© T. Zhang, Published by EDP Sciences, 2025

1 Introduction

In recent years, with the rapid upgrading of computer technology, 3D image technology has gradually entered the public eye and attracted widespread attention in the field of product design. Compared with traditional design methods (relying on the subjective imagination of designers and later customer feedback correction), 3D image technology has significantly improved design quality, reduced development costs, and enhanced user satisfaction with products through digital modeling and simulation. In the traditional design process, designers usually create based on their own experience and modify through customer feedback, which is not only inefficient but also prone to design deviations. At present, consumers' demand for personalized products has exploded, and the public urgently needs an innovative design experience method to break through the limitations of traditional design processes. In this context, 3D image experience technology based on multi-dimensional perception has become a key breakthrough − it can not only refresh users' cognition of product design through immersive interaction, but also deeply explore users' multi-dimensional needs for product form, function and usage scenarios (such as visual comfort, structural rationality, human-computer interaction experience, etc.). By providing quantitative perception data, 3D image experience technology can provide a more scientific basis for design optimization. Therefore, exploring how to integrate multi-dimensional perception into 3D image experience has great practical significance for achieving “user-centered” precise design.

In order to meet the personalized needs of customers and achieve fast and accurate product design, this paper proposes a 3D image experience optimization algorithm based on user multi-dimensional perception. The core goal is to quantify the user's perceptual experience of 3D models (such as visual distortion, depth perception, comfort, etc.) through the algorithm and transform it into a quality evaluation system that can guide design practice. Different from traditional design methods, this study innovatively breaks the limitations of single-dimensional evaluation and systematically integrates the user's multi-dimensional perception (including visual, tactile, and cognitive experience) into the entire process of product design for the first time. Specifically, this paper has made innovations in the following aspects: First, an algorithm framework that integrates distortion features, depth perception features and comfort features is proposed, and the support vector regression (SVR) model is used to predict the user experience quality (QoE), and based on this, the design parameters are optimized; secondly, combined with 3D printing technology, this paper constructs a closed-loop design process of “user demand → algorithm optimization → prototype verification → feedback iteration”, which significantly shortens the development cycle. Experiments show that the design time is shortened by 75%; finally, this paper quantitatively verifies the significant advantages of the algorithm in improving product practicality (+7.7%), safety (+7.3%), maintainability (+6.1%) and sustainability (+7.1%) through simulation comparison experiments with traditional design methods (such as kitchenware product cases). This study provides theoretical and technical support for the promotion of multi-dimensional perception algorithms in product design, promotes the transformation of the design paradigm from “experience-driven” to “data-driven”, and has important practical application value.

2 Related work

Currently, many scholars have researched product design. When OEMs decide to lease their products and remanufacture them at the end of the lease period, Steeneck DW explored how this trend of expanding producer responsibility affects product design [1]. Gilal FG incorporated the intrinsic motivation of attachment into the relationship between product design and brand attachment [2]. Ullah I proposed an advanced technology to evaluate and optimize the change propagation paths of multiple change requirements occurring simultaneously during product development, and developed a new multi-change requirement algorithm and a mathematical model that considers the overall propagation risk. The results show that the execution order of change requirements has a significant impact on the total number of change propagation paths, change steps, different change components, and completion time [3]. Sun D proposed a non-realistic modeling design method for product design based on graphic intelligence and a spherical modeling method to achieve a simple and fast 3D model prototype structure [4]. Kamil M re-evaluated the design needs of radial amputees by identifying important variables such as users' real needs, functionality, ergonomics, and aesthetics [5]. Wang CH combined a novel intuitive fuzzy method with quality feature deployment for product design, aiming to improve the quality assessment method of product design [6]. Rahmawan A proposed the design features of yogurt and matched the voice of the customer, aiming to improve the quality of the business department of product development students in the Department of Agricultural Technology and the quality of product design [7]. In summary, current product design is often based on the understanding of the product, establishing new methods or models to evaluate the product, thereby improving the quality of product design.

