number of workers, and a dispatch position; the worker category score indicates a score of a category of workers required for the dispatch position; the dispatch position indicates a position of a worker to be dispatched in a current worker position image; Obtaining a staff position image at a current time point as a current staff position image; According to the current worker position image, the distance between the worker's position at the current time point and the scheduling position is calculated to obtain the worker distance; According to the worker category score, worker distance and worker status rating, the worker category score, worker distance and worker status rating are multiplied according to a fixed ratio to obtain a scheduling value; Sort the scheduling values ​​from large to small to obtain the sequence to be scheduled; Divide the sequence to be dispatched by the number of staff members to obtain the dispatch sequence; Associate the staff corresponding to the scheduling sequence with the scheduling position and send a scheduling signal.

[0015] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects: The embodiment of the present invention also provides a method for intelligent scheduling of activities and events for a sports complex.

[0016] In the present invention, since many staff members in a sports complex activity and competition are temporarily hired and have not gone through a running-in period with the work they are doing, the burden of the current position on the staff member is judged according to the changes in the key points of the staff member. If the burden is large, it means that the staff member is not adapted to the current position and needs to be scheduled. The key point change feature is judged by detecting the position of the part in contact with the goods and the part in the sinking posture. In order to accurately detect, a change detection network is also used. Within the key point boundary box, the key points are adjusted using the changes in the key points of a working cycle. If a cycle of the staff member's position change is less than the position change threshold, it means that the staff member can be scheduled in a short time. If the length of the position cycle is greater than or equal to the position change threshold, it takes a long time for the staff member to be scheduled.

[0017] In summary, through the posture key point network, the posture and position of the staff are used to judge the busyness of the staff and whether the staff is suitable for the current work, and the staff is scored to obtain the staff status rating. By using the staff category rating, staff status rating and staff location map, the technical effect of real-time and accurate staff scheduling can be achieved based on the distance from the staff to the scheduling location, the staff's adaptability to the current work, the staff's busyness, and the matching degree between the staff and the category of the staff required by the scheduling location as the judgment conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of a method for intelligent scheduling of activities and events for a sports complex provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be described in detail below in conjunction with the accompanying drawings. Example

[0020] like Figure 1 As shown, an embodiment of the present invention provides a method for intelligent scheduling of activities and events for a sports complex, the method comprising: S101: Acquire a staff location map and corresponding staff monitoring images at multiple time points; the staff location map represents the position of a staff member corresponding to a time point; the staff monitoring image represents an image containing staff members taken by multiple cameras at a time point.

[0021] Among them, one time point corresponds to one staff location map and corresponds to multiple staff monitoring images.

[0022] S102: Through the posture key point network, based on the staff monitoring images at multiple time points, the staff status is judged by the staff posture, and multiple staff key point graphs and posture cycle lengths are obtained; the key points in the staff key point graph represent the positions of the parts in contact with the goods and the parts of the sinking posture; one staff monitoring image at one time point corresponds to multiple staff key point graphs at one time point.

[0023] S103: Based on the length of the posture cycle, the key point graph of the worker at multiple time points and the worker position graph, determine whether the worker is busy or idle, and obtain a worker status rating; S104: Performing personnel scheduling based on the worker status rating and the worker location map at the current time point.

[0024] Optionally, the posture key point network includes a human body detection network, a key point detection network, a work change detection network and a change detection network.

[0025] Optionally, the method of using a posture key point network to judge the status of a worker based on the worker monitoring images at multiple time points and to obtain multiple worker key point graphs and posture cycle lengths includes: The worker monitoring image is input into a human body detection network, the position of the human body is segmented, and a plurality of single human body images are obtained.

[0026] In this embodiment, yolov5 is used to perform human body detection to find the position of a human body, and the human body at this position is extracted to obtain a single human body image.

[0027] If a worker monitoring image contains p workers, p single human images are obtained. P is a natural number greater than 0.

