Apparatus for defining sequences of movements in a generic model - Patents.com

Abstract

The present invention relates to an apparatus for defining a generic motion sequence (Tx) in a generic model, comprising: - means for acquiring the position of a reference element moving over a surface, said reference element being configured to perform an actual movement sequence; - means (41) for recording the sequence (Tr) of the actual movement; - means (14) for obtaining a three-dimensional representation (Re) of said surface; - means for adapting said generic model to said three-dimensional representation (Re) of said surface; - means (42) for defining a generic motion sequence in the generic model by applying the adaptation between the generic model and the three-dimensional representation (Re) of the surface to the actual motion sequence.

Description

[Technical field]

[0001] The present invention relates to the field of defining at least one movement sequence in a generic model, i.e. in a digital model having pre-set parameters to represent a group of objects.

[0002] The present invention may be applicable in many technical fields where the working surface is variable among several individuals in a group of subjects. Typically, the present invention finds particularly advantageous application for defining one or more massage trajectories on a generic human body. Alternatively, the individuals of the group of subjects may correspond to physical objects. For example, the group of subjects may correspond to porcelain dishes, and the present invention may find application in defining trajectories for painting on these porcelain dishes.

[0003] The use of this motion sequence of the reference element in the generic model is also variable. In fact, by recording several motion sequences associated with different objects and defining them in the generic model, it is possible to compare two movements performed on different objects. Furthermore, this motion sequence defined using the generic model can also be used to control a robotic means for adapting the generic motion sequence to a specific object. [Background technology]

[0004] Unlike industrial robots, which are programmed to follow preset trajectories, robots operating on unknown surfaces must include a means of managing their motion that can analyze the surface and determine the trajectory.

[0005] For example, a reconnaissance robot typically incorporates at least one camera and image processing means that analyzes the exploration surface over time and determines the trajectory to be followed for the robot.

[0006] This method of analysing an unknown surface requires a lot of computational power to precisely guide the robot's motion over time, and it therefore follows that the exploration robot should move slowly to allow the motion management device to optimise the robot's motion according to the information acquired by the camera and processed by the image processing means.

[0007] Additionally, for a massage robot or an art porcelain painting robot, the movements of the robot must be extremely precise in order to massage the desired areas of an individual's body or apply layers of paint to the desired locations.

[0008] For this purpose, the known process of scanning three-dimensional surfaces allows the operator to program the robot's movements by using a three-dimensional digital modeling of the surface to be treated by the robot. For example, document WO2015 / 187092 describes a massage robot incorporating a three-dimensional scanner to scan the patient's body and enable the practitioner to determine the massage trajectory of the robot using the projection of a three-dimensional model of the patient's body on a touchpad.

[0009] It is also possible to use a robot for which at least one motion sequence is known for a generic model, which can then adapt the generic model to the subject, modifying the motion sequence to apply the treatment to a specific subject.

[0010] It is also necessary for this kind of robot to define at least one movement sequence in the generic model. With respect to the definition given in document WO2015 / 187092, the definition of the at least one movement sequence in the generic model is generally performed manually by an operator by drawing a plot in the generic model.

[0011] Defining this movement sequence may therefore take particularly long depending on the required precision.

[0012] The technical problem of the present invention consists of facilitating the process of defining at least one movement sequence in a generic model. [Prior art documents] [Patent documents]

[0013] [Patent Document 1] WO2015 / 187092 [Non-patent literature]

[0014] [Non-Patent Document 1] "KinectFusion: Real-time 3D Reconstruction and Interaction Using a Moving Depth Camera*" UIST'11, October 16, 2011 [Non-Patent Document 2] "Building Statistical Shape Spaces for 3D Human Modeling," Pishchulin et al., March 19, 2015 in the journal Pattern Recognition 2017. [Non-Patent Document 3] Suzuki, S. and Abe, K., Topological Structural Analysis of Digitized Binary Images by Border Following, CVGIP 30 1, pp. 32-46 (1985) [Non-Patent Document 4] Sklansky, J., Finding the Convex Hull of a Simple Polygon. PRL 1 $number, pp. 79-83 (1982) [Non-Patent Document 5] Visual Pattern Recognition by Moment Invariants, IRE Transactions on Information Theory, 8:2, pp. 179-187 (1962) Summary of the Invention [Problem to be solved by the invention]

[0015] The present invention aims to meet this technical problem by acquiring the motion of a reference element evolving on a surface of variable geometry and by fitting the captured motion onto a generic model. [Means for solving the problem]

[0016] In the sense of the present invention, a "reference element" may correspond to a point or a set of reference points corresponding to an effector or physical element. For example, a reference element may correspond to a glove or a set of points representing the practitioner's hand. Two methods may be used to adapt the real motion to the generic model.

