Clothing cloth dynamic simulation method and system based on convolutional neural network

Abstract

The invention discloses a clothing fabric dynamic simulation method and system based on a convolutional neural network, and relates to the technical field of computer simulation, and the method comprises the following steps: obtaining motion change data of a plurality of joints of a person in a dynamic simulation process of the person executing a squatting action or a stooping action, and an action compression mark is generated according to the joint movement change speed. According to the method, through action compression marking and refreshing rhythm control, the time continuity of the form change of the cloth in actions such as squatting and stooping is kept, and the local change is reasonably dispersed to adjacent wrinkle areas in combination with the area keeping constraint, so that the form of the cloth at the bent part is stabilized; and meanwhile, staged springback limitation is implemented based on a contact position sequence, so that the cloth is gradually recovered to a natural wrinkle state from inside to outside, local agglomeration and collapse are avoided, and the naturalness and visual credibility of dynamic simulation are improved.

Description

Technical Field

[0001] This invention relates to the field of computer simulation technology, specifically to a method and system for dynamic simulation of clothing fabrics based on convolutional neural networks. Background Technology

[0002] Dynamic simulation of clothing fabric based on convolutional neural networks (CNNs) refers to combining the concept of twin simulation and using CNNs to learn and represent the dynamic appearance patterns of clothing fabric during movement, stress, or posture changes, such as deformation, wrinkles, and swaying. Based on this, a virtual mapping corresponding to the real fabric state is constructed, enabling continuous deduction and reproduction of fabric changes over time. This method extracts a large number of temporal features formed by the fabric under different actions, environmental conditions, and interaction scenarios. Within the twin simulation framework, it simultaneously characterizes the response relationship between the physical and virtual fabrics, capturing the intrinsic correlation between local texture changes and overall morphological evolution. This allows the simulation process to quickly generate dynamic representations highly consistent with actual fabric behavior based on input actions or state changes. Overall, this simulation method emphasizes the learning and reproduction of fabric dynamic response patterns supported by a twin simulation mechanism, focusing on visual continuity, morphological coordination, and stable expression of follow-up features. It is suitable for applications requiring high realism of fabric dynamics, such as virtual try-on, digital clothing display, and animation production.

[0003] The existing technology has the following shortcomings: In existing technologies, when a person performs multi-joint coordinated movements such as squatting or bending over in clothing fabric dynamic simulation based on convolutional neural networks, the fabric state changes rapidly at multiple joints simultaneously, easily leading to local response imbalances during the simulation. Specifically, certain areas of the fabric abnormally shrink within a very short time, rapidly concentrating their shape locally and forming a clumpy structure that does not conform to the stress and deformation patterns of real clothing. This anomaly disrupts the original continuous folds of the fabric, causing the overall appearance of the garment to appear abruptly collapsed, especially noticeable during transitions between movements. Because this type of problem often occurs at bending points on the body, it is more easily magnified and identified under close-up observation or close-up shots, thus severely affecting the naturalness and credibility of the clothing dynamic simulation.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for dynamic simulation of clothing fabrics based on convolutional neural networks, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic simulation of clothing fabric based on convolutional neural networks, comprising the following steps: During the dynamic simulation of a character performing a squatting or bending motion, motion change data of multiple joints of the character are acquired, and motion compression markers are generated based on the speed of joint motion changes to identify the time segments in which the fabric is prone to rapid contraction during the dynamic simulation. Based on motion compression markers, the time segments corresponding to the motion compression markers are subdivided, and the cloth screen refresh rhythm is unified within the subdivided time segments. At the same time, continuous transition processing is introduced between adjacent screens to reduce the degree of abrupt changes in cloth form in the time dimension. Based on a unified refresh rate for the fabric screen, an area maintenance constraint is introduced for the fabric area located near the bending position of the human body. When the area change of the fabric area exceeds the preset range, the amount of change exceeding the preset range is distributed to the adjacent fold area to suppress abnormal shrinkage of the fabric area. While maintaining the area constraint, determine the order of contact positions between the fabric and the human body, and adjust the displacement of the fabric closest to the human body according to the contact position order, while unfolding the fabric area on the outside to maintain the overall coordination of the fabric shape. Based on the contact position sequence, a phased rebound restriction process is implemented for fabric areas prone to morphological aggregation during dynamic simulation, so that the fabric shape gradually returns to a natural wrinkled state, thereby avoiding the formation of abnormal aggregation.

[0007] Preferably, the steps for generating action compression markers are as follows: Multiple joints involved in lower limb flexion and extension and trunk rotation are continuously tracked, and spatial displacement change information of each joint in consecutive frames is collected to form a joint motion change data sequence. Based on the joint motion change data sequence, the velocity of joint motion change is calculated, and the velocity of joint motion change is mapped to the corresponding time position to form the temporal distribution of joint motion change. Based on the temporal distribution of joint motion changes, extract time segments that change synchronously and at concentrated rates, and refine these time segments to obtain sub-time segments. The sub-time segments with concentrated rate of change are associated with the cloth coverage position during the motion process, and motion compression markers are generated to identify the time segments in which the cloth is prone to rapid shrinkage during dynamic simulation.

[0008] Preferably, the sub-time segments in which the speed of joint movement changes are concentrated are determined by analyzing the synchronicity of the speed of joint movement changes at adjacent time positions, and motion compression marks are generated only in time segments in which synchronous changes continue to exist, so that the motion compression marks correspond to the stage of multi-joint coordinated change of the character.

[0009] Preferably, the steps for processing time segments based on action compression tags are as follows: Read the action compression markers and extract the boundary information of the corresponding time segments. Map the time segment boundary information to the time axis of the cloth dynamic simulation and establish an index relationship between the time segments and the cloth image sequence. Based on the index relationship, the time interval between adjacent frames within the time segment is counted, and the time segment is further subdivided according to the distribution of time intervals to form a continuously arranged subdivided time segment. Set a uniform cloth screen refresh rhythm within the subdivided time segments, and perform time alignment processing on the original cloth screens within the subdivided time segments so that the cloth screens are arranged according to the refresh rhythm. Based on a unified cloth screen refresh rhythm, a continuous transition process is introduced between adjacent cloth screens, so that the cloth form forms a continuous and stable change process within the subdivided time segments, thereby reducing the degree of abrupt changes in the cloth form in the time dimension.

