CN104183000A - Full-automatic multi-source heterogeneous motion redirecting method of quasi-man character - Google Patents
Full-automatic multi-source heterogeneous motion redirecting method of quasi-man character Download PDFInfo
- Publication number
- CN104183000A CN104183000A CN201410399330.2A CN201410399330A CN104183000A CN 104183000 A CN104183000 A CN 104183000A CN 201410399330 A CN201410399330 A CN 201410399330A CN 104183000 A CN104183000 A CN 104183000A
- Authority
- CN
- China
- Prior art keywords
- bone
- source
- target
- node
- bounding box
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Processing Or Creating Images (AREA)
Abstract
The invention provides a full-automatic multi-source heterogeneous motion redirecting method of a quasi-man character. The full-automatic multi-source heterogeneous motion redirecting method comprises the following steps: according to the joint point topological structure characteristics of a target bone, determining a middle bone, and carrying out normalization alignment processing to the middle bone, the target bone and a source bone; then, carrying out simplification processing, independently carrying out hip, chest and lobe child node matching to the simplified target bone and source bone with the simplified middle bone, and carrying out wrist and ankle matching; independently carrying out semantic information adding to the target bone and the source bone, and according to a selected matching strategy, carrying out node mapping processing from the source bone to the target bone; and transferring the node motion data of the source bone to the nodes of the target bone, reducing a motion track by the target bone according to the node motion data of the source bone, and carrying out multi-source motion fusion. The invention provides a method which gives consideration to the features of a quasi-man cartoon character and can realize a great quantity of automatic redirection, and multi-source motion is effectively integrated.
Description
Technical field
The present invention relates to Computer Animated Graph field, in particular the full-automatic multi-source heterogeneous motion retargeting method of a kind people role.
Background technology
Motion retargeting is to take human skeletal's animation as main study subject, is intended to the motion of source case to transfer in target roles, and keeps the feature of its motion in vision and constraint.If wherein source bone is identical with target bone topological structure, and just bone length is different, is called isomorphism to be redirected; And if source bone is different with target bone topological structure, is called isomery and is redirected.For isomorphism role, because topological structure is identical, motion can be transferred by joint easily, and the redirected difficult point place that is only of isomery.
The articulation point mapping that the main flow process that isomery is redirected has comprised source bone and target bone, transfer and the feature reparation of exercise data.The object of articulation point mapping is, the articulation point of source case and target roles is set up one to one or the mapping relations of one-to-many, and its technical essential is efficiency and the effect of mapping policy; And in exercise data transfer process, each node of target roles will be inherited the movable information of its mapping node, its technical essential is the inheritance strategy of data and the processing that bone is demarcated attitude difference; Feature is mainly studied certain integral body or the reproduction of local feature in target roles of source case in repairing, and its technical essential is feature extraction and feature reparation.
Existing bone articulation point matching process mainly be take the manual coupling of user as main, well-known animation soft motion builder for example, when processing motion retargeting, allow user manually to set up the mapping graph between source bone and target bone, then according to these mapping relations transfer movement data.And in the field of study, automatically be redirected is one of difficult point of research always, has formed and manually source and target bone has been carried out the semantic semi-automatic reorientation method of demarcating, by middle skeleton, isomery skeleton data shifted and by automatic root node and bone is divided into root mates one by one to a plurality of branches of endpoint node as needs.
For class people cartoon role, the upper creativity due to art designing of general design, role's bone is also varied, and there will be big-ear, long-snouted, wing, tail, three folding legs etc. nature other biological and feature that non-human has.Therefore for these roles, semantic method needs artificial cognition key node, obviously cannot reach for various role in real time and be redirected; The design of skeleton in the middle of middle skeleton technique depends on to a certain extent, if centre is too simply, great deal of nodes cannot be mated, if coupling dislocation probably appears in skeleton redundancy; Automatically the root node extracting is also because the Local Complexity of bone has larger error.Therefore above method does not have a kind of feature that can take into account class people cartoon role, thereby cannot realize a large amount of being automatically redirected.
Therefore, prior art has yet to be improved and developed.
Summary of the invention
The technical problem to be solved in the present invention is, above-mentioned defect for prior art, one kind people role is provided full-automatic multi-source heterogeneous motion retargeting method, can effectively solve the feature that can not take into account class people cartoon role in prior art, thereby realize a large amount of automatic redirected defects.
The technical scheme that technical solution problem of the present invention adopts is as follows:
The full-automatic multi-source heterogeneous motion retargeting method of one kind people role, wherein, described method comprises step:
A, according to the articulation point topological structure feature of target bone, determine in the middle of bone, and bone, described target bone and source bone in the middle of described are normalized to registration process;
B, middle bone, target bone and source bone through normalization registration process are simplified to processing, source bone by the target bone after simplifying and after simplifying respectively with simplify after middle bone carry out hip, chest and leaf node and mate, and according to simplification matching result, target bone and the source bone of process normalization registration process are carried out to mating of wrist and ankle with middle bone respectively;
C, that described target bone and described source bone are carried out respectively to semantic information is additional, and according to selected matching strategy, carries out source bone and process to the node mapping of target bone;
D, the node exercise data of described source bone is passed to the node of described target bone, described target bone is according to the node exercise data reduction movement locus of described source bone, and carries out the fusion of multi-source motion.