[0017] The first method consists of transforming the generic model to fit the object before acquiring the motion sequence. The acquisition of the motion sequence is then recorded directly on the transformed generic model. Once the motion sequence is completed, the generic model is transformed again to regain these initial parameters, and the motion sequence is then transformed by regaining the same transformation that was applied to the generic model so that the real motion sequence is transformed into the generic motion sequence.

[0018] The second method is to obtain the motion sequence independent of the generic model, calculate the difference between the object and the generic model, and apply this difference to the motion sequence to convert the real motion sequence to the generic motion sequence. This second method still requires a correct repositioning of the generic motion sequence in the generic model.

[0019] This second method is only possible for translation, rotation and scaling type transformations, whereas the first method can be applied to other kinds of transformations. In either case, the invention provides an apparatus for defining a generic movement sequence in a generic model, comprising: - means for acquiring the position of a reference element moved over a surface, said reference element being configured to perform an actual sequence of movements; - means for recording the sequence of the actual movement; - means for obtaining a three-dimensional representation of said surface; - means for adapting said generic model to said three-dimensional representation of said surface; - means for defining a generic motion sequence in the generic model by applying the adaptation between the generic model and the three-dimensional representation of the surface to the actual motion sequence.

[0020] In the first case, the adaptation means is configured to adapt the generic model to the three-dimensional representation of the surface, the recording means is configured to record a sequence of the actual movements in the generic model while the generic model is adapted to the dimensions of the three-dimensional representation of the surface, and the definition means is configured to transform the generic model in which the movement sequence has been recorded so that the generic model recovers these initial parameters.

[0021] In the second case, the adaptation means is configured to calculate the difference between the generic model and the three-dimensional representation of the surface, the recording means is configured to record the actual movement sequence independently of the generic model, the definition means is configured to transform the recorded movement sequence according to the difference calculated by the adaptation means, and the apparatus includes means for positioning the generic movement sequence in the generic model.

[0022] Whatever the adaptation method used, the present invention makes it possible to define a movement sequence in a generic model without the need to actually visualize a three-dimensional representation of the surface or the generic model.

[0023] It is therefore possible to capture several movement sequences and fit them in the same generic model in order to compare several trajectories performed on said same surface or on several different surfaces.

[0024] In the example of art porcelain painting, the invention thus makes it possible to know the exercises most frequently performed by painters, and by studying different drawings, it is possible to determine which exercises are most effective.

[0025] Similarly, in the example of tracing massage trajectories, it is possible to know the points most frequently requested by the practitioner, or to record one or several massage trajectories so that the robot can reproduce them. These massage trajectories can also be recorded for different individuals. In addition, several motion sequences can be digitized to perform several different types of massage. Several generic models can also be created to improve the adaptation of the generic models with the patient's body, for example by using three generic models for each gender: large, small and average person, and for different age types: children, teenagers and adults, and / or for each recording position: sitting, standing and lying down.

[0026] According to one embodiment, said means for obtaining said reference element position are configured to capture an orientation of a reference element and to transfer this orientation to different points of a generic motion sequence.

[0027] According to one embodiment, the means for obtaining the position of a reference element is configured to capture movements performed or constraints experienced by the reference element and to report these movements or these constraints at different points of a generic motion sequence.

[0028] This embodiment makes it possible to control the operation of actuators, for example during the movement of the robot. In the example of a handicraft porcelain painting robot, the robot can perform a specific surface treatment at certain locations. In the example of a massage robot, a certain position of the robot can control a means of inducing vibrations to improve the comfort and / or effectiveness of the massage. Moreover, a generic movement sequence may thus include some trajectories performed with a palpatory rolling movement, while other movements are performed with another type of movement. The stress experienced by the reference element may thus correspond to a physical stress, for example pressure or temperature, or an external stress, for example the sound emitted during a massage or the diffusion of essential oils.

