Clothing style intelligent construction method and system based on multi-modal image analysis

Abstract

The invention is suitable for the technical field of intelligent costume design and computer vision image analysis, and provides an intelligent costume style construction method and system based on multi-modal image analysis, and the method comprises the steps: carrying out the design optimization of a target pattern region of a to-be-optimized performance costume worn by a target performance subject in a target stage, and obtaining historical performance data of the target stage. According to the invention, by introducing the multi-modal image analysis technology, the quantitative association relationship between the display effect of the key pattern area of the costume and the support metric value of the corresponding body part is established, and the limitation that design optimization is carried out only by experience or single visual judgment in the prior art is broken through.

Description

Technical Field

[0001] This invention belongs to the field of intelligent clothing design and computer vision image analysis technology, and particularly relates to a method and system for intelligent construction of clothing styles based on multimodal image analysis. Background Technology

[0002] In the current field of smart clothing design, research on stage performance costumes mostly focuses on aspects such as style aesthetics, material comfort, and lighting coordination. For example, designers typically rely on experience or subjective aesthetic judgment to determine how key patterned areas in the costumes are displayed, and then evaluate the effect through human try-ons or limited stage demonstrations. While this approach can enhance the stage presence of costumes to some extent, its process depends on subjective experience, lacks quantitative evidence, and often fails to accurately predict the actual visual effect of the costumes under different body types of performers.

[0003] However, stage performances are diverse and dynamic, and the display effect of key patterned areas can vary significantly when performers of different body types wear the same costume. For example, insufficient support in certain areas of the costume may cause the pattern to distort or wrinkle, while excessive support may cause bulging or deformation, reducing the integrity and clarity of the pattern. Current technology has failed to establish a quantitative relationship between the pattern display effect and the differences in the performer's body type, and lacks effective technical means to identify the optimal display area under different support conditions, thus hindering scientific and objective adjustments during intelligent design optimization. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for intelligent construction of clothing styles based on multimodal image analysis, aiming to solve the problems mentioned in the background art.

[0005] This invention is implemented as follows: a method for intelligently constructing clothing styles based on multimodal image analysis, the method comprising: When optimizing the design of the target pattern area of ​​the performance costume worn by the target performer on the target stage, historical performance data of the target stage is obtained. Extract several historical performance samples from historical performance data that are consistent with the target performance subject in performance style but differ in body shape parameters of the corresponding body parts in the target pattern area; Determine the pattern visualization effect index of the contrasting performance subject in each historical performance sample, and determine the pattern area support metric value of the corresponding body part of each contrasting performance subject in the target pattern area. Several historical performance samples were sorted according to the support metric values, and the trend of the pattern visualization effect index with the support metric values ​​was analyzed to determine whether there is an optimal range for the pattern visualization effect index. If an optimal interval is determined, a reference performance sample is selected within the optimal interval. Based on the difference in the support metric of the pattern area between the main performance subject and the target performance subject, the target pattern area of ​​the performance costume to be optimized of the target performance subject is optimized or reduced.

[0006] As a further limitation of the technical solution of the present invention, the consistency of the performance form means that the historical performance sample and the target performance subject have the same type of action, action trajectory and action rhythm in the performance process, or their similarity is within the corresponding preset threshold range.

[0007] As a further limitation of the technical solution of the present invention, the body part shape parameter refers to the parameter used to characterize the geometric features of the body part corresponding to the target pattern area, and the geometric features include, but are not limited to, perimeter, width and thickness.

[0008] As a further limitation of the technical solution of this embodiment of the invention, the quantification process of the pattern visualization effect index includes: Analyze the historical performance data and extract data collected by multi-view cameras on the target stage; Based on image analysis technology, the local multi-view camera data corresponding to each historical performance sample is analyzed to extract the visible pixel ratio, pattern integrity and clarity indicators of the target pattern area under each viewer's perspective. The visible pixel ratio, pattern integrity, and clarity indicators are weighted and combined to obtain the pattern visualization effect index of the target pattern area of ​​the comparative performance subject in the historical performance sample.

[0009] As a further limitation of the technical solution of this embodiment of the invention, the quantization process of the pattern region support metric value includes: Based on image analysis technology, historical performance data is processed to locate and track the target pattern area of ​​the main performance subject. Extract at least one of the following indicators from the target pattern area during the performance: deformation, wrinkle, relative sway, and bulge. The indicators are normalized and weighted to obtain the support metric value of the pattern region.

