29results about How to "Address underutilization" patented technology

The invention discloses a personalized recommendation system based on semantic analysis, and relates to the technical field of personalized recommendation systems.The personalized recommendation system comprises a data acquisition module, a dynamic situational semantic modeling module and a recommendation engine module, the data acquisition module transmits acquired user behavior data and situational data to the dynamic situational semantic modeling module, and the dynamic situational semantic modeling module transmits the acquired user behavior data and situational data to the recommendation engine module; the dynamic situational semantic modeling module extracts sequence features of user behavior data through a behavior context analyzer, generates context embedding through a situational semantic embedding network, completes cross-modal semantic alignment of multi-modal data through a multi-modal fusion layer, and generates multi-level semantic embedding features. According to the method, the dynamic situational semantic modeling module is designed, the semantic alignment function of the situation and the behavior is achieved, the problem that a traditional recommendation system is insufficient in situation data utilization is solved, the adaptability of the recommendation result to the dynamic situation is enhanced, and the method is suitable for dynamically changing user demand scenes.

The present invention relates to a growth device and a growth method of a silicon carbide single crystal by a liquid phase method, and belongs to the technical field of silicon carbide single crystal production. The growth device includes a growth crucible, a seed crystal rod fixed with a seed crystal, a laser rangefinder for determining the position of the growth raw material, a movable graphite block, a braking device, and a blocking graphite; a reserved hole is provided in the crucible cover of the growth crucible corresponding to the laser rangefinder directly below the light exit hole; at least four rows of holes are provided on the side wall of the growth crucible, and each row of holes is evenly spaced from top to bottom, and a hole is provided at the center of the bottom, and the movable graphite block is filled in the hole, and one end of the blocking graphite is embedded in the hole, and the other end is sleeved on the outer wall of the growth crucible; an opening is provided in the center of the blocking graphite for the braking device to pass through and connected to the movable graphite block, and the braking device is used to control the movable graphite block to move forward and backward relative to the growth raw material. The device of the present invention can realize the determination and adjustment of the high temperature line position before crystal growth, can weaken the self-circulating convection in the growth solution, and improve the utilization rate of carbon.

The invention discloses a vehicle body grid parameter dynamic optimization method and system based on model simulation, and belongs to the crossing field of computer-aided engineering and artificial intelligence technology. According to an existing vehicle body grid optimization scheme, deep mining and analysis of grid data cannot be achieved, and consequently the accuracy and reliability of a grid optimization result are not high. According to the vehicle body grid parameter dynamic optimization method based on model simulation, a vehicle body historical database module, a vehicle body grid optimization module and a domain knowledge base module are utilized to match vehicle body grid data to be optimized with vehicle body historical data to obtain the most similar vehicle body historical data; the information is used as experience information and verification basis of vehicle body grid parameter optimization; and then experience information and a verification basis are reused, the to-be-optimized grid data of the vehicle body are analyzed and processed, and optimization suggestion information is generated, so that historical data can be deeply mined and analyzed, and the accuracy and reliability of a grid optimization result can be effectively improved.

The invention discloses a computer multi-process fair scheduling method, and relates to the technical field of multi-process resource allocation in a parallel computer architecture, the processes are dynamically divided into an I / O intensive type and a CPU intensive type through process classification and feature recognition, and the types of the processes are judged by monitoring the CPU use time and the I / O waiting time of each process. A scheduling strategy is set for each process, different priority adjustment strategies are formulated for different types of processes in dynamic priority adjustment, the length of time slices is dynamically adjusted according to the types of the processes, short time slices are allocated for I / O intensive processes, frequent I / O operation is allowed, returning is conducted in time, and waiting time is shortened. Compared with the prior art, the method has the advantages that long time slices are distributed for CPU intensive processes, frequent context switching expenses are reduced, the CPU utilization rate is increased, the execution requirements of different types of processes are better balanced by dynamically adjusting the lengths of the time slices, and the resource utilization efficiency of a system is improved.