[0029] According to one embodiment, the generic model and the three-dimensional representation of the surface are formalized into the shape of a point cloud, and the adaptation means comprises: - means for calculating a normal direction to each point of said three-dimensional representation of said surface; - means for searching for, for each point of the cloud of points of said three-dimensional representation, a nearby point of the generic model which has the smallest difference between the normal direction of the generic model point and the normal direction of the interest point; - means for determining the distance between said detection points and said interest points of a generic model; - means for searching a global transformation of a generic model as a function of the determined distances for all point cloud points of said three-dimensional representation.

[0030] The normal directions allow to obtain information about the orientation of the surfaces of a generic model and the 3D representation of the surfaces. Unlike a simple point-to-point coordinate comparison, the surface comparison provides a more efficient recognition.

[0031] In addition, the adaptation of the generic model or the motion sequence is performed stepwise by modifying the generic model or the motion sequence step by step according to the average distance.

[0032] It follows that the present embodiment makes it possible to effectively adapt a generic model or a movement sequence by comparing the normal direction of each point of the generic model or movement sequence with the normal direction of the three-dimensional representation of the surface.

[0033] According to one embodiment, said searching means are adapted to search for points of the generic model within a preset sphere around the point of interest.

[0034] This embodiment is intended to limit the search range of the generic model points to limit the calculation time. In addition, the limitation of the search area also makes it possible to limit the magnitude of the correction of the generic model between two comparisons, and thus improve the accuracy of the correction of the generic model.

[0035] According to one embodiment, the normal direction is determined by constructing a surface using the coordinates of the three or four points closest to the interest point.

[0036] The present embodiment allows for efficient construction of 3D representations of surfaces and surfaces of generic models.

[0037] According to one embodiment, the adaptation means comprises: - means for detecting feature points on said three-dimensional representation; - means for transforming, by rotation and / or translation, the generic model such that the positions of the feature points correspond to positions of feature points of a generic model.

[0038] This embodiment allows for creating a first rough adaptation of a generic model to improve the speed of the precise adaptation done through normals. In the example of a handicraft porcelain painting robot, the feature points may correspond to the top and bottom edges of the porcelain.

[0039] In the massage robot example, the feature points may correspond to the top of the skull, the armpit location, and the crotch location.

[0040] According to one embodiment, said obtaining means comprise means for capturing multiple 3D representations and pre-processing said 3D representations by means of averaging the coordinates of points between the different 3D representations. This embodiment makes it possible to improve the accuracy of the 3D representation and therefore of the adaptation of the generic model.

[0041] According to one embodiment, said pre-processing means perform a filtering of said averages of coordinates of points between different representations of the three dimensions. This embodiment also makes it possible to improve the accuracy of the three-dimensional representation and therefore of the adaptation of the generic model.

[0042] According to one embodiment, said reference element corresponds to a glove and said movement sequence corresponds to the movements made by said glove during a massage.

[0043] Methods of practicing the invention as well as resulting advantages thereof will become apparent from the following embodiments, which are given by way of guidance and not limitation, with reference to supporting FIGS. 1-3. [Brief description of the drawings]

[0044] [Figure 1] 4 is a flow chart illustrating steps for determining a transformation of a generic model according to an embodiment of the present invention. [Diagram 2] 4 is a flowchart showing the operational steps of an apparatus for defining a generic motion sequence in a generic model according to the first embodiment. [Diagram 3] 10 is a flowchart showing the operational steps of an apparatus for defining a generic motion sequence in a generic model according to a second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0045] In the following description, the invention is described with respect to defining massage sequences, however, the invention is not limited to this particular application and can be used for various movement sequences relating to surfaces that do not have a predefined geometry.

[0046] As illustrated in Figure 1, the surface analysis is carried out by acquisition means 14 capable of providing a three-dimensional representation Re of the surface. The three-dimensional representation Re takes the form of a cloud of points, each point having three orthonormal coordinates: x, y and z.

[0047] The acquisition means 14 may correspond to a set of photographic sensors, a set of infrared sensors, a tomographic sensor, a stereo sensor or any other known sensor making it possible to acquire a three-dimensional representation of a surface. For example, a Kinect® camera from Microsoft® may be used to obtain this three-dimensional representation Re.