[0010] As a further limitation of the technical solution of this embodiment of the invention, the step of sorting several historical performance samples according to the support metric value and analyzing the trend of the pattern visualization effect index with the support metric value to determine whether there is an optimal range for the pattern visualization effect index includes: The historical performance samples are sorted from smallest to largest according to the support metric values ​​to form a support metric value sequence; Obtain the pattern visualization effect index sequence corresponding to the support metric value sequence, and analyze the changing trend of the pattern visualization effect index sequence; When the pattern visualization effect index sequence exhibits a distribution characteristic of first rising and then falling, the interval near the peak is determined as the optimal interval.

[0011] As a further limitation of the technical solution of this invention embodiment, if an optimal interval is determined to exist, then a reference performance sample is selected within the optimal interval, and based on the difference in the support metric value of the pattern area between the performance subject and the target performance subject, the steps of optimizing the support or reducing the target pattern area of ​​the performance costume to be optimized of the target performance subject include: When there is an optimal interval in the pattern visualization effect index sequence, the historical performance sample with the highest pattern visualization effect index is selected from the optimal interval as a reference performance sample; Calculate the relative deviation ratio of the pattern area support metric value of the comparison performance subject in the reference performance sample relative to the pattern area support metric value of the target performance subject; Based on the relative deviation ratio, the local thickness, filling amount, cutting tightness, elastic tension, or stiffness of the supporting components of the target pattern area of ​​the performance costume to be optimized for the target performance subject are optimized or reduced accordingly.

[0012] A smart clothing style construction system based on multimodal image analysis, the system comprising: The data acquisition module is used to acquire historical performance data of the target stage when designing and optimizing the target pattern area of ​​the performance costume worn by the target performer on the target stage. The sample extraction module is used to extract several historical performance samples from historical performance data that are consistent with the target performance subject in performance form but differ in the body shape parameters of the corresponding body parts in the target pattern area. The indicator determination module is used to determine the pattern visualization effect indicator of the comparative performance subject in each historical performance sample, and to determine the pattern area support metric value of the corresponding body part of each comparative performance subject in the target pattern area. The interval analysis module is used to sort several historical performance samples according to the support metric values ​​and analyze the trend of the pattern visualization effect index with the support metric values ​​in order to determine whether there is an optimal interval for the pattern visualization effect index. The optimization processing module is used to select a reference performance sample within the optimal interval if an optimal interval is determined, and to perform optimization support processing or reduction processing on the target pattern area of ​​the performance costume to be optimized for the target performance subject based on the difference in the support metric value of the pattern area between the performance subject and the target performance subject.

[0013] As a further limitation of the technical solution of the present invention, the consistency of the performance form means that the historical performance sample and the target performance subject have the same type of action, action trajectory and action rhythm in the performance process, or their similarity is within the corresponding preset threshold range.

[0014] As a further limitation of the technical solution of the present invention, the body part shape parameter refers to the parameter used to characterize the geometric features of the body part corresponding to the target pattern area, and the geometric features include, but are not limited to, perimeter, width and thickness.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention introduces multimodal image analysis technology to establish a quantitative correlation between the display effect of key pattern areas on clothing and the support measurement values ​​of corresponding body parts, overcoming the limitations of existing technologies that rely solely on experience or single visual judgment for design optimization. Through comparative analysis of different body type samples in historical performance data, this invention can automatically identify the trend of pattern visualization effect changing with support measurement values, thereby determining the existence of an optimal support interval and selecting reference samples to make differentiated corrections to the target clothing. This enables fine-grained adjustments to local thickness, padding amount, cutting tightness, and elastic tension, ensuring the scientific nature and controllability of the optimization range.

[0016] This method not only significantly enhances the visualization of costume patterns and the viewing experience in stage performances, but also has strong universality and scalability, making it suitable for intelligent costume design and stage costume management in different scenarios, and possessing outstanding innovation and application value. Attached Figure Description

[0017] Figure 1 A flowchart of the method provided in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the process of determining the optimal range of pattern visualization effect indicators in the method provided in this embodiment of the invention; Figure 3 This is a flowchart illustrating the process of optimizing or reducing clothing based on a reference performance sample in the method provided in this embodiment of the invention. Figure 4 The application architecture diagram of the system provided in the embodiments of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Figure 1A flowchart of the method provided by an embodiment of the present invention is shown.

