Parachute posture recognition method based on simulation platform

Abstract

The present invention relates to the field of posture recognition, and in particular to a parachute posture recognition method based on a simulation platform. The method comprises: obtaining a template sequence and the position data of each data point of a parachutist during a parachuting process and constructing a position sequence; matching the pre-acquired template sequence and position sequence using a DTW algorithm; performing weighted adjustment on the distances between matching point pairs to obtain adjusted distances and constructing a distance sequence; segmenting the distance sequence to obtain multiple subsequences, calculating the mean of the adjusted distances in the subsequences, multiplying the mean by a preset correlation coefficient as the optimal distance in the subsequence, and further obtaining an updated distance sequence; and obtaining the shortest path between the template sequence and the position sequence based on the updated distance sequence. When the length of the shortest path is greater than a preset threshold, the parachuting posture is deemed unqualified. The present invention has the effect of improving the accuracy of parachutist posture assessment.

Description

Technical Field

[0001] The present invention relates to the field of posture recognition, and in particular to a parachute posture recognition method based on a simulation platform. Background Art

[0002] With the development of virtual reality technology, based on the simulation of parachute operation with electromechanical equipment, virtual reality equipment is combined to build a skydiving scene, providing parachutes with realistic airborne scenes, and providing a new method for skydiving training. It can effectively shorten the training cycle, reduce training consumption and safety risks, and achieve significant training benefits.

[0003] The Chinese patent application document with publication number CN113947810A discloses a Tai Chi evaluation method and system based on posture recognition. The method includes the following steps: obtaining a Tai Chi video to be evaluated, extracting the posture joint skeleton diagram of the person to be evaluated through a deep learning method, and generating a first joint point sequence diagram; based on a dynamic time adjustment algorithm, by assigning different evaluation weight values ​​to different joint points, the first joint point sequence diagram is compared with the standard joint point sequence diagram for similarity to obtain an evaluation result; wherein, the evaluation weight value of the limb joints is higher than that of the trunk joints and the head joints.

[0004] During the parachutist training process, the parachutist's posture needs to be identified and evaluated to determine whether the parachutist's movements are standard. Existing parachutist posture recognition methods usually involve obtaining the parachutist's human posture during parachuting (i.e., the position information of each joint), establishing a feature template database of human motion, and calculating the similarity between the human posture during parachuting and the posture template to determine whether the parachutist's movements meet the standard.

[0005] The DTW (Dynamic Time Warping) algorithm is usually used to calculate the similarity between human posture and posture templates. However, the DTW algorithm only considers the position information of each part in the action process during the calculation process. Skydiving also involves rotation or tilting movements. Therefore, considering only position information leads to low accuracy of the final recognition result. Summary of the Invention

[0006] In order to solve the problem of low accuracy of calculation results when using the DTW algorithm to calculate the similarity between human posture and posture template, the present invention provides a parachute posture recognition method based on a simulation platform.

[0007] The present invention provides a parachute posture recognition method based on a simulation platform, which adopts the following technical solutions:

[0008] Obtain the template sequence and the position data of each data point of the skydiver during the skydiving process and construct a position sequence;

[0009] Use the DTW algorithm to match the pre-acquired template sequence and position sequence to obtain matching point pairs and the initial distance between the matching point pairs;

[0010] Perform weighted adjustment on the distance between matching point pairs to obtain the adjusted distance, and use the adjusted distance to construct a distance sequence;

[0011] The distance sequence is segmented to obtain multiple subsequences, the mean of the adjusted distances in the subsequences is calculated, and the product of the mean and the preset correlation coefficient is used as the optimal distance in the subsequence to further obtain an updated distance sequence;

[0012] The shortest path between the template sequence and the position sequence is obtained according to the updated distance sequence. When the length of the shortest path is greater than a preset threshold, the parachuting posture is unqualified.

