Three-dimensional reconstruction analysis system for item operation trajectory based on skill operation video

Abstract

The application discloses a three-dimensional reconstruction analysis system of an article operation track based on skill operation video, relates to the technical field of three-dimensional reconstruction analysis, and solves the technical problem that the track curve analysis of the limbs of an operator cannot be performed, and the accuracy of track analysis is reduced during curvature monitoring due to the inaccurate calculation of curvature in the prior art, specifically, a behavior marking unit performs operator behavior recognition marking on a video selected part to obtain limb point position coordinates; an article marking unit performs operator corresponding execution article recognition marking on the video selected part to obtain article coordinates; a language marking unit acquires speech from the selected video to obtain a movement trigger word; a two-dimensional data conversion processing unit synchronously analyzes and compares the limb point position coordinates, the article coordinates and the movement trigger word, and models the operation process of the operator in stages; and a three-dimensional data analysis unit analyzes data of the current three-dimensional model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional reconstruction analysis, in particular to a three-dimensional reconstruction analysis system for operation trajectory of an object based on skill operation video. BACKGROUND

[0002] The three-dimensional reconstruction analysis system realizes the conversion and analysis from two-dimensional video to three-dimensional space trajectory by digital modeling of the object motion trajectory in the skill operation video through computer vision, three-dimensional reconstruction and trajectory tracking technology.

[0003] However, in the prior art, the operation video cannot be recognized and analyzed from multiple angles, effective data cannot be provided for three-dimensional model reconstruction, three-dimensional data processing cannot be performed through two-dimensional data, after the three-dimensional model is constructed, the trajectory curve of the operator's limbs cannot be analyzed, the curvature cannot be accurately calculated, the trajectory analysis accuracy is reduced during curvature monitoring, in addition, the deviation of the object trajectory curve cannot be analyzed, and the supervision efficiency of the three-dimensional model on the object operation trajectory is reduced.

[0004] In view of the above technical defects, a solution is proposed. SUMMARY

[0005] The purpose of the present application is to solve the above-mentioned problems, and a three-dimensional reconstruction analysis system for operation trajectory of an object based on skill operation video is proposed.

[0006] The purpose of the present application can be achieved by the following technical solutions:

[0007] The three-dimensional reconstruction analysis system for operation trajectory of an object based on skill operation video comprises a three-dimensional reconstruction analysis platform, the three-dimensional reconstruction analysis platform statistically stores the skill operation video, and analyzes the skill operation video when the real-time statistical storage amount reaches a set storage threshold, and sends the selected part of the skill operation video to an identification and marking module:

[0008] After receiving, the behavior marking unit, the language marking unit and the object marking unit analyze the video selected part;

[0009] The behavior marking unit identifies and marks the operator behavior of the video selected part to obtain the limb point coordinate;

[0010] The object marking unit identifies and marks the object corresponding to the operator of the video selected part to obtain the object coordinate;

[0011] The language marking unit acquires the voice of the selected video, and identifies the semantics of the acquired voice to obtain the moving trigger word;

[0012] The two-dimensional data conversion processing unit synchronously analyzes and compares the limb point coordinate, the article coordinate and the movement trigger word, and models the operation process of the operator in stages;

[0013] The three-dimensional data analysis unit analyzes the current three-dimensional model.

[0014] As a preferred embodiment of the present application, in the selected video, the contour of the operator in the video is collected, and the limb part of the operator is marked according to the current operation type, and the marked point is in the display contour of the operator in the video; the position is marked according to the marked point in the display contour and the two-dimensional coordinate is set with the set origin.

[0015] As a preferred embodiment of the present application, the real-time coordinate position of the article is determined by combining the real-time coordinate analysis of the limb marked point of the operator, and the same set origin is taken as the position standard, when the operator does not touch the article, the real-time coordinate position of the article is marked as the pre-placed position; and when the operator touches the article, the current time is marked as the article contact time, and the article coordinate position without movement after the article contact time is marked as the preparation time and the execution time, and the article coordinate position of the execution time is recorded in real time.