[0048] To obtain this three-dimensional representation Re without capturing the environment, it is possible to capture a first cloud of points corresponding to the environment and a second cloud of points corresponding to the surfaces in that environment. Only the different points between the two clouds of points are kept to extract the points corresponding to the surfaces from the environment. This method allows to separate from the standardized environment for recording and to adapt to any environment.

[0049] As illustrated in Fig. 1, these sensors 14 are often implemented with pre-processing means 15 in order to provide a three-dimensional representation Re of improved quality or accuracy. For example, the pre-processing means 15 may correspond to algorithms for equalizing the histogram, filtering, averaging the representation over several consecutive representations, etc.

[0050] For example, to obtain a better three-dimensional representation, it is possible to use the technique described in the scientific publication "KinectFusion: Real-time 3D Reconstruction and Interaction Using a Moving Depth Camera*", published on October 16, 2011 at UIST'11. The device then performs a computational process to adapt the three-dimensional representation Re and the generic models m1, m2, m3. The generic models m1, m2, m3 are also formalized in the shape of a point cloud, each point having three coordinates in an orthonormal system: x, y and z. Preferably, the generic models include a mean model ModMoy of N vertices in three coordinates, and a transformation matrix ModSigma of M shape components with 3N coordinates, i.e. three coordinates for N vertices. Many different people are needed to enrich each generic model m1, m2, m3, for example 1,000 people.

[0051] Principal component analysis is applied to reduce the dimensionality of the data. By applying principal component analysis to these data, it is possible to determine the variance of the data and associate the common variance with the components. Thus, instead of keeping one component per person, each generic model m1, m2, m3 stores about 20 components, explaining most of the variance over the 1,000 people. This method is described in more detail in the scientific publication "Building Statistical Shape Spaces for 3D Human Modeling", Pishchulin et al., published on March 19, 2015 in the journal "Pattern Recognition 2017".

[0052] Preferably, the generic models m1, m2, m3 are stored together with the three-dimensional representation Re in a memory accessible by the image processing means of the device to which the generic models m1, m2, m3 can be adapted.

[0053] To do this, once the three-dimensional representation Re is obtained, the device performs a detection of feature points Pref of this three-dimensional representation Re by means of digital processing means 16. In the example of figure 1, the feature points Pref correspond to the top of the skull, the position of the armpits and the position of the crotch. These digital processing means 16 can implement all known methods for detecting elements of an image, such as for example the Viola-Jones method.

[0054] Preferably, in order to detect the feature points Pref, the point cloud is converted into a depth image, i.e. an image of grey levels, for example coded with 12 bits, making it possible to code depths ranging from 0 to 4095 mm. This depth image is then thresholded and binarized to highlight only the pixels corresponding to the object / body of interest with a value of 1 and those corresponding to the environment with a value of 0. An edge detection is then applied to this binarized image, for example using the method described in Suzuki, S. and Abe, K., Topological Structural Analysis of Digitized Binary Images by Border Following, CVGIP 30 1, pp. 32-46 (1985). Finally, the corner points of the contour and its convex defects (determined for example using the method Sklansky, J., Finding the Convex Hull of a Simple Polygon. PRL 1 $number, pp. 79-83 (1982)) are used as feature points Pref.

[0055] The means 17 for selecting generic models m1, m2, m3 is then executed to select the generic model m1, m2, m3 that is closest to the three-dimensional representation Re.

[0056] For example, this selection may be done by calculating the distance between the feature points Pref of the top of the skull and the feature points of the crotch to roughly estimate the height dimension of the three-dimensional representation Re, and by selecting the generic models m1, m2, m3 with the closest height dimension. Similarly, the selection of the generic models m1, m2, m3 may be performed by using the width of the three-dimensional representation Re by calculating the distance between the feature points Pref of the axilla.

[0057] Furthermore, the generic models m1, m2, m3 may be articulated by virtual bones that represent the most important bones of the human skeleton. For example, 15 virtual bones may be modeled in the generic models m1, m2, m3 to define the position and shape of the spine, femur, tibia, ulna, humerus and skull. Furthermore, the orientation of these virtual bones makes it possible to define the pose of the generic models, i.e. the generic models m1, m2, m3 raise one arm, the legs move apart...

[0058] This pose of the generic models m1, m2, m3 may also be determined by selection, and the means 17 inputs the depth image contours of the object / body of interest into a depth image contour database of several thousand poses of generic models by comparing the distances (e.g., calculated using the Hu method Visual Pattern Recognition by Moment Invariants, IRE Transactions on Information Theory, 8:2, pp. 179-187 (1962)). The depth image of the articulated generic models of m1, m2, m3 that is closest to the depth image of the object / body of interest is selected, and the rotation values ​​of the virtual bones are saved.