[0020] Specifically, a method for intelligently constructing clothing styles based on multimodal image analysis includes the following steps: Step S100: When optimizing the design of the target pattern area of ​​the performance costume worn by the target performer on the target stage, historical performance data of the target stage is obtained.

[0021] In this embodiment of the invention, the target stage can be a performance venue for stage plays, musicals, song and dance performances, or large-scale variety shows. This stage is typically equipped with multiple cameras positioned at different angles to capture the performance from all angles, allowing for subsequent playback and analysis of the performance effects. Using multiple cameras to collect data from different audience angles is a mature existing technology, and this method is commonly used in current stage environments for broadcasting, archiving, or comprehensive evaluation of stage effects.

[0022] The target performance subject refers to the actor, singer, or dancer performing on the target stage, that is, the individual wearing the performance costume to be optimized and performing the corresponding performance actions.

[0023] The performance costumes to be optimized refer to those worn by the target performers during the performance, which have specific stage design and display requirements. These costumes will be worn and performed by different performers on the target stage. These costumes typically feature important stage display elements in specific areas, such as unique patterns, symbols, designs, or thematic designs, which carry artistic expression or narrative functions. The target pattern area refers to the specific costume area containing this important display element. This area needs to be clearly and completely perceived by the audience during the performance. Its design can be improved through optimization methods such as thickening, reducing, adding padding, or adjusting the cutting method to enhance the stage visualization effect.

[0024] The historical performance data originates from multi-angle video and image data collected by cameras during past performances on the target stage. It may also include 3D human posture data collected by depth cameras, motion trajectory data recorded by the stage management system, and sensor data related to clothing effects (such as pressure and tension sensors); as well as body characteristic data of different performers belonging to the target stage. Combining these data forms historical performance data, providing a foundation for subsequent analysis and comparison.

[0025] The core technical research point of this invention lies in the fact that existing technologies, when designing and optimizing key pattern areas of costumes, often only focus on the aesthetics of the pattern itself or stage lighting conditions, failing to fully consider the impact of differences in the body shapes of different performers on the display effect of the pattern area. Since body shape differences are difficult to quantify uniformly, existing methods lack objective means of measuring such factors, resulting in the inability to accurately assess the actual stage visual effect of the pattern area during the intelligent design process.

[0026] Furthermore, the intelligent construction method for clothing styles based on multimodal image analysis also includes the following steps: Step S200: Extract several historical performance samples from historical performance data that are consistent with the target performance subject in performance form but differ in body shape parameters of the corresponding body parts in the target pattern area.

[0027] The consistency of performance form means that the historical performance sample and the target performance subject have the same type of action, action trajectory and action rhythm in the performance process, or the similarity is within the corresponding preset threshold range.

[0028] The body part shape parameters refer to parameters used to characterize the geometric features of the corresponding body part within the target pattern area. These geometric features include, but are not limited to, perimeter, width, and thickness. In this embodiment of the invention, the consistency of performance style refers to the consistency or high similarity between the historical performance sample and the target performer in terms of the type of movement, movement trajectory, and movement rhythm performed during the stage performance. Consistency here does not require complete identicality, but rather is limited to a controllable and relatively narrow similarity threshold range to ensure that the different subjects being compared are comparable in terms of stage presence and movement amplitude. For the similarity judgment of movement trajectory, image analysis techniques based on skeleton point tracking or pose recognition can be used. For example, keypoint detection algorithms in videos (such as OpenPose-like techniques) can be used to extract body joint coordinates, and spatial similarity of the trajectory can be calculated based on temporal analysis methods. For the similarity of movement rhythm, temporal signal processing and frequency domain analysis methods can be used to extract movement frequency features and compare them with the rhythm features of the target performer. By setting a similarity threshold, it is ensured that the movement pattern of the sample remains within a reasonable range compared to the target performer, thereby guaranteeing the scientific validity of the comparison.

[0029] In practice, the number of historical performance samples should be appropriately increased. Too few samples will weaken the comparative significance and make it difficult to find the optimal range for the display effect of the pattern area. Therefore, when selecting samples, as many historical performance samples as possible should be collected that are consistent with the performance style of the target performance subject but differ in body part shape parameters to ensure the robustness of the analysis results.