[0016] As a preferred embodiment of the present application, according to the movement process of the limb point coordinate and the article coordinate, the language of the selected video is collected, when the limb point left or the article coordinate, the coordinate after real-time movement is taken as the original longitudinal coordinate or horizontal coordinate as the movement direction, the collected voice in the current selected video is analyzed, and the keyword of the direction type is identified, if the movement direction is consistent with the direction keyword appearing in the current operation scene, the corresponding keyword is marked as a two-dimensional movement trigger word; otherwise, it is not marked; in addition to the coordinate after real-time movement divided by the original longitudinal coordinate or horizontal coordinate as the movement direction, the movement direction dimension is increased, if the movement direction is consistent with the direction keyword appearing in the current operation scene, the corresponding keyword is marked as a three-dimensional movement trigger word.

[0017] As a preferred embodiment of the present application, the limb point coordinate and the article coordinate are compared, when they are not overlapped, the current limb point and the article point are set as the starting point and the transfer point, the two-dimensional model of the starting stage is obtained, and the two-dimensional model evolution is carried out according to the movement trigger word combined with the change of the limb point coordinate.

[0018] When the new dimension does not appear, the current two-dimensional model records the point-to-point trajectory, and after the new dimension is added, the two-dimensional model evolves into a three-dimensional model, and the three-dimensional model records the current surface-to-surface trajectory; during the model evolution, the decision position of the moving trigger word is consistent with the coordinate moving position, which indicates that the model construction is consistent with the actual operation; otherwise, the decision position of the moving trigger word is not consistent with the coordinate moving position, which indicates that the operator performs abnormally or the operation expression is abnormal, and timely tracing is performed at the current time to adjust the operation.

[0019] As a preferred embodiment of the present application, the process of the three-dimensional data analysis unit is as follows:

[0020] The algorithm process of the trajectory curvature is as follows: the curvature is a quantity for describing the bending degree of a curve, and for a parameterized curve , wherein s is an arc length parameter, and the curvature K is defined as , wherein is a unit tangent vector;

[0021] According to the three-dimensional model, the limb point position coordinates are marked as discrete trajectory points, and in actual discrete trajectory point calculation, , i=0, 1, …, N-1;

[0022] The curvature is calculated as follows:

[0023] , wherein Δs is an arc length step; represents the tangent vector of the curve at the i-th point, reflecting the trend of the curve at the point; represents the position vector of the i+1-th point in the sequence of discrete points of the curve; is the position vector of the i-1-th point in the sequence of discrete points of the curve, which also contains the spatial position information of the corresponding point; i is the index of the discrete point, used to identify the i-th point in the sequence of discrete points of the curve, and by changing the value of i, the tangent vector at different positions of the curve can be calculated;

[0024] , represents the curvature vector of the curve at the i-th point;

[0025] is the tangent vector of the curve at the i+1-th point;

[0026] is the tangent vector of the curve at the i-1-th point.

[0027] As a preferred embodiment of the present application, in the three-dimensional model, the curve constructed by the trajectory is set in a two-dimensional environment, and in the two-dimensional environment, the curvature is: ; K i represents the curvature of the curve at the i-th point. is the tangent vector In the two-dimensional plane, the coordinate components correspond to the components of the x-axis and y-axis directions respectively. is the curvature vector In the two-dimensional plane, the coordinate components correspond to the components of the x-axis and y-axis directions.

[0028] As a preferred embodiment of the present application, after obtaining the curvature, the coordinates of each limb point position constructed in the three-dimensional model are subjected to curvature analysis, if the current curvature is within the set current range, and the point position coordinates increase the curvature which is still constant within the range, the floating span of the curvature value in this scenario is used as the judgment parameter of the limb point position trajectory, if the floating appears, it indicates that the execution operation of the operator exists trajectory deviation, otherwise, if the floating does not appear, it indicates that the execution operation of the operator is normal.

