Group dance intelligent scoring method and system based on posture recognition and tracking matching

Abstract

The present application discloses a method and system for intelligent group dance scoring based on posture recognition and tracking matching, which relates to the field of artificial intelligence technology. The method includes real-time shooting of a group to be tested to obtain captured image information, extracting key posture points of each dancer in the group to be tested based on the captured image information to obtain human posture key point coordinate information, and obtaining human skeleton information based on the human posture key point coordinate information; judging whether the tracking of each dancer in the group to be tested is accurate based on the human posture key point coordinate information and human skeleton information, and outputting accurate character tracking results if accurate; and analyzing the posture characteristics of each dancer during the group dance based on the human posture key point coordinate information and human skeleton information to obtain posture feature parameter information. The present application has the effect of improving the accuracy of intelligent group dance scoring.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a group dance intelligent scoring method and system based on posture recognition and tracking matching. Background Art

[0002] Currently, group dances are the most common form of Chinese dance. While these works primarily consist of group dances, they sometimes incorporate solos, duets, and trios to highlight individual characters within the group, enhancing their aesthetic appeal through artistic variation and contrast. Because group dance competitions can lead to unfairness, intelligent scoring of group dances based on posture recognition and tracking is crucial.

[0003] In the related technology, patent application number CN202011281312.6 discloses a dance analysis and guidance method and device for group dance. In this patent application document, group dance video data collected by preset dual cameras is obtained, and the group dance video data is compared with preset standard group dance video data according to the music time node to generate a comparison result. According to the comparison result, a dancer's position and posture adjustment guidance video is generated. However, this patent application document does not take into account that during the group dance process, the actors inevitably interact. In complex scenes where the actors are close to each other and intertwined, it is easy to make errors in the recognition of the actors' postures or errors in matching the actors' faces and limbs. The accuracy of group dance posture recognition and matching tracking is low, and there is room for improvement. Summary of the Invention

[0004] In order to improve the accuracy of group dance intelligent scoring, this application provides a group dance intelligent scoring method and system based on posture recognition and tracking matching.

[0005] In the first aspect, the present application provides a group dance intelligent scoring method based on posture recognition and tracking matching, which adopts the following technical solutions:

[0006] The group dance intelligent scoring method based on posture recognition and tracking matching includes the following steps:

[0007] The group to be measured is photographed in real time to obtain photographic image information, and the human posture key points of each dancer in the group to be measured are extracted according to the photographic image information to obtain human posture key point coordinate information, and human skeleton information is obtained based on the human posture key point coordinate information;

[0008] Based on the coordinate information of the key points of human posture and the human skeleton information, it is determined whether the tracking of each dancer in the group to be tested is accurate. If it is accurate, the accurate tracking result of the person is output;

[0009] According to the coordinate information of the key points of the human posture and the human skeleton information of each dancer, the posture characteristics of each dancer are analyzed during the group dance to obtain posture characteristic parameter information;

[0010] A standard group dance database is created based on a standard group dance video of a standard group dance dancer group, and human posture key points are extracted based on the standard group dance database to obtain standard human posture key point coordinate information and standard human skeleton information, and whether the tracking of each dancer in the standard group dance dancer group is accurate is determined, and the posture characteristics of each dancer are analyzed to obtain standard posture feature parameter information;

[0011] Compare the posture feature parameter information with the standard posture feature parameter information to determine the similarity between the posture features of the group to be tested and the standard group of dancers, and obtain the posture feature similarity coefficient W1;

[0012] According to the standard group dance video information of the standard group dance dancers in the standard group dance database, the human posture key point coordinate information and human skeleton information of each dancer in the test group, the alignment degree between the dance movements of the test group and the music rhythm during the group dance is judged to obtain the group time rhythm feature matching coefficient W2;

[0013] According to the standard group dance video information and human body bounding box of the standard group dance dancer group in the standard group dance database, the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance dancer group during the group dance is judged, and the group motion trajectory feature matching coefficient W3 is obtained;

[0014] According to the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, the sequence scoring coefficient W of the group to be tested is obtained, and the group to be tested is scored according to the sequence scoring coefficient W of the group to be tested.

[0015] Preferably, an automatic tracking camera is obtained, and a signal connection is established between the automatic tracking camera and the group to be measured;

[0016] The automatic tracking camera is used to capture the group to be measured in real time to obtain captured image information;

[0017] Extracting the key points of the body posture of each dancer in the group to be measured based on the captured image information based on the neural network model to obtain the coordinate information of the key points of the body posture of each dancer;

[0018] The human skeleton information is obtained based on the coordinate information of the key points of the human posture, wherein the connection between the key points of the human posture is the human skeleton.

[0019] Preferably, a plurality of human body bounding boxes are created in the captured image information based on the human body posture key point coordinate information and the human body skeleton information, wherein the human body bounding box is a minimum frame for selecting the human body posture key point coordinate information and the human body skeleton information of each dancer in the group to be tested, and the human body bounding box corresponds one-to-one to each dancer in the group to be tested;

[0020] Detect the coordinate information of the key points of the human body posture and the number of dancers corresponding to the human skeleton information in each human body bounding box;

[0021] If the number of dancers corresponding to the coordinate information of the key points of human posture and the human skeleton information in the human body bounding box is only one, there is no need to judge the tracking accuracy. If the number of dancers corresponding to the coordinate information of the key points of human posture and the human skeleton information in the human body bounding box is multiple, then it is judged whether the human skeleton information of each tracked and located in the human body bounding box is consistent and whether the faces of the dancers in the human body bounding box match. If the human skeleton information is consistent and the faces of the dancers match, then the first person tracking success result is output;

[0022] Acquire a historical human body posture key point sequence database, wherein the historical human body posture key point sequence database includes human body posture key point sequence sample information of sample dancers;

[0023] Based on the human body posture key point coordinate information in each human body bounding box and the human body posture key point sequence sample information of the sample dancer in the historical human body posture key point sequence database, it is determined whether the human body posture key point sequence in each human body bounding box is smooth. If the match is consistent, it means that the human body posture key point sequence in the human body bounding box is smooth, and a second person tracking success result is output;

[0024] Based on a logic AND gate, a signal connection link is established between the logic AND gate and the group to be measured;

[0025] When the logic AND gate receives the first person tracking success result and the second person tracking success result, it outputs a person tracking accuracy result.