[0013] By updating the distance between matching points to obtain the updated distance, and using the updated distance to obtain the length of the shortest path, the influence of the skydiver's movement rhythm on the matching degree is comprehensively considered. The matching degree between the skydiver's real-time movement and the template movement can be accurately evaluated, thereby improving the accuracy of the skydiver's posture assessment.

[0014] Preferably, the expression for adjusting the distance is:

[0015] ;

[0016] Where, Represents the adjusted distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Represents the distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Indicates the number of data points in the position sequence that match the nth data point in the template sequence.

[0017] The Euclidean distance between matching point pairs is weighted and adjusted according to the number of matching points, which improves the accuracy of the distance calculation results.

[0018] Preferably, the expression for adjusting the distance is:

[0019] ;

[0020] Where, Represents the adjusted distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Indicates the maximum distance between the nth data point in the position sequence and the matching point pair in the template sequence.

[0021] Preferably, the correlation coefficient is calculated as follows: obtaining the real-time quaternary data corresponding to each position data of the skydiver during the skydiving process and the standard quaternary data corresponding to the corresponding standard position data in the template sequence; calculating the angle difference between the real-time quaternary data and the standard quaternary data in the matching point pairs; and taking the average of the angle differences of the matching points in the subsequence as the correlation coefficient.

[0022] Preferably, the correlation coefficient is calculated by obtaining the real-time quaternary data corresponding to each position data of the skydiver during the skydiving process and the standard quaternary data corresponding to the corresponding standard position data in the template sequence, and calculating the angle difference between the real-time quaternary data and the standard quaternary data in the matching point pairs. The expression of the correlation coefficient is:

[0023] ;

[0024] Where, represents the correlation coefficient of the corresponding subsequences, represents the mean of the angle differences of the matching point pairs corresponding to the adjusted distance in the j-th subsequence, and exp represents the exponential function with e as the base.

[0025] By calculating the angle difference between the matching point pairs, the angle difference reflects the difference between the skydiver's actual skydiving action and the standard skydiving action, thereby further understanding the accuracy of the skydiver's skydiving action.

[0026] Preferably, the method of segmenting the distance sequence to obtain multiple subsequences is: detecting the distance sequence to obtain multiple change points, and segmenting the distance sequence using the change points as segmentation points to obtain multiple subsequences.

[0027] By segmenting the distance sequence, the corresponding adjustment distances of the matching point pairs in each step of the parachuting process can be obtained, thereby achieving a holistic analysis of each step, simplifying the evaluation results and improving computational efficiency.

[0028] Preferably, the PELT algorithm is used to detect the distance sequence to obtain multiple change points.

[0029] The PELT algorithm is used to detect the distance sequence, which improves the accuracy of the detection results.

[0030] Preferably, before using the DTW algorithm to match the pre-acquired template sequence and position sequence, the method further includes: constructing a three-dimensional space coordinate system and mapping the template sequence and position sequence into the three-dimensional coordinate system.

[0031] The present invention has the following technical effects:

[0032] 1. The Euclidean distance between matching point pairs is weighted by the number of matching points. The effect of action rhythm on matching degree is comprehensively considered. This can accurately evaluate the matching degree between the skydiver's real-time action and the template action, thereby improving the accuracy of the skydiver's posture assessment.

[0033] 2. The parachuting action is segmented and analyzed based on the mean of the adjusted distances in the subsequences. The adjusted distances are then adjusted again using the quaternion angle differences corresponding to the matching point pairs. This comprehensively considers the impact of the differences in the corresponding position and posture on the matching degree of the matching point pairs, accurately measuring the spatial posture differences of the parachutists' actions and improving the rigor and accuracy of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] By reading the detailed description below with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding numbers represent the same or corresponding parts.

[0035] Figure 1 It is a flow chart of the parachute posture recognition method based on the simulation platform of the present invention. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0037] It should be understood that when the terms "first," "second," and the like are used in the claims, description, and drawings of the present invention, they are merely used to distinguish between different objects, rather than to describe a specific order. The terms "comprise" and "comprising" used in the description and claims of the present invention indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.