The invention discloses a peak-valley difference arbitrage energy storage control method considering peak and deep valley electricity prices, which comprises load distribution prediction and charge and discharge starting time distribution based on the load distribution prediction, the load distribution prediction comprises charge prediction, discharge prediction and a prediction mode, the charge prediction comprises preconditions, opportunity and content, and the discharge prediction comprises preconditions, opportunity and content. The precondition is that the local electricity price has a deep valley; the charging prediction opportunity is prediction before the valley period starts; the charging prediction content comprises the amount of electricity capable of being charged in deep valleys and the amount of electricity capable of being charged in valley sections; the discharge prediction comprises opportunity and content, and the discharge prediction opportunity is predicted before the peak segment starts; the discharge prediction content comprises the discharge capacity of the peak and the peak section; the prediction method comprises the following steps: S1, finding an intra-day load change rule; s2, calculating a reference load base line of the day and comparing the reference load base line with the current load base line in each time period; s3, solving in a proper time period to obtain a predicated value of the chargeable and dischargeable electric quantity of each time period; the load prediction distributes charging and discharging starting time, and the charging logic and the discharging logic are included.

The invention discloses a turbofan engine RUL prediction method and system based on multi-scale historical information fusion, and the method comprises the steps: collecting various sensor data during the operation of a turbofan engine, and carrying out the preprocessing, and obtaining a plurality of pieces of time window data with RUL labels; processing data by using three sliding windows of different scales, superposing all historical time window data before the current moment on a third dimension, performing dimension reduction on the superposed data by using a principal component analysis method, and finally fusing the obtained data of three scales to obtain training and testing samples; an ISEformer network framework is adopted, and an RUL prediction model is obtained through training; and outputting a prediction result by the prediction model according to the operation data of the to-be-detected turbofan engine. Through the innovative method combining historical data fusion and the deep learning technology, the RUL prediction performance of the turbofan engine under the complex working condition is remarkably improved.

The invention relates to the technical field of water treatment, in particular to a catholyte-through multi-electrode-group parallel electrochemical reactor which comprises a reaction tank and a plurality of groups of electrode assemblies arranged in the reaction tank, the electrode assembly includes an anode; the cathode is annularly arranged outside the anode; the separation net is arranged between the anode and the cathode and is configured to isolate the catholyte from the anolyte; the base and the cover are provided with grooves for mounting the anode, the separation net and the cathode and are respectively mounted at the bottoms and the tops of the anode, the separation net and the cathode; the positive terminal is inserted into the cover and is connected with the anode; the negative terminal is inserted into the cover and is connected with the cathode; the acid extraction opening is formed in the center of the cover and penetrates through the cover to extend into the anode area of the anode; and the external pump is connected with the acid pumping port through a pipeline. The problem of insufficient utilization of hydroxyl in the existing water softening technology is solved, the number of the electrode groups can be flexibly adjusted according to actual requirements, the industrial-scale water treatment requirements are met, and meanwhile, the operation cost is effectively reduced.

This invention proposes a method and system for monitoring electricity trading risks based on a multi-source signal mechanism and a sparse hybrid expert model. The method includes: collecting historical day-ahead electricity market price data and raw data on influencing factors; performing data preprocessing to construct a multi-source fusion signal from price spread signals and bidding space signals; using a random forest algorithm for feature selection; constructing a set of similar day data based on the bidding space signal and incorporating it into the training set; using a sparse hybrid expert model as the model framework, with the price spread signal and bidding space signal as targets, training different prediction models; and searching for an optimal declaration curve based on the Northern Eagle optimization search algorithm by predicting the price spread signal and bidding space signal, using the short-term power prediction curve as a benchmark. This invention effectively solves the problem of insufficient model generalization ability, realizes risk monitoring in electricity spot trading, and improves the accuracy of electricity spot trading risk assessment while enhancing the stability of the power system.

The invention discloses an oil gas recovery intelligent control system and method based on multi-source data fusion, and relates to the technical field of intelligent control. The system comprises a data acquisition and classification module, an influence intensity judgment module, a priority dynamic adjustment module, a multi-source fusion control module and an equipment execution module. The data acquisition and classification module divides direct and indirect data, the influence strength judgment module dynamically captures data association strength without a preset rule, the priority dynamic adjustment module matches data priorities according to strength sorting, the multi-source fusion control module fuses high-priority data to generate a control instruction, and the equipment execution module adjusts operation parameters. Through dynamic collection of classified data, real-time judgment of influence intensity, dynamic adjustment of priority, multi-source fusion control and adaptation to multi-medium complex working conditions, the control accuracy is improved, the high recovery rate is stably maintained, emission reaches the standard, and the operation and maintenance risk is reduced.