[0030] The body part corresponding to the target pattern area refers to the part of the human body that corresponds to the local area of ​​the pattern on the clothing, such as the chest area, shoulder area, or thigh area. The differences in the shape of these areas directly affect the display effect of the pattern under different body types. The parameters of these body parts were also obtained through analysis of historical performance data and 3D modeling, and are therefore quantifiable data.

[0031] The body part shape parameters refer to quantitative indicators used to characterize the geometric features of the corresponding body parts within the target pattern area. In this embodiment, the perimeter, width, and thickness are examples of intuitive and easily calculated parameters used to quickly reflect body shape characteristics. In other embodiments, these geometric features may also include longitudinal length, local curvature, surface area, surface tension, or relative proportions. These indicators can be obtained through image analysis and 3D reconstruction methods to further improve quantification accuracy and predictive ability for display effects.

[0032] Furthermore, the intelligent construction method for clothing styles based on multimodal image analysis also includes the following steps: Step S300: Determine the pattern visualization effect index of the comparative performance subject in each historical performance sample, and determine the pattern area support metric value of the corresponding body part of each comparative performance subject in the target pattern area.

[0033] The quantification process of the pattern visualization effect index includes: Analyze the historical performance data and extract data collected by multi-view cameras on the target stage; Based on image analysis technology, the local multi-view camera data corresponding to each historical performance sample is analyzed to extract the visible pixel ratio, pattern integrity and clarity indicators of the target pattern area under each viewer's perspective. The visible pixel ratio, pattern integrity, and clarity indicators are weighted and combined to obtain the pattern visualization effect index of the target pattern area of ​​the comparative performance subject in the historical performance sample.

[0034] The quantization process of the metric value supported by the pattern region includes: Based on image analysis technology, historical performance data is processed to locate and track the target pattern area of ​​the main performance subject. Extract at least one of the following indicators from the target pattern area during the performance: deformation, wrinkle, relative sway, and bulge. The indicators are normalized and weighted to obtain the support metric value of the pattern region.

[0035] In this embodiment of the invention, the quantification process of the pattern visualization effect index is based on camera data from multiple audience perspectives. Specifically, the target stage is typically equipped with multiple cameras at different angles, which can capture video footage of the performance from different positions in the audience. By analyzing this multi-view video data, it is possible to obtain the viewing experience of the target pattern area on the stage by different audience members.

[0036] At the technical implementation level, image segmentation and region detection techniques can be used to locate the target pattern area. Further, pixel-level statistics can be used to calculate the proportion of visible pixels from different viewing angles, reflecting whether the area is obscured by other body parts, clothing, or stage props. Pattern integrity can be quantified through region template matching or feature descriptor comparison, determining the degree of complete overlap between the observed pattern and the original design. Clarity metrics can be obtained based on image sharpness evaluation, spectral energy analysis, or contrast measurement, reflecting the perceived resolution of the pattern from the viewer's perspective. By weighting and combining the visible pixel proportion, pattern integrity, and clarity metrics, a comprehensive and objective pattern visualization effect index can be obtained. Existing image processing and multi-view video analysis technologies are relatively mature, so the above calculation process can be fully implemented. Its novelty lies in: not evaluating the display effect of the pattern from a single viewpoint, but combining multiple viewpoints for a comprehensive assessment, thus more closely reflecting the actual stage performance effect.

[0037] The quantification process of the pattern area support metric focuses on the changes in force and shape of the target pattern area during wearing and movement. First, historical performance data needs to be processed using image analysis techniques to locate the target pattern area and track it temporally. This can be achieved through target detection and optical flow tracking algorithms. Subsequently, the following metrics are extracted: Deformation degree is obtained by analyzing the non-equidistant components in the affine or perspective transformation matrix across frames, reflecting the stretching and deformation of the area during movement; Wrinkle degree can be obtained through texture analysis techniques, such as calculating the high-frequency energy of local areas or detecting the density of wrinkle lines; Relative sway degree is quantified based on the relative displacement variance between the region's centroid and adjacent skeletal key points, reflecting whether the pattern area experiences excessive relative shaking with body movements; Bulging degree can be measured by extracting the normal angle or parallax amplitude of the local surface through multi-view reconstruction, depth estimation, or lighting and shadow analysis to measure the bulging or collapsing of the area. All four types of metrics are based on existing computer vision and video analysis technologies and have clear implementation methods, thus enabling practical implementation. Its novelty lies in abstracting the local support of clothing through a visually quantifiable index system and establishing a quantitative link with the final stage display effect, which has not been considered in existing technologies.