[0029] Compared with the prior art, the present application has the following beneficial effects:

[0030] 1、In the present application, the operator behavior recognition mark is performed on the video selected part, the operator is marked by behavior recognition, the video is analyzed, the behavior change process of the operator is recognized, so as to provide a reference standard for three-dimensional reconstruction analysis; the operator corresponding execution article recognition mark is performed on the video selected part, on the basis of the operator behavior mark, the behavior coordinates of the article are marked in combination with the coordinate position of the behavior point, so as to be able to combine the behavior coordinate position for data analysis, provide reference data for the three-dimensional reconstruction analysis process, be able to combine the inference current operation efficiency of the operator, so as to facilitate the detection of the operation process; the voice is acquired from the selected video, and the acquired voice is subjected to semantic recognition, so as to mark the language of the operator, be able to combine the language feedback to identify the operation of the operator, so as to be able to predict the current operation according to the language recognition, be able to recognize the language action synchronously, improve the accuracy of the three-dimensional modeling of the article operation trajectory.

[0031] 2、In the present application, the data analysis is performed on the current three-dimensional model, whether the execution process of the operator in the current selected video exists abnormity is inferred through three-dimensional model data analysis, the space path complexity is analyzed according to the evolution of the three-dimensional model, the trajectory curvature and repeated operation consistency of the space path are evaluated through algorithm, the intelligent detection accuracy of the three-dimensional model is improved, so as to facilitate the execution trajectory recognition of the operator.

[0032] 3、The three-dimensional model in the application, the trajectory intelligent identification of the operator in the three-dimensional model is carried out, the real-time coordinate points are counted and the two-dimensional trajectory curve is constructed in the detection process of the three-dimensional model through the trajectory intelligent identification, the curve is intelligently identified, the operation process of the operator is detected and analyzed according to the identification of the two-dimensional trajectory curve of the article in the three-dimensional model. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to facilitate those skilled in the art to understand, the application will be further described below in conjunction with the drawings.

[0034] Figure 1 It is the principle block diagram of example one in the application;

[0035] Figure 2 It is the principle block diagram of example two in the application. DETAILED DESCRIPTION

[0036] In order to make the person in the art better understand the application scheme, the technical scheme in the embodiment of the application will be described clearly and completely below in conjunction with the drawings in the embodiment of the application. Obviously, the described embodiment is only a part of the embodiment of the application, not all. Based on the embodiment in the application, all other embodiments obtained by the person skilled in the art without creative labor belong to the protection scope of the application.

[0037] In this paper, the phrase "example" means that the specific features, structures or characteristics described in conjunction with the example can be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment to other embodiments. The person skilled in the art explicitly and implicitly understands that the embodiments described herein can be combined with other embodiments.

[0038] Example one

[0039] Please refer to Figure 1 As shown, the three-dimensional reconstruction analysis system of the article operation trajectory based on skill operation video includes a three-dimensional reconstruction analysis platform, wherein the three-dimensional reconstruction analysis platform is communicatively connected with an identification mark module, and the identification mark module is communicatively connected with a behavior marking unit, a language marking unit and an article marking unit;

[0040] Meanwhile, the three-dimensional reconstruction analysis platform is communicatively connected with a three-dimensional data processing unit and a three-dimensional data conversion unit;

[0041] The three-dimensional reconstruction analysis platform statistically stores the skill operation video, and analyzes the skill operation video when the real-time statistical storage amount reaches the set storage threshold, and sends the selected part of the skill operation video to the identification mark module.

[0042] After the recognition mark module receives, the behavior mark unit, the language mark unit and the article mark unit perform video analysis on the video excerpt part;

[0043] The behavior mark unit marks the operator's behavior by recognizing the operator's behavior and analyzing the video to identify the operator's behavior change process, thereby providing a reference standard for three-dimensional reconstruction analysis;

[0044] In the video excerpt, the operator's contour in the video is collected, and the body part of the operator is marked according to the current operation type, and the marked point is in the display contour of the operator in the video; the position of the marked point in the display contour is marked and the two-dimensional coordinate is set with a set origin;