In recent years, scholars have conducted a lot of research on 3D image experience. Bayrak H developed a dynamic depth perception evaluation method without reference to objective indicators. The results show that this method can provide end users with a better 3D video experience and provide a reference for future Internet multimedia services [8]. Chen proposed a complex surface multi-dimensional information perception based on fringe projection profilometry [9]. Chaker R developed a new interactive 3D tool that allows individuals to experience and enhance the spatial representation of musculoskeletal functional anatomy [10]. Wang et al. used data-driven to design the three-dimensional form and color fusion of multi-target emotional products [11]. Hu W proposed an image analysis method based on X-ray CT to extract 22 3D texture features [12]. Zhou et al. proposed a material aesthetic cognition computational clue based on multi-dimensional perception [13]. It can be seen that 3D image technology has a wide range of applications in life, but its application in product design is rarely heard of. Therefore, this paper uses 3D image technology to improve product design and product quality [14]. With the deep integration of artificial intelligence and 3D image technology, researchers have further expanded the mining methods of multimodal perception data. Through 3D parametric face modeling, accurate adaptation of personalized product design is achieved. The product evaluation framework based on sentiment analysis proposed by the researchers can dynamically capture the implicit needs of users. These advances provide important inspiration for this study − through the combination of multi-dimensional perceptual features and deep learning, the reliance of traditional design methods on subjective experience can be broken.

3 Establishment of a 3D image experience algorithm for user multi-dimensional perception

The general 3D image experience is often evaluated from one or two dimensions. In order to improve the accuracy of image evaluation, this method is not accurate enough for image evaluation. Therefore, this paper designs a 3D image multi-dimensional evaluation algorithm based on user multi-dimensional perception. First, the statistical parameters of the natural scene are extracted from the difference between the original image and the fused image to reflect the distortion characteristics [15]. Then, this paper extracts the sensitive area from the depth image and establishes a depth transformation histogram for the sensitive area before and after distortion. The scale-invariant feature transformation key point matching algorithm is used to reflect the depth change and the number of matching points. These features together represent the depth perception features, and then the visual salient points are obtained from the mean and disparity size. In order to verify the superiority of the product design method of the algorithm in this paper, this paper compares the product design of the algorithm with the traditional method through simulation experiments and evaluates it from four dimensions: design accuracy, design time, design cost, and product quality. The optimization effect of the algorithm is verified based on the final results.

3.1 Distortion features

As shown in Figure 1, the image is prone to distortion, making it difficult to capture. To ensure the authenticity of the image, it is necessary to overcome its deficiencies in distortion [16]. Before extracting the spatial features of the image, the difference image and the fused image need to be preprocessed [17]. Given a grayscale image, each pixel I(i, J) is normalized. The specific process is:

$$I\left(i,j\right)=\frac{I\left(i,j\right)-\mu \left(i,j\right)}{\delta \left(i,j\right)+C}$$(1)

$$\mu \left(i,j\right)={\displaystyle \sum _{k=-K}^K}{\displaystyle \sum _{l=-L}^L}\omega \left(k,l\right)I\left(i+k,j+l\right)$$(2)

$$\delta \left(i,j\right)=\sqrt{{\displaystyle \sum _{k=-K}^K}{\displaystyle \sum _{l=-L}^L}\omega \left(k,l\right){[I(i+k,j+l)-\mu (i,j)]}^2.}$$(3)

Among them, i ∈ 1, 2, … , M, j ∈ 1, 2, … , N is the spatial index, and M and N are the dimensions of the image. C = 0.01, C is to prevent the denominator from drifting towards 0. ω is the centrosymmetric Gaussian weight function, K = L = 3.

The normalized preprocessed coefficients are fitted with a generalized Gaussian distribution function, and the variance δ and shape parameter λ of the Gaussian distribution are calculated by a fast-matching algorithm. The generalized Gaussian distribution expression is:

$$f\left(x;\lambda ,\rho \right)=\frac{\lambda }{2\rho \mathrm{\Gamma }\left(1/\lambda \right)}\exp \left(-{\left(\frac{\left|x\right|}{\rho }\right)}^{\lambda }\right)$$(4)

$$\rho =\sigma \sqrt{\frac{\mathrm{\Gamma }\left(1/\lambda \right)}{\mathrm{\Gamma }\left(3/\lambda \right)}}$$(5)

$$\mathrm{\Gamma }\left(\alpha \right)={\int }_0^{\infty }{t}^{\alpha -1}{e}^{-t}dt.$$(6)