[0038] In addition to the above-described embodiments, this invention can also employ a three-dimensional modeling-based approach to quantify the support metric of the patterned area. For example, by performing a three-dimensional reconstruction of the main performance subject, a three-dimensional surface model corresponding to the target patterned area is obtained. Then, geometric parameters such as the perimeter, width, thickness, surface area, and local curvature of this area are extracted and compared with reference values ​​to obtain the support metric. The advantage of this method is that it can more accurately reflect the geometric support of the clothing area and has strong physical interpretability. The method based on image time-series indicators is more direct, can realistically reflect the dynamic display effect from the audience's perspective, and is relatively simple to calculate. The two methods can complement each other; one is biased towards stage visual evaluation, and the other towards structural geometric quantification. Combining them can comprehensively support the technical solution of this invention.

[0039] Furthermore, the intelligent construction method for clothing styles based on multimodal image analysis also includes the following steps: Step S400: Sort several historical performance samples according to the support metric value, and analyze the trend of the pattern visualization effect index with the support metric value to determine whether there is an optimal range for the pattern visualization effect index.

[0040] Specifically, Figure 2 A flowchart is shown to determine the optimal range for pattern visualization performance metrics.

[0041] The process of sorting several historical performance samples according to their support metrics and analyzing the trend of pattern visualization effect index with the support metrics to determine whether there is an optimal range for the pattern visualization effect index includes the following steps: Step S401: Sort the historical performance samples in ascending order of the support metric values ​​to form a support metric value sequence; Step S402: Obtain the pattern visualization effect index sequence corresponding to the support metric value sequence, and analyze the changing trend of the pattern visualization effect index sequence; Step S403: When the pattern visualization effect index sequence shows a distribution characteristic of first rising and then falling, the interval near the peak is determined as the optimal interval.

[0042] In this embodiment of the invention, step S401 is specifically implemented as follows: the several historical performance samples are sorted from smallest to largest according to their pattern area support metric values, forming an ordered sequence of support metric values. Technically, this can be achieved through database retrieval and sorting algorithms, ensuring that the relative size relationship of different samples in terms of support metric values ​​is clearly presented.

[0043] The implementation process of step S402 is as follows: Based on the support metric value sequence, the pattern visualization effect index of the corresponding sample is obtained one by one, thereby forming a set of effect index sequences that correspond one-to-one with the support metric values. Technically, the previously calculated pattern visualization effect indexes and support metric values ​​can be stored in correspondence through data binding and serialization, and then statistical and plotting algorithms (such as curve fitting or time series analysis tools) can be called to analyze the overall changing trend of the effect index sequence.

[0044] The implementation process of step S403 is as follows: The overall trend of the pattern visualization effect index sequence is detected. When the sequence exhibits a distribution characteristic of first rising and then falling, it is determined that the pattern visualization effect index has a peak interval. Specifically, this characteristic can be identified through trend analysis and inflection point detection algorithms. For example, the inflection point of rising-falling can be detected based on the change in the sign of the first derivative, or a local maximum value can be found by fitting a curve. The optimal interval here refers to a small fluctuation interval near the peak, which can cover the continuous range of the peak point and its surrounding area where a high effect index is maintained. If no interval meeting the small fluctuation condition is detected, the support metric value corresponding to the peak can be directly used as the representative point of the optimal interval.

[0045] This trend reflects the following principle: as the support metric increases, the display effect of the target pattern area improves; that is, when the support is more substantial, the pattern is more easily perceived clearly and completely by the audience. However, when the support metric is too high, i.e., the clothing area is overstretched, it may cause the pattern to deform, distort, or become inaccurate, thus reducing the pattern's visual appeal. In other words, moderate support can optimize the stage display effect, but excessive support will lead to negative results.

[0046] The significance of identifying this trend lies in providing a clear quantitative basis for costume design: there exists an optimal support range for the patterned area, which allows for the best stage presentation. By identifying and determining this optimal range, a precise target range can be provided for subsequent costume optimization. This echoes the core technical challenge of this invention: existing technologies fail to quantify the impact of different performers' body shapes on the pattern's display effect. This invention, however, reveals the principle of "optimal moderate support" by establishing a trend between support metrics and display effect, thus breaking through the limitations of traditional reliance on subjective aesthetics or experience, and achieving intelligent optimization based on objective quantitative indicators.

