A skipping counting detection method based on laser radar region recognition

Abstract

The application discloses a skipping counting detection method based on laser radar area recognition, and relates to the technical field of motion monitoring, which comprises the following steps: continuously scanning each effective detection ROI area to obtain original point cloud data, and acquiring time sequence point cloud data through preprocessing; training a multiple hidden Markov model based on the time sequence point cloud data, and independently modeling the state of a rope in parallel, and solving the optimal state sequence of each rope through a Viterbi algorithm; based on the optimal state sequence, applying Bayesian filtering to dynamically optimize cycle parameters, and acquiring a complete counting sequence through interpolation counting; performing event counting statistics on the complete counting sequence, aligning through a time window, mapping the effective detection ROI area, and outputting the independent counting value of each person. Through the parallel independent modeling mechanism of the multiple hidden Markov model, the application realizes accurate separation and recognition of a multi-person skipping scene; through the Viterbi algorithm, the optimal state sequence is solved, and the recognition accuracy and stability are maintained.

Description

Technical Field

[0001] This invention relates to the field of motion monitoring technology, and in particular to a jump rope counting detection method based on lidar area recognition. Background Technology

[0002] In recent years, motion monitoring technology has seen continuous development in the field of rope skipping counting, with various sensing solutions and algorithms constantly being innovated. Early methods primarily relied on inertial measurement to capture acceleration and angular velocity data to estimate the number of rope skips, and were widely used due to their low cost and ease of deployment. With technological advancements, cameras are now used to capture motion images, combined with computer vision algorithms to extract trajectory features, exhibiting good spatial resolution under suitable lighting conditions. LiDAR technology, due to its high-precision ranging capabilities and anti-interference characteristics, has been introduced into the field of motion monitoring in recent years. It generates three-dimensional point cloud data by emitting and receiving laser beams, providing richer feature information for motion analysis. With the development of multi-sensor fusion and intelligent algorithms, rope skipping counting technology is evolving towards higher accuracy and greater adaptability.

[0003] Existing rope skipping counting technologies face two key limitations in practical applications. Firstly, methods based on inertial measurement units (IMUs) are susceptible to motion noise and environmental interference, leading to data drift or loss during high-speed or variable-speed rope skipping, thus reducing counting reliability. Secondly, visual solutions are highly sensitive to lighting conditions and struggle to handle occlusion and overlap issues in multi-rope interlacing scenarios, limiting their practicality and accuracy in complex environments. These limitations restrict the effective application of existing technologies in group rope skipping or competitive scenarios. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a rope skipping counting detection method based on lidar area recognition to solve the problems of inaccurate multi-target separation and insufficient counting reliability in complex environments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a rope skipping counting detection method based on lidar area recognition. The method includes: fixing a lidar in the rope skipping area; dividing the area into multiple effective detection areas (ROIs) by adjusting the intersection relationship between the scanning plane and the movement trajectories of multiple ropes; continuously scanning each effective ROI to obtain raw point cloud data, and obtaining temporal point cloud data through preprocessing; training a multi-hidden Markov model based on the temporal point cloud data, and modeling the rope states independently and in parallel, solving for the optimal state sequence of each rope using the Viterbi algorithm; dynamically optimizing the period parameters based on the optimal state sequence using Bayesian filtering, and obtaining a complete counting sequence through interpolation counting; performing event counting statistics on the complete counting sequence, and outputting an independent count value for each person through time window alignment and effective ROI mapping.

[0008] As a preferred embodiment of the jump rope counting and detection method based on lidar region recognition described in this invention, the step of fixing the lidar in the jump rope area and dividing multiple effective detection ROI regions by adjusting the intersection relationship between the scanning plane and the movement trajectories of multiple ropes is as follows:

[0009] By identifying and analyzing the spatial layout of the jump rope area, a lidar was fixed and its installation parameters were obtained.

[0010] Based on the radar installation parameters, laser scanning is initiated to capture the rope's movement trajectory in real time and generate the spatial distribution characteristics of each rope's trajectory.

[0011] Analyze the spatial distribution characteristics of the trajectory and dynamically adjust the configuration parameters of the scanning plane to obtain the intersection relationship between the plane and the trajectory;

[0012] Based on the intersection relationship between the plane and the trajectory, a spatial segmentation algorithm is used to divide multiple independent detection regions and generate effective detection ROI regions.