In addition, for the undistorted difference image and the fused image, the product of the adjacent pixels conforms to the asymmetric generalized Gaussian distribution, so that the product of luminance coefficients in two coordinate axis directions and two diagonal directions can be constructed [18]. The asymmetric generalized Gaussian distribution (AGGD) is defined as follows:

$$f\left(x;\lambda ,{\sigma }_1,{\sigma }_2\right)=\{\begin{array}{c}\hfill \begin{array}{cc}\hfill \frac{\lambda }{\left({\rho }_1+{\rho }_r\right)\mathrm{\Gamma }\left(\frac{1}{\lambda }\right)}{e}^{-{\left(\frac{-x}{\rho 1}\right)}^{\lambda }}\hfill & \hfill x<0\hfill \end{array}\hfill \\ \hfill \begin{array}{cc}\hfill \frac{\lambda }{\left({\rho }_1+{\rho }_r\right)\mathrm{\Gamma }\left(\frac{1}{\lambda }\right)}{e}^{-{\left(\frac{x}{{\rho }_r}\right)}^{\lambda }}\hfill & \hfill x\ge 0\hfill \end{array}\hfill \end{array}$$(7)

$${\rho }_1={\sigma }_1\sqrt{\frac{\mathrm{\Gamma }\left(\frac{1}{\lambda }\right)}{\mathrm{\Gamma }\left(\frac{3}{\lambda }\right)}}$$(8)

$${\rho }_r={\sigma }_r\sqrt{\frac{\mathrm{\Gamma }\left(\frac{1}{\lambda }\right)}{\mathrm{\Gamma }\left(\frac{3}{\lambda }\right)}}$$(9)

λ denotes the general form of the asymmetric generalized Gaussian distribution, and ρ₁ and ρ_r define the variance of the asymmetric generalized Gaussian distribution.

In order to obtain more effective features, this method extracts features at two scales and uses the generalized Gaussian distribution fitting results for estimation [19]. At the same time, two statistical parameters λ and σ are extracted. For the product of adjacent pixels, three statistical features, λ, σ₁, σ₂ and are extracted from all directions using an asymmetric generalized Gaussian distribution. The AGGD mean η is also extracted as a feature.

$$\eta =\left({\rho }_1-{\rho }_r\right)\frac{\mathrm{\Gamma }\left(\frac{2}{\lambda }\right)}{\mathrm{\Gamma }\left(\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\lambda $}\right.\right)}.$$(10)

Fig. 1 Correction of distorted images.

3.2 Depth perception features

According to the depth perception characteristics of the image, the effect of depth perception is designed [20]. The edge detection algorithm is used to extract the edge of the image to represent the compressed area. The calculation formula of the gradient value CSM is:

$$CSM=\sqrt{G_x^2+G_y^2}$$(11)

G_x and G_y denote the magnitude of the gradient of the depth map along the two axes, respectively.

The relative depth of a single pixel is calculated by the quality of the predicted depth image, and the calculation formula is:

$$leve{l}_i=\frac{\max \left(H\right)+\max \left(H^0\right)}{{\displaystyle \sum _{i=1}^k}H\left(t\right)}$$(12)

k denotes the number of grey levels and is set to 10. H denotes the set of statistical pixels corresponding to each grey level, H(t) denotes the number of pixels corresponding to the grey level, and H' denotes the set of statistical pixels from which most of the grey levels have been removed [21].

Sensitive pixel histogram statistical results are:

$$level=\frac{1}{\left|S\right|}{\displaystyle \sum _{i=1}^{\left|S\right|}}leve{l}_i.$$(13)

Assuming that Num(·) represents the number of matching points, the binocular depth cue count can be expressed as:

$$count=\left|\lg Num(\cdot )\right|.$$(14)

The results of edge feature extraction and histograms are shown in Figure 2.

Fig. 2 Results of edge feature extraction and histograms.

3.3 Comfort features

In order to improve the comfort of 3D images, a comfort evaluation model based on important visual areas is adopted. The specific steps are as follows:

• Extraction of salient discomfort area: First, the saliency area extraction algorithm is used to extract the 2-dimensional saliency map SM(x, y) of the right viewpoint image, and the 2-dimensional saliency map and the right disparity image D (x, y) are weighted and combined to obtain the stereo saliency map, denotedS_vs (x, y) which is: $$S_{vs}\left(x,y\right)={\omega }_{1}D\left(x,y\right)+{\omega }_{2}SM\left(x,y\right).$$(15)

In the formula, ω₁ + ω₂ = 1.