[0026] Preferably, the coordinate information of the dancer's body posture key points and the body skeleton information in each body bounding box are obtained and cached;

[0027] The position of each dancer in the group dance is determined based on the position of each body bounding box in the picture to obtain the overall picture position information;

[0028] Based on the coordinate information of the dancer's human posture key points and the human skeleton information in each human body bounding box, the angles of the key points of the human skeleton are calculated to obtain the joint angle feature information;

[0029] Based on the human body posture key point coordinate information, human body skeleton information and joint angle feature information of each dancer in each human body bounding box, the relative position between each dancer's human body joints is determined to obtain joint relative position information;

[0030] The overall screen position information, the joint angle feature information and the joint relative position information are combined to form posture feature parameter information.

[0031] Preferably, a standard group dance database is created, wherein the standard group dance database includes standard group dance video information of a standard group dance dancer group;

[0032] Extracting the human body posture key points of each dancer in the standard group dance dancer group according to the standard group dance database to obtain standard human body posture key point coordinate information, and obtaining standard human body skeleton information based on the standard human body posture key point coordinates;

[0033] According to the standard human posture key point coordinate information and standard human skeleton information, it is judged whether the tracking of each dancer in the standard group dance is accurate. If it is accurate, the standard character tracking accuracy result is output;

[0034] According to the standard human posture key point coordinate information and standard human skeleton information of each dancer in the standard group dance, the posture characteristics of each dancer are analyzed during the group dance to obtain standard posture feature parameter information, which includes standard overall picture position information, standard joint angle feature information and standard joint relative position information.

[0035] Preferably, the overall picture position information of each dancer in the group to be tested is compared with the standard overall picture position information of each dancer in the standard group dance group to obtain an overall picture comparison result;

[0036] According to the overall picture comparison result, the degree of similarity between the overall picture position information of each dancer in the tested group and the standard overall picture position information of each dancer in the standard group dance group is determined to obtain the overall picture similarity coefficient AZ;

[0037] Comparing the joint angle feature information of each dancer in the group to be tested with the standard joint angle feature information of each dancer in the standard group dance group to obtain a joint angle feature comparison result;

[0038] Based on the joint angle feature comparison results, the similarity between the joint angle feature information of each dancer in the test group and the standard joint angle feature information of each dancer in the standard group dance group is determined to obtain the joint angle feature similarity coefficient AJ;

[0039] Comparing the joint relative position information of each dancer in the group to be tested with the standard joint relative position information of each dancer in the standard group dance group to obtain a joint relative position comparison result;

[0040] Based on the joint relative position comparison result, the similarity between the joint relative position information of each dancer in the tested group and the standard joint relative position information of each dancer in the standard group dance group is determined to obtain the joint relative position similarity coefficient AX;

[0041] According to the overall picture similarity coefficient AZ, joint angle feature similarity coefficient AJ and joint relative position similarity coefficient AX, based on the posture feature relationship function The posture feature similarity coefficient W1 is calculated, where a1, a2, and a3 are proportional factors and are all greater than 0.

[0042] Preferably, based on the standard group dance video information of the standard group dance dancer group in the standard group dance database, the dance movements of the standard group dance dancer group corresponding to different music rhythms during the group dance process are judged to obtain the standard rhythm dance matching information;

[0043] Based on the human posture key point coordinate information and human skeleton information of each dancer in the group to be tested, the dance movements corresponding to different music rhythms of each dancer in the group dance process are judged to obtain multiple rhythm dance matching feature information;

[0044] Comparing the multiple rhythm dance matching feature information with the standard rhythm dance matching information to obtain the individual time rhythm matching coefficient of each dancer in the test group;

[0045] Detecting the standing position of each dancer in the group to be tested during the group dance to obtain the actor's standing position information, and judging the importance of each dancer's standing position based on the actor's standing position information to obtain the standing position influence weight ratio p of each dancer;

[0046] The individual time rhythm matching coefficient of each dancer in the test group is multiplied by the position influence weight ratio p of each dancer to obtain the individual time rhythm characteristic matching coefficient of each dancer in the test group;

[0047] The individual time rhythm feature matching coefficients of each dancer in the tested group are averaged to obtain the degree of alignment between the dance movements and the music rhythm of the tested group during the group dance, that is, the group time rhythm feature matching coefficient W2.

[0048] Preferably, based on the standard group dance video information of the standard group dance dancer group in the standard group dance database, the motion trajectory of each dancer in the standard group dance dancer group is determined and a standard motion trajectory diagram is drawn;

[0049] Positioning and tracking the motion trajectory of the human body bounding box corresponding to each dancer in the group to be measured based on the automatic tracking camera to obtain actor motion trajectory information, and drawing the motion trajectory of each dancer in the group to be measured based on the actor motion trajectory information to obtain a target motion trajectory diagram;

[0050] Detecting the standing positions of standard group dance dancers in a standard group dance database to obtain dancer standing position information;

[0051] Based on the dancer position information and the actor position information, each dancer in the standard group dance dancer group is matched with each dancer in the test group one by one to obtain personnel matching information;

[0052] Based on the personnel matching information, the standard motion trajectory diagram and the target motion trajectory diagram, the similarity between the motion trajectory of each dancer in the test group and the motion trajectory of the dancers in the corresponding standard group dance group is determined to obtain the individual trajectory similarity coefficient;

[0053] The individual trajectory similarity coefficient of each dancer in the test group is multiplied by the position influence weight ratio p of each dancer to obtain the individual motion trajectory feature matching coefficient of each dancer in the test group;

[0054] The individual motion trajectory feature matching coefficients of each dancer in the tested group are averaged and the similarity between the motion trajectory of the tested group during group dance and the motion trajectory of the standard group dance dancers during group dance is determined to obtain the group motion trajectory feature matching coefficient W3.