[0028] The single human body image is input into the key point detection network to detect the key parts of the human body, and a first key point image to be determined and a key point boundary box are obtained; the key point boundary box represents a rectangular frame containing the key parts of the human body; the first key point image to be determined is an image containing key points; the key point is the center point of the key point boundary box.

[0029] The first key point image to be determined contains one or more key points.

[0030] The first force-generating key point image to be determined is an image with key points of the human body marked. The key points represent positions in contact with the goods or positions that move downward in posture due to supporting the goods. In this embodiment, the key points are elbows, wrists, ankles and knees.

[0031] A single human body image at multiple time points corresponds to obtaining a first force key point map to be determined at multiple time points.

[0032] Among them, if there are n staff monitoring images at a time point, then p*n first key force point images to be determined are obtained at a time point. A single human body image at m time points corresponds to obtaining the first key force point images to be determined at m time points, that is, the number of first key force point images to be determined is p*n*m. n and m are natural numbers greater than 0.

[0033] Based on the first key force point diagram to be determined at the multiple time points, changes in the key points of the staff are detected to obtain the posture cycle length.

[0034] Based on the first key point graph to be determined at multiple time points, the key point bounding box and the posture cycle length, the key points of the continuous working state are determined through the change detection network to obtain multiple key point graphs of the staff.

[0035] Among them, the number of the staff key point graphs is p*n*m.

[0036] Optionally, judging whether a worker is busy or idle based on the posture cycle length, the worker key point graph and the worker position graph at multiple time points, and obtaining a worker status rating includes: The absolute value of the difference between the vertical coordinates of the corresponding key points in the staff key point map at the multiple time points is used as the key point change value; and multiple key point change values ​​are correspondingly obtained for multiple key points in a staff member's key point map at a time point.

[0037] Among them, if there are q key points on a staff member's key point map, such as the four key points of the elbow, wrist, ankle and knee in this embodiment. Then a staff member's key point map at m time points corresponds to obtaining q key point change values. P*n staff member key point maps at m time points correspond to obtaining p*n*q key point change values. That is, one staff member corresponds to q key point change values, and p*n staff members correspond to p*n*q key point change values. q is a natural number greater than 0.

[0038] The change values ​​of multiple key points are averaged to determine the burden of the work on the staff and obtain the first staff rating.

[0039] The vertical coordinate of the key point can represent the change of the position of the key point of the human body caused by the force.

[0040] Based on the staff position graphs at multiple time points, the changes in staff positions are detected to obtain the position cycle length.

[0041] The method for obtaining the position cycle length is the same as the method for obtaining the attitude cycle length.

[0042] The method of detecting the change of the position of the staff based on the staff position graph at multiple time points and obtaining the position cycle length includes: Gets multiple position segmentation lengths.

[0043] According to the time points from early to late, and according to the position segmentation length, the staff position graphs at the multiple time points are put into different sets to obtain multiple segmentation position key point graph sets; the segmentation position key point graph sets include staff key point graphs of the number corresponding to the position segmentation length; The similarity between the staff key point maps corresponding to the subscripts of multiple segmentation position key point map sets is calculated to obtain the similarity values ​​of multiple position key points.

[0044] Among them, the Euclidean distance is used to calculate the similarity of the positions of key points in the staff's key point map.

[0045] The similarity values ​​of multiple key points at different positions are averaged to obtain the average similarity value of key points at different positions; Multiple position segmentation lengths correspond to multiple average position key point similarity values; Among the multiple average position key point similarity values, an average position key point similarity value that is greater than other average position key point similarity values ​​is used as the position period similarity value; The position segmentation length corresponding to the position cycle similarity value is taken as the position cycle length.

[0046] obtaining a second staff rating based on the attitude cycle length and the position cycle length; The first staff rating is added to the second staff rating to obtain a staff status rating.