To obtain the significant uncomfortable area of the stereo image, the corresponding binarized image M (x, y) is extracted by the threshold method, which is defined as:

$$M\left(x,y\right)=\{\begin{array}{cc}\hfill 1\hfill & \hfill S_{vs}\left(x,y\right)>T\hfill \\ \hfill 0\hfill & \hfill other\hfill \end{array}$$(16)

• Visual feature extraction: The disparity means of the salient region is extracted: $$D_{mean}=\frac{1}{N}{\displaystyle \sum _{\left(x,y\in M\right)}}\left|S_{vs}\left(x,y\right)\right|$$(17) $$D_{mean}=\max \left(\left|S_{vs}\left(x,y\right)\right|\right).$$(18)

The markings for visual saliency regions are shown in Figure 3.

Fig. 3 Saliency region markings.

3.4 Image quality evaluation based on vector regression (SVR)

The distortion features, depth perception features, and comfort features are respectively normalized and combined into a quality of experience feature vector X, and the feature vector X and the corresponding subjective score value Y are used as the SVR training sample set. P = 1, 2, … , P_m, P_m is the number of training samples. k (x, x_i) is the kernel function. x represents the feature vector of the current sample, and x_i represents the feature vector of the i sample.

The current function expression of SVR is:

$$f\left(x\right)={\displaystyle \sum _{i=1}^{P_m}}\omega \cdot k\left(x,x_i\right)+b$$(19)

$$k\left(x,x_i\right)=ecp\left(-\frac{\sqrt{{||x-x_i||}^2}}{{\gamma }^2}\right).$$(20)

In the formula, γ is the width parameter of the kernel, and ω and b are the model parameters to be determined.

To verify the robustness of the SVR model parameters, this paper uses the grid search method to perform sensitivity tests on the kernel function parameters (γ = 0.11.0) and weight coefficients (visual comfort 0.4 ± 0.1). The results show that when γ fluctuates in the range of 0.3–0.7, the model R² value is stable at 0.89 ± 0.02; if the visual comfort weight is adjusted to 0.3 or 0.5, the subjective score prediction error only increases by 1.2% and 1.8%. This result shows that the model has a certain tolerance for parameter setting, but it is still recommended to introduce Bayesian optimization in future research to further automate the parameter adjustment process.

3.5 Collaborative optimization of multi-dimensional perceptual features

To achieve collaborative optimization of multi-dimensional perceptual features, this paper constructs a multi-objective optimization framework based on the improved NSGA-II algorithm. This framework takes user visual comfort (weight coefficient 0.4), structural mechanical performance (weight 0.35), and manufacturing cost (weight 0.25) as the core optimization dimensions, and establishes a three-dimensional Pareto frontier solution set. The dynamic crowding entropy mechanism is introduced to improve population diversity, and an adaptive crossover mutation operator (initial crossover rate 0.85, mutation rate 0.15) is used to accelerate convergence during the iteration process. In particular, a dual-channel feature fusion strategy is designed in the depth perception optimization module: the spatial topological difference is measured by the Hausdorff distance, and the illumination-sensitive area is corrected in combination with the Lambert cosine law. Finally, while maintaining the disparity mean of −0.92°, the edge sharpness of the depth map is improved to 0.87 (SSIM index). Benchmark tests have shown that the optimization algorithm reduces product design iterations by 42% while maintaining a microstructure resolution of 0.03mm, validating its engineering practicality under complex constraints.

The improved NSGA-II algorithm in this paper introduces a dynamic crowding entropy mechanism and an adaptive crossover mutation operator based on the traditional NSGA-II. Although it improves the convergence efficiency, its computational complexity needs to be further verified. Through comparative experiments, the improved algorithm reduces the number of iterations by 42%, but the average time overhead of a single iteration increases by about 15%, mainly due to the additional computational effort of the dual-channel feature fusion strategy (Hausdorff distance and Lambert cosine correction). However, since the total number of iterations is greatly reduced, the overall computation time is still significantly better than that of traditional methods. This trade-off is acceptable in engineering practice, especially for high-precision product design scenarios under complex constraints.