[0047] Furthermore, the intelligent construction method for clothing styles based on multimodal image analysis also includes the following steps: In step S500, if an optimal interval is determined, a reference performance sample is selected within the optimal interval, and based on the difference in the support metric of the pattern area between the performance subject and the target performance subject, the target pattern area of ​​the performance costume to be optimized of the target performance subject is optimized or reduced.

[0048] Specifically, Figure 3 The flowchart illustrates the process of optimizing or reducing clothing based on a reference performance sample.

[0049] If an optimal interval is determined, a reference performance sample is selected within the optimal interval. Based on the difference in support metric values ​​of the pattern area between the main performance subject and the target performance subject, the target pattern area of ​​the performance costume to be optimized of the target performance subject is optimized or reduced. Specifically, the following steps are included: Step S501: When there is an optimal interval in the pattern visualization effect index sequence, select the historical performance sample with the highest pattern visualization effect index from the optimal interval as a reference performance sample. Step S502: Calculate the relative deviation ratio of the pattern area support metric value of the comparison performance subject in the reference performance sample relative to the pattern area support metric value of the target performance subject; Step S503: Based on the relative deviation ratio, perform corresponding optimization support treatment or reduction treatment on the local thickness, filling amount, cutting tightness, elastic tension or support component stiffness of the target pattern area of ​​the target performance costume of the target performance subject.

[0050] In this embodiment of the invention, the specific process of step S501 can not only select the historical performance sample with the highest pattern visualization effect index within the optimal range as the reference performance sample, but also select a historical performance sample within the optimal range that is most consistent with the target performer in terms of performance style as the reference. The former has the advantage of maximizing the final stage visualization effect, directly corresponding to the best visual performance; the latter has the advantage of maintaining maximum consistency with the target performer in terms of movement type, trajectory, and rhythm, reducing deviations caused by differences in performance movements, thus making the optimization result more consistent with the actual performance of the target performer. The two selection methods can be used in combination, flexibly chosen according to actual design needs.

[0051] In step S502, the relative deviation ratio is used to reflect the difference in the pattern area support metric value between the target performance subject and the reference performance sample. This ratio can be obtained by comparing the pattern area support metric value of the reference performance sample with the support metric value of the target performance subject. The calculation method can be (reference sample metric value minus target metric value) divided by the reference sample metric value. This method can intuitively reflect the insufficient proportion of the target relative to the reference, which is more conducive to guiding optimization and adjustment. In addition, the deviation can also be represented by the ratio of the difference between the target and the reference to the target metric value, or by the ratio of the difference between the two to the average value. These methods each have advantages in numerical characteristics and can all be used to quantitatively describe the degree of deviation.

[0052] In step S503, the specific advantage of using the relative deviation ratio to guide the optimization of support processing or reduction processing is that it avoids blindly thickening or thinning, but rather makes targeted adjustments according to the actual degree of difference, so that the optimization range is neither insufficient nor excessive. For example, when the relative deviation ratio is large, the support effect can be enhanced by increasing the local thickness or filling material; when the relative deviation ratio is negative, over-support can be avoided by reducing the thickness, decreasing the tightness, or weakening the stiffness of the support component. Specific measures include: adding flexible filling material to the target pattern area to enhance fullness; changing the tightness by adjusting the cutting line or sewing tension; or achieving fine control by changing the elastic modulus of the material or the hardness of the support skeleton.

[0053] Understandably, when making corrections based on the relative deviation ratio, a preset control amplitude coefficient should be introduced to avoid the correction amplitude being too large or too small.

[0054] To illustrate with a concrete example: Suppose a stage costume has an important pattern on its chest area. The support measurement value for this area in the reference performance sample is 100, while the corresponding measurement value for the target performer is 85. The relative deviation is (100-85) / 100 = 0.15. If the selected control amplitude coefficient is 1, this means the target is 15% lower than the reference in this area. In this case, the thickness of the chest area of ​​the target performer can be increased by 15% based on this ratio, for example, by adding a 3mm inner lining padding layer. Simultaneously, the tightness of the cut can be adjusted appropriately to maintain comfort. If the deviation is negative, for example, the target is 110 while the reference is 100, it indicates that the target area is over-supported. This can be corrected to an optimal state by reducing the corresponding thickness or decreasing the elastic tension.