[0047] Optionally, the first force key point map to be determined based on multiple time points, the key point boundary box and the posture cycle length, through a change detection network, determines the key points of the continuous working state to obtain multiple staff key point maps, including: According to the time from early to late and the length of the posture cycle, the first key force point diagrams to be determined at multiple time points are sequentially filled into different sets to obtain multiple sets of first key force point diagrams to be determined.

[0048] Among them, the number of elements in the set of key force point graphs to be determined is the same.

[0049] According to the time from morning to night, a plurality of first key point graphs to be determined in the set of first key point graphs to be determined are sequentially input into the work change detection network to detect the changes of key points in the work process and obtain the key point change characteristics.

[0050] Wherein, the working change detection network is a temporal convolutional network (TCN).

[0051] A plurality of sets of key point graphs to be identified correspond to obtaining a plurality of key point change features.

[0052] Among them, a set of key point graphs to be judged corresponds to obtaining a key point change feature.

[0053] Through the change detection network, based on multiple key point change features, multiple first key point map sets to be determined and corresponding key point bounding boxes, the positions of key points are adjusted to obtain multiple staff key point maps.

[0054] Optionally, obtaining a second staff rating based on the posture cycle length and the position cycle length includes: According to the time from morning to night and the length of the posture cycle, the staff position maps at multiple time points are sequentially filled into different sets to obtain multiple periodic posture image sets; the periodic posture image sets include multiple staff key point maps of a repeated action.

[0055] Among them, the number of worker position maps in the multiple periodic posture image sets is the same.

[0056] The distance between the key points corresponding to the key point graphs of the staff at two adjacent time points in the periodic posture image set is calculated to obtain the key point change value; the w key point graphs of the staff in the periodic posture image set correspond to w-1 key point change values.

[0057] In this embodiment, the time length between two time points is 5 seconds.

[0058] Cluster the w-1 key point change values ​​to obtain the clustered key point change values; Wherein, w is a natural number greater than 0.

[0059] Among them, the k-means method is used for clustering.

[0060] Based on the cluster key point change values ​​and location cycle lengths, a second staff rating is obtained.

[0061] Optionally, obtaining a second staff rating based on the cluster key point change value and the position cycle length includes: Get the position change threshold and key point change threshold.

[0062] In this embodiment, the position change threshold is 1.5 meters, and the key point change threshold is 1 centimeter.

[0063] If the position cycle length is less than the position change threshold, and the cluster key point change value is less than the key point change threshold, the second staff rating is set to 0; the second staff rating of 0 indicates that the staff is resting; If the cluster position change value is less than the position change threshold, and the cluster key point change value is greater than or equal to the key point change threshold, the second worker rating is set to 0.5; the second worker rating of 0.5 indicates that the scheduling can be completed in a short time; If the position cycle length is greater than or equal to the position change threshold, and the cluster key point change value is greater than or equal to the key point change threshold, the second staff rating is set to 0; the second staff rating of 0 indicates that the staff is not suitable for the current position and needs to be dispatched to other positions; If the position cycle length is greater than or equal to the position change threshold, and the cluster key point change value is less than the key point change threshold, the second staff rating is set to 1; the second staff rating of 1 indicates that the staff is suitable for the current position and scheduling requires a long time.

[0064] Optionally, the position of the key point is adjusted through the change detection network based on multiple key point change features, multiple first force key point map sets to be determined and corresponding key point bounding boxes to obtain multiple staff key point maps, including: The first key point graph to be determined in the first set of key point graphs to be determined is input into a first convolutional network to obtain a first convolutional feature.

[0065] Among them, the first convolutional network is a convolutional neural network (CNN).

[0066] The first convolution feature, the corresponding key point change feature and the corresponding key point bounding box are input into a first neural network to obtain an adjustment feature.

[0067] Among them, the first convolution feature and the corresponding key point change feature are one-dimensional vectors that are stretched from small to large in terms of the number of rows, the number of columns, and the number of channels.

[0068] Among them, the first neural network is a fully-connected neural network (FCNN).

