Applied Mathematics and Nonlinear Sciences

This paper used a new interior graphic modeling research based on CAD and depth enhancement teaching models. A massive database for graphic design has been established. An optimization method is proposed based on intelligent decision making, intelligent monitoring, panoramic vision, professional cooperation and intelligent planning. This system can make many systems of different dimensions share and integrate horizontally. The graphic design of CAD is introduced into 3D CAD. The Boolean method is introduced into the smooth grid instruction to obtain the smooth surface of the target surface. Combining the object of plane decomposition with other geometric shapes by form-fitting instruction achieves object control. Experiments show the effectiveness of the method. The system has good running performance, stability and safety.


Introduction
AutoCAD is a standard drawing method in traditional interior graphic design.CAD drawing technology can not meet the requirements of complex interior three-dimensional plane effects.Presently, CAD drawing technology still faces problems such as poor communication of information, disunity of data and poor design quality.It is widely used in games, animation and 3D special effects.The paper introduces this method to study modeling and automatic learning [1].This method can deal with the high dimension problem in a graphic design well.Based on the deep reinforcement learning model, this paper proposes a new idea of using computers for interior plane modeling.The system was simulated using 3DMax stereoscopic renderings to improve its fidelity and user satisfaction.

Use CAD technology for graphic research
3DSMax technology is the most widely used computer technology at present.It combines stratified subdivision surfaces with polygonal geometric models.The integrated technology of dynamic color and element rendering effect gives the 3D model of the space plane get a better design scheme [2].This makes a reasonable plan and section.At the same time, with CAD software for data interaction, DXF documents and AutoCAD data exchange.
The interior design of the building is modeled, visualized and designed with CAD technology.Models include hard-mounted models and soft-mounted models.The software process is shown in Figure 1.Enter furniture, appliances, and placement into the software model [3].The concept of visualization is the concrete realization of materials, maps, lighting, etc., of internal objects.In the overall decoration, beautify simultaneously to achieve the comprehensive decorative effect analysis and regulation.BREP form visualization is a parallel operation carried out by the image processing device.The system can trace and draw the light of the target shape to enhance the fidelity of the target shape.

Three-dimensional walls are constructed by the tensile method
Draw two plans based on the introduced interior design drawings to get a closed line.Concatenate vertex nodes by submitting CAD internal 3D CAD plan.This prevents errors in the picture.Use the Extrude command to pull out the actual size of the wall.The wall obtained by Boolean operation has irregular, irregular slender surfaces.Because the system's illumination cannot be transmitted well, "dark spots" will appear after the presentation [4].Before cutting, the wall should be converted into a multilateral shape to ensure a smooth border.Figure 2 shows the stretched wall model.

Grid smoothing and arbitrary deformation
When the simple object in the room is completed, the smooth mesh instructions are fused to the object's surface using a Boolean algorithm to get a tricky thing.The raster smoothing instruction is completed before finishing the contour of the target.This instruction replaces an uneven part of the model with a fusion surface.A smoother surface can be obtained by fine machining the character.
When shaping indoor furniture such as tables, chairs and beds, a line frame composed of control points is added to the exterior of the furniture by using F-type free deformation instruction.The shape of control points and line frames is achieved by arbitrary deformation [5].This changes the appearance of the furniture.Finally, the furniture modeling will be beautifully completed.

Rotation and shape combination
This paper takes 3DMax as an example to model indoor furniture with rotation instruction.This command can be rotated according to the two-dimensional plane of the object.At this point, the paper obtains the diorama of the target.Geometric shapes can be separated from the grid when patterns and text are incorporated into the interior furniture model.This way, the target surface shape control is achieved [6].After the shape fusion is completed, the model's surface will generate new lines.The internal target's spatial structure is adjusted by adjusting each node.

Research on the architecture of 3DMax technology
The functional components of the internal graphic design of the 3DMax technology architecture are shown in Figure 3.The system consists of six main components.The parts are separate from each other.The call interface is used for interaction.The CPU host simulates the internal plane model.The GPU performs many computing processes, such as ray tracing and physics engines for indoor objects.The system realizes the search of the whole internal environment [7].The paper uses a lot of lighting manipulation and offline methods to achieve the actual image.When the drawing idea is unrealistic, The paper will gradually improve to achieve a high-resolution image.Can input light into the optical domain network document.This way can improve the quality of brightness regulation.They are using grating technology to enhance indoor 3D environment fluency.Users can control the indoor environment through a mouse and keyboard.This allows it to operate from location to location in a three-dimensional environment.

