48 results about "Level set" patented technology

The invention discloses a self-adaptive point cloud implicit curved surface reconstruction method based on a truncation level B spline, and the method comprises the steps: constructing a first-layer truncation level B spline primary function space, and obtaining an initial implicit function through solving an optimization model; evaluating the fitting error of each point of the input point cloud, and determining the fitting error as a region which needs to be further refined; establishing a new truncation hierarchy B-spline primary function set, and expanding a current function space; performing truncation operation on the THB spline primary function which is completely positioned in the refined region; combining the truncated primary function with the primary function set generated by the new layer, and updating the whole primary function library; re-optimizing the model, and solving the updated implicit function; refining and optimizing layer by layer, and outputting a final implicit function and a zero level set thereof; and performing triangular meshing on the final zero level set to finish curved surface reconstruction. According to the method, an extra training process is not needed, excellent point-to-surface distance and curvature indexes can be kept at the noise level up to 10%, and the memory and calculation overhead can be greatly reduced.

The invention discloses a multi-objective topological optimization method for aircraft design, and belongs to the technical field of aerospace. The method comprises the steps that a three-dimensional geometric model is established, and an optimization area and an aerodynamic reservation area are delimited; describing a structural boundary by adopting a level set function; with rigidity maximization and inherent frequency maximization as targets, volume and manufacturing constraints are applied; utilizing a compromise planning method to construct a comprehensive objective function for minimizing the ideal point distance; transforming the constraint problem through an augmented Lagrangian method; calculating shape sensitivity and exporting a normal velocity field; updating a level set function to drive topology evolution; iterating until convergence, and generating a Pareto frontier solution set; and selecting an optimal scheme and reconstructing the optimal scheme into a three-dimensional model capable of being manufactured. On the premise of keeping the aerodynamic configuration, the problem of multi-target collaborative optimization is solved, the boundary is clear, the manufacturability is high, and the structural performance and the engineering applicability are remarkably improved.

The embodiment of the invention discloses a bill key field identification method and device. The method comprises the following steps: acquiring a bill image; defining the key field, wherein the key field comprises at least one of an invoice number, an amount and a date; fractional order differential calculation is carried out on the bill image, an enhanced image and fractional order differential features are generated, and the fractional order differential features comprise a fractional order gradient field and a multi-scale feature vector; semantic features of the bill image are extracted according to a predefined bill semantic template, semantic priori is generated, and the semantic priori represents the relative position and expected content of the key field; a level set function is initialized based on the fractional order differential features and the semantic priori, an energy function is constructed, the energy function is minimized to drive the level set function to evolve, and a segmentation result is generated; and performing feature extraction on the segmentation result to generate structured data. According to the invention, the accuracy and efficiency of key field identification are improved.

The invention discloses a level set field reconstruction method and system based on a three-dimensional finite element grid and a medium. Relates to the technical field of geometric reconstruction. Positioning a space Cartesian grid area intersected with the boundary surface grid based on a separation axis theory; calculating a geometric space point-surface position relation, calculating a directed projection distance from a space Cartesian grid center to a two-dimensional boundary surface grid, and correcting the positive and negative characteristics of a Cartesian grid level set value based on a three-dimensional eight-direction rapid scanning method, so that the grid level set value is consistent with a region boundary value; according to the method, the geometric reconstruction problem of the deformable body which is complex in appearance and free of explicit analytic expression can be efficiently and accurately solved, a coupling approach between a finite element grid and a Cartesian grid is established under a non-conformal grid system, and the flexibility and universality of explicit calculation of a level set function under the Cartesian grid are improved.

The application belongs to the technical field of image segmentation, and particularly relates to a complex scene target segmentation method and system based on image recognition, which comprises the following steps: performing superpixel adaptive division on an input image, fusing gray scale and texture features to determine a superpixel boundary, mapping the superpixel boundary into a graph node and calculating an edge weight, and constructing an undirected weighted graph; adaptively encoding a graph signal, utilizing a hybrid graph wavelet-Fourier joint transform to optimize and separate features, and obtaining purified graph frequency domain features; sparsely reconstructing features through adaptive super-complete dictionary learning, and obtaining target enhanced features; extracting topological parameters based on an improved persistent homology, constructing a topological constraint feature graph, mapping the topological constraint feature graph into an initial contour field, iteratively optimizing a level set and a contour through an adaptive partial differential equation, and obtaining a high-fidelity coarse segmentation result; extracting geometric features to construct a joint constraint model to repair an occluded area, and outputting a precise segmentation result. In the application, sparse topological modeling is adopted, weak features are strengthened, and target discrimination accuracy is improved.