[0059] A first adaptation is then performed by the adaptation means 18 by transforming the selected generic model to approximate the three-dimensional representation Re. For example, this first adaptation may simply transform the width and height of the selected generic model so that the spacing of the feature points Pref of the selected generic model corresponds to the spacing of the feature points Pref of the three-dimensional representation Re. This first adaptation may also determine the pose of the virtual bones of the generic models m1, m2, m3.

[0060] Following this first, rather rough adaptation, it is possible to use a second, more precise adaptation by using the normal directions formed by each surface defined between the points of the three-dimensional representation Re. To do this, the device incorporates means 19 for calculating the normals of each surface of the three-dimensional representation Re and of the selected generic model.

[0061] For example, the normal direction may be determined by constructing each surface of the three-dimensional representation Re using the coordinates of the three or four points closest to the point of interest. Alternatively, the normal direction of the generic model may be calculated during the step of defining the generic model.

[0062] The device then uses a search means 20 capable of finding, for each point of the point cloud of the three-dimensional representation Re, a point of the selected generic model nearby for which the difference between the normal direction of the generic model point and the normal direction of the point of interest is minimal. If virtual bones are components of the selected generic model, the search means 20 adapts the positions and dimensions of the virtual bones by modifying the characteristics of each virtual bone to adapt the virtual bones to the positions of the body elements present in the three-dimensional representation Re.

[0063] For example, the searching means 20 may be configured to search for points of the generic model within a pre-defined sphere around the point of interest, preferably the radius of this sphere being determined according to the number of vertices of the generic model and the dimensions of the object / body of interest, such that the sphere contains about 10 points.

[0064] Using all these normal directions, the device can then calculate the difference between the selected generic model and the three-dimensional representation Re using a determination means 21 capable of calculating the distance between the interest point and the point detected by the search means on the selected generic model. All these distances form a vector of transformations that should be applied to the interest point to correspond to the detected point. The search means 22 are designed to determine the average of these transformation vectors to obtain the overall transformation of the selected generic model.

[0065] In other words, by considering a new transformation vector CompVec of M components, it is possible to know the 3D configuration of the Pts3D vertices by applying the following formula: Pts3D=ModAv+CompVec*ModSigma

[0066] For an unknown Pts3D configuration, e.g., a new patient, the goal is to find the values ​​of the morphological components CompVec that correspond to this person knowing the average model ModAverage and the transformation matrix ModSigma.

[0067] To do this, the search means 22 calculates the difference DiffMod between the three-dimensional configuration of vertices Pts3D and the pseudo-inverse matrix ModSigmaInv of the mean model ModMoyen and ModSigma.

[0068] For example, the pseudo-inverse matrix ModSigmaInv may be calculated by decomposing the ModSigma matrix into singular values ​​using the following formula: ModSigma=VU*, ModSigmaInv=VE t U*, Here, E t corresponds to the transpose of E, V* is the transpose conjugate of V, U* is the transpose conjugate matrix of U.

[0069] Using these data, the searching means 22 calculates the morphological components CompVec using the following formula: DiffMod*ModSigmaInv=CompVec*ModSigma*ModSigmaInv

[0070] That is, CompVec=DiffMod*ModSigmaInv, which also makes it possible to obtain the CompVec morphological components for a particular patient.

[0071] The CompVec transformation vectors are then applied to the selected generic model. The pose is estimated again as before, the generic model is adjusted if necessary, and a new search is performed until the generic model is sufficiently close to the 3-step representation dimension Re. Finally, the loop stops when the average Euclidean distance between all vertices of the generic model and their counterparts on the point cloud is below a threshold, e.g. 2 mm, defined according to the number of vertices of the generic model and the dimensions of the object / body of interest, or when a maximum number of iterations, e.g. 100 iterations, is reached while still not reaching an average distance below the threshold.

[0072] A calibration phase between the sensor 14 and the robot must be performed frequently. To calibrate the visual sensor 14 and the robot, it is possible to record the coordinates of at least three common points in the two signs. In practice, it is preferable to use a number N of points greater than or equal to four. The robot is moved over its working range and stopped N times.