The application discloses a kind of nonlinear carrier phase removal methods and systems of fringe projection image.The method obtains the fringe projection image of first and second view angle with nonlinear carrier;Including the following steps: (1) the fringe projection image of first and second view angle is fused using image splicing algorithm;(2) using singular value decomposition, the first principal component obtained is regarded as background component, and other components are regarded as foreground component;(3) detect foreground region and background region;(4) polynomial curved surface fitting is carried out to the background region of first and / or second view angle, and the first and / or second nonlinear phase is obtained;(5) subtract respective carrier phase and the respective nonlinear phase obtained in step (4).The application obtains global fusion fringe projection image without obstruction by double camera fringe imaging system, and detects background region based on singular value decomposition, so that nonlinear carrier phase is accurately removed.

The invention provides a large-scale generation method and system for interpretable question and answer pairs for the vertical field, and relates to the technical field of data processing, and the method comprises the steps: 1, constructing a cross-modal heterogeneous meteorological knowledge graph; heterogeneous graph neural network modeling is carried out on the cross-modal heterogeneous meteorological knowledge graph, an embedded vector of each node is learned, a knowledge graph with node embedded vectors is obtained, semantic community division is carried out on an entity based on the embedded vectors, and an optimized knowledge graph is generated. According to the method, through knowledge graph optimization, conical semantic space construction, structured evidence chain retrieval and pre-training large language model generation, large-scale generation of interpretable meteorological question and answer pairs in the vertical field is realized, and interpretability and credibility of question and answer results are improved.

The invention relates to the technical field of flue gas waste heat recovery, and provides an aluminum-silicon flue gas waste heat cooperative power generation system and an energy saving method.The system comprises a gas-water heat exchanger, a hot water pump, a waste heat boiler, a steam turbine, a power generator, a condenser, a circulating water pump, an aluminum electrolysis cell group, an ore-smelting electric furnace group, a dust removal system and a flue gas purification system; the aluminum electrolysis cell group is connected with the gas-water heat exchanger; hot water subjected to heat exchange of the gas-water heat exchanger is pumped into a deaerator through a hot water pump to be deoxidized and then enters a waste heat boiler, high-temperature flue gas of the ore-smelting electric furnace group is introduced into the waste heat boiler to heat the hot water, and then the hot water is discharged after being treated by a dust removal system; high-temperature steam generated by the waste heat boiler enters the steam turbine, and the steam turbine is matched with the power generator to generate power; after being condensed by the condenser, steam discharged by the steam turbine enters the gas-water heat exchanger through the circulating water pump for water replenishing and heat exchange. The system saves energy and reduces consumption, waste heat waste is reduced, greenhouse gas emission is reduced, and the requirement for green and low-carbon production is met.

The invention relates to a building energy calibration and prediction method based on a multi-task Gaussian process proxy model. The building energy calibration and prediction method is characterized by comprising the following steps: S1, data acquisition; s2, constructing an initial model; s3, sensitivity analysis and parameter screening; s4, sampling parameters; s5, performing batch simulation; s6, carrying out data sub-sampling; s7, constructing an agent model; s8, Bayesian parameter calibration is carried out; s9, predicting a physical model; s10, judging whether a preset standard is met or not; s11, correcting model deviation; and S12, final output is carried out. The method has the advantages that efficient and accurate calibration of building parameters and effective correction of model deviation are realized by constructing a weather decoupling agent model and a sequential calibration framework, and processing of high-dimensional data and construction of digital twinning can be supported.

This invention belongs to the field of bio-fermentation feed technology, specifically involving a two-step fermentation method for producing microbial protein bio-feed from corn husks. The invention provides a method for preparing microbial protein bio-feed, specifically: inoculating a first fermentation substrate with Aspergillus niger inoculum for a first fermentation to obtain a starter culture; and inoculating a second fermentation substrate with a compound microbial agent for a second fermentation to obtain the microbial protein bio-feed. The preparation method described in this invention can maximize the degradation of cellulose and hemicellulose in corn husk raw materials while considering cost, and increase its protein content through secondary fermentation, providing the feed industry with a method for preparing high-quality protein feed.