The present application belongs to the technical field of wind power generation, and specifically relates to a wind power tower material crack propagation prediction method and system, which is suitable for fatigue life evaluation of large supporting structures in the wind energy industry. The method comprises: deploying a multi-source heterogeneous sensor network to collect strain, acoustic emission and environmental load data; constructing a digital twin high-fidelity finite element submodel to obtain a dynamic stress intensity factor; constructing a deep neural network, embedding the Paris law of fracture mechanics in the form of residual into the loss function to form a physical information neural network, and realizing crack propagation rate prediction; using the extended finite element method combined with the level set function and the maximum circumferential stress criterion to dynamically track the three-dimensional space expansion path of the crack; combining Monte Carlo simulation and recursive Bayesian update to evaluate the remaining life and trigger graded early warning. The present application solves the problems of insufficient fusion of physical mechanisms and data-driven in the prior art, low prediction accuracy under small samples, and lack of crack space path tracking.

The present invention belongs to the field of image detection technology and is a SAR image coastline detection method based on level sets. The level set function is initialized as a signed distance function, the grayscale means of the foreground and background are calculated according to the initialized level set function, and each point of the level set function is evolved using the grayscale means of the foreground and background estimated previously. The iterative step size of the level set function in the evolution process is adaptively changed by a step size control function. If the level set function value of the current point is greater than 100 or less than -100, the step size is changed from the original 0.1 to 1, otherwise it remains unchanged. The level set function is then continuously evolved. According to the definition, foreground>0 and background<0, if the grayscale value of the current point is close to the foreground valuation, the value of the level set is correspondingly increased, otherwise it is decreased. The level set function is reinitialized according to the penalty term to make it conform to the signed distance function. Finally, it is checked whether the level set function converges, and if not, it is re-evolved.

This invention discloses a multi-objective topology optimization method for aircraft design, belonging to the field of aerospace technology. The method includes: establishing a three-dimensional geometric model and defining the optimization region and aerodynamic preservation region; describing the structural boundary using level set functions; applying volume and manufacturing constraints with the objectives of maximizing stiffness and natural frequency; constructing a comprehensive objective function that minimizes the distance between ideal points using a compromise programming method; transforming the constraint problem using the augmented Lagrangian method; calculating shape sensitivity and deriving the normal velocity field; updating the level set functions to drive topology evolution; iterating until convergence to generate a Pareto front solution set; selecting the optimal solution and reconstructing it into a manufacturable three-dimensional model. This invention solves the multi-objective collaborative optimization problem while preserving the aerodynamic shape, resulting in clear boundaries, strong manufacturability, and significantly improved structural performance and engineering applicability.

The present application relates to the field of additive manufacturing, and specifically discloses a kind of generation method of manufacturing three-dimensional Chinese character cup model based on energy driving.The present application constructs the smooth time evolution path from multiple outline initial region to single target outline, adopts level set function to describe the outline evolution process, and establishes energy functional including curve length regularization term, adaptive inflation term and shape prior term;Through introducing the dynamic coupling control mechanism based on area difference of region, shape error and boundary length, the adaptive adjustment of inflation coefficient is realized;Then the level set evolution equation is obtained by variational derivation, and after numerical discrete solution, two-dimensional evolution sequence is added to three-dimensional volume field along time axis;Finally, through isosurface extraction and thick shell processing, the 3D printable three-dimensional Chinese character cup model with complete structure and uniform wall thickness is generated.

This application belongs to the field of structural optimization technology and specifically discloses a parameterized level set topology optimization method for porous structures that takes geometric nonlinearity into account. The method comprises: constructing a level set function to implicitly describe the structural boundary, and using compactly supported radial basis function spline interpolation to obtain the parameterized level set function; establishing an equilibrium equation for geometric nonlinear topology optimization, and performing geometric nonlinear finite element analysis to solve the structural response under large deformation conditions; approximating and simplifying local volume constraints through local constraint domains and p-norm functions; establishing a topology optimization model based on maximizing the geometric nonlinear porous filling stiffness; performing sensitivity analysis through the adjoint method, and calculating the partial derivatives of the objective function of the topology optimization model with respect to the expansion coefficient to determine the optimization direction of each finite element in maximizing the geometric nonlinear porous filling stiffness; and performing iterative optimization to obtain the optimized topology structure. This application can realize the topology optimization of porous fillings under geometric nonlinear conditions.