The full-automatic multi-source heterogeneous motion retargeting method of described class people role, wherein, described steps A specifically comprises:
A1, according to the topological structure feature of the articulation point in target bone, determine in the middle of bone;
A2, by bone convergent-divergent in the middle of described and be placed in that length and width are unit length, height and the width are carried out the first bounding box of convergent-divergent in proportion, obtain bone in the middle of normalization; By described target bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the second bounding box of convergent-divergent in proportion, obtain normalization target bone; By source bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the 3rd bounding box of convergent-divergent in proportion, obtain normalization source bone;
A3, described the first bounding box, the second bounding box and described the 3rd bounding box are all placed in to rectangular coordinate system in space, and the bottom surface that makes the bottom surface of described the first bounding box, the bottom surface of the second bounding box and described the 3rd bounding box all with xoz planar registration, the right flank that makes the right flank of described the first bounding box, the right flank of the second bounding box and described the 3rd bounding box all with yoz planar registration, make described the 3rd bounding box before, the second bounding box above and described the 3rd bounding box above all with xoy planar registration.
The full-automatic multi-source heterogeneous motion retargeting method of described class people role, wherein, described step B specifically comprises:
B1, the node that in bone, described normalization target bone and described normalization source bone in the middle of described normalization, all degree are 2 and two adjacent sides thereof are deleted, add again a limit and connect two consecutive point, until only remaining degree is 1 leaf node and spends the node of divergence that is greater than 2, be simplified middle bone, simplify target bone and simplify source bone;
B2, according to the matching algorithm that is communicated with the total Least-cost of constraint, by simplifications target bone and simplify source bone respectively with simplify middle bone and carry out hip, chest and leaf node and mate;
B3, normalization target bone and normalization source bone are carried out to mating of wrist and ankle with bone in the middle of normalization respectively.
The full-automatic multi-source heterogeneous motion retargeting method of described class people role, wherein, all nodes in the bone of normalization source described in described step B3 are once mapped.
The full-automatic multi-source heterogeneous motion retargeting method of described class people role, wherein, described middle bone is the moving middle bone of catching role's bone, the simplest middle bone or having expanded tail wing of standard.
The full-automatic multi-source heterogeneous motion retargeting method of described class people role, wherein, described middle bone is extendible middle bone.
The full-automatic multi-source heterogeneous motion retargeting method of a kind people role provided by the present invention, method comprises: bone in the middle of determining according to the articulation point topological structure feature of target bone, and bone, described target bone and source bone in the middle of described are normalized to registration process; To simplifying processing through middle bone, target bone and the source bone of normalization registration process, source bone by the target bone after simplifying and after simplifying respectively with simplify after middle bone carry out hip, chest and leaf node and mate, and according to simplification matching result, target bone and the source bone of process normalization registration process are carried out to mating of wrist and ankle with middle bone respectively; Described target bone and described source bone are carried out respectively to semantic information additional, and according to selected matching strategy, carry out source bone and process to the node mapping of target bone; The node exercise data of described source bone is passed to the node of described target bone, described target bone is according to the node exercise data reduction movement locus of described source bone, and carries out the fusion of multi-source motion.The invention provides a kind of feature of taking into account class people cartoon role, realize a large amount of automatic redirected methods, effectively integrated multi-source motion.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the full-automatic multi-source heterogeneous motion retargeting method of class people role of the present invention preferred embodiment.
Fig. 2 is the particular flow sheet of normalization registration process in the full-automatic multi-source heterogeneous motion retargeting method of the described class people role of invention.
Fig. 3 a, Fig. 3 b, Fig. 3 c are respectively the moving schematic diagram of catching role's bone, the simplest middle bone and having expanded the middle bone of tail wing of Plays of the present invention.
Fig. 4 is the normalized schematic diagram of bone in the full-automatic multi-source heterogeneous motion retargeting method of the described class people role of invention.
The particular flow sheet of simplification and matching treatment in the full-automatic multi-source heterogeneous motion retargeting method of Fig. 5 class people role of the present invention.
Fig. 6 is the schematic diagram of right-hand man's subtree and left and right bottom sediment tree in the present invention.
Fig. 7 is bone semantic marker schematic diagram in the present invention.
Fig. 8 a, Fig. 8 b is respectively source bone and target bone topological structure schematic diagram, Fig. 8 c is the mapping coupling schematic diagram of source bone head and target bone head, Fig. 8 d is the mapping coupling schematic diagram of source bone right arm and target bone right arm, Fig. 8 e is that the source bone right hand refers to the mapping coupling schematic diagram referring to the target bone right hand, Fig. 8 f is the mapping coupling schematic diagram of source bone backbone and target bone backbone, Fig. 8 g is the mapping coupling schematic diagram of the right leg of source bone and the right leg of target bone, Fig. 8 h is the mapping coupling schematic diagram of source bone right crus of diaphragm toe and target bone right crus of diaphragm toe, Fig. 8 i is the mapping coupling schematic diagram of source bone tail and target bone tail.
Fig. 9 is the structured flowchart of the full-automatic multi-source heterogeneous motion retargeting system of class people role of the present invention preferred embodiment.
Embodiment
For making object of the present invention, technical scheme and advantage clearer, clear and definite, referring to accompanying drawing, developing simultaneously, the present invention is described in more detail for embodiment.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
Refer to Fig. 1, Fig. 1 is the process flow diagram of the full-automatic multi-source heterogeneous motion retargeting method of class people role of the present invention preferred embodiment.As shown in Figure 1, the full-automatic multi-source heterogeneous motion retargeting method of described class people role, comprises the following steps:
Step S100, according to the articulation point topological structure feature of target bone, determine in the middle of bone, and bone, described target bone and source bone in the middle of described are normalized to registration process.