[0013] As a preferred embodiment of the jump rope counting detection method based on lidar region recognition described in this invention, the step of continuously scanning each effective detection ROI region to obtain the original point cloud data refers to obtaining the original point cloud data by dynamically adjusting the scanning frequency and sampling density of each region based on the effective detection ROI region.

[0014] As a preferred embodiment of the jump rope counting and detection method based on lidar region recognition described in this invention, the original point cloud data includes spatial coordinate data, reflection intensity information and timestamp data;

[0015] The preprocessing includes filtering, temporal alignment, and multi-dimensional data fusion.

[0016] As a preferred embodiment of the jump rope counting and detection method based on lidar region recognition described in this invention, the specific steps for training a multi-hidden Markov model based on temporal point cloud data are as follows:

[0017] Based on time-series point cloud data, an independent observation sequence for each rope is extracted using a multi-target separation algorithm, and a dataset of independent observation sequences for the rope is obtained by multi-feature fusion.

[0018] Based on an independent observation sequence dataset, a parameter initialization algorithm is used to independently set the initial probability, transition matrix and emission matrix for each Hidden Markov Model, thereby obtaining the initial model parameter set.

[0019] The model parameters are updated by calculating the detection confidence based on the initial model parameter set, and the trained multi-hidden Markov model is obtained.

[0020] As a preferred embodiment of the jump rope counting and detection method based on lidar region recognition described in this invention, the steps for parallel and independent modeling of the rope state and solving for the optimal state sequence of each rope using the Viterbi algorithm are as follows:

[0021] Based on the trained multi-hidden Markov model and the independent observation sequence of each rope, the path indication matrix is ​​obtained by calculating the maximum log probability;

[0022] Extract the sequence endpoint from the path indication matrix and obtain the globally optimal termination state by comparing the maximum log probability of each state;

[0023] Based on the path indication matrix and the global optimal termination state, the complete optimal path is solved in reverse, and the optimal state sequence is obtained by analyzing the spatiotemporal continuity characteristics of the state sequence.

[0024] As a preferred embodiment of the jump rope counting and detection method based on lidar region recognition described in this invention, the step of dynamically optimizing the period parameters based on the optimal state sequence using Bayesian filtering involves the following specific steps.

[0025] Extract state transition time points from the optimal state sequence and extract the time interval data for each valid state transition to obtain the state transition time interval sequence;

[0026] Based on the state transition time interval sequence, the Bayesian filtering method is applied to dynamically monitor the changing trend of the period estimation parameters and obtain the period parameters.

[0027] As a preferred embodiment of the jump rope counting detection method based on lidar area recognition described in this invention, the step of obtaining a complete counting sequence by interpolation counting refers to obtaining a complete counting sequence by detecting missed scans and inserting compensation counts based on the optimal state sequence and period parameters.

[0028] As a preferred embodiment of the jump rope counting and detection method based on lidar region recognition described in this invention, the specific steps for performing event counting statistics on the complete counting sequence, through time window alignment and effective ROI region mapping, are as follows:

[0029] Based on the complete counting sequence, valid rope skipping events are identified by analyzing time regularity, and valid event sequences are obtained.

[0030] Within the valid event sequence, time windows are aligned across the event sequences of multiple ropes to obtain a time-synchronized event sequence.

[0031] The time-synchronized event sequence is associated with the effective detection ROI region, and personal associated event data is obtained by analyzing the spatial distribution of motion features.

[0032] As a preferred embodiment of the jump rope counting detection method based on lidar area recognition described in this invention, the output of each person's independent count value refers to the separate statistical analysis of individual related event data in each valid detection ROI area to generate each person's independent count value.

[0033] The beneficial effects of this invention are as follows: by using the parallel and independent modeling mechanism of multiple hidden Markov models, the precise separation and recognition of the motion states of each rope in a multi-person rope skipping scenario is achieved; by using the Viterbi algorithm to solve for the optimal state sequence, it ensures that extremely high recognition accuracy and stability can still be maintained under complex motion modes. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart of a rope skipping counting detection method based on lidar area recognition.