[0055] Preferably, based on the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, the scoring conversion function Calculate and obtain the sequence scoring coefficient W of the population to be tested, where: 、 、 is the proportional factor and is greater than 0;

[0056] The group dance performance of the group to be tested is scored according to the sequence scoring coefficient W of the group to be tested, wherein the larger the sequence scoring coefficient W of the group to be tested is, the higher the group dance performance score of the group to be tested is.

[0057] Secondly, this application provides a group dance intelligent scoring system based on posture recognition and tracking matching, which adopts the following technical solutions:

[0058] The group dance intelligent scoring system based on posture recognition and tracking matching includes:

[0059] A human posture feature extraction module is configured to capture the group to be tested in real time to obtain captured image information, extract the human posture key points of each dancer in the group to be tested based on the captured image information to obtain human posture key point coordinate information, and obtain human skeleton information based on the human posture key point coordinate information;

[0060] The character tracking judgment module is configured to judge whether the tracking of each dancer in the group to be tested is accurate based on the coordinate information of the key points of human posture and human skeleton information, and output the accurate character tracking result if it is accurate;

[0061] The posture feature analysis module is configured to analyze the posture features of each dancer in the group dance process according to the human posture key point coordinate information and human skeleton information of each dancer to obtain posture feature parameter information;

[0062] a standard data acquisition module configured to create a standard group dance database based on a standard group dance video of a standard group dance dancer group, extract human posture key points based on the standard group dance database to obtain standard human posture key point coordinate information and standard human skeleton information, determine whether the tracking of each dancer in the standard group dance dancer group is accurate, and analyze the posture characteristics of each dancer to obtain standard posture feature parameter information;

[0063] The posture feature similarity analysis module is configured to compare the posture feature parameter information with the standard posture feature parameter information, determine the similarity between the posture features of the group to be tested and the standard group of dancers, and obtain the posture feature similarity coefficient W1;

[0064] The time rhythm matching analysis module is configured to determine the alignment between the dance movements of the group to be tested and the music rhythm during the group dance process based on the standard group dance video information of the standard group dance dancer group in the standard group dance database, the human posture key point coordinate information and human skeleton information of each dancer in the group to be tested, and obtain the group time rhythm feature matching coefficient W2;

[0065] The motion trajectory similarity analysis module is configured to determine the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance dancer group during the group dance based on the standard group dance video information and the human body bounding box of the standard group dance dancer group in the standard group dance database, and obtain the group motion trajectory feature matching coefficient W3;

[0066] The group dance scoring module is configured to obtain the sequence scoring coefficient W of the group to be tested based on the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, and perform a scoring operation on the group to be tested based on the sequence scoring coefficient W of the group to be tested.

[0067] In summary, this application includes at least one of the following beneficial technical effects:

[0068] By detecting the human posture key points of each dancer in the test group, the coordinate information of the human posture key points and the human skeleton information are obtained. Based on the human posture key point coordinate information and the human skeleton information, it is judged whether the shooting, tracking and positioning of the dancers in the test group by the test group is accurate. If accurate, the accurate result of character tracking is output, and further judgment is made on the personnel identification, tracking and positioning, which reduces the risk of personnel identification, tracking and positioning errors and improves the accuracy of group dance intelligent scoring. According to the human posture key point coordinate information and the human skeleton information of each dancer, the posture characteristics of each dancer are analyzed to obtain posture feature parameter information, the standard posture feature parameter information of the standard group dance dancer group is detected, the posture feature parameter information is compared with the standard posture feature parameter information, and the similarity of the posture characteristics is judged to obtain the posture feature parameter information. The posture feature similarity coefficient W1 is obtained, and the group time rhythm feature matching coefficient W2 is obtained according to the standard group dance video information of the standard group dance dancer group, the human posture key point coordinate information of each dancer in the test group, and the human skeleton information. The standard group dance video information of the standard group dance dancer group and the human body bounding box group motion trajectory feature matching coefficient W3 are obtained. According to the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, the sequence scoring coefficient W of the test group is obtained through multi-dimensional analysis, which improves the detection accuracy of the sequence scoring coefficient W of the test group, and scores the group dance performance of the test group according to the sequence scoring coefficient W of the test group, which further improves the accuracy of the group dance performance scoring of the test group. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of a group dance intelligent scoring method based on posture recognition and tracking matching, which is mainly reflected in this embodiment;

[0070] Figure 2 This is a module diagram of a group dance intelligent scoring system based on posture recognition and tracking matching, which is mainly embodied in this embodiment.

[0071] Figure numerals: 1. Human body posture feature extraction module; 2. Person tracking and judgment module; 3. Posture feature analysis module; 4. Standard data acquisition module; 5. Posture feature similarity analysis module; 6. Time rhythm matching analysis module; 7. Motion trajectory similarity analysis module; 8. Group dance scoring module. DETAILED DESCRIPTION

[0072] The present application is further described in detail below with reference to the accompanying drawings.

[0073] The embodiment of the present application discloses a group dance intelligent scoring method based on posture recognition and tracking matching.