The present invention relates to the field of image safety supervision, and provides a wildfire anomaly detection method based on multi-frame dynamic processing of positive samples. The purpose is to improve the accuracy and stability of the wildfire detection algorithm through innovative dynamic processing and positive sample strategies, especially in the application of power transmission scenarios, so as to effectively reduce false alarms and improve detection efficiency. The main scheme includes obtaining a fine data set and training a detection model by annotating wildfire data. The video frame is differentially processed to identify the changed area, and the specific size area is extracted and mapped to N equally divided masks to achieve dynamic target detection. The sky segmentation model and OpenCV process the image, save the outline of the sky area, and filter out false alarms of clouds. The scoring mechanism determines the alarm, and an alarm is triggered when the continuous score exceeds the specified threshold. After the alarm, the fire alarm score is reset, the cooling period is started, and the alarm is suspended.

This invention discloses a method and system for reusing discarded convolutional information in small-sample fine-grained image classification, belonging to the field of computer vision. The method includes the following steps: acquiring images of a support set and a query set, and extracting the backbone feature descriptors of the images using a convolutional neural network; after at least one convolutional block in the convolutional neural network, retrieving the discarded feature information of the convolutional block through a feature retrieval path to generate a retrieved feature descriptor; fusing the backbone feature descriptor with the retrieved feature descriptor to obtain a high-dimensional feature descriptor in a unified format; calculating the similarity between the images of the query set and the support set based on the high-dimensional feature descriptor, and performing small-sample fine-grained image classification based on the similarity, thereby realizing the reuse of discarded convolutional information. This invention can effectively retrieve discarded feature information in convolutional neural networks, improving the performance of small-sample fine-grained image classification.

The present invention belongs to the technical field of material preparation and environmental pollution control, and discloses a method for preparing hydrophobic confined microbeads based on Ni-based bimetallic biochar and its Fenton catalytic application. The steps are: (1) culturing sweet osmanthus; (2) preparing Ni-based bimetallic biochar Ni / Fe-MSBC; (3) preparing Ni / Fe-MSBC@hydrophobic confined microbead reactor based on phase inversion technology. The Ni-based bimetallic biochar prepared by utilizing the property of sweet osmanthus that it is easy to enrich metals, and the porosity of the biochar and the dispersion of the metal are further improved by the molten salt method, so that the active sites are more evenly dispersed inside and on the surface of the biochar. In addition, the hydrophilic polymer and the hydrophobic polymer are used as carriers of the photo-Fenton catalyst, and the porous microbead reactor HCBRs with a hydrophobic confined microenvironment is prepared by regulating the blending of the hydrophilic polymer and the hydrophobic polymer. The HCBRs have excellent flexibility and stability, can effectively improve the catalytic activity and catalytic stability, and efficiently degrade antibiotic pollutants.

This application discloses a method and related apparatus for human behavior recognition based on graph-structured skeletal data using the replay method, relating to the field of computer vision technology. The method includes inputting target skeletal data into a trained action recognition classifier to obtain human behavior recognition results. The process of determining the trained action recognition classifier involves: dividing the graph-structured skeletal dataset into multiple disjoint subsets based on a partitioning criterion; treating each subset as a task; and utilizing a combination of buffer compression, composite pruning, and mask-based autoencoder reconstruction tasks, processing each task sequentially in chronological order to ultimately obtain the trained action recognition classifier. This application significantly improves the model's ability to retain memories of old tasks and its learning and adaptation capabilities for new tasks under fixed buffer resource constraints, effectively enhancing the accuracy and robustness of the final action recognition.

The invention relates to the technical field of experimental animal monitoring, in particular to a feeding device for experimental rat vital sign observation, which comprises a placement frame, a plurality of breeding assemblies placed in the placement frame, a separation assembly fixedly mounted at the lower end of a breeding box, and a limiting assembly movably arranged in the breeding box, the overturning assembly is fixedly installed on the surfaces of one sides of the multiple push plates. According to the invention, a grating cover plate above the breeding box is turned over, so that the grating cover plate drives a plurality of partition plates below the grating cover plate to move upwards, the breeding box is divided into a plurality of spaces, a plurality of white rats are separated, and meanwhile, the partition plates move upwards to drive a push plate to turn over to a vertical state and pull the push plate to move, so that the activity space of the white rats is limited; and when the push plate moves, the stirring roller rotates, strikes and stirs the padding at the bottom of the breeding box and scatters and turns over the padding, the situation that the padding on the lower layer is not fully utilized due to the fact that the padding on the upper layer is bonded into blocks is avoided, and the utilization rate of the padding is increased.