[0036] Figure 2 A flowchart for generating valid RIO regions.

[0037] Figure 3 A flowchart for obtaining the optimal state sequence.

[0038] Figure 4 A flowchart for generating independent count values. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0042] Reference Figures 1-4 This is one embodiment of the present invention, which provides a jump rope counting and detection method based on lidar area recognition, including the following steps:

[0043] S1: Fix a lidar in the jump rope area, and divide multiple effective detection ROI areas by adjusting the intersection relationship between the scanning plane and the movement trajectory of multiple ropes.

[0044] S1.1: By identifying and analyzing the spatial layout of the jump rope area, fix the lidar and obtain the lidar installation parameters;

[0045] It should be noted that lidar is used to scan and locate the coordinates of the human body area, obtain the cluster center and distribution range of the spatially distributed location points, and record the height parameters and horizontal deflection angle parameters of the lidar to form complete lidar installation parameters.

[0046] S1.2: Based on the radar installation parameters, initiate laser scanning to capture the movement trajectory of the rope in real time and generate the spatial distribution characteristics of the trajectory of each rope.

[0047] It should be noted that, based on the radar installation parameters, the lidar scanning is initiated; the angle sampling interval is controlled according to the horizontal deflection angle parameters to perform periodic laser scanning. During the laser scanning process, the lidar emits a laser beam and receives reflected signals, converting the received reflected signals into distance and angle measurements to obtain the coordinate position of each scan point in three-dimensional space, forming a set of scan point coordinates.

[0048] Motion point identification is performed on the set of scan point coordinates. By comparing the set of scan point coordinates in consecutive scan frames, motion scan points whose positions change are identified. These motion scan points are organized according to a time sequence to form a motion scan point sequence. By connecting the motion scan points in adjacent time frames within the motion scan point sequence, a complete motion trajectory is constructed. Morphological features (e.g., spatial curvature features, velocity change features, periodicity features, and distribution density features) are extracted from the trajectory points in the rope's motion trajectory. Trajectory segments that conform to these morphological features are then identified. These trajectory segments representing the motion characteristics of each rope are categorized and organized according to their spatial distribution characteristics, generating data records containing the spatial distribution characteristics of each rope's motion trajectory. These records include the spatial coordinate range of the motion trajectory, a description of the morphological features of the motion trajectory, and spatiotemporal characteristic parameters of the motion trajectory, forming a complete spatial distribution feature of each rope's trajectory.

[0049] Furthermore, rope motion characteristics refer to the unique physical and statistical features exhibited by the rope in space during rope skipping that can be captured and quantified by lidar. The rope's trajectory is a regular, periodic waveform in space; the projected trajectory of the rope on the lidar scanning plane has a relatively fixed shape and spatial scale.

[0050] S1.3: Analyze the spatial distribution characteristics of the trajectory and dynamically adjust the scanning plane configuration parameters to obtain the intersection relationship between the plane and the trajectory;

[0051] It should be noted that, based on the spatial distribution characteristics of each rope trajectory, the spatial coordinate range parameters and centerline position parameters of the motion trajectory are extracted to determine the area to be covered by the scanning plane, and the ideal intersection position between the scanning plane and each motion trajectory is determined by combining the centerline position parameters.

[0052] Based on the spatial distribution range of the motion trajectory, the centerline position of the motion trajectory, the fluctuation amplitude of the motion trajectory, and the periodic characteristics of the motion trajectory, adjust the azimuth and elevation components in the scanning plane configuration parameters so that the scanning plane and each motion trajectory reach the ideal intersection position; record the final value of the scanning plane configuration parameters at this time to obtain the intersection relationship between the plane and the trajectory.

[0053] S1.4: Based on the intersection relationship between the plane and the trajectory, a spatial segmentation algorithm is used to divide multiple independent detection regions and generate effective detection ROI regions.

[0054] Specifically, the intersection relationship between the plane and the trajectory includes the set of intersection point coordinates and the intersection area range parameter. The intersection point coordinate set contains the spatial location information of all intersection points between the plane and the motion trajectory; the intersection area range parameter defines the spatial area range that needs to be covered.