[0045] The article mark unit marks the article executed by the operator according to the behavior coordinate of the article based on the behavior mark of the operator and the coordinate position of the behavior point, so as to analyze the data combined with the behavior coordinate position, provide reference data for the three-dimensional reconstruction analysis process, and infer the article execution efficiency of the operator to facilitate the detection of the operation process;

[0046] The real-time coordinate position of the marked point of the operator's body is analyzed to determine the real-time coordinate of the article, and the same set origin is used as the position standard. When the operator does not touch the article, the real-time coordinate position of the article is marked as the pre-placed position. When the operator touches the article, the current time is marked as the article contact time, and the article coordinate position without movement after the article contact time is marked as the preparation time and the execution time, respectively. The article coordinate position of the execution time is recorded in real time. It should be explained that the body point coordinate of the operator and the article coordinate are the same origin;

[0047] The language mark unit acquires the voice of the video excerpt and performs semantic recognition on the acquired voice, thereby marking the language of the operator. The operation of the operator can be recognized combined with the language feedback, so as to predict the current operation according to the language recognition, and the language action can be recognized synchronously, thereby improving the accuracy of the three-dimensional modeling of the article operation track;

[0048] According to the moving process of the limb point coordinate and the article coordinate, the language of the selected video is collected by voice, when the limb point coordinate or the article coordinate, the corresponding coordinate after real-time movement is taken as the moving direction of the original longitudinal coordinate or the horizontal coordinate, the collected voice in the current selected video is analyzed, and the keywords of the direction type are identified, if the moving direction is consistent with the direction keyword appearing in the current operation scene, the corresponding keyword is marked as a two-dimensional moving trigger word; otherwise, it is not marked; in addition to the moving direction of the corresponding coordinate after real-time movement divided by the original longitudinal coordinate or the horizontal coordinate, the dimension of the moving direction is increased, if the moving direction is consistent with the direction keyword appearing in the current operation scene, the corresponding keyword is marked as a three-dimensional moving trigger word;

[0049] The limb point coordinate, the article coordinate and the moving trigger word are sent to the three-dimensional reconstruction analysis platform;

[0050] After receiving, the three-dimensional reconstruction analysis platform generates a two-dimensional data conversion processing signal and sends it to the two-dimensional data conversion processing unit;

[0051] The two-dimensional data conversion processing unit synchronously analyzes and compares the limb point coordinate, the article coordinate and the moving trigger word, and models the operation process of the operator in stages;

[0052] The limb point coordinate and the article coordinate are compared, and when they are not overlapped, the current limb point and the article point are set as the starting point and the transfer point to obtain a two-dimensional model in the starting stage, and the two-dimensional model is evolved according to the change of the limb point coordinate combined with the moving trigger word, when no new dimension appears, the current two-dimensional model is recorded as a point-to-point trajectory, then after the new dimension appears, the two-dimensional model evolves into a three-dimensional model, and the three-dimensional model is recorded as a current surface-to-surface trajectory; during the model evolution, if the decision direction of the moving trigger word is consistent with the coordinate moving direction, it indicates that the model construction is consistent with the actual operation; otherwise, if the decision direction of the moving trigger word is not consistent with the coordinate moving direction, it indicates that the operator performs abnormally or the operation expression is abnormal, and timely tracing is performed at the current time to adjust the operation;

[0053] After completion, a three-dimensional data analysis signal is generated and sent to a three-dimensional data analysis unit;

[0054] After receiving, the three-dimensional data analysis unit analyzes the data of the current three-dimensional model, infers whether the execution process of the operator in the current selected video is abnormal through three-dimensional model data analysis, analyzes the spatial path complexity according to the evolution of the three-dimensional model, evaluates the trajectory curvature and repeated operation consistency of the spatial path through an algorithm, improves the intelligent detection accuracy of the three-dimensional model, and facilitates the execution trajectory recognition of the operator;

[0055] Algorithmic process of trajectory curvature: curvature is a quantity describing the bending degree of a curve, for a parameterized curve where s is the arc length parameter, and the curvature K is defined as where, is the unit tangent vector;

[0056] According to the three-dimensional model, the coordinates of the limb points are marked as discrete trajectory points. In the actual calculation of discrete trajectory points, , i = 0, 1, …, N-1;