A modified deep reinforcement learning method
Traditional depth enhancement algorithm has many inherent defects.This paper presents a learning method based on depth enhancement according to the practical situation of object-oriented highdimensional judgment mode in graphic design.There are significant improvements in feedback prediction, data utilization, training space adaptability, target value evaluation, autonomous perception and autonomous perception [8].At the same time, this method can also improve the matching of high-dimensional decision modes in graphic design.Figure 4 shows the optimization of an algorithm used to enhance learning.

Introduction of probabilistic decision-making method based on action space selection
This paper proposes a new depth enhancement method for the discrete nonlinear coupling problem with few scales in the numerical space of motion.This method cannot solve the nonlinear coupling problem of the multi-dimensional and multi-variable wells.A gradient mechanism of deterministic strategy based on behavior space selection is proposed.The non-synchronous advantage Actor-Critic method based on the offline process is a high-dimensional decision model in graphic design.This paper uses the differential timing method to synchronize multiple threads asynchronously [9].The new technique can effectively improve the use of sampling.The paper applies deep neural networks to approximation.The approach combines depth with a defined strategic gradient mechanism over the same lifetime.This method overcomes the shortcomings of traditional feedback prediction, such as low utilization rate of data lag and poor adaptability of training space.The paper defines the decisive strategy as .Guided by the decision policy parameter , which represents the current state and the current behavior , the transformation of is completed.Define pattern expected reward : ( During iteration, each step needs to integrate all the operations as a whole, thus dramatically increasing the operation speed.This paper adopts a PG-based determination strategy.It's the function that determines a behavior.is the best action strategy .The performance objectives of DPGS are: (2) Then the gradient of the decision strategy is: (3)

Learning mechanism of multiple asynchronous enhancements using object-based assessment
DQN is a classical DQN algorithm.The destination value is the sum of the reward received immediately after the action and the value gained in the next state transition.Based on the traditional deep reinforcement learning method, the network parameters selected are very different to effectively control the best search route.There are cumulative problems with learning and convergence [10].An objective-based evaluation method based on multiple asynchronous enhancements is proposed.DDP uses the deterministic approach of to select behavior . is a parameter of a decision network.It generates a definite behavior.Use policy network as the role.function is fitted with a value net.This will act as a critique.The purpose function of DDPG is: (4) In this case, function is expressed as the expected return of the behavior using a deterministic strategy .The DQN schema in DDPG is inherited.
! is used to represent  parameter in A.  " (, ()) is the expected income obtained by using  strategy to select behavior in  pattern.Since we're in continuous space, the paper needs to get an available integral [11].The evaluation function representing the advantages and disadvantages of strategy  is as follows: (6) In the introductory paper, an object-based evaluation method based on multiple asynchronous enhancements is proposed to redefine the loss function: (7) can be expressed in equation (7) as follows: (8)

Adopt appropriate generation mechanism of noise network based on modeling to expand function
The greedy epsilon strategy is adopted to improve the detection performance of the model in the new environment.The institution is inherently flawed.It does not solve the persistent state problem well.
The convergence rate and training data utilization rate of this method is poor.This paper proposes a way of generating a spread spectrum noise network based on modeling ability to study the above problems.Adaptive noise with unknown detection and adjustment functions is generated by adding the proper generation mechanism of the noise network to the fully connected hierarchy of deep reinforcement learning [12].This improves the unknown detection performance of the system.The implementation procedure is as follows: The neural network that determines appropriate noise generation is determined by weighting and deviation.The definition of represents the result of a network generated by low noise.refers to a network input. is a physical map: The parameters  and  in equation ( 9) conform to normal distribution. is used to determine the normal allocation parameters.Other noises, including random noises, are denoted by .Assume that the initial noise is subject to the standard normal distribution (0,1): The adaptive noise generation performed according to equation ( 10) is:   ( , ) ) The function mapping ℎ in equations ( 11) and ( 12) is expressed as follows: (13)

Experimental results and analysis
The plan layout of the house designed by this design scheme is shown in Figure 5.It can be seen from the attached figure 5 that the designer effectively divides the spatial order and the internal plan effect with a strong rhythm by repeated combination.The value reflection of banquet restaurants is studied using the research method of this paper and the traditional research method.It can be seen from Table 1 that the overall impression score of this study is 9.0, while the conventional evaluation method is 8.3.Operating profit was 9,6.The overall score of this study was 8.91.The regular 8.52.It can be seen that the results of this study are better than those of conventional studies.

Figure 1 .
Figure 1.Graphic design flow chart

Figure 4 .
Figure 4. Improvement process of deep reinforcement learning algorithm

Figure 5 .
Figure 5. Graphic design effect of house

Table 1 .
Scoring results of the two methods