[0055] It should be noted that the extracted set of intersection point coordinates is used as input data for the spatial segmentation algorithm. Based on the point location information in the set of intersection point coordinates, an initial region division centered on the intersection point coordinates is generated. Each initial region contains all spatial locations closest to the corresponding intersection point.

[0056] The initial region division is adjusted based on the intersection area range parameters. By expanding or shrinking the region boundaries, it is ensured that each divided region fully covers the corresponding trajectory movement range, while maintaining clear boundaries between regions.

[0057] The adjusted region divisions are validated to ensure their effectiveness. Each region is checked to ensure it fully encompasses the trajectory of a single rope and that there is no overlap between regions. All parameters of the ROI regions are effectively detected, including geometric boundary coordinates, area size, and spatial relationships, forming a complete and valid ROI detection framework.

[0058] S2: Continuously scan each valid detection ROI region to obtain raw point cloud data, and obtain time-series point cloud data through preprocessing.

[0059] S2.1: Based on the effective detection of the ROI region, the original point cloud data is obtained by dynamically adjusting the scanning frequency and sampling density of each region;

[0060] It should be noted that the same scanning frequency and sampling density parameters are applied to all valid detection ROI areas. For example, the scanning frequency parameter is set to a fixed value of 40Hz, and the sampling density parameter is set to 84,000 scan points per square meter.

[0061] The lidar scanning function is activated, and the scanning parameters perform periodic scanning operations on each effective detection ROI area. During the scanning process, the lidar emits a laser beam and receives the reflected signal, recording the measurement data for each scan point.

[0062] The collected scan point data includes spatial coordinate data, reflection intensity information, and timestamp information. Spatial coordinate data records the position coordinates of the scan point in three-dimensional space, reflection intensity information records the intensity value of the laser echo, and timestamp data records the acquisition time of the scan point. All scan point data are organized in chronological order to form a complete raw point cloud dataset.

[0063] S2.2: The raw point cloud data includes spatial coordinate data, reflection intensity information, and timestamp data;

[0064] Specifically, spatial coordinate data constitutes the spatial skeleton of the point cloud, which is used to accurately describe the motion trajectory and position changes of the rope in space, providing basic spatial information for subsequent motion trajectory analysis and state recognition.

[0065] The reflection intensity information reflects the physical characteristics of the surface where the scanning point is located. It can be used to distinguish objects of different materials, help identify the differences between ropes and other objects, and assist in data filtering and feature enhancement in the preprocessing stage.

[0066] Timestamp data ensures the temporal order and continuity of the data, providing a time reference for constructing time-series point cloud data. This enables the point cloud data to accurately reflect the motion process and temporal evolution of the rope, providing time dimension support for subsequent time-series analysis and state sequence solutions.

[0067] S2.3: Preprocessing includes filtering, time-series alignment, and multi-dimensional data fusion;

[0068] Specifically, the spatial coordinate data is smoothed by filtering to eliminate outliers caused by environmental interference and measurement errors; the reflection intensity information is threshold filtered to remove noise data caused by stray reflections, and the spatial coordinates and reflection intensity values ​​are obtained.

[0069] All scan points are rearranged according to time sequence based on timestamp data to establish a strict time order; interpolation processing is performed on spatial coordinate data in the time dimension to fill the time gaps that may occur due to scanning frequency limitations; time synchronization calibration is performed on reflection intensity information to obtain the timestamp of spatial coordinate data in the time dimension.

[0070] By integrating the filtered and time-aligned spatial coordinate data, reflection intensity information, and timestamp data in three dimensions, a complete spatial-intensity-time mapping relationship is established to generate time-series point cloud data containing spatial coordinates, reflection intensity values, and precise timestamps.

[0071] S2.4: Temporal point cloud data includes spatial coordinates, reflection intensity values, and timestamps.

[0072] Specifically, the spatial coordinate data records the precise position information of each scan point in three-dimensional space after preprocessing. This spatial coordinate data, after filtering and time alignment, can accurately describe the continuous trajectory of the rope's motion, providing the spatial position features required for state recognition of the multi-hidden Markov model, and is used to identify the rope's motion state and spatial change patterns.