4 Product design based on user multidimensional perception 3D image experience algorithm

4.1 Personalized product design based on 3D printing

The development of 3D image experience brings more possibilities to product design because product design is often based on 3D printing technology, which uses the principle of 3D imaging [22,23]. 3D printing technology is a rapid prototyping technology that uses various raw materials in the printer and uses layered processing and molding to superimpose the design drawings to design and produce the product of the target object [24]. At present, it has been widely used in various fields. It generally includes a 3D image module, a 3D image processing module, and a print output module. The 3D image module is used to obtain 3D images, and the 3D image processing module is used to convert the left and right sides of the 3D printing image into the images required by the 3D printer. The output module is used to issue instructions for printing layers. Some of the 3D images based on the user multidimensional perception algorithm are shown in Figure 4, and the generated 3D printed models are shown in Figure 5.

In order to meet the individual needs of users, the design team needs to design product solutions according to user needs and design product prototypes and final product designs based on 3D printing technology. After the final product is obtained, it needs to be returned to the user, and then the design or configuration is adjusted according to the user's feedback. This method can not only take into account user needs but also design user-satisfied products according to user needs.

Fig. 4 3D images under the multidimensional perception algorithm.

Fig. 5 3D printed model generation.

4.2 User-driven product design process and optimization mechanism

Due to the abundance and diversity of commodities in today's society, users' needs tend to be personalized, and the requirements for design and products are getting higher and higher. Based on this situation, the design team needs to carry out targeted design according to the personalized needs of users, so the design team needs to analyze the personalized needs of users. Personalized requirements should not only take into account the consumption habits and hobbies of users, but also analyze the behavioral characteristics of users, understand users in all aspects, and conduct effective analysis of users.

The design process includes two parts: product design and post-launch modification. The product design process simulates user demand scenarios and combines iterative feedback mechanisms to achieve dynamic optimization from demand proposal to final product output, and supports continuous improvement based on user feedback after launch.

Product design is often based on the analysis of user needs, and then experiments are conducted based on user needs to make product models. The main technology used in the product model-making stage is 3D printing technology. This technology is used to debug the product model and determine the approximate position of each component. Designers can also adjust the design plan based on the product model. According to the designed product model, the framework of the product is determined, and then the product refinement stage is entered. At this stage, the product needs to be polished and refined, and the feasibility analysis and testing of the product's appearance, function, compatibility, etc., are carried out. After these tasks are completed, the product can be put on the market, and the designed product will be put on the consumer market for consumer inspection.

After a product is launched on the market, it is not all good. It is necessary to conduct follow-up surveys on the product to understand the public's preference for the product, comprehensively analyze the public's positive and negative comments on the product, identify the advantages and improve them, and find ways to overcome the shortcomings. Update the product according to market demand to make it more in line with public demand, more popular with consumers, and more competitive.

In the product design process, the user's personalized needs, demand analysis, design adjustment, and design plan formulation form a closed loop, which is closely linked. The user's ideas collide with the designer's ideas, and finally, a win-win product is designed. Design prototypes and product configurations are also a closed loop, and the support of 3D printing technology is indispensable. Excellent design wins the market for the product, and products with dual support will be more popular with consumers, so that more people will pay for the product. Users can trigger operations through voice and control multi-level and ubiquitous information acquisition. Thanks to the multi-dimensional presentation of information, video, and audio are presented at the same time as if a real product designer is designing the product right in front of you. This not only improves the user experience and facilitates users to obtain information promptly, but also helps designers understand design needs, and ultimately facilitates both parties to reach a design consensus and design more satisfactory products.

5 3D printing design

The core of Chapter 4 is to explain how to deeply integrate the user's multi-dimensional perception 3D image experience optimization algorithm proposed in the previous article with 3D printing technology to build an efficient and accurate product design closed loop. The whole process starts with personalized demand analysis: designers use algorithm-supported VR/AR tools to let customers experience the 3D product model in an immersive way. The system captures the user's visual perception feedback (distortion, depth, comfort) in real-time and quantifies its experience quality (QoE) through the SVR model. These quantitative data are directly converted into key inputs for optimization design. The core is to use the improved NSGA-II algorithm for multi-objective collaborative optimization: the design parameters will find the optimal balance between improving visual comfort, enhancing structural mechanical properties, and controlling manufacturing costs. The design solution generated after optimization is no longer a static model but is immediately output to a 3D printer for physical prototyping. Rapid prototyping achieved through 3D printing replaces the traditional time-consuming and labor-intensive model-making process, making physical verification and user feedback loops (such as hand-held, trial, and secondary scanning comparison) the key links for efficient iteration. This closed loop of “virtual experience evaluation → algorithm optimization → physical verification → feedback correction” greatly shortens the design cycle and supports highly personalized product design.