[0074] The group dance intelligent scoring method based on posture recognition and tracking matching includes the following steps:

[0075] Reference Figure 1 In step S1, the group to be tested is photographed in real time to obtain photographic image information, and the key points of the body posture of each dancer in the group to be tested are extracted based on the photographic image information to obtain the coordinate information of the key points of the body posture, and the human skeleton information is obtained based on the coordinate information of the key points of the body posture. Step S1 specifically includes the following sub-steps:

[0076] Step S11: Acquire an automatic tracking camera and establish a signal connection between the automatic tracking camera and the group to be measured.

[0077] Step S12: Using the automatic tracking camera to shoot the group to be measured in real time to obtain shooting image information.

[0078] Step S13 , based on the neural network model, extracting the key points of the body posture of each dancer in the group to be measured according to the captured image information to obtain the coordinate information of the key points of the body posture of each dancer.

[0079] Step S14, obtaining human skeleton information based on the coordinate information of the human body posture key points, wherein the connection between the human body posture key points is the human skeleton.

[0080] Reference Figure 1 In step S2, based on the coordinate information of the key points of the human body posture and the human skeleton information, it is determined whether the tracking of each dancer in the group to be tested is accurate. If it is accurate, the accurate tracking result of the person is output. Step S2 specifically includes the following sub-steps:

[0081] In step S21, a plurality of human body bounding boxes are created in the captured image information based on the coordinate information of the key points of human body posture and the human body skeleton information. The human body bounding box is the smallest frame used to select the coordinate information of the key points of human body posture and the human body skeleton information of each dancer in the group to be tested. The human body bounding box corresponds one-to-one to each dancer in the group to be tested.

[0082] Step S22: Detect the number of dancers corresponding to the coordinate information of the key points of the human body posture and the human skeleton information in each human body bounding box.

[0083] Step S23: If the number of dancers corresponding to the human body posture key point coordinate information and human body skeleton information in the human body enclosing frame is only one, there is no need to judge the tracking accuracy. If the number of dancers corresponding to the human body posture key point coordinate information and human body skeleton information in the human body enclosing frame is multiple, then judge whether the human body skeleton information tracked and located in the human body enclosing frame is consistent and whether the faces of the dancers in the human body enclosing frame match. If the human body skeleton information is consistent and the dancers' faces match, then output the first person tracking success result.

[0084] Step S24: obtaining a historical human body posture key point sequence database, wherein the historical human body posture key point sequence database includes human body posture key point sequence sample information of sample dancers.

[0085] Step S25, based on the human body posture key point coordinate information in each human body bounding box and the human body posture key point sequence sample information of the sample dancer in the historical human body posture key point sequence database, it is determined whether the human body posture key point sequence in each human body bounding box is smooth. If the match is consistent, it means that the human body posture key point sequence in the human body bounding box is smooth, and the second person tracking success result is output.

[0086] Step S26: Based on the logic AND gate, a signal connection link is established between the logic AND gate and the group to be tested.

[0087] Step S27 , when the logic AND gate receives the first person tracking success result and the second person tracking success result, outputs a person tracking accuracy result.

[0088] Reference Figure 1 In step S3, the posture characteristics of each dancer in the group dance process are analyzed based on the coordinate information of the key points of the dancer's body posture and the human skeleton information to obtain posture characteristic parameter information. Step S3 specifically includes the following sub-steps:

[0089] Step S31 , obtaining the coordinate information of the dancer's body posture key points and the body skeleton information in each body bounding box and performing buffering processing.

[0090] Step S32: determining the position of each dancer in the group dance process based on the position of each body enclosing frame in the picture to obtain the overall picture position information.

[0091] Step S33 , based on the coordinate information of the dancer's body posture key points and the body skeleton information in each body bounding box, the angles of the key points of the body skeleton are calculated to obtain joint angle feature information.

[0092] Step S34 , based on the body posture key point coordinate information, body skeleton information and joint angle feature information of each dancer in each body bounding box, the relative positions between the body joints of each dancer are determined to obtain joint relative position information.

[0093] In step S35, the overall screen position information, the joint angle feature information, and the joint relative position information are combined to form posture feature parameter information.

[0094] Reference Figure 1 In step S4, a standard group dance database is created based on the standard group dance video of the standard group dance dancer group. The human body posture key points are extracted based on the standard group dance database to obtain the standard human body posture key point coordinate information and standard human body skeleton information. The accuracy of the tracking of each dancer in the standard group dance dancer group is determined, and the posture characteristics of each dancer are analyzed to obtain the standard posture feature parameter information. Step S4 specifically includes the following sub-steps:

[0095] Step S41: creating a standard group dance database, wherein the standard group dance database includes standard group dance video information of a group of standard group dance dancers.

[0096] Step S42: extracting the human body posture key points of each dancer in the standard group dance dancer group according to the standard group dance database to obtain standard human body posture key point coordinate information, and obtaining standard human body skeleton information based on the standard human body posture key point coordinates.

[0097] Step S43, judging whether the tracking of each dancer in the standard group dance is accurate based on the standard human posture key point coordinate information and the standard human skeleton information, and outputting the standard person tracking accuracy result if it is accurate.

[0098] Step S44, based on the standard human posture key point coordinate information and standard human skeleton information of each dancer in the standard group dance, the posture characteristics of each dancer in the group dance process are analyzed to obtain standard posture feature parameter information, and the standard posture feature parameter information includes standard overall picture position information, standard joint angle feature information and standard joint relative position information.

[0099] Reference Figure 1 In step S5, the posture feature parameter information is compared with the standard posture feature parameter information to determine the similarity between the posture features of the group to be tested and the standard group of dancers, and obtain the posture feature similarity coefficient W1. Step S5 specifically includes the following sub-steps:

[0100] Step S51 : comparing the overall picture position information of each dancer in the group to be tested with the standard overall picture position information of each dancer in the standard group dance to obtain an overall picture comparison result.