[0073] The reflection intensity value reflects the reflection characteristics of the object surface after preprocessing. The reflection intensity value after threshold filtering and multi-dimensional data fusion can effectively distinguish the material differences between the rope and other objects, provide auxiliary feature information for state recognition, and enhance the recognition robustness of the multi-hidden Markov model in complex environments.

[0074] Timestamps provide precise time information after time-series alignment. Maintaining a strict time order ensures the temporal continuity of point cloud data, provides a time reference for the Viterbi algorithm to solve for the optimal state sequence, and provides time dimension support for Bayesian filtering to dynamically optimize periodic parameters.

[0075] S3: Train a multi-hidden Markov model based on time-series point cloud data, and model the state of the ropes in parallel and independently. Solve for the optimal state sequence of each rope using the Viterbi algorithm.

[0076] S3.1: Based on time-series point cloud data, the independent observation sequence of each rope is extracted through a multi-objective separation algorithm, and the independent observation sequence dataset of the rope is obtained by multi-feature fusion.

[0077] It should be noted that scan points with similar spatial distances in spatial coordinates are grouped into the same category, thus dividing the temporal point cloud data into multiple independent categories, each corresponding to the motion trajectory of a rope. Based on timestamps, the scan points within each category are arranged in chronological order, forming a preliminary independent observation sequence for each rope.

[0078] A multi-feature fusion method is used to process the initial independent observation sequence of each rope. The spatial coordinates, reflection intensity values ​​and timestamps are mapped to a unified feature space through multi-layer feature transformation. The spatial coordinates, reflection intensity values ​​and timestamps in the unified feature space are integrated into a unified feature representation to generate a comprehensive feature vector.

[0079] By integrating the comprehensive feature vectors of all ropes, a dataset of independent observation sequences of ropes containing complete spatiotemporal features is formed.

[0080] S3.2: Based on the independent observation sequence dataset, a parameter initialization algorithm is used to independently calculate the initial probability, transition matrix and emission matrix for each Hidden Markov Model to obtain the initial model parameter set;

[0081] Specifically, the statistical characteristics of the independent observation sequence dataset for each rope include the distribution of state duration, the frequency of state transitions, and the range of observation distribution.

[0082] It should be noted that the initial probability distribution is obtained by statistically analyzing the frequency of occurrence of each state in the observation sequence, specifically the frequency of occurrence of the initial state for each rope's observation sequence; the number of transitions between states is counted to obtain the state transition probability matrix; and the distribution characteristics of the observed values ​​under each state are analyzed to obtain the observation emission probability matrix. By processing the observation sequence data of all ropes, an independent set of initialization model parameters is generated.

[0083] S3.3: Update the model parameters by calculating the detection confidence based on the initial model parameter set to obtain the trained multi-hidden Markov model;

[0084] It should be noted that the variance and covariance matrices of the state transition probability matrix and the observation emission probability matrix are extracted. Based on these matrices, the step size and convergence threshold are adjusted. The gradient descent optimization algorithm is used to iteratively adjust the parameters, initializing the hyperparameters of the algorithm, including the learning rate, regularization coefficient, and momentum factor.

[0085] During parameter optimization, the detection confidence level under the current model parameters is calculated. The parameter update step size is dynamically adjusted based on the detection confidence level: the step size is increased when the detection confidence is low and decreased when the detection confidence is high, thus obtaining the trained multi-hidden Markov model. The expression for calculating the detection confidence level is:

[0086] ;

[0087] in, To test the confidence level, The length of the observation sequence, It is the natural logarithm function. The number of hidden states. The index value represents the number of hidden states. Indicates the first The joint probability of each hidden state.

[0088] S3.4: Based on the trained multi-hidden Markov model and the independent observation sequence of each rope, the path indication matrix is ​​obtained by calculating the maximum log probability;

[0089] It should be noted that, by extracting the observations from the independent observation sequence of each rope, and using the initial probabilities in the trained multi-hidden Markov model, the log probability of the observations for each state is calculated, expressed as:

[0090] ;

[0091] in, To observe the logarithmic probability, It is the natural logarithm function. The variance of the observed values, The first observation in an independent observation sequence. The mean of the observed values, Let be the initial state probability. It is an exponential function.