The invention belongs to the technical field related to structure optimization design, and discloses a lattice structure optimization design method and device combining a multivariable cutting level set and progressive optimization, and the method comprises the steps: (1) building a lattice structure optimization model by taking structure flexibility as a target function and cutting height as a design variable; (2) performing finite element analysis based on the lattice structure optimization model to obtain a structure volume fraction; (3) if the current number of iterations is an odd number or the structural volume fraction is lower than the set threshold value, calculating the sensitivity of the target function and the constraint function to the design variable, and updating the design variable by adopting a moving asymptote algorithm based on the obtained sensitivity; if the number of iterations is an even number and the volume fraction of the structure exceeds a set threshold value, calculating a unit sensitivity degree so as to obtain a sensitivity degree of the virtual microstructure, and removing a part of the virtual microstructure based on the sensitivity degree of the virtual microstructure; and a final lattice structure is further obtained. According to the invention, the optimization efficiency is improved.

The application provides a kind of marine structure flow field simulation method, the method comprises: generating calculation grid and object surface grid for flow field simulation;Initialize object surface grid level set function field;According to flow field information, calculate the force condition of object surface, calculate object translation vector and rotation vector;Complete data mapping of object surface grid to calculation grid, obtain the level set function field of calculation grid;According to the obtained calculation grid level set function field, divide cell type;Carry out geometric calculation to cut cell, based on cut cell flow field reconstruction strategy, complete fluid cell, cut cell estimated speed update;Solve pressure Poisson equation, obtain pressure field, pressure gradient, update velocity field according to the obtained pressure gradient, obtain the final velocity field satisfying continuity condition.The marine structure flow field simulation method of the application greatly improves the numerical simulation efficiency of complex structure flow field, and reduces the implementation difficulty.

The present application belongs to the technical field of structure optimization design, and discloses a structure topology optimization design method and device, steps are as follows: (1) dividing a physical design domain into a background grid, based on coordinate information of background nodes of the background grid, a sparse filtering matrix is constructed by using a Gaussian kernel function with row normalization; (2) using the sparse filtering matrix to perform front filtering on design variables on the background nodes to obtain a smooth density field, and then obtaining a level set function field; (3) using an adaptive grid subdivision algorithm to cut and locally adaptively reconstruct the level set function field to obtain an analysis grid; (4) interpolating the level set function field to the analysis grid through a sparse global interpolation matrix from the background grid to the analysis grid to obtain a level set function field; (5) projecting the level set function field to obtain a pseudo-density field on the analysis grid nodes, and then obtaining an optimal topology configuration. The present application realizes good boundary smoothness while improving optimization efficiency.

The present application provides a training method for a mask pattern correction model and a mask pattern correction method. The method includes: obtaining a pre-OPC mask pattern and a post-OPC mask pattern; calculating a first level set function and a second level set function based on the pre-OPC mask pattern and the post-OPC mask pattern, respectively; and training the first level set function as the input of a neural network model and the second level set function as the output of the neural network model to obtain a mask pattern correction model. The method provided in the present application utilizes the first level set function and the second level set function to represent the images of the pre-OPC mask pattern and the post-OPC mask pattern on a discrete grid, eliminating the need to use a large number of feature vectors to encode the mask pattern's proximity, thereby improving the efficiency of OPC.

The invention relates to a high-order interface capturing method for a coupling level set and a fluid volume based on mixing precision, which comprises the following steps of: firstly, dividing a calculation area into a grid comprising a plurality of control body units, initializing, then identifying interface units, and reconstructing a THINC interface function by adopting a high-order polynomial based on an LS value; thirdly, introducing a mixed precision iterative algorithm to solve an interface position nonlinear equation, and adaptively switching to multi-precision operation to obtain a robust solution when double precision cannot converge; then, flux is calculated through an accurate interface, a VOF transport equation is solved, and mass conservation is guaranteed; and finally, on the basis of the predicted interface geometry, an LS field which does not need to be reinitialized is directly reconstructed through pure geometric methods such as interface point cloud generation through projection and nearest point searching in a narrow band. According to the method, the capture precision and geometric fidelity of complex interfaces with high curvature, sharp corners and the like on the unstructured grids are remarkably improved, and the method has high precision and strong robustness.