[0057] The curvature calculation process is as follows:

[0058] where Δs is the arc length step; represents the tangent vector (the first derivative corresponding to the vector) of the curve at the i-th point, which describes the tangent direction of the curve at the point and the "rate" of the change of the tangent direction with the arc length (combined with the arc length step), reflecting the trend of the curve at the point; represents the position vector of the i+1-th point in the sequence of discrete points of the curve, which contains the position information of the point in space (such as the (x, y, z) coordinates in three-dimensional space); is the position vector of the i-1-th point in the sequence of discrete points of the curve, which also contains the spatial position information of the corresponding point; i is the index of the discrete point, used to identify the i-th point in the sequence of discrete points of the curve, and by changing the value of i, the tangent vector at different positions of the curve can be calculated;

[0059] , represents the curvature vector (the second derivative corresponding to the vector) of the curve at the i-th point, which is related to the bending degree of the curve, and its modulus and other characteristics can be used to calculate the curvature and other quantities describing the bending characteristics of the curve, reflecting the change of the tangent vector direction with the arc length;

[0060] is the tangent vector of the curve at the i+1-th point (calculated from the tangent vector formula above), which contains the tangent direction of the point and other information;

[0061] is the tangent vector of the curve at the i-1-th point;

[0062] In the three-dimensional model, the curve constructed by the trajectory is set in a two-dimensional environment. In the two-dimensional environment, the curvature is: ; K i represents the curvature of the curve at the i-th point; is the tangent vector The coordinate components of the two-dimensional plane correspond to the x-axis and y-axis directions, respectively;

[0063] is the curvature vector In the two-dimensional plane, the coordinate components correspond to the components of the x-axis and y-axis directions.

[0064] After obtaining the curvature, the coordinates of each limb point position constructed in the three-dimensional model are subjected to curvature analysis. If the current curvature is within the set current range, and the point position coordinates increase the curvature which is still constant within the range, the floating span of the curvature value in this scenario is taken as the judgment parameter of the limb point position trajectory. If there is a floating, it indicates that the operator's execution operation has a trajectory deviation. Conversely, if there is no floating, it indicates that the operator's execution operation is normal.

[0065] Embodiment Two

[0066] Please refer to Figure 2 On the basis of the previous embodiment, the operation trajectory for completing three-dimensional data analysis is intelligently identified. The three-dimensional reconstruction analysis platform generates a trajectory intelligent identification signal and sends it to the trajectory intelligent identification unit.

[0067] After receiving, the trajectory intelligent identification unit intelligently identifies the trajectory of the operator in the three-dimensional model. Through trajectory intelligent identification, a two-dimensional trajectory curve is constructed by statistically analyzing the real-time coordinate points in the detection process of the three-dimensional model, and the curve is intelligently identified. According to the identification of the two-dimensional trajectory curve of the object in the three-dimensional model, the operation process of the operator is detected and analyzed.

[0068] The curvature of the two-dimensional trajectory curve of the object in the three-dimensional model is marked according to the corresponding adjacent coordinate points. According to the curvature, the trend of each two-dimensional trajectory curve of the object is obtained, and the coordinate of the point where the corresponding object is located in the two-dimensional trajectory curve of the object is obtained. When the curve continues to be generated, the distance deviation value of the non-moving dimension in the corresponding three-dimensional coordinate system is obtained, and the floating span of the corresponding ratio of the distance of the moving dimension in the three-dimensional coordinate system to the actual distance is obtained, and the distance deviation value of the non-moving dimension in the corresponding three-dimensional coordinate system and the floating span of the corresponding ratio of the distance of the moving dimension in the three-dimensional coordinate system to the actual distance are analyzed. The non-moving dimension represents the dimension that does not produce displacement in the three-dimensional model.