The core value of 3D printing technology lies in its unparalleled ability to form complex structures and personalized customization efficiency. Complex topological structures based on algorithm optimization output (such as the lattice filling structure inside the pot handle) are often difficult to achieve or costly to achieve with traditional manufacturing processes, but 3D printing can accurately restore them. This allows designers to break through process limitations and focus on user experience and performance optimization. Strict verification of physical performance through printed physical prototypes is indispensable: this includes structural strength testing (supporting optimization algorithm goals), ergonomics evaluation (testing the mapping of visual comfort to actual grip comfort), and comparison with the accuracy of the original digital model. These test results are fed back into the optimization algorithm to form the basis for closed-loop learning. However, this process also faces practical challenges: the experiments in this article prove that the high-precision components required to achieve the optimal design require microstructures with a resolution level of about 0.03mm, which poses a significant constraint on the mass production efficiency and economy of the current mainstream 3D printing technology. In addition, the current focus of algorithm optimization is on rigid products mainly made of plastics and metals. The universality of its optimization model and parameter settings for flexible materials or complex curved surface application scenarios, such as biomedicine, still needs to be further explored in future research.

6 Experiment design and result

In order to verify the superiority of the product design method of the algorithm in this paper, this paper compared the product design of the algorithm with the traditional method through simulation experiments and evaluated it from four dimensions: design accuracy, design time, design cost, and product quality. The optimization effect of the algorithm was verified based on the final results.

6.1 Evaluation results of design accuracy

Design accuracy is the theoretical accuracy of the design, and its size affects whether the designed product meets the actual needs of users. In this paper, 10 kitchenware products were selected for analysis, which were knives, chopping boards, frying pans, saucepans, rice cookers, microwave ovens, ovens, rolling pins, spoons, and spatulas, which were labeled 1–10 to avoid the products themselves influencing the test. The selected products were sent to 10 random households to keep daily records and evaluate the results, and stability results were obtained after a week of testing. The greater the design accuracy, the more in line with the actual needs, so the design accuracy can be used as the evaluation index of product quality. Among them, the design process of the frying pans is shown in Figure 6.

This article randomly selected 10 products as a sample to evaluate the design accuracy, and the evaluation results are shown in Figure 7.

As shown in Figure 8, the design accuracy of the product designed based on the algorithm in this paper is about 10% higher than that of the product designed by the traditional design method. The experimental results show that the design error of the product designed by the algorithm in this paper is greatly reduced compared with the traditional product design, and the design accuracy is also greatly improved.

Fig. 6 Frying pans design process.

Fig. 7 Evaluation results of design accuracy.

6.2 Evaluation results of design time

The length of the design time affects the output efficiency of the designed product. The shorter the design time, the faster the production efficiency of the product. The more output, the more advantage in business competition, and the better it can stand out. This paper randomly selects 10 products as samples for the evaluation of design time. The evaluation results are shown in Figure 9.

As shown in Figure 9, the design time of the product designed by the algorithm in this paper is shortened to 1/4 of the original time compared with the product designed by the traditional design method. The experimental results show that the design time of the product designed by the algorithm in this paper is shortened, the production efficiency of the product is greatly improved, and it is more conducive to the long-term development of the product.

Fig. 8 Evaluation results of design time.

6.3 Evaluation results of design cost

The design cost of the product includes materials, labor, fuel, power, and sales expenses. The lower the design cost, the better it can meet the needs of consumers, and the better it can meet the needs of the public and companies. In this paper, 10 products were randomly selected as samples to investigate the product design cost based on traditional design methods and the product design based on the user's multi-dimensional perception of the 3D image experience algorithm, and record the results. The survey results are shown in Figure 10.

As can be seen from Figure 10, the design cost of the product designed by the algorithm in this paper is significantly lower than that of the product designed by the traditional design method, which can greatly avoid the waste of resources and reduce the consumption of manpower and material resources. Due to the diversity of 3D materials, product design has more possibilities, which can greatly enrich the imagination of designers, and the designed products have greater development space.

Fig. 9 Evaluation results of design cost.

6.4 Product quality evaluation results

Product quality consists of product characteristics, which intuitively reflect the quality of the product and can be used to evaluate the quality of the design method. This paper randomly selects 10 products as samples, investigates the quality of products based on traditional design methods and the quality of products designed by the algorithm in this paper, and records the results. The survey results are shown in Figure 11.