The invention belongs to the technical field of aero-engine cooling design and flow heat exchange optimization, and provides a turbine blade cooling turbulence rib construction method, system and configuration based on level set topological optimization and a blade. The method comprises the steps that a fluid-solid coupling calculation model containing a cooling channel fluid domain and a blade wall surface solid domain is constructed, defining a level set function to implicitly represent the two-phase region; establishing a unified control equation set covering a full computational domain based on the function, and introducing a flow velocity suppression mechanism into a solid domain to carry out single-domain coupling solution of fluid flow and solid heat conduction; according to the physical field solution, constructing an optimization model taking the cooling performance as a target function and the volume fraction of the turbulent flow rib as a constraint, and calculating the sensitivity of the target function to a level set function; using the sensitivity to drive a level set function update to evolve a spoiler rib boundary; iterating until convergence and outputting the final geometric configuration of the turbulent flow rib. According to the invention, the irregular turbulent flow structure with high heat exchange efficiency and low flow resistance can be automatically generated without presetting a rib type.

The present invention relates to an active contour model image segmentation method based on K-medoids clustering, comprising: using a K-medoids clustering algorithm to perform binarization processing on an image to be segmented to obtain a fitting result, and acquiring a foreground image and a background image; constructing a pre-segment fitting function based on the binarization fitting result; using a zero level set that meets the Lipschitz condition to replace the contour curve of the pre-segment fitting function to obtain a KM pre-segment fitting function; using a gradient descent method to minimize the energy functional of the KM pre-segment fitting function to obtain a gradient flow equation; using an adaptive regularization function to regularize a data-driven term of the gradient flow equation; using a rulesig function to regularize the zero level set function, using a kernel function to smooth and shorten the curve of the regularized zero level set function, and outputting a level set function as a segmentation result of the image to be segmented.

The invention provides a reactive infiltration process numerical simulation method based on COMSOL multi-physics field coupling, which comprises the following steps: constructing a simplified liquid silicon infiltration model, and simulating a pore structure of a porous body by adopting a two-dimensional single capillary equivalent model; based on the liquid silicon infiltration model, coupling a laminar flow-level set two-phase flow field, a chemical reaction field, a dilute substance transfer field and a fluid heat transfer field to form a multi-physical field numerical model; and a liquid silicon permeation height curve, concentration distribution and pore structure evolution data are obtained through transient solution. According to the method, the time cost required by calculation is saved by simplifying the model, meanwhile, the accuracy of the model can be ensured, and the pore structure change in the reaction infiltration process is predicted according to the reaction layer change rule obtained through simulation; the problem that in the prior art, pore closing critical conditions and a reaction-permeation dynamic coupling mechanism are lack of systematic revelation is effectively solved, theoretical support is provided for directional regulation and control of prefabricated body pore parameters and optimization of process parameters, and the method has remarkable innovativeness and practicability.

The invention discloses a shot peening forming process planning method based on a level set algorithm, and the method comprises the steps: S1, obtaining the features of a target wallboard, and selecting the type and number of level set functions according to the features of the target wallboard, so as to describe the boundary between a shot peening region and a non-shot peening region of the target wallboard; s2, establishing a shot peening forming process optimization mathematical model based on the level set function; s3, calculating the sensitivity of an objective function of the optimization model, and deducing an iterative formula of the level set movement velocity field; s4, the optimization model is initialized, the level set function is solved, the shot blasting area boundary is updated, if the boundary is not convergent, the step S3 is executed, and if the boundary is convergent, the step S5 is executed; and S5, discretely generating a shot blasting impact point set based on the converged shot blasting area boundary. According to the method, the problems existing in shot peening forming of complex wallboards in an existing shot peening process are solved, the process planning cost is remarkably reduced, and the forming precision is improved.

This invention belongs to the field of aero-engine cooling design and flow heat transfer optimization technology. It provides a method, system, configuration, and blade for constructing turbine blade cooling fringe based on level set topology optimization. The method includes: constructing a fluid-structure interaction (FSI) computational model encompassing the fluid domain of the cooling channel and the solid domain of the blade wall; defining a level set function to implicitly represent the two-phase region; establishing a unified set of governing equations covering the entire computational domain based on this function; introducing a velocity suppression mechanism in the solid domain to perform single-domain coupling solution of fluid flow and solid heat conduction; constructing an optimization model based on the physical field solution, with cooling performance as the objective function and fringe volume fraction as the constraint; calculating the sensitivity of the objective function to the level set function; using this sensitivity to drive the level set function update to evolve the fringe boundary; iterating until convergence to output the final fringe geometry. This invention does not require pre-setting the fringe type and can automatically generate irregular fringe structures with high heat transfer efficiency and low flow resistance.