[0010] Preferably, when introducing continuous transition processing between adjacent cloth images, the cloth form state corresponding to the adjacent images is used as the transition boundary. Under a unified cloth image refresh rhythm, the form changes between adjacent images are gradually unfolded, and the continuous transition processing and the processing within the segment are kept consistent at the boundary of the subdivided time segment, so that the cloth images form a continuous sequence of changes throughout the entire time segment.

[0011] Preferably, the steps for introducing area preservation constraints while maintaining a consistent cloth screen refresh rate are as follows: At each refresh time position, the bending angle of the joint and the curvature distribution of the human body surface are combined to determine the adjacent position of the human body bending, and the adjacent position of the human body bending is mapped to the corresponding cloth area to extract the geometric description of the cloth area. Based on the geometric description of the cloth area, the area change between adjacent refresh time positions is calculated, and the area is constrained by constructing an area with reference area, while the preset range is determined according to the speed of joint movement. When the area change exceeds the preset range, the change exceeding the preset range is converted into a shape adjustment amount and distributed to the adjacent pleated areas according to the adjacency relationship of the fabric areas. At the same time, a corresponding reverse adjustment is applied to the fabric areas. By keeping the area constraint consistent with the refresh rhythm of the uniform cloth screen, the allocated shape adjustment amount is continuously introduced between adjacent refresh time positions, thereby suppressing abnormal shrinkage of the cloth area and forming a stable shape evolution process.

[0012] Preferably, the preset range is adjusted synchronously according to the speed of joint movement. The amount of change exceeding the preset range is allocated according to the shared boundary length between the fabric area and the adjacent fold area and the consistency of the fold direction, so that the shape adjustment is transmitted along the fold extension direction and maintains continuous change between adjacent refresh time positions.

[0013] Preferably, the steps for determining the contact position sequence and adjusting the fabric displacement while maintaining area constraints are as follows: At each refresh time position, the set of contact positions between the fabric and the human body is determined based on the fabric shape and the human body surface shape, and the contact duration and contact area changes of the contact positions are recorded to form contact description information. Based on the contact description information, the contact position sequence is determined according to the contact duration, contact area change and human body bending direction, so that the contact position sequence is arranged from the position closest to the human body to the outer position. Based on the order of contact positions, the displacement of the fabric closest to the human body is adjusted first, so that the displacement of the fabric closest to the human body unfolds step by step along the direction of the human body surface and remains continuous. After the fabric close to the human body has shifted and changed position, the outer fabric area is unfolded to create a continuous spatial hierarchy between the outer fabric area and the area close to the human body, thereby maintaining the overall harmony of the fabric shape.

[0014] Preferably, the steps for implementing phased rebound limiting treatment based on the contact position sequence are as follows: After adjusting the fabric displacement changes under the contact position sequence, the fabric morphology is analyzed along the contact position sequence, the fabric areas prone to morphology aggregation are marked, and the corresponding area tracking relationship is established. Based on the fabric area where morphology aggregation is likely to occur, the springback restriction treatment is divided into stages according to the contact position sequence, and the springback target morphology and springback restriction range are set for each stage. Under a unified cloth screen refresh rhythm, the rebound restriction process is applied to the cloth areas prone to form aggregation in stages, so that the rebound process is advanced from the area close to the human body to the outer area. Maintaining continuity of phased rebound between adjacent refresh time points allows the fabric shape to gradually recover to a natural wrinkled state under the phased rebound restriction process, thereby avoiding the formation of abnormal clustering.

[0015] A dynamic simulation system for clothing fabric based on convolutional neural networks includes a motion compression and marking module, a timing refresh control module, an area preservation constraint module, a contact sequence control module, and a springback limit recovery module. The motion compression tagging module acquires motion change data of multiple joints of the character during the dynamic simulation of the character performing a squatting or bending motion, and generates motion compression tags based on the speed of joint motion change. These tags are used to identify the time segments in which the fabric is prone to rapid contraction during the dynamic simulation. The timing refresh control module, based on the action compression mark, subdivides the time segment corresponding to the action compression mark, and unifies the cloth screen refresh rhythm within the subdivided time segment. At the same time, it introduces continuous transition processing between adjacent screens to reduce the degree of abrupt changes in cloth form in the time dimension. The area retention constraint module introduces an area retention constraint for the fabric area located near the bending position of the human body, based on the unified refresh rhythm of the fabric screen. When the area change of the fabric area exceeds the preset range, the amount of change exceeding the preset range is distributed to the adjacent fold area to suppress the abnormal shrinkage of the fabric area. The contact sequence control module determines the contact position sequence between the fabric and the human body while maintaining the area constraint. Based on the contact position sequence, it prioritizes adjusting the displacement of the fabric closest to the human body, while unfolding the fabric area on the outside to maintain the overall coordination of the fabric shape. The springback limitation and recovery module, based on the contact position sequence, performs phased springback limitation processing on fabric areas prone to morphological aggregation during dynamic simulation, so that the fabric shape gradually returns to a natural wrinkled state, thereby avoiding the formation of abnormal aggregation.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention introduces motion compression markers during the dynamic simulation of a character performing a squatting or bending motion. By unifying the cloth screen refresh rhythm and ensuring continuous transitions around the time segments corresponding to these compression markers, the change in cloth form over time is transformed from abrupt adjustments to a continuous unfolding process. This effectively alleviates the problems of abrupt changes and imbalances in cloth form during the transition phase of the action. Combined with area preservation constraints and a mechanism for distributing changes to adjacent fold areas, the cloth near the bending position of the human body maintains a stable form distribution during rapid posture changes. The cloth folds naturally extend along their existing direction, and the overall appearance remains consistent across continuous frames, significantly improving the smoothness and continuity of cloth form evolution during dynamic simulation.