In embodiments of the invention, first, from storing a plurality of motion roles' motor pool, choose a bone with athletic performance as source bone, then choose a bone without athletic performance as target bone.Then bone in the middle of determining according to the articulation point topological structure feature of target bone, and source bone, middle bone and target bone are all normalized and registration process to the intermediary that the middle bone of usining mates with target bone as source bone.By choosing bone in the middle of, as the medium of both couplings, make coupling more accurate.
Step S200, middle bone, target bone and source bone through normalization registration process are simplified to processing, source bone by the target bone after simplifying and after simplifying respectively with simplify after middle bone carry out hip, chest and leaf node and mate, and target bone and the source bone of process normalization registration process are carried out to mating of wrist and ankle with middle bone respectively.
In step S100, source bone, middle bone and target bone are all normalized with registration process after, can not directly mate.But first to simplifying processing through middle bone, target bone and the source bone of normalization registration process, after simplification, again source bone and target bone are carried out to hip, chest and leaf node with middle bone respectively and mate, and at target bone and the source bone through normalization registration process only, carry out mating of wrist and ankle with middle bone according to simplifying matching result respectively.
Step S300, that described target bone and described source bone are carried out respectively to semantic information is additional, and according to selected matching strategy, carries out source bone and process to the node mapping of target bone.
Step S400, the node exercise data of described source bone is passed to the node of described target bone, described target bone is according to the node exercise data reduction movement locus of described source bone, and carries out the fusion of multi-source motion.
Embodiment further, as shown in Figure 2, in described step S100, normalization registration process specifically comprises:
Step S101, according to the topological structure feature of the articulation point in target bone, determine in the middle of bone.
The standard that middle bone is chosen is to need coverage goal bone node of divergence how as far as possible, and guarantees the coupling of these nodes.Although bone in the middle of using in embodiments of the invention, in the middle of experimental results show that, bone before only participating in two-layer bone coupling in the situation that, need to possess basic semantic structure can obtain good effect.Therefore can choice criteria in the most situations moving role's bone of catching is as middle bone (comprise hip, chest, two wrists, two ankles totally six can divaricable node), equally, if role need to carry out semantic description to improper limbs (as wing of angel etc.), can suitably increase corresponding node.During concrete enforcement, as shown in Fig. 3 a-3c, described in the middle of bone optional select from standard as shown in Figure 3 a moving catch role's bone, the simplest as shown in Figure 3 b in the middle of bone or expanded as shown in Figure 3 c the middle bone of tail wing and chosen the middle bone of a kind of conduct.In the middle of described, bone is not limited to above-mentioned three kinds of middle bones, is a kind of extendible middle bone.Middle bone does not require determines unique structure, but can expand according to the form of source and target bone, increases adaptability.
Step S102, by bone convergent-divergent in the middle of described and be placed in that length and width are unit length, height and the width are carried out the first bounding box of convergent-divergent in proportion, obtain bone in the middle of normalization; By described target bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the second bounding box of convergent-divergent in proportion, obtain normalization target bone; By source bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the 3rd bounding box of convergent-divergent in proportion, obtain normalization source bone.
Role itself to be matched is very strange, also not of uniform size in size, can cause very large impact to node matching.First need all roles to be normalized alignment.As shown in Figure 4, it is the normalized schematic diagram of bone in the full-automatic multi-source heterogeneous motion retargeting method of class people role of the present invention.First stipulate all roles crown sensing y axle positive dirction, towards z axle positive dirction; Secondly by all role's convergent-divergents, making its bounding box width on x axle and y direction of principal axis is unit 1, and z direction of principal axis and y axle carry out convergent-divergent in proportion.In step S102, bone, described target bone and described source bone in the middle of described are normalized respectively, make it in same size class, be more conducive to node matching.
Step S103, described the first bounding box, the second bounding box and described the 3rd bounding box are all placed in to rectangular coordinate system in space, and the bottom surface that makes the bottom surface of described the first bounding box, the bottom surface of the second bounding box and described the 3rd bounding box all with xoz planar registration, the right flank that makes the right flank of described the first bounding box, the right flank of the second bounding box and described the 3rd bounding box all with yoz planar registration, make described the 3rd bounding box before, the second bounding box above and described the 3rd bounding box above all with xoy planar registration.
Embodiment further, as shown in Figure 5, described step S200 simplifies and matching treatment specifically comprises:
Step S201, the node that in bone, described normalization target bone and described normalization source bone in the middle of described normalization, all degree are 2 and two adjacent sides thereof are deleted, add again a limit and connect two consecutive point, until only remaining degree is 1 leaf node and spends the node of divergence that is greater than 2, be simplified middle bone, simplify target bone and simplify source bone.
For each node of bone in the middle of simplifying, in the role to be matched (source bone or target bone) who simplifies, be all the mapping node of its designated displacement, the mapping node of different nodes can not be identical.
Step S202, according to the matching algorithm that is communicated with the total Least-cost of constraint, by simplifications target bone and simplify source bone respectively with simplify middle bone and carry out hip, chest and leaf node and mate.
For all adjacent nodes pair in the middle bone of simplifying, after being mapped to bone to be matched, if connected in the right path of its mapping node, comprised other mapped nodes, think that it does not meet this connection constraint, rejects this scheme;
Secondly, for meeting the mapping scheme that is communicated with constraint, calculate Euclidean distance the summation of all nodes and its mapping node, as total cost (distance) of this scheme.