This invention belongs to the field of energy-absorbing metamaterial design, specifically to a method for inverse design of energy-absorbing metamaterials based on a diffusion model and parameterized level sets. The invention generates a metamaterial given a desired stress-strain curve. The diffusion model learns the conditional distribution of metamaterials with given properties, achieving a one-to-many mapping from properties to geometry. The use of a level set representation improves the precision and design flexibility of the metamaterial topology, ensuring that the generated metamaterial possesses high-quality geometry. This invention offers the advantages of fast design speed and high precision, providing a new solution path for more intelligent and customized metamaterial design.

A solution search device (1) comprises: a setting unit (12) for setting a first desired level, which is a target value for a plurality of objective functions, in a multi-objective optimization problem; a first-optimal-solution computation unit (13) for computing a first optimal solution using the first desired level set by the setting unit and a bound, which is a solution set set for optimizing the plurality of objective functions; and a first-optimal-solution fitting unit (15) for receiving an input of an adjustment amount to fit the first optimal solution from a user when an instruction not to use the first optimal solution computed by the first-optimal-solution computation unit is received from the user, and for fitting the first optimal solution within a range of the bound based on the received adjustment amount.

Techniques for level-set based shape optimization with shape constraints include simulating a state of a current shape of a component, calculating a sensitivity of an objective function to with respect to modification to the current shape based on the simulated shape based on the simulated state, generating a source function for an optimization equation based on the sensitivity, adding one or more constraint terms to the optimization equation based on one or more constraints on an interface of the shape to generate a constrained optimization equation, solving the constrained optimization equation to generate a solution to the constrained optimization equation, updating a velocity field based on the solution to the constrained optimization equation, solving a transport function based on the updated velocity field to generate a solution of the transport equation, and updating the current shape based on the solution to the transport equation to generate an updated shape.

The invention relates to the technical field of image processing, and discloses a target segmentation method for a complex background image, and the method comprises the steps: carrying out the pre-fitting of foreground and background features of a to-be-segmented filling image through fuzzy C-means clustering based on parallelization and down-sampling strategy improvement, and reconstructing a pre-fitting denoising image; jS divergence and Euclidean distance are introduced to carry out joint multiplicative measurement analysis on three components of the foreground, the background and the de-noised image obtained through pre-fitting calculation, and difference energy is obtained; and constructing a data driving item containing a global fidelity fitting item and a local fitting item based on the filling image to be segmented, the pre-fitting component and the difference energy, performing iterative updating on the initial level set function to obtain a final contour line, and segmenting the image to be segmented to obtain an image segmentation result. According to the method, the accuracy and robustness of target segmentation under a complex background image are improved through fuzzy C-means clustering based on parallelization and down-sampling strategy improvement and difference energy of JS divergence introduction.

The invention relates to the technical field of computer vision and three-dimensional reconstruction, particularly discloses a neural unsigned distance field (UDF) learning method and a surface reconstruction method based on cross product geometric prior, and aims to solve the technical problems that the gradient direction is unstable and the reconstructed surface is fragmented and discontinuous due to the fact that an existing neural UDF cannot be differentiated at a zero level set. The method is characterized in that a geometric prior constraint mechanism based on an original point cloud cross product is introduced; a local surface normal vector is estimated by calculating the cross product robustness of three adjacent points in an input point cloud, and a neural UDF gradient direction learned by constructing a loss function forced constraint is strictly aligned with a normal vector direction; in the optimal scheme, near surface optimization is focused in combination with an adaptive weight strategy, and a network is jointly driven to learn a more accurate and smoother UDF in combination with unsupervised chamfering distance loss and surface distance constraint. According to the method, the fuzziness of the UDF gradient near a zero level set is effectively overcome, the continuity and integrity of a reconstructed surface are remarkably improved, the method is particularly suitable for high-quality reconstruction of an open surface and an object with a complex internal structure, and the key indexes of the method are superior to those of an existing advanced technology through verification of multiple standard data sets.