[0073] At each stop, the position of the robot is recorded by calculating the movements implemented by the robot's movement commands, and detection makes it possible to know the position of this stop in three dimensions using the visual sensor 14.

[0074] At the end of these N stops, the coordinates of the N points are known in the two reference systems. The centroid of the distribution of the N points in the two systems is determined using the following formula: The center of gravity A=1 / N sum(PA(i)), where i=1 to N, PA(i) is a point in the sensor's reference frame 14, and The center of gravity B = 1 / N sum(PB(i)), where i = 1 to N, and PB(i) is a point in the robot's system.

[0075] The covariance matrix C is then determined by the following formula: C=sum((PA(i)-center of gravity A)(PB(i)-center of gravity B) t ), where i = 1 to N

[0076] This covariance matrix C is then decomposed into singular values: C=UEV*

[0077] The following formula then gives the rotation matrix R between the two references: R=VU t , if the determinant of R is negative, it is possible to multiply the third column of the rotation matrix R by -1.

[0078] The following formula determines the transformation that should be applied between the two tags: T=-R*center of gravity A+center of gravity B

[0079] It is therefore possible to transform any point Pa in the reference system of the sensor 14 into the reference system Pb of the robot by applying the following formula: Pb=R*Pa+T

[0080] In a first embodiment of the invention illustrated in FIG. 2, adaptation of the selected generic model makes it possible to obtain the position Pr of the reference element 45 directly in the adapted generic model.

[0081] The acquisition is performed in one step 40 by determining the position Pr of a reference element 45 on the three-dimensional representation Re of the surface. The reference element may correspond to a reference point or a set of reference points corresponding to an effector or physical element. For example, the reference element may correspond to a glove or a set of points representing the hand of the practitioner. The position Pr of the reference element 45 on the three-dimensional representation Re of the surface may be determined by a position triangulation module or by an image processing analysis similar to that used to capture the three-dimensional representation Re of the surface. In addition to the position Pr of the reference element 45 on the three-dimensional representation Re of the surface, the acquisition may also make it possible to capture the orientation of the reference element 45 or a movement performed on the reference element 45, such as heating or a specific movement.

[0082] The acquisition is repeated several times in step 41 to form a sequence Tr of records illustrating the actual movements performed by the reference element 45. For example, an acquisition may be performed every 0.1 s.

[0083] Once a motion sequence Tr is completed, the points of the sequence Tr are projected onto the generic model transformed to the human form. The generic model is then transformed again to resume these initial parameters.

[0084] To do this, a transformation of the generic model is calculated from the initial parameters to the parameters of the transformed generic model. The motion sequence Tr is then transformed using the same transformation that was applied to the generic model.

[0085] Thus, the real motion sequence of Tr is transformed into a generic motion sequence Tx associated with the generic model.

[0086] In a second embodiment of the present invention illustrated in Fig. 3, the acquisition 45 is performed independently of the generic model. In this embodiment, step 23 of Fig. 1 merely determines the transformations of the generic model without actually applying them. To fit the motion sequence Tx to the generic model, the motion sequence may pass through the feature points of the surface morphology. For example, the acquisition 45 may be performed by moving the reference element 45 to the top and axilla of the subject's skull.

[0087] Once the motion sequence Tr has been recorded in step 46, the real motion sequence Tr is converted in step 47 into a generic motion sequence Tx by applying the difference between the generic model and the 3D representation Re of the surface. In a final step, at 48, the generic motion sequence is repositioned in the selected generic model. Preferably, this step 48 is performed by trying to match the feature points passed by the reference element 45 and the position of these feature points on the generic model. This step 48 may also be performed by taking into account the position of the object.

[0088] The invention therefore allows to define a generic movement sequence Tx in a generic model in a practical way, i.e. without the operator having to use a computer screen or a digital tablet. The invention therefore allows to greatly simplify the process of defining a movement sequence, as the operator is often more efficient during real recording in a real situation.