[0069] The adjustment features are input into a reconstruction network, and the positions of the key points are adjusted within the key point boundary box to obtain a staff key point map; the key points in the staff key point map represent the positions where the human body exerts force.

[0070] The reconstructed network is a network that performs deconvolution using a convolutional neural network (CNN).

[0071] Among them, the first convolutional network, the first neural network and the reconstruction network are trained by using the points under the pressure of heavy objects marked at multiple time points as key points.

[0072] A first key point diagram of force to be determined corresponds to a key point diagram of a staff member.

[0073] Optionally, the first force key point map to be determined based on the multiple time points detects changes in key points of the staff member to obtain the posture cycle length, including: Get multiple split lengths; Among them, in this embodiment, the position segmentation length is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10.

[0074] According to the time points from early to late, and according to the segmentation length, the first key point graphs to be determined at the multiple time points are put into different sets to obtain multiple segmentation key point graph sets; the segmentation key point graph sets include the first key point graphs to be determined corresponding to the segmentation length; Calculate the similarity between the first force key point graphs to be determined corresponding to the subscripts of the plurality of segmentation key point graph sets to obtain a plurality of key point similarity values; The average of multiple key point similarity values ​​is calculated to obtain an average key point similarity value; Multiple segmentation lengths correspond to multiple average key point similarity values; Among multiple average key point similarity values, an average key point similarity value that is greater than other average key point similarity values ​​is used as a posture cycle similarity value; The segmentation length corresponding to the posture cycle similarity value is taken as the posture cycle length.

[0075] Optionally, the performing of personnel scheduling based on the personnel status rating and the personnel location map at the current time point includes: Obtain a worker category score, the number of workers required, and a dispatch position; the worker category represents the score of the category of workers required for the dispatch position; and the dispatch position represents the position of the worker to be dispatched in the current worker position image.

[0076] In this embodiment, if the category of the staff member is the same as that required by the dispatching position, the staff category score is set to 1. If the category of the staff member is different from that required by the dispatching position, the staff category score is set to 0.5.

[0077] Obtaining a staff position image at a current time point as a current staff position image; According to the current worker position image, the distance between the worker's position at the current time point and the scheduling position is calculated to obtain the worker distance.

[0078] Among them, the Euclidean distance method is used to calculate the distance.

[0079] The staff category score, staff distance and staff status rating are multiplied according to a fixed ratio to obtain a scheduling value.

[0080] Among them, in this embodiment, the fixed ratios corresponding to the staff category score, staff distance and staff status rating are 0.3:0.4:0.3, which means that the product of the staff category score multiplied by 0.3 plus the product of the staff distance multiplied by 0.4 plus the product of the staff status rating multiplied by 0.3 is taken as the scheduling value.

[0081] Sort the scheduling values ​​from large to small to obtain the sequence to be scheduled; The sequence to be scheduled is divided by the number of staff members to obtain the scheduling sequence.

[0082] In this embodiment, the subscript of the sequence to be scheduled starts from 0. The scheduling sequence is a sequence after the sequence to be scheduled is divided with the subscript 0 as the starting point and the subscript corresponding to the difference between the number of staff members minus one as the end point. The number of elements in the scheduling sequence is equal to the number of staff members.

[0083] Associate the staff corresponding to the scheduling sequence with the scheduling position and send a scheduling signal.

[0084] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0085] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components in the apparatus according to an embodiment of the present invention. The present invention may also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

Claims

1. A method for intelligent scheduling of activities and events for sports complexes, characterized in that: include: Obtaining worker location maps and corresponding worker monitoring images at multiple time points; The staff position graph represents the position of the staff corresponding to a time point; The staff monitoring image refers to an image containing staff taken by multiple cameras at a time point; Through the posture key point network, based on the staff monitoring images at multiple time points, the staff status is judged by the staff posture, and multiple staff key point graphs and posture cycle lengths are obtained; the key points in the staff key point graph represent the positions of the cargo contacting parts and the sinking posture parts; one staff monitoring image at one time point corresponds to multiple staff key point graphs at one time point; Based on the length of the posture cycle, the key point diagram of the staff at multiple time points and the staff position diagram, it is judged whether the staff is busy or idle, and the staff status rating is obtained; Personnel scheduling is performed based on the worker status rating and the worker location map at the current time point.