The invention relates to the field of ship navigation, and provides a ship path planning method, a ship path planning device, electronic equipment and a computer readable storage medium, the ship path planning method comprises the following steps: constructing a level set equation based on dynamic environment data of a ship, the level set equation comprising dynamic environment data items; dynamically constructing a reachable region boundary of the ship based on a nonlinear development partial differential equation by taking the level set equation as a state variable; and solving the level set equation based on the reachable region boundary and the dynamic environment data to obtain a planned path. According to the ship path planning method, the ship path planning device, the electronic equipment and the computer readable storage medium provided by the invention, the technical problem in the prior art that dynamic sea conditions greatly influence ship navigation is solved.

The invention relates to the technical field of pipeline layout, in particular to a multi-communication pipeline layout optimization design method based on a worm simulation characteristic, and can solve the problems that the directional matching requirement of multiple groups of inlets and outlets of complex pipelines is difficult to meet and multi-pipeline three-dimensional space interference cannot be effectively avoided in the prior art to a certain extent. The method comprises the steps that a design domain is given, finite element grid division is carried out, the number of simulated worm features is determined, an initial layout is generated, an implicit level set function is calculated, the implicit level set function of a pipeline system is obtained through operation, a non-interference constraint function is defined, constraint condition data is generated, a density variable field is obtained through mapping of a fairing Heaviside function, and the density variable field is obtained. And establishing a flow resistance coefficient interpolation model, solving a Navier-Stokes equation of fluid in the pipeline, selecting fluid flow dissipated energy as an optimization target, taking a non-interference constraint condition as a constraint, constructing an optimization problem model, iteratively optimizing a design variable through a mathematical programming algorithm until a convergence condition is met, and outputting a pipeline layout scheme.

The invention discloses a transient synchronous stability analysis method and system for a multi-type converter grid-connected system based on coordinate transformation and combined Lyapunov. According to the method, a combined Lyapunov function theory is taken as a core, a trainable coordinate transformation layer is introduced into a Lyapunov neural network structure, decoupling and dimension reduction of a high-dimensional coupling system are realized through implicit coordinate transformation, a complex multi-type converter grid-connected system is decomposed into a plurality of low-dimensional subsystems for independent modeling, and a combined Lyapunov function is constructed on the basis. And estimating the stability domain boundary of the system based on the level set structure of the function. The method effectively overcomes the dimension disaster problem of a high-dimensional coupling system, has calculation high efficiency and theoretical interpretability, is especially suitable for transient synchronization stability analysis of a large-scale grid-connected system with multiple fans and multiple converters, and provides a systematic solution for practical engineering application.

The present application proposes a shell-based metamaterial design strategy based on level set function weighted combination. The framework is inspired by three-periodic minimal surfaces, and adopts a level set function weighted combination strategy to construct a shell-based metamaterial database. The effective elastic properties of the metamaterial are evaluated by the homogenization theory of voxel modeling. A backpropagation neural network model is developed, which can realize the reverse mapping between the geometric parameters and material properties of the shell-based unit. In addition, the present application verifies the reliability of the homogenization theory by using quasi-static compression tests, and evaluates the effective elastic properties and compression response of the proposed shell-based structure. The research results show that the shell-based metamaterial not only shows isotropic elasticity and high stiffness beyond the upper limit of Hashin-Shtrikman theory, but also has significant improvement in yield strength and energy absorption capacity under compression. The design finally forms a series of lightweight, high-stiffness structures, and is 3D printed by PolyJet technology.

This invention relates to the field of image processing technology and discloses a method for polyp image segmentation and a computer-readable storage medium. First, the polyp image to be segmented is input into a backbone network to extract multi-layer initial feature maps. Then, the initial feature maps are sequentially input into a level set edge enhancement module. Learnable parameters and the geometric properties of the level set function are used to explicitly model and dynamically enhance the edges to suppress feature confusion and obtain an optimized feature map. The optimized feature map is then input into a dual-branch attention module, which coordinates global and local information to adaptively calibrate channel weights. This optimized feature map is then input into a cross-layer fusion module for multi-scale feature complementarity, outputting an attention-enhanced feature map. Based on the cross-layer fusion module, deep feature aggregation, and layer-by-layer feature aggregation, layer-by-layer fusion is performed to obtain the segmentation result map. This invention effectively solves the problems of over-segmentation and under-segmentation of polyp images caused by edge blurring and feature confusion, significantly improving the accuracy of polyp segmentation tasks.