Choose and meet the coupling that is communicated with constraint and total Least-cost, determine hip and the pereonite point of bone to be matched.Can obtain the matching relationship of one group of leaf node, for instructing next step simultaneously.
Step S203, normalization target bone and normalization source bone are carried out to mating of wrist and ankle with bone in the middle of normalization respectively.
In embodiments of the invention, except row hip, chest and leaf node are mated, also need wrist and ankle node to mate.Coupling in this step is between bone, described normalization target bone and described normalization source bone, to mate in the middle of unreduced described normalization.
Take left hand as example: according to the mapping relations that obtain in step S202, the mapping node of bone left hand leaf node in the middle of finding.Search and take all subtrees that the child node of skeleton pereonite point to be matched is root, if this subtree comprises above-mentioned mapping node, think that this is left hand subtree, as shown in Figure 6, it is the schematic diagram of right-hand man's subtree in the present invention and left and right bottom sediment tree.Path using root node in left hand subtree to each leaf node, as a paths, for each node of each paths, calculates distance function as follows: length ratio, and root node is to the path of this node and the ratio of path total length; Regional percentage, node to the path of neighbors and with the ratio of path total length; Degree, the i.e. degree of node.Three values are weighted on average, and weights are (2,1,1), then in all results, select left hand wrist with intermediate node apart from minimum some the left hand wrist as role to be matched.Right hand wrist is in like manner calculated, and left leg and right leg also in like manner calculate, and distinguishes the root node of setting for left and right hired thug and chooses from the child node of hip node.
Embodiment further, as shown in Figure 7, it is bone semantic marker schematic diagram in the present invention, in described step S300, described target bone and described source bone are carried out respectively to semantic information additional hours, be not target bone after simplification and simplify after source bone on to carry out semantic information additional, but in step S200, carry out after the coupling of hip, chest and leaf node coupling and wrist and ankle, determined after the above-mentioned key node of source bone and target bone, according to a default bone semantic marker rule, the mark that carries out following order arranges:
Left/right finger: each stalk tree that each child node of left/right wrist of take is root node is exactly a finger is finger one, two, three according to the z axial coordinate sort descending of its leaf node As shown in Figure 7, the right hand in the subtree that the right finesse 2100 of take is root node refers to 1, the right hand refer to 2 2120 and the right hand refer to that 3 2130 left hands refer to, and the left hand of take in the subtree that left finesse 2200 is root node refers to 1, left hand refer to 2 2220 and left hand refer to 3 2230;
Left/right arm: pereonite is put all nodes (not comprising chest, wrist) on left/right carpopodium point path; As shown in Figure 7, the chest 1000 of take is right arm 2300 to all nodes on right finesse 2100 paths, and the chest 1000 of take is left arm 2400 to all nodes on left finesse 2200 paths;
Head: the child node that the pereonite point of take is not labeled is root, and each stalk tree is a stature, is first according to its leaf node to the horizontal range sort ascending of pereonite point, two, three As shown in Figure 7, it is first 1100 that each subtree that the child node that the chest 1000 of take is not labeled is root is labeled as respectively, and 2 1200 and 3 1300, wherein first 1100 leaf node is the shortest to the horizontal range of chest 1000,2 1200 take second place, 3 1300 maximum;
Left/right toe: each stalk tree that each child node of left/right ankle of take is root node is exactly a toe, is toe one, two to the x wheelbase of hip from sort ascending according to its leaf node, three As shown in Figure 7, the right crus of diaphragm toe 1 in the subtree that the right ankle 4100 of take is root node and right crus of diaphragm toe 2 4120, and left foot toe 1 and the left foot toe 2 4220 of take in the subtree that left ankle 4200 is root node;
Left/right leg: hip node is to all nodes (not comprising hip, ankle) on left/right ankle node path; As shown in Figure 7, the hip 3000 of take is right leg 3100 to all nodes on right ankle 4100 paths, and the hip 3000 of take is left leg 3200 to all nodes on left ankle 4200 paths;
Backbone: the child node of hip node of take is root, the path that pereonite point is leaf; As shown in Figure 7, the child node of hip 3000 of take is root, and the path that the chest 1000 of take is leaf is backbone 1400;
Tail: the child node that the hip node of take is not labeled is root, and every stalk tree is a tail, is tail one, two according to its frondesce node to the horizontal range sort ascending of hip node, three As shown in Figure 7, each subtree that the child node that the hip 3000 of take is not labeled is root is labeled as respectively tail 1, tail 2 3020 and tail 3 3030, and wherein the leaf node of tail 1 is the shortest to the horizontal range of hip 3000, tail 2 1200 takes second place, tail 3 1300 maximums;
Other annexes: all unmarked nodes are all inherited the semanteme of its father node, support the special semantemes such as wing as needs, can carry out semantic marker by bone in the middle of expanding.
In described step S300, carry out matching strategy that source bone processes time institute's basis to the node mapping of target bone and comprise intermediate value 1-1 matching strategy, intermediate value 1-n matching strategy, adjacent end matching strategy and sequence number matching strategy, wherein:
Intermediate value 1-1 matching strategy, for mating in the Liang Ge branch of intermediate value (can be locus, plane projection position or the intermediate value of coordinate axis numerical value, be used the latter conventionally) in selected source case and target roles, is applicable to the situation of head, tail;
Intermediate value 1-n matching strategy mates with n branch of target roles for the branch in intermediate value on selected source case, is applicable to the situation of head, tail;
Adjacent end matching strategy is for all branches of target roles, gets the immediate source case of its leaf node branch coupling, is applicable to the multiple situations such as head, tail, arm, hand, leg, pin;
Sequence number matching strategy is for when m >=n, and front n branch of the n of a target bone branch and source bone sequentially number mated; When m<n, m branch of front m branch of target bone and source bone sequentially number mated, and the m+1 of target bone is mated with m branch of source bone to n branch.This strategy is applicable to the multiple situations such as head, tail, arm, hand, leg, pin.