[0017] This invention introduces a sequence of contact positions based on the relationship between fabric and the human body, and combines this with phased rebound control. It regulates the rebound of fabric areas prone to clumping from the inside out, releasing it in stages, ensuring the fabric's shape follows a reasonable spatial hierarchy during recovery. By uniformly constraining the displacement changes in areas close to the human body, the unfolding process of the outer fabric areas, and the rebound release rhythm, the fabric avoids localized clumping and abnormal collapse as it recovers to its natural wrinkled state. The overall shape exhibits a dynamic response effect that coordinates with human movements, thereby enhancing the naturalness, stability, and visual credibility of the clothing fabric's dynamic simulation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method for dynamic simulation of clothing fabric based on convolutional neural networks according to the present invention.

[0020] Figure 2 This is a schematic diagram of the modules of the clothing fabric dynamic simulation system based on convolutional neural networks of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] This invention provides, for example Figure 1 The illustrated method for dynamic simulation of clothing fabric based on convolutional neural networks includes the following steps: During the dynamic simulation of a character performing a squatting or bending motion, motion change data of multiple joints of the character are acquired, and motion compression markers are generated based on the speed of joint motion changes to identify the time segments in which the fabric is prone to rapid contraction during the dynamic simulation. During the character's continuous movements, including lower limb flexion or trunk flexion, in order to accurately depict the temporal positions of rapid shape changes in the cloth during the dynamic simulation phase, the character's joint motion information is hierarchically processed according to the chronological order, and motion compression marks that can be used for subsequent cloth timing control are gradually formed. The specific implementation steps are as follows: In the initial stage of motion simulation, multiple joints in the human skeletal structure involved in lower limb flexion and extension and trunk rotation are continuously tracked and recorded. Spatial displacement change information of each joint in consecutive frames is collected according to a uniform time sampling scale, and the collected spatial displacement change information is converted into corresponding joint motion change data sequences. During this process, the displacement change of the same joint between adjacent time nodes is compared and calculated to obtain continuous change data reflecting the amplitude of joint motion change. This allows the motion state of each joint throughout the entire motion cycle to be described in time series form, and provides a unified data foundation for subsequent analysis of joint motion change speed.

[0023] After obtaining the complete joint motion change data sequence, the data sequence is processed in chronological order. Based on the relationship between the change amplitudes of adjacent time nodes, the joint motion change velocity is calculated, and the obtained joint motion change velocity is mapped to the corresponding time position, thus forming a temporal distribution relationship reflecting the changes in the speed of movement of each joint during the action. On this basis, the joint motion change velocities of multiple joints are comprehensively analyzed. When multiple joints show synchronous changes and concentrated fluctuations in change velocity within the same time period, this time period is marked as the joint coordinated change time period, so that a clear correspondence is formed between the joint motion change velocity and the specific time segment.

[0024] After identifying the time periods of joint coordination changes, the spatial movement direction and velocity distribution of the joints within these time periods are combined to further refine the joint coordination change time periods. Long-duration joint coordination change time periods are divided into several continuous sub-time segments, and the trend of joint motion velocity changes is analyzed within each sub-time segment to distinguish different stages where the joint motion velocity changes are rising, maintaining, or falling. After completing the sub-time segment division, the sub-time segments where the joint motion velocity changes are concentrated in a rising or falling state are extracted, so that the process of joint motion velocity changes can be finely characterized on the time axis, laying the foundation for the generation of motion compression markers.

[0025] After obtaining the sub-time segments where the speed of joint movement changes is concentrated, the sub-time segments are correlated with the cloth coverage position during the character's movement. Based on the concentration and duration of the speed of joint movement changes, the corresponding sub-time segments are assigned motion compression marks, which clearly indicate the time segments in which the cloth is prone to rapid contraction during dynamic simulation. After the motion compression marks are generated, they are aligned with the original action timeline in chronological order, so that the motion compression marks can serve as the time basis for subsequent cloth screen refresh rhythm control and form transition processing. This provides clear, continuous, and stable time guidance for fine-tuning specific time segments during cloth dynamic simulation.

[0026] Based on motion compression markers, the time segments corresponding to the motion compression markers are subdivided, and the cloth screen refresh rhythm is unified within the subdivided time segments. At the same time, continuous transition processing is introduced between adjacent screens to reduce the degree of abrupt changes in cloth form in the time dimension. To ensure that the time positions corresponding to the motion compression markers can be precisely expanded and form a stable temporal presentation, the time segments marked by the motion compression markers are processed according to the consistency of the time axis and the rhythm of the scene, resulting in three types of results: subdivided time segments, a unified cloth scene refresh rhythm, and continuous transition processing between adjacent scenes. These three types of results are kept in correspondence within the same time axis. The specific implementation steps are as follows: Read the motion compression markers and extract the time segment boundary information corresponding to the motion compression markers. Map the time segment boundary information to the time axis coordinates of the cloth dynamic simulation and establish an index relationship between the time segment and the cloth image sequence. The index relationship includes the start time position, the end time position, and the set of existing image sequence numbers within the time segment. After establishing the index relationship, the time interval between adjacent images within the time segment is statistically analyzed to obtain the time interval distribution within the time segment. At the same time, the connection position between the time segment boundary and the adjacent image is recorded so that subsequent subdivision processing can maintain continuous connection at the adjacent position of the boundary and avoid boundary breaks. Based on index relationships and time interval distribution, time segments are subdivided. Each time segment is divided into multiple subdivided time segments with a fixed time granularity. The fixed time granularity is taken from the representative value of the time interval distribution within the time segment, which is determined by the median position of the time interval distribution. This ensures that the fixed time granularity is on the same order of magnitude as the original frame time interval and maintains temporal consistency. After the subdivided time segments are formed, subdivided segment identifiers are assigned to them. These identifiers maintain a one-to-one correspondence with motion compression markers. Each subdivided segment identifier includes a subdivision sequence number, subdivision start time position, and subdivision end time position. This results in the time segment corresponding to the motion compression marker being split into consecutively arranged subdivided time segments, providing a clear segmentation basis for unifying the cloth frame refresh rhythm. Based on the subdivided time segments, a unified cloth screen refresh rhythm is established. This unified refresh rhythm includes three parameters: refresh start point, refresh interval, and refresh end point. The refresh start point is the starting time position of the subdivided time segment, the refresh end point is the ending time position of the subdivided time segment, and the refresh interval is a fixed time granularity. This ensures that each refresh time position within the subdivided time segment falls between the refresh start point and the refresh end point, forming an evenly spaced arrangement. After establishing the unified cloth screen refresh rhythm, the original cloth screens within the subdivided time segments undergo time alignment processing. The time positions of the original cloth screens are aligned to the nearest refresh time position, and the mapping relationship between the original cloth screen sequence set and the refresh time position set is maintained during the alignment process. This ensures that the cloth screen refresh rhythm within the subdivided time segments is unified and predictable, while preserving the original cloth screen sequence on the timeline, providing a stable input sequence for introducing continuous transition processing between adjacent screens. Based on the unified refresh rhythm and time alignment of the cloth screen, a continuous transition process is introduced between adjacent screens. This continuous transition process uses the corresponding cloth form states of adjacent screens as constraints at both ends, gradually unfolding these constraints within the refresh interval. This gradual transition unfolding includes four elements: the transition start form, the transition end form, the number of transition steps, and the transition weight sequence. The transition start form is taken from the cloth form state of the previous screen, and the transition end form is taken from the cloth form state of the next screen. The number of transition steps is determined by both the refresh interval and a fixed time granularity. The transition weight sequence monotonically increases within the number of transition steps, causing the contribution of the transition start form to gradually decrease and the contribution of the transition end form to gradually increase, thus forming multi-level transition screens within the refresh interval. The transition between adjacent frames is achieved by introducing a boundary connection transition near the boundary of the subdivided time segment. This boundary connection transition incorporates adjacent frames outside the boundary into the constraints at both ends, ensuring that the continuous transition processing at the boundary of the time segment maintains a consistent refresh interval and transition weight sequence with the continuous transition processing within the segment. Through the combined effect of a unified cloth frame refresh rhythm and continuous transition processing, the degree of abrupt change in cloth form over time is reduced, resulting in a transformation of form change between adjacent frames from a single jump to a multi-level progressive change. This forms a continuous and stable cloth frame sequence within the time segment corresponding to the action compression mark, providing a consistent time scale and frame rhythm as a foundation for the subsequent introduction of area maintenance constraints and contact position sequence processing.