The present disclosure provides a mobile sales inspection method, a scheduling method and an apparatus, the mobile sales inspection method comprising: at a mobile sales inspection end: according to a preset automatic driving strategy, driving along a predetermined route; obtaining real-time monitoring data of the route area during driving, and sending the real-time monitoring data to a remote platform; simultaneously receiving sales task information sent by the remote platform and detecting purchase identifications of users in the route area during driving; and selecting a parking location for sales according to the sales task information, or selecting a parking location for sales according to the purchase identification.

The utility model relates to a waste steam heat energy recycling structure adopting two-stage heat exchange treatment, which belongs to the technical field of waste steam heat energy utilization and comprises a water tank body, a partition plate dividing the water tank body into two inner cavities is arranged in the water tank body, and a spray tower I used for being communicated with the water tank body is arranged at the upper end of one side of the water tank body. A waste steam conveying pipe communicated with the spraying tower I is arranged on the side wall of the lower portion of the spraying tower I. A spraying tower II communicated with the other side of the water tank body is arranged on the side edge of the spraying tower I. A steam-water separator communicated with the upper end of the spraying tower I, the lower portion of the spraying tower II and the water tank body through pipelines is arranged between the upper end of the spraying tower I and the lower portion of the spraying tower II. Enhanced heat exchange pieces positioned above the waste steam conveying pipe are arranged in the spray tower I and the spray tower II; and exhaust pipes are arranged at the upper ends of the spray tower I and the spray tower II. The device has the characteristics of simple structure, good cooling treatment effect and high operation stability. The problem that low-pressure waste steam heat energy is not fully utilized is solved.

The present invention discloses a micro-expression recognition method based on multimodal fusion, which relates to intelligent micro-expression recognition, including obtaining input data of a model from a micro-expression database, extracting features from each input data using ViT and a convolutional neural network CNN, and obtaining a feature vector V a and the eigenvector V f , for the eigenvector V a and the eigenvector V f Similarity measurement is performed to obtain output features. By fusing peak frames, optical flow maps, and key point classification micro-expression data, and utilizing data features of different modalities, the problem of insufficient utilization of micro-expression data in the existing technology is solved; for data of different modalities, a network architecture is adopted: Vision Transformer (ViT) is used to extract global features of peak frames, and convolutional neural network (CNN) is used to extract local features of optical flow maps. The fused features are optimized through contrast loss and classification loss, which improves the similarity of features of the same expression sample and enhances the accuracy and robustness of recognition.

This invention relates to the field of intelligent decision-making technology and discloses an intelligent decision-making system for agricultural irrigation based on artificial intelligence. The system constructs a collaborative architecture comprising a field perception layer, an edge computing layer, a cloud-based decision-making layer, and an irrigation execution layer, forming an irrigation task context through multi-source perception data. Based on this, a meta-reinforcement learning-driven multi-round agent-based search decision-making mechanism is introduced, combined with explicit self-reflection to achieve iterative optimization of the decision-making process. Simultaneously, a preference-aware reward evaluation model is constructed, achieving multi-objective trade-off modeling through latent variable mirror constraints and inverse autoregressive flow transformation, and globally optimizing the strategy using a cross-round benefit attribution mechanism to generate an irrigation scheduling scheme that meets water-saving, crop water requirement, and energy consumption constraints. This invention can improve the accuracy, stability, and adaptability of irrigation decisions.

The utility model relates to a waste steam heat energy recycling structure adopting three-stage heat exchange treatment, which belongs to the technical field of waste steam heat energy utilization and comprises a water tank body, a first-stage spray tower communicated with the water tank body is arranged at the upper end of the water tank body, an exhaust pipe is arranged at the upper end of the first-stage spray tower, and a second-stage spray tower communicated with the exhaust pipe is arranged at the lower end of the exhaust pipe. An enhanced heat exchange piece is arranged in the middle of the first-stage spray tower, a steam exhaust pipe communicated with the first-stage spray tower is arranged on the outer side of the first-stage spray tower, and a heat exchanger I communicated with the first-stage spray tower and a tank body of the water tank through pipelines is arranged between the steam exhaust pipe and the first-stage spray tower; and a waste steam conveying pipe communicated with the lower part of the enhanced heat exchange piece is arranged between the heat exchanger I and the primary spray tower. The device has the characteristics of simple structure, good cooling treatment effect and high operation stability. The problem that low-pressure waste steam heat energy is not fully utilized is solved.