Generally for head, choose intermediate value 1-1 coupling, tail is chosen intermediate value 1-n coupling, and sequence number coupling is chosen at other positions, and the compensation process of adjacent end coupling when indefinite as the meaning of one's words used.Can append other strategies in addition.Coupling branch pair to selecting, adopts the mapping policy identical with strand.
In step S300, the schematic diagram that mates target bone and source bone according to above-mentioned matching principle is as shown in Fig. 8 a-8i, wherein Fig. 8 a and Fig. 8 b are respectively source bone and target bone topological structure schematic diagram, Fig. 8 c is the mapping coupling schematic diagram of source bone head and target bone head, Fig. 8 d is the mapping coupling schematic diagram of source bone right arm and target bone right arm, Fig. 8 e is that the source bone right hand refers to the mapping coupling schematic diagram referring to the target bone right hand, Fig. 8 f is the mapping coupling schematic diagram of source bone backbone and target bone backbone, Fig. 8 g is the mapping coupling schematic diagram of the right leg of source bone and the right leg of target bone, Fig. 8 h is the mapping coupling schematic diagram of source bone right crus of diaphragm toe and target bone right crus of diaphragm toe, Fig. 8 i is the mapping coupling schematic diagram of source bone tail and target bone tail.
Embodiment further, in described step S400, process the differences in motion that tri-kinds of pose arms of N T A cause, node for arm and finger, each root bone need to be rotated under the attitude identical with the corresponding bone of source case under demarcation posture, after transfer action, rotate back to original position again and process its child node, other joint rotation angles are directly transferred to target roles frame by frame.The attitude that wherein N-pose refers to uprightly, both hands hang down naturally, the attitude that T-pose refers to uprightly, both hands side direction horizontal opens, A-pose refers to uprightly, both hands side is opened and the attitude of trunk folder 45 degree left and right.
In step S400, described target bone first records root node and the left and right pin track of source action while reducing movement locus according to the node exercise data of described source bone; Again according to the hip node terrain clearance proportional zoom track of source and target bone; Finally the pressure track using the track after convergent-divergent as target bone, calculates IK(Inverse Kinematic, i.e. inverse kinematics), obtain new leg exercise.
Meanwhile, while carrying out the fusion of multi-source motion in step S400, adopt a kind of in Crossover Strategy, combined strategy or Mutation Strategy, wherein:
Crossover Strategy refers to carries out region division to the bone of target roles, and the action in each region derives from different source cases, thereby is combined into new motion.For example source case A does not have wing and tail, and source case B has wing and source case C has tail, at this moment for a target roles T who has wing and tail simultaneously, can from A, obtain to obtain in body work, B and in wing action, C, obtain tail action; And for example, for source movement A and a source movement B who runs of the shooting of squatting down, can respectively get lower part of the body motion, the target travel T of the synthetic shooting of running.
Mutation Strategy refers to increases delicate disturbance to source movement, makes target roles have new motion feature.For example original common walking action is increased to a layback position, generate the motion of layback walking; And for example, by increasing the both hands amplitude of fluctuation of original running action, generate a kind of walking action of joy.
Combined strategy refers to a plurality of source movements and is spliced to form a set of new target roles action.For example, for a running, and the action of a shooting, by splicing and calculating middle transition, generate an action sequence of running → shooting.
Step S100-step S400 all can automatically complete, and wherein middle skeleton and matching strategy can manually also can be set automatically.Suppose that motor pool has N motion role, wherein having comprised m has motion bone, has the motion of s cover, and n without motion bone.Complete after the full-automatic multi-source heterogeneous motion retargeting of described class people role, each role of motor pool will all have s action, become the role movement storehouse with N*s motion fragment, get through the redirected key link of isomery role multi-source.After this by carrying out multi-source Motion fusion on each role self, the motion of fusion can act on remaining role again through being redirected, thereby realizes the expansion of motor pool geometric progression.
Based on above-described embodiment, the present invention also provides a kind people role full-automatic multi-source heterogeneous motion retargeting system, and as shown in Figure 9, the full-automatic multi-source heterogeneous motion retargeting system of described class people role comprises:
Normalizing alignment module 100, for bone in the middle of determining according to the articulation point topological structure feature of target bone, and is normalized registration process by bone, described target bone and source bone in the middle of described; As detailed above.
Middle matching module 200, for middle bone, target bone and the source bone of process normalization registration process are simplified to processing, source bone by the target bone after simplifying and after simplifying respectively with simplify after middle bone carry out hip, chest and leaf node and mate, and target bone and the source bone of process normalization registration process are carried out to mating of wrist and ankle with middle bone respectively; As detailed above.
Mapping matching module 300, additional for described target bone and described source bone being carried out respectively to semantic information, and according to selected matching strategy, carry out source bone and process to the node mapping of target bone; As detailed above.
Module 400 is transmitted in motion, and for the node exercise data of described source bone being passed to the node of described target bone, described target bone is according to the node exercise data reduction movement locus of described source bone, and carries out the fusion of multi-source motion; As detailed above.