[0092] For adjacent time steps from the beginning of the previous time step to the end of the independent observation sequence, obtain the maximum log probability of the observation log probabilities of all states transitioning from the previous time step to the current state, and record the state index corresponding to the maximum log probability in the path indicator matrix.

[0093] Specifically, the probability matrix contains the maximum log probability value for each time step and state, and the path indication matrix contains the index of the optimal predecessor state for each time step and state.

[0094] S3.5: Extract the sequence endpoint from the path indication matrix and obtain the globally optimal termination state by comparing the maximum log probability of each state;

[0095] It should be noted that all state indices at the last time step are extracted from the path indication matrix; these indices represent the predecessor state information of each state at the end of the sequence. The maximum logarithmic probability values ​​of each state at the last time step are extracted from the probability matrix; these probability values ​​reflect the likelihood of different states being the end of the sequence.

[0096] The state node with the highest logarithmic probability is identified by comparing the maximum logarithmic probability values ​​of each state in the last time step of the probability matrix. The state node with the highest probability is the termination position of the sequence with the highest probability among the independent observation sequences of the rope. The index of this state node is recorded as the globally optimal termination state.

[0097] S3.6: Based on the path indication matrix and the global optimal termination state, the complete optimal path is solved in reverse, and the optimal state sequence is obtained by analyzing the spatiotemporal continuity characteristics of the state sequence.

[0098] It should be noted that the state index corresponding to the globally optimal termination state is extracted from the path indication matrix as the backtracking starting point. Based on the predecessor state index information recorded in the path indication matrix, backtracking is performed step by step from the end of the sequence towards the starting point, and the optimal state selection at each time step is obtained in turn.

[0099] During backtracking, the optimal state path is derived step by step from the preceding time steps based on the state transition relationships stored in the path indicator matrix (e.g., the proportion of transitions from the cordless state to the corded state to the total number of transitions). For each time step, the optimal state selection for the current time step is determined using the state index recorded at the corresponding position in the path indicator matrix, and the state index is added to the state sequence to obtain the optimal state sequence.

[0100] S4: Based on the optimal state sequence, apply Bayesian filtering to dynamically optimize the periodic parameters, and obtain the complete counting sequence through interpolation counting.

[0101] S4.1: Extract state transition time points from the optimal state sequence and extract the time interval data for each valid state transition to obtain the state transition time interval sequence;

[0102] It should be noted that the state change patterns in the optimal state sequence are extracted, and all state transition events are identified, such as the transition from cordless state to corded state and the transition from corded state to cordless state. For the identified state transition events, the timestamp information of the occurrence is extracted to form the state transition time point.

[0103] The timestamps at the state transition points are sorted sequentially to ensure that the time points are arranged in the order of occurrence; the time difference between adjacent state transition points is obtained as the state transition time interval, and the state transition time intervals of all state transition events are integrated to generate a time interval sequence.

[0104] S4.2: Based on the state transition time interval sequence, apply Bayesian filtering to dynamically monitor the changing trend of the period estimation parameters and obtain the period parameters;

[0105] It should be noted that, based on the state transition time interval sequence, the total number of state transition time interval sequence data is obtained by statistically analyzing the total value of the state transition time interval; the ratio of the total value of the state transition time interval to the total number of state transition time interval sequence data is used as the initial period estimation parameter.

[0106] Based on the initial period estimation parameters, the sum of the observed time interval data and the period estimation parameters is used as the period estimation parameters for the next time interval, and the period estimation parameters are adjusted accordingly. By monitoring the changing trend of the period estimation parameters in the state transition time interval sequence, the parameter update intensity is dynamically adjusted; when a change in rope skipping speed is detected, the parameter update intensity is increased to quickly adapt to the change, and when the speed is stable, the update intensity is decreased to maintain stability, and the period estimation parameters are output.

[0107] S4.3: Based on the optimal state sequence and periodic parameters, a complete counting sequence is obtained by detecting missed scans and inserting compensation counts.

[0108] It should be noted that all state transition events from the cordless state to the corded state and from the corded state to the cordless state are extracted from the optimal state sequence. The proportion of all state transition events from the cordless state to the corded state and from the corded state to the cordless state is recorded as the state transition relationship in the optimal state sequence. By comparing the transition time points of all transition events, the actual state jump time interval sequence is obtained.