[0038] The embodiment of the present invention discloses a parachute posture recognition method based on a simulation platform, referring to Figure 1 , including the following steps, as follows:

[0039] S1: Obtain the template sequence and the position data of each data point of the skydiver during the skydiving process and construct a position sequence.

[0040] Position sensors are installed at each joint of the skydiver, and IMU sensors are installed on the skydiver's upper body. The position sensors are used to collect the spatial position of the corresponding joints in real time during the skydiver's jump, while the IMU sensors are used to measure quaternary data during the jump. The quaternary data represents the skydiver's torso posture during the jump. The position sensors and IMU sensors collect data at the same frequency, so each position data point corresponds to a quaternary data point. The position data of multiple data points are used to construct a position sequence, where the position data are arranged in descending order of spatial position.

[0041] S2: Use the DTW algorithm to match the pre-acquired template sequence and position sequence to obtain matching point pairs and the initial distance between the matching point pairs.

[0042] A template sequence is constructed based on actual conditions. The template sequence represents the standard spatial position of the corresponding joints of the skydiver during the skydiving process. A three-dimensional spatial coordinate system is constructed, in which the z-axis represents height and the x-axis and y-axis represent distance. It can be understood that the three-dimensional coordinates of the data points in the three-dimensional spatial coordinate system represent the spatial position of the corresponding data points. The data in the template sequence and the data in the position sequence are mapped into the three-dimensional coordinate system. The data points in the template sequence and the data points in the position sequence are matched using the DTW algorithm to obtain multiple matching point pairs. It can be understood that a matching point pair includes a data point in the template sequence and a data point in the position sequence. At the same time, the initial distance between the two data points in each matching point pair is obtained, where the initial distance is the Euclidean distance. During the skydiving training process, the smaller the distance between the matching point pairs, the more standardized the movement of the corresponding joints, which further indicates that the skydiver's skydiving posture is more standard.

[0043] S3: Perform weighted adjustment on the distances between the matching point pairs to obtain adjusted distances, and use the adjusted distances to construct a distance sequence.

[0044] When performing key actions such as pulling the parachute and lowering the pitch angle during skydiving, accurately performing standard actions at the corresponding time points can ensure that the corresponding actions are completed smoothly. If there are time deviations or changes in the skydiver's actions, it will affect the stability and safety of the flight.

[0045] In the process of matching the template sequence with the position sequence, if multiple data points in the position sequence match one data point in the template sequence, it indicates that there is a difference between the parachutist's action at the corresponding time point and the standard action. It can also be understood that there is a difference between the parachutist's action at the corresponding spatial position and the standard action.

[0046] Therefore, it is necessary to perform weighted adjustment on the distance between the matching point pairs to obtain the adjusted distance.

[0047] In one embodiment, the expression for adjusting the distance is:

[0048] ;

[0049] Where, Represents the adjusted distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Represents the distance between the matching point pair formed by the mth data point in the position sequence and the nth data point in the template sequence. The distance here is the Euclidean distance. Represents the number of data points in the position sequence that match the nth data point in the template sequence. One matching point pair corresponds to one adjusted distance, and multiple matching point pairs correspond to multiple adjusted distances. The distance sequence is constructed using the obtained multiple adjusted distances.

[0050] For example, for the 5th data point in the template sequence , there are 3 data points in the position sequence 、 、 With data points correspond, The value of is 3, forming a matching point pair ( , ),( , ),( , ); where the matching point pair ( , )The distance between two data points is , then adjust the distance 3 , similarly, matching point pairs ( , )The distance between two data points is , then adjust the distance 3 , matching point pairs ( , )The distance between two data points is , then adjust the distance 3 .