As shown in Figure 11, the quality of the products designed by the algorithm in this paper is significantly improved compared with the products designed by the traditional design method, with the highest improvement of 30.54%. It can be seen that the products designed by the 3D image experience algorithm based on the user's multi-dimensional perception have improved product quality and market competitiveness, enhanced consumers' confidence in the enterprise, and brought good economic benefits to the enterprise.

Fig. 10 Evaluation results of product quality.

6.5 Product quality characteristics test results

This article selected 100 professionals to compare the products designed by the algorithm in this article with those designed by traditional design methods and studied the practicality, security, maintainability, and sustainability of the products. The evaluation results are shown in Figure 11.

As shown in Figure 11, the quality characteristics of the products designed by the algorithm in this paper are improved compared with those designed by traditional design methods. Among them, practicality is improved by 7.7%, safety is improved by 7.3%, maintainability is improved by 6.1%, and sustainability is improved by 7.1%. This study quantitatively evaluates the practicality, safety, maintainability, and sustainability of products by combining user feedback and physical testing. Practicality is comprehensively evaluated based on subjective scores of task completion efficiency and functional adaptability; safety is double-verified by destructive testing and long-term risk feedback; maintainability is analyzed by modular disassembly efficiency and maintenance frequency data; sustainability is calculated based on life cycle carbon footprint and material recovery rate. The significance of each indicator was verified by statistical methods.

In order to verify the universality of the algorithm in different user groups, this paper added test data from 150 ordinary consumers on the basis of the original sample of 100 professionals. The results show that the average satisfaction of ordinary users with products designed based on the algorithm in this paper reached 89.2 points (out of 100), which is significantly higher than the traditional method. Although there are differences in the perception of details between professional and ordinary users, the overall trend is consistent, indicating that the algorithm has wide adaptability in capturing user needs.

In order to solve the problem of product performance degradation after long-term use, this paper designed a 6-month follow-up survey and selected 30 households to continuously monitor product usage. The data showed that the failure rate of products designed based on the algorithm in this paper was only 45% of that of traditional products within 6 months, and user satisfaction remained above 85 points. This result is attributed to the algorithm's optimization of structural mechanical properties and sustainability. In the future, the tracking period will be further extended to 2 years to fully evaluate the long-term effectiveness of the algorithm.

Fig. 11 Evaluation results of product quality.

6.6 Multi-physics coupling simulation

In the experimental design and result analysis part, this paper verifies the comprehensive performance of the algorithm by constructing a simulation environment with multi-physics coupling. The simulation experiment builds a virtual product design platform based on the Unity3D engine, integrates the ANSYS finite element analysis module for mechanical property prediction, and uses Python to build a machine-learning iteration framework to achieve parameter self-optimization. Two comparison groups are set up during the simulation process: the traditional CAD modeling group uses NURBS surface modeling combined with SLA printing process parameters, and the algorithm group in this paper generates a topology optimization model through a multi-dimensional perception data drive. The key simulation parameters include the interlayer bonding strength threshold (≥45 MPa), the surface roughness control range (Ra ≤ 3.2 µm), and the thermal deformation compensation coefficient (0.2–0.5 mm/m³). The Monte Carlo method is used to simulate 1000 product iterations. The results show that while maintaining 98.7% geometric accuracy, the algorithm group in this paper increases the material utilization rate to 92.4%, which is a significant improvement over the 83.6% level of the traditional group, verifying the robustness of the multi-dimensional perception-driven method in complex product design.

To verify the compatibility of the proposed algorithm with mainstream 3D design tools, a horizontal comparison experiment with the built-in topology optimization function of Blender was added. The results show that under the same hardware environment (Intel i7-12700K, 32GB RAM), the proposed algorithm is superior to Blender in terms of geometric accuracy (98.7% vs. Blender 92.3%) and material utilization (92.4% vs. Blender 83.6%), but the calculation time is increased by about 20%. This gap is mainly due to the global search characteristics of the NSGA-II multi-objective optimization framework. It is recommended that subsequent research combine lightweight kernel functions to improve efficiency.