[0101] Step S52: Based on the overall picture comparison result, the degree of similarity between the overall picture position information of each dancer in the group to be tested and the standard overall picture position information of each dancer in the standard group dance is determined to obtain an overall picture similarity coefficient AZ. The higher the degree of similarity between the overall picture position information of each dancer in the group to be tested and the standard overall picture position information of each dancer in the standard group dance, the greater the overall picture similarity coefficient AZ.

[0102] Step S53: Compare the joint angle feature information of each dancer in the group to be tested with the standard joint angle feature information of each dancer in the standard group dance group to obtain a joint angle feature comparison result.

[0103] Step S54: Based on the joint angle feature comparison results, the degree of similarity between the joint angle feature information of each dancer in the test group and the standard joint angle feature information of each dancer in the standard group dance group is determined to obtain a joint angle feature similarity coefficient AJ. The higher the degree of similarity between the joint angle feature information of each dancer in the test group and the standard joint angle feature information of each dancer in the standard group dance group, the greater the joint angle feature similarity coefficient AJ.

[0104] Step S55 , comparing the joint relative position information of each dancer in the group to be tested with the standard joint relative position information of each dancer in the standard group dance group to obtain a joint relative position comparison result.

[0105] Step S56: Based on the joint relative position comparison result, the degree of similarity between the joint relative position information of each dancer in the test group and the standard joint relative position information of each dancer in the standard ensemble dance group is determined to obtain a joint relative position similarity coefficient AX. The higher the degree of similarity between the joint relative position information of each dancer in the test group and the standard joint relative position information of each dancer in the standard ensemble dance group, the greater the joint relative position similarity coefficient AX.

[0106] Step S57, based on the overall picture similarity coefficient AZ, the joint angle feature similarity coefficient AJ and the joint relative position similarity coefficient AX, the posture feature relationship function The posture feature similarity coefficient W1 is calculated, where a1, a2, and a3 are proportional factors and are all greater than 0.

[0107] Reference Figure 1In step S6, based on the standard group dance video information of the standard group dance dancers in the standard group dance database, the human posture key point coordinate information and human skeleton information of each dancer in the group to be tested, the alignment degree between the dance movements of the group to be tested and the music rhythm during the group dance is determined to obtain the group time rhythm feature matching coefficient W2. Step S6 specifically includes the following sub-steps:

[0108] Step S61, based on the standard group dance video information of the standard group dance dancer group in the standard group dance database, determining the dance movements of the standard group dance dancer group corresponding to different music rhythms during the group dance process to obtain standard rhythm dance matching information.

[0109] Step S62: Based on the human posture key point coordinate information and human skeleton information of each dancer in the group to be tested, the dance movements corresponding to different music rhythms of each dancer in the group dance process are determined to obtain multiple rhythm dance matching feature information.

[0110] Step S63 : Compare the plurality of rhythm dance matching feature information with the standard rhythm dance matching information to obtain the individual time rhythm matching coefficient of each dancer in the group to be tested.

[0111] Step S64: detecting the standing position of each dancer in the group to be tested during the group dance to obtain the dancer's standing position information, and judging the importance of each dancer's standing position based on the actor's standing position information to obtain the standing position influence weight ratio p of each dancer.

[0112] Step S65 , multiplying the individual time rhythm matching coefficient of each dancer in the group to be tested by the position influence weight ratio p of each dancer to obtain the individual time rhythm feature matching coefficient of each dancer in the group to be tested.

[0113] In step S66, the individual time rhythm feature matching coefficients of the dancers in the group to be tested are averaged to obtain the degree of alignment between the dance movements and the music rhythm of the group to be tested during the group dance, that is, the group time rhythm feature matching coefficient W2.

[0114] Reference Figure 1 In step S7, based on the standard group dance video information and the human body bounding box of the standard group dance group in the standard group dance database, the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance group during the group dance is determined, and the group motion trajectory feature matching coefficient W3 is obtained. Step S7 specifically includes the following sub-steps:

[0115] Step S71, judging the motion trajectory of each dancer in the standard group dance dancer group and drawing a standard motion trajectory diagram based on the standard group dance video information of the standard group dance dancer group in the standard group dance database.

[0116] Step S72: The automatic tracking camera locates and tracks the motion trajectory of the human body bounding box corresponding to each dancer in the group to be measured to obtain the actor motion trajectory information; based on the actor motion trajectory information, the motion trajectory of each dancer in the group to be measured is plotted to obtain a target motion trajectory diagram.

[0117] Step S73: detecting the standing positions of standard group dance dancers in the standard group dance database to obtain dancer standing position information.

[0118] Step S74: Based on the dancer position information and the actor position information, each dancer in the standard group dance dancer group is matched with each dancer in the group to be tested to obtain personnel matching information.

[0119] Step S75: Based on the personnel matching information, the standard motion trajectory diagram and the target motion trajectory diagram, the similarity between the motion trajectory of each dancer in the test group and the motion trajectory of the dancers in the corresponding standard group dance group is determined to obtain the individual trajectory similarity coefficient.

[0120] Step S76: Multiply the individual trajectory similarity coefficient of each dancer in the test group by the position influence weight ratio p of each dancer to obtain the individual motion trajectory feature matching coefficient of each dancer in the test group.

[0121] Step S77, calculate the mean of the individual motion trajectory feature matching coefficients of each dancer in the group to be tested, determine the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance dancer group during the group dance, and obtain the group motion trajectory feature matching coefficient W3.

[0122] Reference Figure 1 In step S8, a sequence scoring coefficient W of the group to be tested is obtained based on the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2, and the group motion trajectory feature matching coefficient W3, and a scoring operation is performed on the group to be tested based on the sequence scoring coefficient W of the group to be tested. Step S8 specifically includes the following sub-steps:

[0123] Step S81, based on the score conversion function, the similarity coefficient W1 of the posture feature, the matching coefficient W2 of the group time rhythm feature and the matching coefficient W3 of the group motion trajectory feature are calculated. Calculate and obtain the sequence scoring coefficient W of the population to be tested, where: 、 、 are scaling factors and are all greater than 0.