[0051] The greater the number of data points in the position sequence that match the nth data point in the template sequence, the greater the diver's posture and movement at the corresponding spatial location deviates from the standard movement. Therefore, the distance between the corresponding matching point pairs is adjusted to reduce the matching degree. It should be noted that if a data point in the position sequence corresponds to multiple data points in the template sequence, forming multiple matching point pairs, the distance between the matching point pairs remains unchanged.

[0052] In one embodiment, the expression for adjusting the distance is:

[0053] ;

[0054] Where, Represents the adjusted distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Indicates the maximum distance between the nth data point in the position sequence and the matching point pair in the template sequence.

[0055] For example, for the 5th data point in the template sequence , there are 3 data points in the position sequence 、 、 With data points Corresponding to, forming a matching point pair ( , ),( , ),( , ); where the matching point pair ( , )The distance between two data points is , matching point pairs ( , )The distance between two data points is , matching point pairs ( , )The distance between two data points is ,exist 、 、 middle, The value of is the largest, so the matching point pair ( , ),( , ),( , ) are adjusted to .

[0056] S4: Segment the distance sequence to obtain multiple subsequences.

[0057] The method for segmenting the distance sequence is as follows: using the PELT algorithm to detect multiple change points in the distance sequence, segmenting the distance sequence using the change points as segmentation points to obtain multiple subsequences. It is understood that the data in a subsequence is the adjusted distance between two data points in the corresponding matching point pair. During the segmentation process, a change point can be assigned to the head or tail of the corresponding subsequence, or it can be discarded. There is no specific restriction on which subsequence a change point belongs to.

[0058] During a skydiving process, the action can be divided into several specific steps. For example, the action is mainly divided into takeoff, free fall, parachute deployment, and landing. The change point in the distance sequence indicates that the skydiver's skydiving step has changed at the corresponding spatial position. In each step, the accuracy requirements for the skydiver's posture vary. Therefore, the distance needs to be segmented into multiple subsequences, where each step corresponds to a subsequence. For example, the takeoff step corresponds to one subsequence, the free fall step corresponds to one subsequence, the parachute deployment step corresponds to one subsequence, and the landing step corresponds to another subsequence.

[0059] S5: Calculate the mean of the adjusted distances in the subsequence, and use the product of the mean and the preset correlation coefficient as the optimal distance in the subsequence, to further obtain an updated distance sequence.

[0060] Compute the correlation coefficient of subsequences:

[0061] In one embodiment, the correlation coefficient is calculated by obtaining real-time quaternary data corresponding to each position data of the skydiver during the skydiving process and standard quaternary data corresponding to the corresponding standard position data in the template sequence, and calculating the angular difference between the real-time quaternary data and the standard quaternary data in the matching point pairs. The correlation coefficient is expressed as:

[0062] ;

[0063] Where, represents the correlation coefficient of the corresponding subsequences, represents the mean of the angle differences of the matching point pairs corresponding to the adjusted distance in the j-th subsequence, and exp represents the exponential function with e as the base.

[0064] In one embodiment, the correlation coefficient is calculated by obtaining real-time quaternary data corresponding to each position data of the skydiver during the skydiving process and standard quaternary data corresponding to the corresponding standard position data in the template sequence; in the matching point pairs, the angular difference between the real-time quaternary data and the standard quaternary data is calculated; and the average of the angular differences of the matching points in the subsequence is used as the correlation coefficient.

[0065] The expression for the optimal distance is:

[0066] ;

[0067] Where, represents the best distance in the j-th subsequence, represents the mean of the adjusted distances in the j-th subsequence, represents the correlation coefficient of the j-th subsequence.

[0068] The correlation coefficient indicates the differences in a skydiver's body posture during corresponding jump steps. A larger correlation coefficient indicates a greater difference between the skydiver's body posture and the standard posture during the corresponding jump, further indicating a greater difference between the corresponding jump and the standard. Therefore, the corresponding adjusted distances in the subsequences are adjusted twice based on the angle difference to obtain the optimal distance, thereby updating the distance sequence. It should be understood that after the second adjustment, the adjusted distances in each subsequence are identical, facilitating a comprehensive assessment of the differences in each jump step.