The invention relates to the field of coal mine three-dimensional geological modeling, in particular to a multi-interpolation three-dimensional geological modeling method based on drill hole data, which comprises the following steps: determining a three-dimensional geological modeling range and exploration data of a plurality of drill holes in the three-dimensional geological modeling range, determining plane coordinates of different drill holes, and converting the same stratum demarcation point of the different drill holes into a space z coordinate; performing interpolation operation by adopting different interpolation operation methods to obtain point cloud data of each stratum; comprehensive evaluation analysis is carried out to determine a proper interpolation operation method for each stratum; exporting the point cloud data to rhinoceros modeling software; extruding an entity in rhinoceros modeling software, performing grid division according to a Griddle plug-in, and generating a multi-stratum three-dimensional geologic model; and importing the generated multi-stratum three-dimensional geologic model into other software for analysis. According to the method, advantages and disadvantages of various interpolation operation methods can be considered through the established evaluation indexes, and it is ensured that the generated model better conforms to actual geological conditions; the method is especially suitable for modeling work in a complex geological environment.

The invention discloses a pointing object text positioning method based on iterative semantic visual association. The method comprises the following steps: acquiring a document image containing a pointing object and a natural language instruction; visual and language features are extracted, a pointing object bounding box is predicted, and a pointing mask is generated to extract pointing context features; the pointing context features and instruction semantics are spliced to generate a channel scaling coefficient, and space and channel modulation is performed on the visual features; using instruction semantics to generate FiLM parameters to modulate the enhanced visual features, and predicting an initial target bounding box through cross-modal attention fusion features; and starting an iterative refinement process, generating a target mask according to the current prediction frame and calculating a semantic-visual consistency score, modulating fusion features in combination with geometric difference and semantic information to predict a better bounding box, and outputting a result with the highest score after iteration is performed until a termination condition is met. According to the method, fine-grained spatial semantic understanding and high-precision iterative positioning capabilities are enhanced.

This invention provides a method and system for dynamic optimization of UAV flight trajectories based on neural networks, belonging to the field of data processing technology. The method includes: calculating the geometric moments of a two-dimensional contour; obtaining the central moment of the contour based on the geometric moments; normalizing the central moment and extracting Hu invariant moment features; combining shape fit features and Hu invariant moment features to form a digital environment feature vector; processing the digital environment feature vector with pose state information and a preset desired trajectory; fusing the results through a neural network model to obtain control parameter adjustment instructions; and, based on the control parameter adjustment instructions, real-time correcting of the UAV flight controller parameters to generate control signals driving the UAV actuators, thereby achieving dynamic tracking of the desired trajectory. This invention enables accurate dynamic tracking of complex trajectories by UAVs.

The application discloses a kind of real-time semantic segmentation method and system based on multi-resolution branch, it is related to graphics processing technical field.By obtaining image data, using depth multi-resolution network to carry out semantic segmentation to image data, obtain semantic segmentation feature map;Wherein, semantic segmentation process includes: to image data is carried out convolution feature extraction, obtains the spatial information and semantic information of image;Using different resolution branch network to the spatial information and semantic information of image respectively learns to obtain the feature map of each branch network;Fusion is the feature map of each branch network, obtains the final semantic segmentation feature map.The application extracts the semantic and spatial information of image from different resolution respectively, solves the problem of insufficient utilization of semantic and spatial information.The application also adds semantic enhancement module and feature fusion module, improve the effect of semantic information extraction while also solve the problem of spatial information being covered under multi-branch.

The application discloses a medical image segmentation method combining a double attention mechanism and a U-Net++, and related devices. The method comprises the following steps: acquiring a medical image and inputting the medical image into a segmentation network; extracting multi-scale coding features through an encoder, and generating multi-scale decoding features through a decoder; inputting the same-scale features into a double attention module, and sequentially executing channel attention and spatial attention; and finally, fusing the spatially weighted coding features to the decoder, and outputting a segmentation result. The application can be deployed in an artificial intelligence optimization operating system or an artificial intelligence middleware platform, high-efficiency calculation is realized by using a related function library, is suitable for computer audio-visual software, biological feature recognition software and other artificial intelligence application software, and the double attention mechanism and dense skip connection are combined, so that the feature expression capability is effectively enhanced, and the precision and robustness of medical image segmentation are improved.