Embodiment further, in the full-automatic multi-source heterogeneous motion retargeting system of described class people role, described normalizing alignment module 100 specifically comprises:
Middle bone determining unit, for bone in the middle of determining according to the topological structure feature of the articulation point of target bone; As detailed above.
Normalization unit, for by bone convergent-divergent in the middle of described and be placed in that length and width are unit length, height and the width are carried out the first bounding box of convergent-divergent in proportion, obtains bone in the middle of normalization; By described target bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the second bounding box of convergent-divergent in proportion, obtain normalization target bone; By source bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the 3rd bounding box of convergent-divergent in proportion, obtain normalization source bone; As detailed above.
Alignment unit, for described the first bounding box, the second bounding box and described the 3rd bounding box are all placed in to rectangular coordinate system in space, and the bottom surface that makes the bottom surface of described the first bounding box, the bottom surface of the second bounding box and described the 3rd bounding box all with xoz planar registration, the right flank that makes the right flank of described the first bounding box, the right flank of the second bounding box and described the 3rd bounding box all with yoz planar registration, make described the 3rd bounding box before, the second bounding box above and described the 3rd bounding box above all with xoy planar registration; As detailed above.
Embodiment further, in the full-automatic multi-source heterogeneous motion retargeting system of described class people role, described in the middle of matching module 200 specifically comprise:
Simplified element, for the node that in the middle of described normalization, bone, described normalization target bone and all degree of described normalization source bone are 2 and two adjacent sides thereof are deleted, add again a limit and connect two consecutive point, until only remaining degree is 1 leaf node and spends the node of divergence that is greater than 2, be simplified middle bone, simplify target bone and simplify source bone; As detailed above.
The first matching unit, for according to the matching algorithm that is communicated with the total Least-cost of constraint, carries out hip, chest and leaf node with bone in the middle of simplification by simplification target bone respectively with simplification source bone and mates; As detailed above.
The second matching unit, for carrying out mating of wrist and ankle with bone in the middle of normalization respectively by normalization target bone and normalization source bone; As detailed above.
Embodiment further, in the full-automatic multi-source heterogeneous motion retargeting system of described class people role, all nodes described in described the second matching unit in the bone of normalization source are once mapped; As detailed above.
Embodiment further, in the full-automatic multi-source heterogeneous motion retargeting system of described class people role, described in the middle of bone be standard moving catch role's bone, the simplest in the middle of bone or expanded the middle bone of tail wing; As detailed above.
In sum, the full-automatic multi-source heterogeneous motion retargeting method of a kind people role provided by the present invention, method comprises: bone in the middle of determining according to the articulation point topological structure feature of target bone, and bone, described target bone and source bone in the middle of described are normalized to registration process; To simplifying processing through middle bone, target bone and the source bone of normalization registration process, source bone by the target bone after simplifying and after simplifying respectively with simplify after middle bone carry out hip, chest and leaf node and mate, and target bone and the source bone of process normalization registration process are carried out to mating of wrist and ankle with middle bone respectively; Described target bone and described source bone are carried out respectively to semantic information additional, and according to selected matching strategy, carry out source bone and process to the node mapping of target bone; The node exercise data of described source bone is passed to the node of described target bone, described target bone is according to the node exercise data reduction movement locus of described source bone, and carries out the fusion of multi-source motion.The invention provides a kind of feature of taking into account class people cartoon role, realize a large amount of automatic redirected methods, effectively integrated multi-source motion.
Should be understood that, application of the present invention is not limited to above-mentioned giving an example, and for those of ordinary skills, can be improved according to the above description or convert, and all these improvement and conversion all should belong to the protection domain of claims of the present invention.
Claims (6)
1. the full-automatic multi-source heterogeneous motion retargeting method of a kind people role, is characterized in that, described method comprises step:
A, according to the articulation point topological structure feature of target bone, determine in the middle of bone, and bone, described target bone and source bone in the middle of described are normalized to registration process;
B, middle bone, target bone and source bone through normalization registration process are simplified to processing, source bone by the target bone after simplifying and after simplifying respectively with simplify after middle bone carry out hip, chest and leaf node and mate, and according to simplification matching result, target bone and the source bone of process normalization registration process are carried out to mating of wrist and ankle with middle bone respectively;
C, that described target bone and described source bone are carried out respectively to semantic information is additional, and according to selected matching strategy, carries out source bone and process to the node mapping of target bone;
D, the node exercise data of described source bone is passed to the node of described target bone, described target bone is according to the node exercise data reduction movement locus of described source bone, and carries out the fusion of multi-source motion.
2. the full-automatic multi-source heterogeneous motion retargeting method of class people role according to claim 1, is characterized in that, described steps A specifically comprises:
A1, according to the topological structure feature of the articulation point in target bone, determine in the middle of bone;
A2, by bone convergent-divergent in the middle of described and be placed in that length and width are unit length, height and the width are carried out the first bounding box of convergent-divergent in proportion, obtain bone in the middle of normalization; By described target bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the second bounding box of convergent-divergent in proportion, obtain normalization target bone; By source bone convergent-divergent and be placed in that length and width are unit length, height and the width are carried out the 3rd bounding box of convergent-divergent in proportion, obtain normalization source bone;
A3, described the first bounding box, the second bounding box and described the 3rd bounding box are all placed in to rectangular coordinate system in space, and the bottom surface that makes the bottom surface of described the first bounding box, the bottom surface of the second bounding box and described the 3rd bounding box all with xoz planar registration, the right flank that makes the right flank of described the first bounding box, the right flank of the second bounding box and described the 3rd bounding box all with yoz planar registration, make described the 3rd bounding box before, the second bounding box above and described the 3rd bounding box above all with xoy planar registration.