[0027] Based on a unified refresh rate for the fabric screen, an area maintenance constraint is introduced for the fabric area located near the bending position of the human body. When the area change of the fabric area exceeds the preset range, the amount of change exceeding the preset range is distributed to the adjacent fold area to suppress abnormal shrinkage of the fabric area. To stably represent the morphological evolution of bending areas within a continuous cloth image sequence formed by a unified cloth image refresh rhythm, and to avoid concentrated contraction in local areas within a short period of time, an executable area preservation constraint is established around the bend of the human body. When the area change exceeds a preset range, the change is distributed along the direction of the folds, thereby forming a synergistic relationship of constraint and mitigation in both the temporal and spatial dimensions. The specific implementation steps are as follows: At each refresh time point corresponding to the unified cloth screen refresh rhythm, the spatial range of the near-bending position of the human body is determined by combining the joint bending angle in the character's posture and the curvature distribution of the human body surface. This spatial range is then mapped to the corresponding cloth area in the cloth form. The cloth area is expressed as a continuous cloth surface patch, and the patch boundary is determined along the direction of the cloth texture and folds, so that the cloth area covers the near position of the abdomen when bending over or the near position of the hip and knee flexion when squatting. After determining the cloth area, the geometric description of the cloth area is extracted based on the cloth form state at the same refresh time point. The geometric description includes the ordered set of boundary points of the cloth area boundary, the set of facets inside the cloth area, and the adjacency relationship between facets. This ensures that the cloth area has a traceable boundary and internal structure at each refresh time point and provides a consistent geometric basis for subsequent area change calculations. After obtaining the geometric description of the cloth area, the area change of the cloth area between adjacent refresh time positions in the unified cloth screen refresh rhythm is calculated as a time reference. The area change is obtained by the difference between the total area of ​​the cloth area's internal facets at the current refresh time position and the total area at the previous refresh time position. The area change and the reference area of ​​the cloth area are used together to construct the area maintenance constraint. The reference area is the total area of ​​the cloth area at the beginning of the time segment corresponding to the action compression mark, or a stable representative value of the total area within that time segment, so that the area maintenance constraint is consistent with the high change period corresponding to the action compression mark. When constructing the area maintenance constraint, the upper and lower limits of the preset range are established based on the reference area. The width of the preset range is related to the rate of change of the human body bending angle. The rate of change of the human body bending angle is the joint movement change speed obtained during the automatic compression mark generation process, so that the preset range is adjusted synchronously with the change of bending intensity, and the area change at the same refresh time position can be directly compared with the preset range. After establishing a correspondence between the area change and the preset range, when the area change exceeds the preset range, the amount of change exceeding the preset range is determined and converted into an allocable shape adjustment amount. This shape adjustment amount is expressed by the outward expansion displacement of the fabric region boundary point set and the unfolding displacement of the fabric region's internal facet set. The outward expansion displacement and unfolding displacement are set along the tangential direction of the fold direction to ensure that the application direction of the shape adjustment amount is consistent with the natural extension direction of the folds. Subsequently, adjacent fold regions are selected from outside the fabric region based on the adjacency relationship of the fabric regions. Adjacent fold regions are expressed as fold areas sharing boundary segments or adjacent facets with the fabric region, and each adjacent fold region is assigned a shape adjustment amount. The weights for assigning folded areas are determined by the shared boundary length between adjacent folded areas and the fabric area, the consistency of fold direction, and the current curvature change trend of adjacent folded areas. This ensures that the weights reflect the ability of adjacent folded areas to absorb shape adjustments. After the weights are determined, changes exceeding the preset range are distributed among adjacent folded areas according to the weights. At the current refresh time, corresponding shape adjustments are applied to each adjacent folded area, while a reverse offset adjustment corresponding to the total amount is applied to the fabric area. This mitigates the shrinkage trend of the fabric area and shifts the shape changes to the surrounding folded areas, thereby preventing the shape of the fabric area from concentrating locally and forming a clustered structure. After allocating changes exceeding a preset range, the area maintenance constraint is bound to a unified cloth screen refresh rhythm. This ensures that each refresh time position performs morphological adjustments according to the same reference area, the same preset range definition, and the same weighting calculation rules. The weighting remains continuously variable between adjacent refresh time positions, updating gradually with changes in fold direction and curvature trend without jumps. Simultaneously, near the time segment boundary corresponding to the action compression marker, the allocated morphological adjustment amount maintains a consistent time interpolation rhythm with the transition screen obtained through continuous transition processing. This allows the allocated morphological adjustment amount to be introduced step-by-step within the transition screen, creating a continuous morphological evolution trajectory between the cloth area and adjacent fold areas in the time dimension. Through this processing, the cloth area near the human body's bending position is subject to area maintenance constraints under a unified cloth screen refresh rhythm. When the area change exceeds a preset range, the excess change is allocated to adjacent fold areas. Abnormal contraction of the cloth area is suppressed, and the abrupt change in fold morphology in the time dimension is reduced. This provides a stable basic input for determining the contact sequence between the cloth and the human body and implementing displacement adjustments.