[0069] If the distance deviation value of the non-moving dimension in the corresponding three-dimensional coordinate system exceeds the distance deviation threshold, or the floating span of the corresponding ratio of the distance of the moving dimension in the three-dimensional coordinate system to the actual distance exceeds the floating span threshold, it is inferred that the two-dimensional trajectory curve of the object in the three-dimensional model is abnormally identified, an object operation deviation signal is generated and sent to the three-dimensional reconstruction analysis platform, and is forwarded to the operator terminal, and the real-time operation is adjusted.

[0070] If the distance deviation value of the non-moving dimension in the three-dimensional coordinate system does not exceed the distance deviation threshold value, and the floating span of the distance corresponding to the coordinate system of the moving dimension in the three-dimensional coordinate system to the actual distance corresponding ratio does not exceed the floating span threshold value, it is inferred that the two-dimensional trajectory curve recognition of the object in the three-dimensional model is normal, and an object operation accuracy signal is generated and sent to the three-dimensional reconstruction analysis platform.

[0071] In use, the three-dimensional reconstruction analysis platform statistically stores the skill operation video, and analyzes the skill operation video when the real-time statistical storage amount reaches the set storage threshold value, and sends the selected part of the skill operation video to the recognition marking module: after receiving, the behavior marking unit, the language marking unit and the object marking unit perform video analysis on the video selected part; the behavior marking unit performs operation personnel behavior recognition marking on the video selected part to obtain the body point coordinate; the object marking unit performs operation personnel corresponding execution object recognition marking on the video selected part to obtain the object coordinate; the language marking unit performs voice acquisition on the selected video, and performs semantic recognition on the acquired voice to obtain the movement trigger word; the two-dimensional data conversion processing unit synchronously analyzes and compares the body point coordinate, the object coordinate and the movement trigger word, and models the operation process of the operation personnel in stages; the three-dimensional data analysis unit performs data analysis on the current three-dimensional model.

[0072] The threshold value or the preset value, the preset range and the like are set for result comparison and analysis, so as to determine whether it is good or bad, and the size value thereof is set by combining large model analysis of sample data and artificial experience, and is also adjusted appropriately through seasonal or common influence conditions;

[0073] And the setting of the weight proportion coefficient, the influence factor and the like is to allocate specific numerical values to finally reflect the influence on the result according to the influence size of each parameter on the result, and is also set by combining large model analysis of sample data and artificial experience, and is also adjusted appropriately through seasonal or common influence conditions.

[0074] The preferred embodiments disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details, and the application is not limited to the specific embodiments. Obviously, many modifications and changes can be made according to the content of the present application. The embodiments are selected and described in detail in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited by the claims and their entire scope and equivalents.