3. the full-automatic multi-source heterogeneous motion retargeting method of class people role according to claim 2, is characterized in that, described step B specifically comprises:
B1, the node that in bone, described normalization target bone and described normalization source bone in the middle of described normalization, all degree are 2 and two adjacent sides thereof are deleted, add again a limit and connect two consecutive point, until only remaining degree is 1 leaf node and spends the node of divergence that is greater than 2, be simplified middle bone, simplify target bone and simplify source bone;
B2, according to the matching algorithm that is communicated with the total Least-cost of constraint, by simplifications target bone and simplify source bone respectively with simplify middle bone and carry out hip, chest and leaf node and mate;
B3, normalization target bone and normalization source bone are carried out to mating of wrist and ankle with bone in the middle of normalization respectively.
4. the full-automatic multi-source heterogeneous motion retargeting method of class people role according to claim 3, is characterized in that, all nodes in the middle of normalization described in described step B3 in bone are once mapped.
5. the full-automatic multi-source heterogeneous motion retargeting method of class people role according to claim 1, is characterized in that, described in the middle of bone be standard moving catch role's bone, the simplest in the middle of bone or expanded the middle bone of tail wing.
6. the full-automatic multi-source heterogeneous motion retargeting method of class people role according to claim 1, is characterized in that, described in the middle of bone be extendible in the middle of bone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410399330.2A CN104183000B (en) | 2014-08-14 | 2014-08-14 | Full-automatic multi-source heterogeneous motion redirecting method of quasi-man character |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410399330.2A CN104183000B (en) | 2014-08-14 | 2014-08-14 | Full-automatic multi-source heterogeneous motion redirecting method of quasi-man character |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104183000A true CN104183000A (en) | 2014-12-03 |
| CN104183000B CN104183000B (en) | 2017-04-19 |
Family
ID=51964016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410399330.2A Active CN104183000B (en) | 2014-08-14 | 2014-08-14 | Full-automatic multi-source heterogeneous motion redirecting method of quasi-man character |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104183000B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274467A (en) * | 2017-06-29 | 2017-10-20 | 厦门游亨世纪科技有限公司 | A kind of model animation interlock method based on Unity3D |
| CN107424203A (en) * | 2017-08-02 | 2017-12-01 | 湖南大学 | The motion retargeting method and device being combined based on Displacement mapping method and Jacobian matrix algorithm |
| CN107633542A (en) * | 2016-07-19 | 2018-01-26 | 珠海金山网络游戏科技有限公司 | One kind pinches face editor and animation fusion method and system |
| CN109887070A (en) * | 2019-01-10 | 2019-06-14 | 珠海金山网络游戏科技有限公司 | A kind of virtual face's production method and device |
| CN110264551A (en) * | 2019-06-20 | 2019-09-20 | 合肥工业大学 | A kind of motion retargeting method and system |
| CN112164129A (en) * | 2020-09-02 | 2021-01-01 | 北京电影学院 | No-pairing action migration method based on deep convolutional network |
| CN112233211A (en) * | 2020-11-03 | 2021-01-15 | 网易（杭州）网络有限公司 | Animation production method and device, storage medium and computer equipment |
| CN112927331A (en) * | 2021-03-31 | 2021-06-08 | 腾讯科技（深圳）有限公司 | Animation generation method and device of character model, storage medium and electronic equipment |
| CN113313794A (en) * | 2021-05-19 | 2021-08-27 | 深圳市慧鲤科技有限公司 | Animation migration method and device, equipment and storage medium |
| CN116012497A (en) * | 2023-03-29 | 2023-04-25 | 腾讯科技（深圳）有限公司 | Animation redirection method, device, equipment and medium |
| CN117710468A (en) * | 2024-02-06 | 2024-03-15 | 天度(厦门)科技股份有限公司 | Gesture redirection method, device, equipment and medium based on joint grid deformation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060139355A1 (en) * | 2004-12-27 | 2006-06-29 | Seyoon Tak | Physically based motion retargeting filter |
| CN102693549A (en) * | 2011-03-25 | 2012-09-26 | 上海日浦信息技术有限公司 | Three-dimensional visualization method of virtual crowd motion |
| CN102708582A (en) * | 2012-05-08 | 2012-10-03 | 电子科技大学 | Character movement redirecting method for heterogeneous topology |
| CN103530897A (en) * | 2013-09-30 | 2014-01-22 | 华为软件技术有限公司 | Movement redirection processing method and device |
| WO2014051584A1 (en) * | 2012-09-27 | 2014-04-03 | Empire Technology Development, Llc | Character model animation using stored recordings of player movement interface data |
-
2014
- 2014-08-14 CN CN201410399330.2A patent/CN104183000B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060139355A1 (en) * | 2004-12-27 | 2006-06-29 | Seyoon Tak | Physically based motion retargeting filter |
| CN102693549A (en) * | 2011-03-25 | 2012-09-26 | 上海日浦信息技术有限公司 | Three-dimensional visualization method of virtual crowd motion |
| CN102708582A (en) * | 2012-05-08 | 2012-10-03 | 电子科技大学 | Character movement redirecting method for heterogeneous topology |
| WO2014051584A1 (en) * | 2012-09-27 | 2014-04-03 | Empire Technology Development, Llc | Character model animation using stored recordings of player movement interface data |