[0028] While maintaining the area constraint, determine the order of contact positions between the fabric and the human body, and adjust the displacement of the fabric closest to the human body according to the contact position order, while unfolding the fabric area on the outside to maintain the overall coordination of the fabric shape. To ensure that the shape adjustment of adjacent bending locations, after maintaining area constraints, aligns with changes in human posture and avoids reverse pulling or accumulation of the outer fabric area at the same refresh time, a contact position sequence is established based on the fit between the fabric and the human body within the time segment corresponding to the motion compression mark. Then, the displacement changes of the fabric closest to the human body are prioritized for adjustment according to this contact position sequence. Once the displacement changes of the fabric closest to the human body stabilize, the outer fabric area is simultaneously expanded. This creates a consistent spatial hierarchy between local constraints and overall extension of the fabric shape. The specific implementation steps are as follows: At each refresh time position after the area constraint and the distribution of changes exceeding the preset range are completed, the set of contact positions between the cloth and the human body is determined based on the cloth shape state and the human body surface shape state. The set of contact positions is obtained through the minimum distance relationship between the cloth surface and the human body surface in spatial position. The cloth position where the minimum distance falls within the contact threshold range is recorded as the contact position. The contact positions are classified according to human body parts as the position near the abdomen when bending over, the position near the torso when bending laterally, the position near the hip when squatting, the position near the knee when bending, and the position near the back of the buttocks, so that the set of contact positions has a clear correspondence with the bending parts related to the action. After the contact position set is classified, the contact duration and contact area change of each contact position between adjacent refresh time positions are recorded in time sequence along the uniform cloth screen refresh rhythm. The contact duration is obtained by the duration of the contact position within the contact threshold range at consecutive refresh time positions, and the contact area change is obtained by the change of the number of contact positions in a certain neighborhood around the contact position, thus forming contact description information that reflects the contact stability and contact expansion trend, and the contact description information is used as the input basis for determining the contact position sequence. After establishing the contact description information, the contact position sequence is determined based on the contact duration, contact area change, and the direction of the human body bending angle change. The contact position sequence consists of multiple contact levels, which are arranged from the inside out according to the degree of proximity to the human body and contact stability. The degree of proximity to the human body is determined by the size relationship of the minimum spacing, and the contact stability is determined by the combination relationship of the contact duration and the contact area change, so that the contact position with the smaller minimum spacing and the longer contact duration enters the earlier contact level. After forming the contact level, the contact positions in each contact level are sorted along the tangential direction of the human body surface. The sorting is determined based on the positional continuity in the tangential direction of the human body surface and the consistency of the fabric fold direction, so that adjacent contact positions are continuously distributed in space and maintain the same or nearly same direction relationship with the fold direction, thereby obtaining the contact position sequence that can be executed at the same refresh time position. The contact position sequence is then linked to the allocation weight in the area retention constraint of the previous step, so that the contact position sequence can inherit the allocated shape adjustment direction and fold direction after the change amount of the adjacent bending position is allocated. After the contact position sequence is established, the fabric displacement closest to the human body is adjusted first according to this sequence. The fabric displacement closest to the human body starts from the foremost contact position in the contact layer, and displacement corrections are applied sequentially according to the contact position sequence. The displacement corrections are formed by both the normal and tangential directions along the human body surface. The normal displacement is used to maintain a stable fit between the fabric and the human body surface within the contact threshold range, while the tangential displacement is used to allow the fabric to slide along the human body surface to accommodate bending posture changes. When applying displacement corrections, a displacement distribution ratio is introduced to constrain the spatial propagation range of the displacement correction. The displacement distribution ratio and area are maintained within the constraint distribution. By maintaining consistent weights, displacement correction is prioritized to propagate in the direction of the allocated change amount while keeping the area constraint, and a displacement attenuation relationship is formed within the contact layer, decreasing from the contact position to the adjacent cloth position. This ensures that the displacement change of cloth close to the human body is continuously diffused in space without forming local jumps. At the same time, the uniform cloth screen refresh rhythm and the transition screen obtained by continuous transition processing are incorporated into the time unfolding process of displacement correction. This allows displacement correction to be introduced step by step within the same refresh interval and to be consistent with the continuous transition processing of adjacent screens. As a result, the displacement change of cloth close to the human body is consistent with the high change rhythm of the time segment corresponding to the motion compression mark in the time dimension. After the fabric displacement changes close to the human body are completed sequentially according to the contact positions, the outer fabric area is unfolded. The outer fabric area is defined by the non-contact fabric patch extending outward from the last contact position in the contact hierarchy. The boundary of the non-contact fabric patch is connected to the boundary of the contact position set and extends along the fold direction, giving the outer fabric area a clear spatial boundary with the area close to the human body. The unfolding process uses the fabric displacement changes close to the human body as boundary constraints, gradually unfolding the outer fabric area along the fold direction. This gradual unfolding includes three continuous operations: outer boundary stretching, fold unfolding, and boundary smoothing. Outer boundary stretching adjusts the outer edge of the outer fabric area synchronously with the displacement changes of the area close to the human body; fold unfolding releases the fold amplitude within the outer fabric area step by step under the action of the allocated changes; and boundary smoothing maintains spatial continuity at the connection between the outer fabric area and the area close to the human body. During the processing, the reference area and preset range that maintain area constraints are used as the constraint benchmarks for the expansion range, ensuring that the expansion amount of the outer fabric area corresponds to the area change of the adjacent bending position. The distribution result of the change amount exceeding the preset range is used as the guide for the expansion direction, so that the outer fabric area expands along the distribution direction of the adjacent fold area when receiving the change amount without producing reverse folding. Through the above processing of prioritizing the adjustment of the fabric displacement change close to the human body according to the contact position sequence and simultaneously expanding the outer fabric area, the area close to the human body and the outer fabric area form a spatial hierarchy consistency from the inside to the outside within the same refresh time position. Within the time frame of unified fabric screen refresh rhythm and continuous transition processing, a continuous morphological evolution trajectory is maintained, thereby maintaining the overall coordination of the fabric shape and providing a stable displacement change basis input for subsequent phased rebound restriction processing of fabric areas prone to shape aggregation.