[0055] The overall beneficial effects of this invention are as follows: By introducing a multimodal image analysis method, a quantitative correlation is established between the support metric of the patterned area and the stage visualization effect index, achieving precise optimization of key patterned areas of stage performance costumes. Existing technologies often rely solely on experience or single parameter judgment, while this invention can automatically identify the optimal support interval through comprehensive evaluation of multi-dimensional indicators, and perform differentiated optimization in conjunction with reference samples, thereby effectively improving the visual presentation effect of stage costumes and avoiding excessive or insufficient support.

[0056] In terms of application prospects, this invention is not only applicable to the intelligent design and customization of stage performance costumes, but can also be extended to film and television costumes, themed performance costumes, and sports performance equipment. With the popularization of multi-view video acquisition and artificial intelligence image analysis technologies, this invention can be widely used in actual performance design, costume prototyping, and virtual simulation, providing new solutions for digital costume design and intelligent stage management.

[0057] Furthermore, Figure 4 An application architecture diagram of the system provided in an embodiment of the present invention is shown.

[0058] In another preferred embodiment of the present invention, a clothing style intelligent construction system based on multimodal image analysis includes: The data acquisition module 100 is used to acquire historical performance data of the target stage when designing and optimizing the target pattern area of ​​the performance costume worn by the target performer on the target stage.

[0059] Furthermore, the intelligent clothing style construction system based on multimodal image analysis also includes: The sample extraction module 200 is used to extract several historical performance samples from historical performance data that are consistent with the target performance subject in performance form but differ in body shape parameters of the corresponding body parts in the target pattern area.

[0060] The consistency of performance form means that the historical performance sample and the target performance subject have the same type of action, action trajectory and action rhythm in the performance process, or the similarity is within the corresponding preset threshold range.

[0061] The body part shape parameters refer to the parameters used to characterize the geometric features of the corresponding body parts in the target pattern area. These geometric features include, but are not limited to, perimeter, width, and thickness.

[0062] Furthermore, the intelligent clothing style construction system based on multimodal image analysis also includes: The indicator determination module 300 is used to determine the pattern visualization effect indicator of the comparative performance subject in each historical performance sample, and to determine the pattern area support metric value of the corresponding body part of each comparative performance subject in the target pattern area.

[0063] Furthermore, the intelligent clothing style construction system based on multimodal image analysis also includes: The interval analysis module 400 is used to sort several historical performance samples according to the support metric and analyze the trend of the pattern visualization effect index with the support metric to determine whether there is an optimal interval for the pattern visualization effect index.

[0064] The optimization processing module 500 is used to select a reference performance sample within the optimal interval if an optimal interval is determined, and to perform optimization support processing or reduction processing on the target pattern area of ​​the performance costume to be optimized of the target performance subject based on the difference in the support metric value of the pattern area between the performance subject and the target performance subject.

[0065] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0066] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0067] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0068] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligently constructing clothing styles based on multimodal image analysis, characterized in that, The method includes: When optimizing the design of the target pattern area of ​​the performance costume worn by the target performer on the target stage, historical performance data of the target stage is obtained. Extract several historical performance samples from historical performance data that are consistent with the target performance subject in performance style but differ in body shape parameters of the corresponding body parts in the target pattern area; Determine the pattern visualization effect index of the contrasting performance subject in each historical performance sample, and determine the pattern area support metric value of the corresponding body part of each contrasting performance subject in the target pattern area. Several historical performance samples were sorted according to the support metric values, and the trend of the pattern visualization effect index with the support metric values ​​was analyzed to determine whether there is an optimal range for the pattern visualization effect index. If an optimal interval is determined, a reference performance sample is selected within the optimal interval. Based on the difference in the support metric of the pattern area between the main performance subject and the target performance subject, the target pattern area of ​​the performance costume to be optimized of the target performance subject is optimized or reduced.

2. The intelligent construction method for clothing styles based on multimodal image analysis according to claim 1, characterized in that, The consistency of performance form means that the historical performance sample and the target performance subject have the same type of action, action trajectory and action rhythm in the performance process, or the similarity is within the corresponding preset threshold range.

3. The intelligent construction method for clothing styles based on multimodal image analysis according to claim 1, characterized in that, The body part shape parameters refer to the parameters used to characterize the geometric features of the corresponding body parts in the target pattern area. These geometric features include, but are not limited to, perimeter, width, and thickness.