| CN103530897A (en) * | 2013-09-30 | 2014-01-22 | 华为软件技术有限公司 | Movement redirection processing method and device |
Non-Patent Citations (1)
| Title |
|---|
| 罗忠祥等: ""基于时空约束的运动编辑和运动重定向"", 《计算机辅助设计与图形学学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633542A (en) * | 2016-07-19 | 2018-01-26 | 珠海金山网络游戏科技有限公司 | One kind pinches face editor and animation fusion method and system |
| CN107274467A (en) * | 2017-06-29 | 2017-10-20 | 厦门游亨世纪科技有限公司 | A kind of model animation interlock method based on Unity3D |
| CN107424203A (en) * | 2017-08-02 | 2017-12-01 | 湖南大学 | The motion retargeting method and device being combined based on Displacement mapping method and Jacobian matrix algorithm |
| CN109887070A (en) * | 2019-01-10 | 2019-06-14 | 珠海金山网络游戏科技有限公司 | A kind of virtual face's production method and device |
| CN110264551B (en) * | 2019-06-20 | 2022-10-11 | 合肥工业大学 | Motion redirection method and system |
| CN110264551A (en) * | 2019-06-20 | 2019-09-20 | 合肥工业大学 | A kind of motion retargeting method and system |
| CN112164129A (en) * | 2020-09-02 | 2021-01-01 | 北京电影学院 | No-pairing action migration method based on deep convolutional network |
| CN112233211A (en) * | 2020-11-03 | 2021-01-15 | 网易（杭州）网络有限公司 | Animation production method and device, storage medium and computer equipment |
| CN112233211B (en) * | 2020-11-03 | 2024-04-09 | 网易（杭州）网络有限公司 | Animation production method, device, storage medium and computer equipment |
| CN112927331B (en) * | 2021-03-31 | 2023-09-22 | 腾讯科技（深圳）有限公司 | Character model animation generation method and device, storage medium and electronic equipment |
| CN112927331A (en) * | 2021-03-31 | 2021-06-08 | 腾讯科技（深圳）有限公司 | Animation generation method and device of character model, storage medium and electronic equipment |
| CN113313794A (en) * | 2021-05-19 | 2021-08-27 | 深圳市慧鲤科技有限公司 | Animation migration method and device, equipment and storage medium |
| CN116012497A (en) * | 2023-03-29 | 2023-04-25 | 腾讯科技（深圳）有限公司 | Animation redirection method, device, equipment and medium |
| CN116012497B (en) * | 2023-03-29 | 2023-05-30 | 腾讯科技（深圳）有限公司 | Animation redirection method, device, equipment and medium |
| CN117710468A (en) * | 2024-02-06 | 2024-03-15 | 天度(厦门)科技股份有限公司 | Gesture redirection method, device, equipment and medium based on joint grid deformation |
| CN117710468B (en) * | 2024-02-06 | 2024-05-17 | 天度(厦门)科技股份有限公司 | Gesture redirection method, device, equipment and medium based on joint grid deformation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104183000B (en) | 2017-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104183000A (en) | Full-automatic multi-source heterogeneous motion redirecting method of quasi-man character | |
| Möller et al. | A survey on human-aware robot navigation | |
| Chen et al. | Knowledge-guided visual perception of 3-D human gait from a single image sequence | |
| CN102708582B (en) | Character movement redirecting method for heterogeneous topology | |
| CN103530897B (en) | Movement redirection processing method and device | |
| US9378589B2 (en) | Apparatus and method for generating digital clone | |
| CN111724459A (en) | Method and system for reorienting movement facing heterogeneous human skeleton | |
| CN108182719A (en) | The traveling animation producing method and device of the adaptive obstacle terrain of artificial intelligence | |
| Haworth et al. | Deep integration of physical humanoid control and crowd navigation | |
| Megaro et al. | Designing cable-driven actuation networks for kinematic chains and trees | |
| CN112650239A (en) | Multi-underwater robot formation obstacle avoidance method and system based on improved artificial potential field method | |
| Cunha et al. | A pig, an angel and a cactus walk into a blender: A descriptive approach to visual blending | |
| Zhou et al. | Learning multiscale correlations for human motion prediction | |
| CN107424203A (en) | The motion retargeting method and device being combined based on Displacement mapping method and Jacobian matrix algorithm | |
| CN113240714A (en) | Human motion intention prediction method based on context-aware network | |
| Liu et al. | Video based human animation technique | |
| Ren et al. | Customize-a-video: One-shot motion customization of text-to-video diffusion models | |
| CN110264551B (en) | Motion redirection method and system | |
| CN111515959B (en) | Programmable puppet performance robot control method and system and robot | |
| Sher et al. | An embodied, platform-invariant architecture for connecting high-level spatial commands to platform articulation | |
| CN103077545B (en) | A kind of human motion edit methods based on path | |
| CN114863562A (en) | Method and device for learning movement of multi-legged robot, electronic device, and storage medium | |
| CN114494341A (en) | Real-time completion method for optical motion capture mark points by fusing time-space constraints | |
| CN100399360C (en) | Lattice simplified restrain method of three-dimensional human model | |
| Sukthankar et al. | Modeling physical variability for synthetic MOUT agents |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220401 Address after: 230000 c3-101 / 103 / 104 / 105 / 106, phase I, innovation industrial park, No. 800, Wangjiang West Road, high tech Zone, Hefei, Anhui Patentee after: HEFEI YAOAN TECHNOLOGY CO.,LTD. Address before: Tunxi road in Baohe District of Hefei city of Anhui Province, No. 193 230009 Patentee before: Hefei University of Technology |