[0109] State transition time intervals that differ from the actual state transition time interval sequence are identified as abnormal time intervals and marked as potential missed scan locations. Missing state transition events are added to the potential missed scan locations in the optimal state sequence to generate a complete count sequence containing the original count and the compensation count.

[0110] S5: Perform event counting statistics on the complete counting sequence, and output the independent count value for each person by aligning the time window and effectively detecting ROI region mapping.

[0111] S5.1: Based on the complete counting sequence, valid rope skipping events are identified by analyzing the time regularity, and a valid event sequence is obtained;

[0112] It should be noted that the timestamp of each state transition event is extracted from the complete counting sequence, the time interval between adjacent state transition events is obtained, and the time interval sequence is obtained by integrating the time intervals between all consecutive state transition events in the complete counting sequence.

[0113] By analyzing each time interval and period parameter in the time interval sequence, events that coincide with the period parameter are identified as part of the rope skipping motion cycle, thus forming a valid event sequence.

[0114] S5.2: In the valid event sequence, align the event sequences of multiple ropes with time windows to obtain the time-synchronized event sequence;

[0115] It should be noted that timestamps of all events are extracted from the valid event sequence, and the period parameter is used as a fixed-length time window. The length of the time window is an integer multiple of the period parameter to ensure that each window covers the complete rope skipping cycle.

[0116] For each rope, a time window is applied to the valid event sequence, segmenting the sequence according to the time window to ensure that each valid event is assigned to its corresponding time window based on its timestamp. Within each time window, time alignment is performed on all ropes' valid event sequences, adjusting the timestamps to synchronize the events across different ropes. Finally, the aligned valid events from all time windows are integrated to form a time-synchronized event sequence.

[0117] S5.3: Associate time-synchronized event sequences with effective detection ROI regions, and obtain individual associated event data by analyzing the spatial distribution of motion features;

[0118] It should be noted that, based on the time synchronization event sequence and the spatial coordinate data of the effective detection ROI region, each time synchronization event in the time synchronization event sequence is matched according to the spatial coordinate data of the ROI region; the spatial distribution characteristics of the time synchronization events in each effective detection ROI region are analyzed, such as the cluster center location and distribution density pattern of the event points, to generate the personal related event data corresponding to each effective detection ROI region.

[0119] S5.4: Collect individual related event data for each valid detection ROI region and generate an independent count value for each person.

[0120] It should be noted that all valid detection ROI regions are traversed. For each valid detection ROI region, the personal related event data associated with the valid detection ROI region is extracted, and the total number of events recorded in the personal related event data is counted. The total number of events obtained is used as the count value of the valid detection ROI region, thereby generating an independent count value for each person.

[0121] This embodiment also provides a computer device applicable to the rope skipping counting detection method based on lidar area recognition, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the rope skipping counting detection method based on lidar area recognition as proposed in the above embodiment.

[0122] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0123] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the jump rope counting detection method based on lidar area recognition as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0124] In summary, this invention achieves accurate separation and recognition of the motion states of each rope in a multi-person rope skipping scenario through a parallel and independent modeling mechanism using multiple hidden Markov models; and ensures extremely high recognition accuracy and stability even under complex motion modes by solving for the optimal state sequence using the Viterbi algorithm.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A jump rope counting and detection method based on lidar area recognition, characterized in that: include, A lidar is fixed in the rope skipping area, and multiple effective detection ROI areas are divided by adjusting the intersection relationship between the scanning plane and the movement trajectory of multiple ropes; Raw point cloud data is obtained by continuously scanning each effective detection ROI region, and time-series point cloud data is obtained through preprocessing. A multi-hidden Markov model is trained based on time-series point cloud data, and the states of the ropes are modeled in parallel and independently. The optimal state sequence of each rope is solved by the Viterbi algorithm. Based on the optimal state sequence, Bayesian filtering is applied to dynamically optimize the periodic parameters, and the complete counting sequence is obtained by interpolation counting. Perform event counting statistics on the complete counting sequence, and output the independent count value for each person by aligning the time window and effectively detecting ROI region mapping.