[0069] S6: Obtain the shortest path between the template sequence and the position sequence according to the updated distance sequence. When the length of the shortest path is greater than a preset threshold, the parachuting posture is unqualified.

[0070] The data in the updated distance sequence is mapped to the distance matrix in the DTW algorithm. The shortest path is then derived from the distance matrix. The length of the shortest path indicates the similarity between the real-time and standard actions during the parachute jump. The shorter the shortest path length, the greater the similarity between the real-time and standard actions. Conversely, the longer the shortest path length, the less similar the real-time and standard actions are. If the shortest path length exceeds a preset threshold, it indicates that the actions performed during the parachute jump differ significantly from the standard actions, indicating that the parachute jump posture is unqualified. The threshold is set manually based on actual conditions; for example, the threshold is 30.

[0071] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

[0072] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A parachute posture recognition method based on a simulation platform, characterized in that: Including steps: Obtain the template sequence and the position data of each data point of the skydiver during the skydiving process and construct a position sequence; Use the DTW algorithm to match the pre-acquired template sequence and position sequence to obtain matching point pairs and the initial distance between the matching point pairs; Adjust the initial distance between the matching point pairs to obtain the adjusted distance, and use the adjusted distance to construct a distance sequence; The distance sequence is segmented to obtain multiple subsequences, the mean of the adjusted distances in the subsequences is calculated, and the product of the mean and the preset correlation coefficient is used as the optimal distance in the subsequence to further obtain an updated distance sequence; The shortest path between the template sequence and the position sequence is obtained based on the updated distance sequence. If the length of the shortest path is greater than a preset threshold, the parachuting posture is unqualified. The correlation coefficient is calculated by obtaining the real-time quaternary data corresponding to each position data of the skydiver during the skydiving process and the standard quaternary data corresponding to the corresponding standard position data in the template sequence. In the matching point pairs, the angular difference between the real-time quaternary data and the standard quaternary data is calculated, and the average of the angular differences of the matching points in the subsequence is used as the correlation coefficient. Alternatively, the correlation coefficient can be calculated by obtaining the real-time quaternary data corresponding to each position data of the skydiver during the skydiving process and the standard quaternary data corresponding to the corresponding standard position data in the template sequence. In the matching point pairs, the angle difference between the real-time quaternary data and the standard quaternary data is calculated. The expression of the correlation coefficient is: ； Where, represents the correlation coefficient of the corresponding subsequences, represents the mean of the angle differences of the matching point pairs corresponding to the adjusted distance in the j-th subsequence, and exp represents the exponential function with e as the base.

2. The parachute posture recognition method based on the simulation platform according to claim 1 is characterized in that: The expression for adjusting the distance is: ； Where, Represents the adjusted distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Represents the distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Indicates the number of data points in the position sequence that match the nth data point in the template sequence.

3. The parachute posture recognition method based on the simulation platform according to claim 1, characterized in that: The expression for adjusting the distance is: ； Where, Represents the adjusted distance between the matching point pair consisting of the mth data point in the position sequence and the nth data point in the template sequence, Indicates the maximum distance between the nth data point in the position sequence and the matching point pair in the template sequence.

4. The parachute posture recognition method based on the simulation platform according to claim 1, characterized in that: The method for segmenting the distance sequence to obtain multiple subsequences is as follows: performing change point detection on the distance sequence to obtain multiple change points, and segmenting the distance sequence using the change points as segmentation points to obtain multiple subsequences.

5. The parachute posture recognition method based on the simulation platform according to claim 4 is characterized in that: The PELT algorithm is used to detect change points in distance sequences.

6. The parachute posture recognition method based on a simulation platform according to claim 1, characterized in that: Before using the DTW algorithm to match the pre-acquired template sequence and position sequence, the method also includes: constructing a three-dimensional space coordinate system and mapping the template sequence and position sequence into the three-dimensional coordinate system.

Images