2. The method for intelligent scheduling of activities and events for sports complexes according to claim 1 is characterized in that: The posture key point network includes a human body detection network, a key point detection network, a work change detection network and a change detection network.

3. The method for intelligent scheduling of activities and events for sports complexes according to claim 2 is characterized in that: The method uses a posture key point network to judge the status of a worker based on the worker monitoring images at multiple time points, and obtains multiple worker key point graphs and posture cycle lengths, including: Inputting the worker monitoring image into a human body detection network, segmenting the position of the human body, and obtaining a plurality of single human body images; Input the single human body image into a key point detection network, detect the key parts of the human body, and obtain a first key point image to be determined and a key point boundary box; the key point boundary box represents a rectangular frame containing the key parts of the human body; the first key point image to be determined is an image containing key points; the key point is the center point of the key point boundary box; A single human body image at multiple time points corresponds to obtaining a first force key point map to be determined at multiple time points; Based on the first force key point graph to be determined at the plurality of time points, detecting changes in the key points of the staff member to obtain the posture cycle length; Based on the first key point graph to be determined at multiple time points, the key point bounding box and the posture cycle length, the key points of the continuous working state are determined through the change detection network to obtain multiple key point graphs of the staff.

4. The method for intelligent scheduling of activities and events for sports complexes according to claim 1 is characterized in that: The method of judging whether a worker is busy or idle based on the posture cycle length, the worker key point graph at multiple time points, and the worker position graph, and obtaining a worker status rating includes: The absolute value of the difference between the ordinates of the corresponding key points in the staff key point graphs at the multiple time points is used as the key point change value; multiple key point change values ​​are correspondingly obtained for multiple key points in a staff key point graph at a time point; The change values ​​of multiple key points are averaged to determine the burden of the work on the staff and obtain the first staff rating; Based on the worker position graphs at multiple time points, detect the change of worker positions and obtain the position cycle length; obtaining a second staff rating based on the attitude cycle length and the position cycle length; The first staff rating is added to the second staff rating to obtain a staff status rating.

5. The method for intelligent scheduling of activities and events for sports complexes according to claim 3 is characterized in that: The first force key point map to be determined based on multiple time points, the key point boundary box and the posture cycle length, through the change detection network, determines the key points of the continuous working state, and obtains multiple key point maps of the staff, including: According to the time from early to late, and according to the length of the posture cycle, the first key point graphs to be determined at multiple time points are sequentially filled into different sets to obtain multiple sets of first key point graphs to be determined; Inputting a plurality of first key point graphs to be determined in the first key point graph set to a work change detection network in order from morning to night, detecting changes of key points in the work process, and obtaining key point change features; A plurality of key point graph sets to be determined correspond to obtaining a plurality of key point change features; Through the change detection network, based on multiple key point change features, multiple first key point map sets to be determined and corresponding key point bounding boxes, the positions of key points are adjusted to obtain multiple staff key point maps.

6. The method for intelligent scheduling of activities and events for sports complexes according to claim 4 is characterized in that: The step of obtaining a second staff rating based on the posture cycle length and the position cycle length comprises: According to the time from morning to night and the length of the posture cycle, the position images of the staff at multiple time points are sequentially filled into different sets to obtain multiple periodic posture image sets; the periodic posture image sets include multiple key point images of the staff for a repeated action; Calculate the distance between the key points corresponding to the key point graphs of the staff at two adjacent time points in the periodic posture image set to obtain the key point change value; obtain w-1 key point change values ​​corresponding to the w key point graphs of the staff in the periodic posture image set; Cluster the w-1 key point change values ​​to obtain the clustered key point change values; Based on the cluster key point change values ​​and location cycle lengths, a second staff rating is obtained.