4. The intelligent construction method for clothing styles based on multimodal image analysis according to claim 1, characterized in that, The quantification process of the pattern visualization effect index includes: Analyze the historical performance data and extract data collected by multi-view cameras on the target stage; Based on image analysis technology, the local multi-view camera data corresponding to each historical performance sample is analyzed to extract the visible pixel ratio, pattern integrity and clarity indicators of the target pattern area under each viewer's perspective. The visible pixel ratio, pattern integrity, and clarity indicators are weighted and combined to obtain the pattern visualization effect index of the target pattern area of ​​the comparative performance subject in the historical performance sample.

5. The intelligent construction method for clothing styles based on multimodal image analysis according to claim 3, characterized in that, The quantization process of the metric value supported by the pattern region includes: Based on image analysis technology, historical performance data is processed to locate and track the target pattern area of ​​the main performance subject. Extract at least one of the following indicators from the target pattern area during the performance: deformation, wrinkle, relative sway, and bulge. The indicators are normalized and weighted to obtain the support metric value of the pattern region.

6. The intelligent construction method for clothing styles based on multimodal image analysis according to claim 5, characterized in that, The steps involved in sorting several historical performance samples according to their support metrics and analyzing the trend of pattern visualization effectiveness index changes with the support metrics to determine whether there exists an optimal range for the pattern visualization effectiveness index include: The historical performance samples are sorted from smallest to largest according to the support metric values ​​to form a support metric value sequence; Obtain the pattern visualization effect index sequence corresponding to the support metric value sequence, and analyze the changing trend of the pattern visualization effect index sequence; When the pattern visualization effect index sequence exhibits a distribution characteristic of first rising and then falling, the interval near the peak is determined as the optimal interval.

7. The intelligent construction method for clothing styles based on multimodal image analysis according to claim 6, characterized in that, If an optimal interval is determined, a reference performance sample is selected within the optimal interval. Based on the difference in support metric values ​​between the target performance subject and the target performance subject in the pattern area, the steps for optimizing or reducing the target pattern area of ​​the target performance subject's costume to be optimized include: When there is an optimal interval in the pattern visualization effect index sequence, the historical performance sample with the highest pattern visualization effect index is selected from the optimal interval as a reference performance sample; Calculate the relative deviation ratio of the pattern area support metric value of the comparison performance subject in the reference performance sample relative to the pattern area support metric value of the target performance subject; Based on the relative deviation ratio, the local thickness, filling amount, cutting tightness, elastic tension, or stiffness of the supporting components of the target pattern area of ​​the performance costume to be optimized for the target performance subject are optimized or reduced accordingly.

8. A smart clothing style construction system based on multimodal image analysis, characterized in that, The system includes: The data acquisition module is used to acquire historical performance data of the target stage when designing and optimizing the target pattern area of ​​the performance costume worn by the target performer on the target stage. The sample extraction module is used to extract several historical performance samples from historical performance data that are consistent with the target performance subject in performance form but differ in the body shape parameters of the corresponding body parts in the target pattern area. The indicator determination module is used to determine the pattern visualization effect indicator of the comparative performance subject in each historical performance sample, and to determine the pattern area support metric value of the corresponding body part of each comparative performance subject in the target pattern area. The interval analysis module is used to sort several historical performance samples according to the support metric values ​​and analyze the trend of the pattern visualization effect index with the support metric values ​​in order to determine whether there is an optimal interval for the pattern visualization effect index. The optimization processing module is used to select a reference performance sample within the optimal interval if an optimal interval is determined, and to perform optimization support processing or reduction processing on the target pattern area of ​​the performance costume to be optimized of the target performance subject based on the difference in the support metric value of the pattern area between the comparison performance subject and the target performance subject.

9. The intelligent clothing style construction system based on multimodal image analysis according to claim 8, characterized in that, The consistency of performance form means that the historical performance sample and the target performance subject have the same type of action, action trajectory and action rhythm in the performance process, or the similarity is within the corresponding preset threshold range.

10. The intelligent clothing style construction system based on multimodal image analysis according to claim 9, characterized in that, The body part shape parameters refer to the parameters used to characterize the geometric features of the corresponding body parts in the target pattern area. These geometric features include, but are not limited to, perimeter, width, and thickness.

Images