Claims

1. A three-dimensional reconstruction analysis system of an article operation trajectory based on a skill operation video, characterized by, The three-dimensional reconstruction analysis platform statistically stores the skill operation video, and analyzes the skill operation video when the real-time statistical storage reaches a set storage threshold, and sends the selected part of the skill operation video to the identification marking module: After receiving, the behavior marking unit, the language marking unit and the article marking unit perform video analysis on the video selected part; The behavior marking unit identifies and marks the operator's behavior in the video selected part to obtain the limb point coordinates; The article marking unit identifies and marks the corresponding executed articles of the operator in the video selected part to obtain the article coordinates; The language marking unit acquires the voice of the selected video, and performs semantic recognition on the acquired voice to obtain the movement trigger word; The two-dimensional data conversion processing unit synchronously analyzes and compares the limb point coordinates, the article coordinates and the movement trigger word, and models the operation process of the operator in stages; In the selected video, the outline of the operator in the video is collected, and the limb parts of the operator are marked according to the current operation type, and the marked points are in the display outline of the operator in the video; The position of the marked point in the display outline is marked and the two-dimensional coordinate of the set origin is set; Combined with the real-time coordinate position analysis of the limb marking point of the operator, the real-time coordinate of the article is determined, and the same set origin is taken as the position standard. When the operator does not touch the article, the real-time coordinate position of the article is marked as the pre-placed position; When the operator touches the article, the current time is marked as the article contact time, and the article coordinate position after the article contact time is marked as the preparation time and the execution time when there is no movement and there is movement, respectively. At the same time, the article coordinate position at the execution time is recorded in real time; According to the movement process of the limb point coordinates and the article coordinates, the language of the selected video is collected, and when the limb point left or the article coordinate, the coordinate after real-time movement is taken as the original longitudinal coordinate or transverse coordinate as the movement direction, the collected voice in the current selected video is analyzed, and the keywords of the direction type are identified. If the movement direction is consistent with the direction keyword appearing in the current operation scene, the corresponding keyword is marked as a two-dimensional movement trigger word; Otherwise, it is not marked; In addition to the movement direction of the coordinate after real-time movement divided by the original longitudinal coordinate or transverse coordinate, the movement direction dimension is increased. If the movement direction is consistent with the direction keyword appearing in the current operation scene, the corresponding keyword is marked as a three-dimensional movement trigger word; Compare the limb point coordinates and the article coordinates. When they are not overlapped, set the current limb point and the article point as the starting point and the transfer point to obtain a two-dimensional model of the starting stage, and evolve the two-dimensional model according to the movement trigger word combined with the change of the limb point coordinates. When the new dimension does not appear, the current two-dimensional model records the point-to-point trajectory, and after the new dimension is added, the two-dimensional model evolves into a three-dimensional model, and the three-dimensional model records the current surface-to-surface trajectory; during the model evolution, if the decision position of the movement trigger word is consistent with the coordinate movement position, it indicates that the model construction is consistent with the actual operation; On the contrary, if the decision position of the movement trigger word is not consistent with the coordinate movement position, it indicates that the operator performs abnormally or the operation expression is abnormal, and timely tracing is performed at the current time to adjust the operation; The three-dimensional data analysis unit analyzes the data of the current three-dimensional model.

2. The three-dimensional reconstruction analysis system of an object operation trajectory based on a skill operation video according to claim 1, characterized by, The process of the three-dimensional data analysis unit is as follows: Algorithmic process for trajectory curvature: Curvature is a quantity that describes the degree of bending of a curve, for a parametrized curve where s is the arc length parameter, the curvature K is defined as where, is the unit tangent vector; According to the three-dimensional model, the limb point position coordinates are marked as discrete trajectory points, and in actual discrete trajectory point calculation, , i = 0, 1, …, N-1; The curvature is calculated as follows: where Δs is the arc length step; represents the tangent vector of the curve at the i-th point, reflecting the trend of the curve at the point; represents the position vector of the i+1-th point in the sequence of discrete points of the curve; is the position vector of the i-1-th point in the sequence of discrete points of the curve, also containing the spatial position information of the corresponding point; i is the index of the discrete point, used to identify the i-th point in the sequence of discrete points of the curve, and by changing the value of i, the tangent vector at different points of the curve can be calculated; , denotes the curvature vector of the curve at the i-th point; is the tangent vector of the curve at the i+1 point; is the tangent vector of the curve at the i-1th point. 3.The three-dimensional reconstruction analysis system of an article operation trajectory based on a skill operation video according to claim 2, wherein a curve constructed by the trajectory is set in a two-dimensional environment in the three-dimensional model, and in the two-dimensional environment, the curvature is: K = (x′ (t) y″ (t) - y′ (t) x″ (t) ) / (x′ (t) 2+y′ (t) 2) 3 / 2. ;K i denotes a curvature of the curve at the i th point; is a tangent vector The coordinate components of the two-dimensional plane correspond to the components of the x-axis and y-axis directions, respectively. is a curvature vector The coordinate components of the two-dimensional plane correspond to the components of the x-axis and y-axis directions.

4. The three-dimensional reconstruction analysis system of the article operation trajectory based on the skill operation video according to claim 3, after obtaining the curvature, the coordinates of each limb point position constructed in the three-dimensional model are subjected to curvature analysis, if the current curvature is within the set current range, and the point coordinate increases the curvature and is still constant within the range, the floating span of the curvature value in this scene is used as the judgment parameter of the limb point position trajectory, if floating appears, it indicates that the operator's operation exists trajectory deviation; on the contrary, if no floating appears, it indicates that the operator's operation is normal.

Images