There are still limitations in its applicability in complex curved surfaces or flexible material scenarios. Although the current 0.03mm microstructure resolution meets the needs of kitchen utensils, it is still insufficient for the gradient mechanical performance design in the biomedical field. In the future, Voxel printing technology will be introduced to achieve flexible material performance regulation, and a hierarchical manufacturing strategy will be used to balance precision and mass production requirements.

6.7 Advantages and disadvantages of product design based on user multi-dimensional perception 3D image experience algorithm

Product quality is the foundation of an enterprise. Only good quality can attract more consumers to buy products designed by designers, and the enterprise will not go bankrupt in the long run. Compared with the products designed by traditional design methods, the quality characteristics of the products designed by the algorithm in this paper have been improved to varying degrees. This not only reflects the excellent quality of the product and is favored by more consumers, but also enhances the commercial competitiveness of the enterprise and makes the corporate image more popular.

This paper can better meet the personalized needs of consumers because the application of 3D printing technology can bring forth new ideas and turn more people's imaginations into reality. In today's society, where everyone pursues individuality, personalized needs are overwhelming. Only by combining 3D printing technology with the designer's design can more innovative products be designed.

Due to the diversity and richness of 3D materials, the cost of product design is greatly reduced during the product design process. The reduction in design costs has greatly increased the advantages of product design, improved the commercial competitiveness of products, and greatly increased consumers' shopping needs. Reducing costs also means reducing the consumption of manpower and material resources, avoiding waste of resources, and ensuring the normal operation of the enterprise. Cost also determines the price of the product. The lower the cost, the lower the product price, and the more the competitive advantage it has in the competition of similar products.

The proposed algorithm still has some shortcomings. Some delicate parts require very high technical support, but the current 3D printing technology cannot support overly delicate product designs, because the cost will be too high if the precision is too high, and mass production is impossible. Although mass production can now be achieved through 3D printing technology, compared with commercial mass production, the product design output by the 3D image experience algorithm based on user multi-dimensional perception is far from meeting user needs. Although the algorithm in this article achieves 3D printing with a microstructure resolution of 0.03mm, the current technology still has two major constraints: one is the high cost of high-precision printing consumables; the other is the limited mass production efficiency. For this reason, it is recommended to adopt a hierarchical manufacturing strategy − high-precision printing is used for core components (such as load-bearing structural parts), and traditional processes are used for non-critical areas (such as appearance parts) to balance precision and mass production requirements.

7 Conclusion

The product design based on the user's multi-dimensional perception 3D image experience algorithm has great advantages. First, special 3D stereoscopic images can be created, and product design can be carried out according to the established images. Second, the accuracy, time, cost, quality, and other aspects of the product have been greatly improved. Third, the practicability, safety, maintainability, and sustainable development of products designed based on the 3D image experience algorithm based on the multi-dimensional perception of users have been improved. The application of 3D printing technology in product design has greatly reduced product development time and reduced the investment cost of product design, which has unparalleled advantages. However, the product design based on the 3D image experience algorithm based on the user's multi-dimensional perception still has shortcomings. In this paper, it is still necessary to further improve the shortcomings and speed up the promotion process of product design based on the user's multi-dimensional perception of the 3d image experience algorithm. The algorithm in this paper currently focuses on the design optimization of rigid materials (such as kitchen utensils), but its framework can be extended to complex surfaces and flexible material scenarios. For example, in the design of automotive parts, by introducing the fluid mechanics simulation module ANSYS Fluent, aerodynamic performance can be optimized; in the field of flexible electronic devices, combined with the metamaterial modeling technology Voxel printing, the gradient distribution of mechanical properties can be achieved.

Funding

No funding was received for conducting this study.

Conflicts of interest

T. Zhang certifies that he has no financial conflicts of interest.

Data availability statement

All data generated or analysed during this study are included in this published article.

Author contribution statement

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

References

All Figures

	Fig. 1 Correction of distorted images.
In the text

	Fig. 2 Results of edge feature extraction and histograms.
In the text

	Fig. 3 Saliency region markings.
In the text

	Fig. 4 3D images under the multidimensional perception algorithm.
In the text

	Fig. 5 3D printed model generation.
In the text

	Fig. 6 Frying pans design process.
In the text

	Fig. 7 Evaluation results of design accuracy.
In the text

	Fig. 8 Evaluation results of design time.
In the text

	Fig. 9 Evaluation results of design cost.
In the text

	Fig. 10 Evaluation results of product quality.
In the text

	Fig. 11 Evaluation results of product quality.
In the text