[0029] Based on the contact position sequence, a phased springback restriction process is implemented for the fabric area that is prone to morphological aggregation during dynamic simulation, so that the fabric shape gradually returns to a natural wrinkled state, thereby avoiding the formation of abnormal aggregation. To suppress the imbalance of shape rebound between the adjacent bending position and the outer unfolding area under the constraint of the contact position sequence, and to avoid local clustering of the fabric during the transition phase corresponding to the motion compression mark time segment, a phased rebound restriction process is introduced around the fabric area prone to shape clustering. This process advances from the inside out and from the front to the back along the contact position sequence, and the rebound amplitude is released gradually within the time frame of a unified fabric screen refresh rhythm and continuous transition processing. This allows the fabric shape to gradually return to a natural wrinkled state and avoids the formation of abnormal clustering. The specific implementation steps are as follows: After adjusting the fabric displacement changes close to the human body and unfolding the outer fabric area under the contact position sequence, the clustering risk of the fabric shape is located at each refresh time position. The clustering risk location is based on the contact position sequence, unfolding sequentially from the contact position at the earlier contact level outwards. Areas along this direction where the local shape converges towards the center, folds are densely overlapping, and the boundary length of fabric surface areas is shortened are marked as fabric areas prone to shape clustering. When marking fabric areas prone to shape clustering, the distribution result of the amount of change exceeding the preset range in the area maintenance constraint is used as a spatial guide, prioritizing the adjacent folds that bear the change. The connecting band between the wrinkled area and the area close to the human body is included in the boundary range of the cloth area where the pattern is likely to gather, along with the connecting band and its extended neighboring area. This ensures that the cloth area where the pattern is likely to gather has a consistent spatial correspondence with the contact position sequence and area constraint distribution direction. At the same time, in order to ensure that the subsequent phased rebound restriction processing can remain continuous in the time dimension, the boundary point set of the cloth area where the pattern is likely to gather is kept consistent with the index number at adjacent refresh time positions. This makes the same boundary point have a traceable correspondence between consecutive frames, and this correspondence is bound to the transition frame sequence generated by the continuous transition processing to form a regional tracking link that runs through the refresh interval.

[0030] After obtaining the fabric areas prone to pattern aggregation and their tracking links, a phased rebound limitation process is constructed, including phase divisions and parameters. The phase divisions are based on the contact position sequence, setting at least three consecutive phases and expandable to more. The phases sequentially include a close-to-human phase, a connecting zone phase, and an outward expansion phase. The close-to-human phase corresponds to the neighborhood of the earlier contact position in the contact layer; the connecting zone phase corresponds to the transition zone between the close-to-human area and adjacent wrinkled areas; and the outward expansion phase corresponds to the outward expansion area of ​​adjacent wrinkled areas. In the phase parameter settings, a rebound target shape and rebound limitation amplitude are set for each phase. The rebound target shape starts from the fabric pattern state at the current refresh time position and is combined with the action compression marker within the corresponding time segment. The direction of the folds and the trend of curvature changes are determined to ensure that the rebound target shape continues the existing fold lines in terms of shape and avoids reversal. The rebound limit is set with the refresh interval of the uniform fabric screen refresh rhythm as the time unit to set the stage release ratio. The stage release ratio increases monotonically in the stage sequence, so that the release ratio is the smallest when close to the human body and the release ratio is the largest when unfolding to the outside. This allows the rebound process to gradually open from the area close to the human body to the unfolding area, and is consistent with the priority adjustment logic of the previous contact position sequence. At the same time, the upper and lower limits of the preset range in the area maintenance constraint are used as the dimensional benchmark of the rebound limit, so that the change in the area of ​​the fabric area during the rebound process remains within the change band corresponding to the preset range, avoiding the rebound release from causing local abnormal contraction again.

[0031] After the phase division and phase parameters are determined, phased rebound restriction processing is implemented to ensure that areas of cloth prone to morphological aggregation rebound sequentially along the phase order within each refresh interval. Specifically, multi-level rebound images with the same phase release ratio are introduced between adjacent images with a unified cloth image refresh rhythm, and these multi-level rebound images are inserted into the transition image sequence generated by continuous transition processing, ensuring that the rebound process and the transition images maintain the same interpolation rhythm. In the phase close to the human body, rebound restrictions are applied to the neighborhood of the contact position at the earlier contact level. The rebound restriction primarily focuses on limiting displacement changes, with the restriction direction combining the normal and tangential directions along the human body surface. The displacement change of the cloth close to the human body is used as a boundary constraint to prevent the area close to the human body from experiencing reverse displacement and detachment during the rebound process. In the connecting strip... In the first stage, a springback constraint is applied to the wrinkled areas within the connecting strip. The springback constraint is mainly released in segments based on the wrinkle amplitude, with the release direction set along the wrinkle direction. The distribution result of the amount of change exceeding the preset range in the area maintenance constraint is used as the guide for the release direction, so that the connecting strip will preferentially unfold along the distribution direction during the springback process and will not gather locally. In the outer unfolding stage, a springback constraint is applied to the outer unfolding area. The springback constraint is mainly released gradually based on the unfolding displacement. The result of the unfolding process of the outer fabric area is used as the starting state, and the range of action of the outer boundary stretching and wrinkle unfolding is gradually expanded according to the stage release ratio in each refresh interval. This allows the outer unfolding area to gradually take over the shape recovery trend from the connecting strip, so that the fabric area prone to shape gathering forms a continuous springback chain from the inside to the outside and avoids local clustering.