7. The method for intelligent scheduling of activities and events for sports complexes according to claim 6 is characterized in that: The second staff rating is obtained based on the cluster key point change value and the position cycle length, including: Get the position change threshold and key point change threshold; If the position cycle length is less than the position change threshold, and the cluster key point change value is less than the key point change threshold, the second staff rating is set to 0; the second staff rating of 0 indicates that the staff is resting; If the position cycle length is less than the position change threshold, and the cluster key point change value is greater than or equal to the key point change threshold, the second worker rating is set to 0.5; the second worker rating of 0.5 indicates that the scheduling can be completed in a short time; If the position cycle length is greater than or equal to the position change threshold, and the cluster key point change value is greater than or equal to the key point change threshold, the second staff rating is set to 0; the second staff rating of 0 indicates that the staff is not suitable for the current position and needs to be dispatched to another position; If the position cycle length is greater than or equal to the position change threshold, and the cluster key point change value is less than the key point change threshold, the second staff rating is set to 1; the second staff rating of 1 indicates that the staff is suitable for the current position and scheduling requires a long time.

8. The method for intelligent scheduling of activities and events for sports complexes according to claim 5 is characterized in that: The change detection network is used to adjust the positions of the key points based on the multiple key point change features, the multiple first force key point graph sets to be determined and the corresponding key point bounding boxes to obtain multiple staff key point graphs, including: Inputting the first key point graph to be determined in the first set of key point graphs to be determined into a first convolutional network to obtain a first convolutional feature; Inputting the first convolution feature, the corresponding key point change feature and the corresponding key point bounding box into a first neural network to obtain an adjustment feature; Inputting the adjusted features into a reconstruction network, adjusting the positions of the key points within the key point boundary box, and obtaining a worker key point map; the key points in the worker key point map represent positions where the human body exerts force; A first key point diagram of force to be determined corresponds to a key point diagram of a staff member.

9. The method for intelligent scheduling of activities and events for sports complexes according to claim 3 is characterized in that: The first force key point map to be determined based on the multiple time points detects changes in the key points of the staff member to obtain the posture cycle length, including: Get multiple split lengths; According to the time points from early to late, and according to the segmentation length, the first key point graphs to be determined at the multiple time points are put into different sets to obtain multiple segmentation key point graph sets; the segmentation key point graph sets include the first key point graphs to be determined corresponding to the segmentation length; Calculate the similarity between the first force key point graphs to be determined corresponding to the subscripts of the plurality of segmentation key point graph sets to obtain a plurality of key point similarity values; The average of multiple key point similarity values ​​is calculated to obtain an average key point similarity value; Multiple segmentation lengths correspond to multiple average key point similarity values; Among multiple average key point similarity values, an average key point similarity value that is greater than other average key point similarity values ​​is used as a posture cycle similarity value; The segmentation length corresponding to the posture cycle similarity value is taken as the posture cycle length.

10. The method for intelligent scheduling of activities and events for sports complexes according to claim 1, characterized in that: The personnel scheduling based on the worker status rating and the worker location map at the current time point includes: Obtaining a worker category score, a required number of workers, and a dispatch position; the worker category score indicates a score of a category of workers required for the dispatch position; the dispatch position indicates a position of a worker to be dispatched in a current worker position image; Obtaining a staff position image at a current time point as a current staff position image; According to the current worker position image, the distance between the worker's position at the current time point and the scheduling position is calculated to obtain the worker distance; According to the worker category score, worker distance and worker status rating, the worker category score, worker distance and worker status rating are multiplied according to a fixed ratio to obtain a scheduling value; Sort the scheduling values ​​from large to small to obtain the sequence to be scheduled; Divide the sequence to be dispatched by the number of staff members to obtain the dispatch sequence; Associate the staff corresponding to the scheduling sequence with the scheduling position and send a scheduling signal.

Images