[0089] This motion sequence can then be used for various applications, such as comparing several motion sequences or controlling a robotic means for adapting a generic motion sequence to a specific subject. [Explanation of symbols]

[0090] 14 Acquisition methods, sensors, visual sensors, reference systems 15 Pretreatment methods 16 Digital Processing Means 17. Methods for selecting a generic model 18 Adaptation Measures 19 Means for calculating normals 20 Search means 21 Decision-making procedures 22 Search means 45 Reference element, get m1, m2, m3 general-purpose models Pref feature points Re 3-dimensional representation, 3-step representation dimension Pr position Tr sequence, movement sequence, real movement sequence Tx General-purpose Exercise Sequence

Claims

1. An apparatus for defining a generic motion sequence (Tx) in a three-dimensional generic model (m1 to m3), comprising: - acquisition means (40, 45) for acquiring a position (Pr) of a reference element (45) moving over a surface, said reference element (45) being configured to perform an actual movement sequence (Tr); recording means (41, 46) for recording said actual movement sequence (Tr); acquisition means (14) for acquiring a three-dimensional representation (Re) of said surface; - adaptation means (23) for adapting said three-dimensional generic model (m1-m3) to said three-dimensional representation (Re) of said surface; definition means (42, 47) for defining a generic motion sequence (Tx) in said three-dimensional generic model (m1-m3) by applying said adaptation between said three-dimensional generic model (m1-m3) and said three-dimensional representation (Re) of said surface to said actual motion sequence (Tr); An apparatus comprising:

2. 2. The apparatus according to claim 1, wherein the adaptation means (23) is configured for adapting the three-dimensional generic models (m1-m3) to the three-dimensional representation (Re) of the surface, the recording means (41) is configured for recording the actual movement sequence (Tr) in the three-dimensional generic models (m1-m3) when the three-dimensional generic models are adapted to the dimensions of the three-dimensional representation (Re) of the surface, and the definition means (42) is configured for transforming the three-dimensional generic models (m1-m3) in which the movement sequence has been recorded, so that the three-dimensional generic models resume their initial parameters.

3. 2. The apparatus according to claim 1, wherein the adaptation means (23) is configured to calculate a difference between the three-dimensional generic model (m1-m3) and the three-dimensional representation (Re) of the surface, the recording means (46) is configured to record the actual movement sequence (Tr) independently of the three-dimensional generic model (m1-m3), the definition means (47) is configured to transform the recorded movement sequence according to the difference calculated by the adaptation means (23), and the apparatus comprises means (48) for positioning the generic movement sequence in the three-dimensional generic model.

4. 4. Apparatus according to claim 2 or 3, wherein the means (40, 45) for acquiring the position (Pr) of the reference element (45) are configured to capture an orientation of the reference element (45) and to transfer this orientation to different points of the generic motion sequence (Tx).

5. 4. The device according to claim 2 or 3, wherein the means (40, 45) for acquiring the position (Pr) of the reference element (45) are configured to capture movements made or constraints experienced by the reference element (45) and to report these movements or these constraints at different points of the generic movement sequence (Tx).

6. The three-dimensional generic models (m1 to m3) and the three-dimensional representation (Re) of the surface are formalized in the form of a point cloud, and the adaptation means (23) means (19) for calculating a normal direction to each point of said three-dimensional representation (Re) of said surface; A search means (20) for a point of the three-dimensional generic model (m1 to m3) in the vicinity of each point of the point cloud of the three-dimensional representation (Re) where the difference between the normal direction of the point of the three-dimensional generic model (m1 to m3) and the normal direction of a point of interest is the smallest; means (21) for determining the distance between the detection points of said three-dimensional generic model (m1-m3) and said points of interest; means (22) for searching a global transformation of said three-dimensional generic model (m1 to m3) as a function of said distances determined for all points of said point cloud of said three-dimensional representation (Re); The apparatus of claim 1 , comprising:

7. 7. The apparatus according to claim 6, wherein the searching means (22) is configured to search for the point of the three-dimensional generic model within a preset sphere around the point of interest.

8. 8. An apparatus according to claim 6 or 7, wherein the normal direction is determined by constructing a surface using the coordinates of the three or four points closest to the point of interest.

9. The adapting means (23) means (16) for detecting feature points (Pref) on said three-dimensional representation (Re); means (18) for transforming, by rotation and / or translation, said three-dimensional generic models (m1 to m3) so that said positions of said feature points correspond to positions of feature points of said three-dimensional generic model; 9. The apparatus according to claim 1 , comprising:

10. 10. Apparatus according to any one of the preceding claims, wherein the reference element (45) corresponds to a glove and the movement sequence (Tr, Tx) corresponds to movements performed by the glove during a massage.

Images