[0124] Step S82: scoring the group dance performance of the group to be tested according to the sequence scoring coefficient W of the group to be tested, wherein the larger the sequence scoring coefficient W of the group to be tested is, the higher the group dance performance score of the group to be tested is.

[0125] The embodiment of the present application also discloses an intelligent group dance scoring system based on posture recognition and tracking matching.

[0126] Reference Figure 2 , an intelligent group dance scoring system based on posture recognition and tracking matching, including:

[0127] The human posture feature extraction module is configured to shoot the group to be tested in real time to obtain shooting image information, extract the human posture key points of each dancer in the group to be tested based on the shooting image information to obtain human posture key point coordinate information, and obtain human skeleton information based on the human posture key point coordinate information.

[0128] The character tracking judgment module is configured to judge whether the tracking of each dancer in the group to be tested is accurate based on the coordinate information of the key points of human posture and human skeleton information, and output the accurate character tracking results if accurate.

[0129] The posture feature analysis module is configured to analyze the posture features of each dancer in the group dance process according to the human posture key point coordinate information and human skeleton information of each dancer to obtain posture feature parameter information.

[0130] The standard data acquisition module is configured to create a standard group dance database based on the standard group dance video of the standard group dance dancer group, extract the key points of human posture based on the standard group dance database to obtain the coordinate information of the key points of standard human posture and standard human skeleton information, judge whether the tracking of each dancer in the standard group dance dancer group is accurate, and analyze the posture characteristics of each dancer to obtain the standard posture feature parameter information.

[0131] The posture feature similarity analysis module is configured to compare the posture feature parameter information with the standard posture feature parameter information, determine the similarity between the posture features of the group to be tested and the standard group dance dancer group, and obtain the posture feature similarity coefficient W1.

[0132] The time rhythm matching analysis module is configured to determine the alignment between the dance movements of the group to be tested and the music rhythm during the group dance based on the standard group dance video information of the standard group dance dancers in the standard group dance database, the human posture key point coordinate information of each dancer in the group to be tested, and the human skeleton information, and obtain the group time rhythm feature matching coefficient W2.

[0133] The motion trajectory similarity analysis module is configured to judge the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance dancer group during the group dance based on the standard group dance video information and human body bounding box of the standard group dance dancer group in the standard group dance database, and obtain the group motion trajectory feature matching coefficient W3.

[0134] The group dance scoring module is configured to obtain the sequence scoring coefficient W of the group to be tested based on the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, and perform a scoring operation on the group to be tested based on the sequence scoring coefficient W of the group to be tested.

[0135] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. An intelligent scoring method for group dance based on posture recognition and tracking matching, characterized by: The following steps are involved: The group to be measured is photographed in real time to obtain photographic image information, and the human posture key points of each dancer in the group to be measured are extracted according to the photographic image information to obtain human posture key point coordinate information, and human skeleton information is obtained based on the human posture key point coordinate information; Based on the coordinate information of the key points of human posture and the human skeleton information, it is determined whether the tracking of each dancer in the group to be tested is accurate. If it is accurate, the accurate tracking result of the person is output; According to the coordinate information of the key points of the human posture and the human skeleton information of each dancer, the posture characteristics of each dancer are analyzed during the group dance to obtain posture characteristic parameter information; A standard group dance database is created based on a standard group dance video of a standard group dance dancer group, and human posture key points are extracted based on the standard group dance database to obtain standard human posture key point coordinate information and standard human skeleton information, and whether the tracking of each dancer in the standard group dance dancer group is accurate is determined, and the posture characteristics of each dancer are analyzed to obtain standard posture feature parameter information; Compare the posture feature parameter information with the standard posture feature parameter information to determine the similarity between the posture features of the group to be tested and the standard group of dancers, and obtain the posture feature similarity coefficient W1; According to the standard group dance video information of the standard group dance dancers in the standard group dance database, the human posture key point coordinate information and human skeleton information of each dancer in the test group, the alignment degree between the dance movements of the test group and the music rhythm during the group dance is judged to obtain the group time rhythm feature matching coefficient W2; According to the standard group dance video information and human body bounding box of the standard group dance dancer group in the standard group dance database, the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance dancer group during the group dance is judged, and the group motion trajectory feature matching coefficient W3 is obtained; according to The posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3 are used to obtain the sequence scoring coefficient W of the group to be tested, and the group to be tested is scored according to the sequence scoring coefficient W of the group to be tested.

2. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 1 is characterized in that: The steps of shooting the group to be tested in real time to obtain image information, extracting the key points of the body posture of each dancer in the group to be tested based on the image information to obtain the coordinate information of the key points of the body posture, and obtaining the human skeleton information based on the coordinate information of the key points of the body posture, specifically include: Acquire an automatic tracking camera and establish a signal connection between the automatic tracking camera and the group to be measured; The automatic tracking camera is used to capture the group to be measured in real time to obtain captured image information; Extracting the key points of the body posture of each dancer in the group to be measured based on the captured image information based on the neural network model to obtain the coordinate information of the key points of the body posture of each dancer; The human skeleton information is obtained based on the coordinate information of the key points of the human posture, wherein the connection between the key points of the human posture is the human skeleton.

3. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 2 is characterized in that: The steps of determining whether the tracking of each dancer in the group to be tested is accurate based on the coordinate information of the key points of the human body posture and the human skeleton information, and outputting the accurate tracking results of the dancers if accurate, specifically include: Creating a plurality of human body bounding boxes in the captured image information based on the human body posture key point coordinate information and the human body skeleton information, wherein the human body bounding boxes are the smallest frames for enclosing the human body posture key point coordinate information and the human body skeleton information of each dancer in the group to be tested, and the human body bounding boxes correspond one-to-one to each dancer in the group to be tested; Detect the coordinate information of the key points of the human body posture and the number of dancers corresponding to the human skeleton information in each human body bounding box; If the number of dancers corresponding to the coordinate information of the key points of human posture and the human skeleton information in the human body bounding box is only one, there is no need to judge the tracking accuracy. If the number of dancers corresponding to the coordinate information of the key points of human posture and the human skeleton information in the human body bounding box is multiple, then it is judged whether the human skeleton information of each tracked and located in the human body bounding box is consistent and whether the faces of the dancers in the human body bounding box match. If the human skeleton information is consistent and the faces of the dancers match, then the first person tracking success result is output; Acquire a historical human body posture key point sequence database, wherein the historical human body posture key point sequence database includes human body posture key point sequence sample information of sample dancers; Based on the human body posture key point coordinate information in each human body bounding box and the human body posture key point sequence sample information of the sample dancer in the historical human body posture key point sequence database, it is determined whether the human body posture key point sequence in each human body bounding box is smooth. If the match is consistent, it means that the human body posture key point sequence in the human body bounding box is smooth, and a second person tracking success result is output; Based on a logic AND gate, a signal connection link is established between the logic AND gate and the group to be measured; When the logic AND gate receives the first person tracking success result and the second person tracking success result, it outputs a person tracking accuracy result.

4. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 3 is characterized in that: The steps of analyzing the posture characteristics of each dancer in the group dance process to obtain posture characteristic parameter information based on the human posture key point coordinate information and human skeleton information of the dancers in each human body bounding box specifically include: Obtain the coordinate information of the dancer's body posture key points and body skeleton information in each body bounding box and cache them; The position of each dancer in the group dance is determined based on the position of each body bounding box in the picture to obtain the overall picture position information; Based on the coordinate information of the dancer's human posture key points and human skeleton information in each human body bounding box, the angles of the key points of the human skeleton are calculated to obtain joint angle feature information; Based on the human body posture key point coordinate information, human body skeleton information and joint angle feature information of each dancer in each human body bounding box, the relative position between each dancer's human body joints is determined to obtain joint relative position information; The overall screen position information, the joint angle feature information and the joint relative position information are combined to form posture feature parameter information.

5. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 4 is characterized in that: The steps of creating a standard group dance database based on a standard group dance video of a standard group dance dancer group, extracting human posture key points based on the standard group dance database to obtain standard human posture key point coordinate information and standard human skeleton information, determining whether the tracking of each dancer in the standard group dance dancer group is accurate, and analyzing the posture characteristics of each dancer to obtain standard posture feature parameter information specifically include: Creating a standard group dance database, wherein the standard group dance database includes standard group dance video information of a standard group dance dancer group; Extracting the human body posture key points of each dancer in the standard group dance dancer group according to the standard group dance database to obtain standard human body posture key point coordinate information, and obtaining standard human body skeleton information based on the standard human body posture key point coordinates; According to the standard human posture key point coordinate information and standard human skeleton information, it is judged whether the tracking of each dancer in the standard group dance is accurate. If it is accurate, the standard character tracking accuracy result is output; According to the standard human posture key point coordinate information and standard human skeleton information of each dancer in the standard group dance, the posture characteristics of each dancer are analyzed during the group dance to obtain standard posture feature parameter information, which includes standard overall picture position information, standard joint angle feature information and standard joint relative position information.

6. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 5 is characterized in that: The steps of comparing the posture feature parameter information with the standard posture feature parameter information, determining the similarity between the posture features of the group to be tested and the group of standard group dancers, and obtaining the posture feature similarity coefficient W1 specifically include: Comparing the overall picture position information of each dancer in the group to be tested with the standard overall picture position information of each dancer in the standard group dance dancer group to obtain an overall picture comparison result; According to the overall picture comparison result, the degree of similarity between the overall picture position information of each dancer in the tested group and the standard overall picture position information of each dancer in the standard group dance group is determined to obtain the overall picture similarity coefficient AZ; Comparing the joint angle feature information of each dancer in the group to be tested with the standard joint angle feature information of each dancer in the standard group dance group to obtain a joint angle feature comparison result; Based on the joint angle feature comparison results, the similarity between the joint angle feature information of each dancer in the test group and the standard joint angle feature information of each dancer in the standard group dance group is determined to obtain the joint angle feature similarity coefficient AJ; Comparing the joint relative position information of each dancer in the group to be tested with the standard joint relative position information of each dancer in the standard group dance group to obtain a joint relative position comparison result; Based on the joint relative position comparison result, the similarity between the joint relative position information of each dancer in the tested group and the standard joint relative position information of each dancer in the standard group dance group is determined to obtain the joint relative position similarity coefficient AX; According to the overall picture similarity coefficient AZ, joint angle feature similarity coefficient AJ and joint relative position similarity coefficient AX, based on the posture feature relationship function The posture feature similarity coefficient W1 is calculated, where a1, a2, and a3 are proportional factors and are all greater than 0.

7. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 6 is characterized in that: The step of determining the alignment between the dance movements of the group to be tested and the music rhythm during the group dance to obtain the group temporal rhythm feature matching coefficient W2 according to the standard group dance video information of the standard group dance dancer group in the standard group dance database, the human posture key point coordinate information and the human skeleton information of each dancer in the group to be tested, specifically includes: According to the standard group dance video information of the standard group dance dancer group in the standard group dance database, the dance movements of the standard group dance dancer group corresponding to different music rhythms in the group dance process are judged to obtain standard rhythm dance matching information; Based on the human posture key point coordinate information and human skeleton information of each dancer in the group to be tested, the dance movements corresponding to different music rhythms of each dancer in the group dance process are judged to obtain multiple rhythm dance matching feature information; Comparing the multiple rhythm dance matching feature information with the standard rhythm dance matching information to obtain the individual time rhythm matching coefficient of each dancer in the test group; Detecting the standing position of each dancer in the group to be tested during the group dance to obtain the actor's standing position information, and judging the importance of each dancer's standing position based on the actor's standing position information to obtain the standing position influence weight ratio p of each dancer; The individual time rhythm matching coefficient of each dancer in the test group is multiplied by the position influence weight ratio p of each dancer to obtain the individual time rhythm characteristic matching coefficient of each dancer in the test group; The individual time rhythm feature matching coefficients of each dancer in the tested group are averaged to obtain the alignment between the dance movements and the music rhythm of the tested group during the group dance, that is, the group time rhythm feature matching coefficient W2.

8. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 7 is characterized in that: The steps of determining the similarity between the motion trajectory of the group to be tested and the motion trajectory of the standard group dance dancer group during the group dance process based on the standard group dance video information and the human body bounding box of the standard group dance dancer group in the standard group dance database, and obtaining the group motion trajectory feature matching coefficient W3 specifically include: According to the standard group dance video information of the standard group dance dancer group in the standard group dance database, the motion trajectory of each dancer in the standard group dance dancer group is determined and a standard motion trajectory diagram is drawn; Positioning and tracking the motion trajectory of the human body bounding box corresponding to each dancer in the group to be measured based on the automatic tracking camera to obtain actor motion trajectory information, and drawing the motion trajectory of each dancer in the group to be measured based on the actor motion trajectory information to obtain a target motion trajectory diagram; Detecting the standing positions of standard group dance dancers in a standard group dance database to obtain dancer standing position information; Based on the dancer position information and the actor position information, each dancer in the standard group dance dancer group is matched with each dancer in the test group one by one to obtain personnel matching information; Based on the personnel matching information, the standard motion trajectory diagram and the target motion trajectory diagram, the similarity between the motion trajectory of each dancer in the test group and the motion trajectory of the dancers in the corresponding standard group dance group is determined to obtain the individual trajectory similarity coefficient; The individual trajectory similarity coefficient of each dancer in the test group is multiplied by the position influence weight ratio p of each dancer to obtain the individual motion trajectory feature matching coefficient of each dancer in the test group; The individual motion trajectory feature matching coefficients of each dancer in the tested group are averaged and the similarity between the motion trajectory of the tested group during group dance and the motion trajectory of the standard group dance dancers during group dance is determined to obtain the group motion trajectory feature matching coefficient W3.

9. The group dance intelligent scoring method based on posture recognition and tracking matching according to claim 8 is characterized in that: According to the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, the sequence scoring coefficient W of the group to be tested is obtained. The steps of scoring the group to be tested according to the sequence scoring coefficient W of the group to be tested specifically include: According to the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, based on the scoring conversion function Calculate and obtain the sequence scoring coefficient W of the population to be tested, where: 、 、 is the proportional factor and is greater than 0; The group dance performance of the group to be tested is scored according to the sequence scoring coefficient W of the group to be tested, wherein the larger the sequence scoring coefficient W of the group to be tested is, the higher the group dance performance score of the group to be tested is.

10. The group dance intelligent scoring system based on posture recognition and tracking matching is characterized by: The group dance intelligent scoring system based on posture recognition and tracking matching is used to implement the group dance intelligent scoring method based on posture recognition and tracking matching described in any one of claims 1 to 9, comprising: A human posture feature extraction module is configured to capture the group to be tested in real time to obtain captured image information, extract the human posture key points of each dancer in the group to be tested based on the captured image information to obtain human posture key point coordinate information, and obtain human skeleton information based on the human posture key point coordinate information; The character tracking judgment module is configured to judge whether the tracking of each dancer in the group to be tested is accurate based on the coordinate information of the key points of human posture and human skeleton information, and output the accurate character tracking result if it is accurate; The posture feature analysis module is configured to analyze the posture features of each dancer in the group dance process according to the human posture key point coordinate information and human skeleton information of each dancer to obtain posture feature parameter information; a standard data acquisition module configured to create a standard group dance database based on a standard group dance video of a standard group dance dancer group, extract human posture key points based on the standard group dance database to obtain standard human posture key point coordinate information and standard human skeleton information, determine whether the tracking of each dancer in the standard group dance dancer group is accurate, and analyze the posture characteristics of each dancer to obtain standard posture feature parameter information; The posture feature similarity analysis module is configured to compare the posture feature parameter information with the standard posture feature parameter information, determine the similarity between the posture features of the group to be tested and the standard group of dancers, and obtain the posture feature similarity coefficient W1; The time rhythm matching analysis module is configured to determine the alignment between the dance movements of the group to be tested and the music rhythm during the group dance process based on the standard group dance video information of the standard group dance dancer group in the standard group dance database, the human posture key point coordinate information and human skeleton information of each dancer in the group to be tested, and obtain the group time rhythm feature matching coefficient W2; The motion trajectory similarity analysis module is configured to determine the similarity between the motion trajectory of the group to be tested during the group dance and the motion trajectory of the standard group dance dancer group during the group dance based on the standard group dance video information and the human body bounding box of the standard group dance dancer group in the standard group dance database, and obtain the group motion trajectory feature matching coefficient W3; The group dance scoring module is configured to obtain the sequence scoring coefficient W of the group to be tested based on the posture feature similarity coefficient W1, the group time rhythm feature matching coefficient W2 and the group motion trajectory feature matching coefficient W3, and perform a scoring operation on the group to be tested based on the sequence scoring coefficient W of the group to be tested.

Images