2. The jump rope counting and detection method based on lidar region recognition as described in claim 1, characterized in that: The process involves fixing a lidar in the jump rope area and dividing the area into multiple effective detection ROI regions by adjusting the intersection relationship between the scanning plane and the movement trajectories of multiple ropes. The specific steps are as follows: By identifying and analyzing the spatial layout of the jump rope area, a lidar was fixed and its installation parameters were obtained. Based on the radar installation parameters, laser scanning is initiated to capture the rope's movement trajectory in real time and generate the spatial distribution characteristics of each rope's trajectory. Analyze the spatial distribution characteristics of the trajectory and dynamically adjust the configuration parameters of the scanning plane to obtain the intersection relationship between the plane and the trajectory; Based on the intersection relationship between the plane and the trajectory, a spatial segmentation algorithm is used to divide multiple independent detection regions and generate effective detection ROI regions.

3. The jump rope counting and detection method based on lidar area recognition as described in claim 2, characterized in that: The phrase "continuously scanning in each effective detection ROI region to obtain raw point cloud data" refers to obtaining raw point cloud data based on the effective detection ROI region by dynamically adjusting the scanning frequency and sampling density of each region.

4. The jump rope counting and detection method based on lidar region recognition as described in claim 3, characterized in that: The raw point cloud data includes spatial coordinate data, reflection intensity information, and timestamp data; The preprocessing includes filtering, temporal alignment, and multi-dimensional data fusion.

5. The jump rope counting and detection method based on lidar area recognition as described in claim 4, characterized in that: The specific steps for training a multi-hidden Markov model based on temporal point cloud data are as follows. Based on time-series point cloud data, an independent observation sequence for each rope is extracted using a multi-target separation algorithm, and a dataset of independent observation sequences for the rope is obtained by multi-feature fusion. Based on an independent observation sequence dataset, a parameter initialization algorithm is used to independently set the initial probability, transition matrix and emission matrix for each Hidden Markov Model, thereby obtaining the initial model parameter set. The model parameters are updated by calculating the detection confidence based on the initial model parameter set, and the trained multi-hidden Markov model is obtained.

6. The jump rope counting and detection method based on lidar area recognition as described in claim 5, characterized in that: The parallel and independent modeling of the rope states, and the solution for the optimal state sequence of each rope using the Viterbi algorithm, are detailed below. Based on the trained multi-hidden Markov model and the independent observation sequence of each rope, the path indication matrix is ​​obtained by calculating the maximum log probability; Extract the sequence endpoint from the path indication matrix and obtain the globally optimal termination state by comparing the maximum log probability of each state; Based on the path indication matrix and the global optimal termination state, the complete optimal path is solved in reverse, and the optimal state sequence is obtained by analyzing the spatiotemporal continuity characteristics of the state sequence.

7. The jump rope counting and detection method based on lidar region recognition as described in claim 6, characterized in that: The specific steps for dynamically optimizing the periodic parameters based on the optimal state sequence using Bayesian filtering are as follows. Extract state transition time points from the optimal state sequence and extract the time interval data for each valid state transition to obtain the state transition time interval sequence; Based on the state transition time interval sequence, the Bayesian filtering method is applied to dynamically monitor the changing trend of the period estimation parameters and obtain the period parameters.

8. The jump rope counting and detection method based on lidar region recognition as described in claim 7, characterized in that: The method of obtaining a complete counting sequence by interpolation count refers to obtaining a complete counting sequence based on the optimal state sequence and period parameters, by detecting missed scan phenomena and inserting compensation counts.

9. The jump rope counting and detection method based on lidar region recognition as described in claim 8, characterized in that: The specific steps for performing event counting statistics on the complete counting sequence, through time window alignment and effective ROI region mapping, are as follows: Based on the complete counting sequence, valid rope skipping events are identified by analyzing time regularity, and valid event sequences are obtained. Within the valid event sequence, time windows are aligned across the event sequences of multiple ropes to obtain a time-synchronized event sequence. The time-synchronized event sequence is associated with the effective detection ROI region, and personal associated event data is obtained by analyzing the spatial distribution of motion features.

10. The jump rope counting and detection method based on lidar region recognition as described in claim 9, characterized in that: The output of each person's independent count value refers to the separate statistical analysis of individual related event data in each valid detection ROI region to generate an independent count value for each person.

Images