[0032] After the phased bounce limitation process is completed, the phased bounce limitation process is continuously associated with the contact position sequence, ensuring that the phased bounce continues along the same contact position sequence at the start of the next refresh time position. A phase transition process is introduced near the boundary of the time segment corresponding to the action compression mark. This transition process smoothly connects the phase release ratio at the end of the previous time segment with the phase release ratio at the beginning of the next time segment, ensuring that the phase release ratio changes continuously on the time axis and maintains a monotonic change trend consistent with the transition weight sequence of the continuous transition process. Simultaneously, the area tracking link for cloth areas prone to pattern aggregation is updated, ensuring that the boundary point set remains consistent after experiencing the bounce limitation. Maintaining consistent index numbers and ensuring the correspondence between the same boundary points across consecutive frames allows subsequent refresh times to continue the phased bounce restriction process along the same spatial path. Through the aforementioned phase division, phase parameter settings, phase execution, and phase connection based on the contact position sequence, the bounce process of cloth areas prone to form aggregation is gradually released in the time dimension by multiple bounce frames, and in the spatial dimension, it spreads outward from the area close to the human body. The cloth form gradually returns to a natural wrinkled state, the abnormal aggregation phenomenon is suppressed, and a consistent collaborative relationship is formed with area constraints, unified cloth frame refresh rhythm, continuous transition processing, and contact position sequence.

[0033] This invention introduces motion compression markers during the dynamic simulation of a character performing a squatting or bending motion. By unifying the cloth screen refresh rhythm and ensuring continuous transitions around the time segments corresponding to these compression markers, the change in cloth form over time is transformed from abrupt adjustments to a continuous unfolding process. This effectively alleviates the problems of abrupt changes and imbalances in cloth form during the transition phase of the action. Combined with area preservation constraints and a mechanism for distributing changes to adjacent fold areas, the cloth near the bending position of the human body maintains a stable form distribution during rapid posture changes. The cloth folds naturally extend along their existing direction, and the overall appearance remains consistent across continuous frames, significantly improving the smoothness and continuity of cloth form evolution during dynamic simulation.

[0034] This invention introduces a sequence of contact positions based on the relationship between fabric and the human body, and combines this with phased rebound control. It regulates the rebound of fabric areas prone to clumping from the inside out, releasing it in stages, ensuring the fabric's shape follows a reasonable spatial hierarchy during recovery. By uniformly constraining the displacement changes in areas close to the human body, the unfolding process of the outer fabric areas, and the rebound release rhythm, the fabric avoids localized clumping and abnormal collapse as it recovers to its natural wrinkled state. The overall shape exhibits a dynamic response effect that coordinates with human movements, thereby enhancing the naturalness, stability, and visual credibility of the clothing fabric's dynamic simulation.

[0035] This invention provides, for example Figure 2 The clothing fabric dynamic simulation system shown is based on a convolutional neural network and includes a motion compression and marking module, a timing refresh control module, an area preservation constraint module, a contact sequence control module, and a springback limitation recovery module. The motion compression tagging module acquires motion change data of multiple joints of the character during the dynamic simulation of the character performing a squatting or bending motion, and generates motion compression tags based on the speed of joint motion change. These tags are used to identify the time segments in which the fabric is prone to rapid contraction during the dynamic simulation. The timing refresh control module, based on the action compression mark, subdivides the time segment corresponding to the action compression mark, and unifies the cloth screen refresh rhythm within the subdivided time segment. At the same time, it introduces continuous transition processing between adjacent screens to reduce the degree of abrupt changes in cloth form in the time dimension. The area retention constraint module introduces an area retention constraint for the fabric area located near the bending position of the human body, based on the unified refresh rhythm of the fabric screen. When the area change of the fabric area exceeds the preset range, the amount of change exceeding the preset range is distributed to the adjacent fold area to suppress the abnormal shrinkage of the fabric area. The contact sequence control module determines the contact position sequence between the fabric and the human body while maintaining the area constraint. Based on the contact position sequence, it prioritizes adjusting the displacement of the fabric closest to the human body, while unfolding the fabric area on the outside to maintain the overall coordination of the fabric shape. The springback limitation and recovery module, based on the contact position sequence, performs phased springback limitation processing on fabric areas prone to morphological aggregation during dynamic simulation, so that the fabric shape gradually returns to a natural wrinkled state, thereby avoiding the formation of abnormal aggregation.

[0036] The clothing fabric dynamic simulation method based on convolutional neural networks provided in this embodiment of the invention is implemented through the above-mentioned clothing fabric dynamic simulation system based on convolutional neural networks. For details of the specific methods and processes of the clothing fabric dynamic simulation system based on convolutional neural networks, please refer to the embodiments of the clothing fabric dynamic simulation method based on convolutional neural networks, which will not be repeated here.

[0037] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for dynamic simulation of clothing fabrics based on convolutional neural networks, characterized in that, Includes the following steps: During the dynamic simulation of a character performing a squatting or bending motion, motion change data of multiple joints of the character are acquired, and motion compression markers are generated based on the speed of joint motion changes. Based on motion compression markers, the time segments corresponding to the motion compression markers are subdivided, and the cloth screen refresh rhythm is unified within the subdivided time segments. At the same time, continuous transition processing is introduced between adjacent screens. Based on a unified cloth screen refresh rate, an area maintenance constraint is introduced for the cloth area located near the bending position of the human body. When the area change of the cloth area exceeds the preset range, the amount of change exceeding the preset range is allocated to the adjacent fold area. While maintaining the area constraint, determine the order of contact positions between the fabric and the human body, and adjust the displacement of the fabric closest to the human body first according to the contact position order, while unfolding the fabric area on the outside. Based on the contact position sequence, a phased rebound restriction process is implemented for fabric areas that are prone to shape aggregation during dynamic simulation, so that the fabric shape gradually returns to a natural wrinkled state.

2. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 1, characterized in that, The steps for generating action compression tags are as follows: Multiple joints involved in lower limb flexion and extension and trunk rotation are continuously tracked, and spatial displacement change information of each joint in consecutive frames is collected to form a joint motion change data sequence. Based on the joint motion change data sequence, the velocity of joint motion change is calculated, and the velocity of joint motion change is mapped to the corresponding time position to form the temporal distribution of joint motion change; Based on the temporal distribution of joint motion changes, extract time segments that change synchronously and at concentrated rates, and refine these time segments to obtain sub-time segments. The sub-time segments with concentrated rate of change are associated with the cloth coverage position during the action, and action compression marks are generated.

3. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 2, characterized in that, The sub-time segments where the speed of joint movement changes is concentrated are determined by analyzing the synchronicity of the speed of joint movement changes at adjacent time positions. Action compression marks are generated only within the time segments where synchronous changes continue, so that the action compression marks correspond to the stage of multi-joint coordinated changes of the character.

4. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 2, characterized in that, The steps for processing time segments based on action compression tags are as follows: Read the action compression markers and extract the boundary information of the corresponding time segments. Map the time segment boundary information to the time axis of the cloth dynamic simulation and establish an index relationship between the time segments and the cloth image sequence. Based on the index relationship, the time interval between adjacent frames within the time segment is counted, and the time segment is further subdivided according to the distribution of time intervals to form a continuously arranged subdivided time segment. Set a uniform cloth screen refresh rhythm within the subdivided time segments, and perform time alignment processing on the original cloth screens within the subdivided time segments so that the cloth screens are arranged according to the refresh rhythm. Based on a unified refresh rate for the cloth image, a continuous transition process is introduced between adjacent cloth images, so that the cloth form forms a continuous and stable change process within the subdivided time segments.

5. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 4, characterized in that, When introducing continuous transition processing between adjacent cloth frames, the cloth form state corresponding to the adjacent frames is used as the transition boundary. Under a unified cloth frame refresh rhythm, the form changes between adjacent frames are gradually unfolded. At the boundary of the subdivided time segment, the time consistency between continuous transition processing and processing within the segment is maintained, so that the cloth frames form a continuous sequence of changes throughout the entire time segment.

6. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 4, characterized in that, Based on a unified cloth screen refresh rate, the following steps are introduced to maintain area constraints: At each refresh time position, the bending angle of the joint and the curvature distribution of the human body surface are combined to determine the adjacent position of the human body bending, and the adjacent position of the human body bending is mapped to the corresponding cloth area to extract the geometric description of the cloth area. Based on the geometric description of the cloth area, the area change between adjacent refresh time positions is calculated, and the area is constrained by constructing an area with reference area, while the preset range is determined according to the speed of joint movement. When the area change exceeds the preset range, the change exceeding the preset range is converted into a shape adjustment amount and distributed to the adjacent pleated areas according to the adjacency relationship of the fabric areas. At the same time, a corresponding reverse adjustment is applied to the fabric areas. By keeping the area constraint consistent with the refresh rhythm of the uniform cloth screen, the allocated shape adjustment amount is continuously introduced between adjacent refresh time positions.

7. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 6, characterized in that, The preset range is adjusted synchronously according to the speed of joint movement. The amount of change exceeding the preset range is allocated according to the shared boundary length between the fabric area and the adjacent fold area and the consistency of the fold direction, so that the shape adjustment is transmitted along the fold extension direction and maintains continuous change between adjacent refresh time positions.

8. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 6, characterized in that, The steps for determining the contact position sequence and adjusting the fabric displacement while maintaining area constraints are as follows: At each refresh time position, the set of contact positions between the fabric and the human body is determined based on the fabric shape and the human body surface shape, and the contact duration and contact area changes of the contact positions are recorded to form contact description information. Based on the contact description information, the contact position sequence is determined according to the contact duration, contact area change and human body bending direction, so that the contact position sequence is arranged from the position closest to the human body to the outer position. Based on the order of contact positions, the displacement of the fabric closest to the human body is adjusted first, so that the displacement of the fabric closest to the human body unfolds step by step along the direction of the human body surface and remains continuous. After the fabric close to the human body has shifted and changed position, the fabric area on the outside is unfolded to create a continuous spatial hierarchy between the outer fabric area and the area close to the human body.

9. The method for dynamic simulation of clothing fabric based on convolutional neural networks according to claim 8, characterized in that, The steps for implementing phased rebound limiting based on the contact position sequence are as follows: After adjusting the fabric displacement changes under the contact position sequence, the fabric morphology is analyzed along the contact position sequence, the fabric areas prone to morphology aggregation are marked, and the corresponding area tracking relationship is established. Based on the fabric area where morphology aggregation is likely to occur, the springback restriction treatment is divided into stages according to the contact position sequence, and the springback target morphology and springback restriction range are set for each stage. Under a unified cloth screen refresh rhythm, the rebound restriction process is applied to the cloth areas prone to form aggregation in stages, so that the rebound process is advanced from the area close to the human body to the outer area. Maintaining continuity of phased rebound between adjacent refresh time positions allows the fabric shape to gradually recover to a natural wrinkled state under the phased rebound restriction process.

10. A clothing fabric dynamic simulation system based on convolutional neural networks, used to implement the clothing fabric dynamic simulation method based on convolutional neural networks as described in any one of claims 1-9, characterized in that, This includes an action compression marking module, a timing refresh control module, an area preservation constraint module, a contact sequence adjustment module, and a springback limit recovery module. The motion compression tagging module acquires motion change data of multiple joints of a person during the dynamic simulation of a person performing a squatting or bending motion, and generates motion compression tags based on the speed of joint motion change. The timing refresh control module, based on the action compression mark, subdivides the time segment corresponding to the action compression mark, unifies the refresh rhythm of the cloth screen within the subdivided time segment, and introduces continuous transition processing between adjacent screens. The area retention constraint module introduces an area retention constraint for the fabric area located near the bending position of the human body, based on the unified refresh rhythm of the fabric screen. When the area change of the fabric area exceeds the preset range, the amount of change exceeding the preset range is distributed to the adjacent fold area. The contact sequence control module determines the contact position sequence between the fabric and the human body while maintaining area constraints. Based on the contact position sequence, it prioritizes adjusting the displacement of the fabric closest to the human body, while simultaneously expanding the fabric area located on the outer side. The springback limitation and recovery module, based on the contact position sequence, performs phased springback limitation processing on fabric areas prone to shape aggregation during dynamic simulation, so that the fabric shape gradually returns to a natural wrinkled state.

Images