78 results about "Enhancement Technologies" patented technology

The invention discloses a building structure crack intelligent detection method, which comprises the steps of S1, constructing a multi-dimensional intelligent sensing array, and obtaining four-dimensional spatio-temporal data including vision, stress, vibration and temperature; s2, performing feature enhancement processing on the multi-modal data; s3, realizing cross-modal crack identification and positioning based on a space-time attention neural network; s4, adopting a sub-pixel edge detection and ultrasonic genetic inversion algorithm; s5, constructing a coupling dynamics prediction model; s6, establishing a dynamic threshold evaluation and multi-dimensional self-calibration mechanism; 0.05 mm micro-crack identification and 3mm depth detection precision are achieved and are improved by 50% compared with a traditional method, the detection accuracy under the complex environment is larger than or equal to 98% through the multi-modal data fusion and GAN enhancement technology, the RMSE is smaller than or equal to 0.03 mm through short-term prediction, the accuracy is larger than or equal to 85% through long-term prediction, double guarantees of a physical mechanism and data driving are established, and the detection accuracy is improved by more than or equal to 98% through the multi-modal data fusion and GAN enhancement technology. And the system has self-calibration, self-adaptive sampling and multi-modal fusion decision-making capabilities, so that the manual intervention cost is greatly reduced.

The invention relates to the technical field of aero-engines, and discloses an automatic design method for aero-engine turbine blades, which analyzes design requirements through a language large model, combines a retrieval enhancement technology and a turbine blade expert knowledge base, performs interdisciplinary coupling verification by utilizing a step-by-step reasoning template, and generates and optimizes part design schemes and parameters. And then calling parametric modeling and an AI simulation agent model to carry out multi-physics field prediction, comparing a result with a requirement, if the result meets the requirement, outputting design, otherwise, automatically iterating and adjusting parameters until the parameters reach the standard or reach the maximum number of iterations. According to the method, end-to-end intelligent closed-loop design from design requirements to design results can be realized, manual intervention is reduced, a design scheme meeting performance requirements can be quickly generated, the design efficiency can be improved, the design cost can be reduced, and the design quality can be improved.

The invention discloses a panoramic video frame insertion method based on a potential diffusion model. The method comprises the following steps: compressing an input image to a potential space by using a panoramic perception vector quantization variational auto-encoder to obtain potential features; constructing initial noise, and fusing the motion features extracted by the panoramic optical flow adapter to obtain condition information; performing iterative denoising operation on the potential features of the intermediate frame to obtain final potential features of the intermediate frame; and restoring the final potential features of the intermediate frame into a frame insertion image of a pixel space through a condition decoder. According to the panoramic video frame interpolation method provided by the invention, the potential diffusion model and the panoramic characteristic enhancement technology are combined, so that the perception quality of a frame interpolation result can be remarkably improved, details and complex textures of a pole region can be better reserved, and a high-quality time interpolation solution is provided for immersive panoramic video application.

The invention provides a cell culture medium aided design system and method based on a large oracle model, and belongs to the technical field of artificial intelligence. A plurality of candidate culture medium formula schemes meeting user requirements are generated by obtaining culture medium design parameters and requirements and designing structured cues and inputting the designed structured cues into a large language model; a substance matched with the structured cue word is retrieved by utilizing a retrieval enhancement technology, a replacement substance is recommended for a user to refer, the user selects and purchases a required experimental material according to own requirements, a candidate culture medium formula scheme is fed back and evaluated in an actual experiment, and the system performs iterative optimization based on the feedback and evaluation, so that the accuracy of the experiment is improved. And a more appropriate candidate culture medium formula scheme is provided. According to the cell culture medium aided design system and method, intelligence, scientization and individuation of culture medium formula scheme design are achieved, experiment cost is reduced, repeatability and stability of cell culture experiments are improved at the same time, and therefore wide application of culture medium design in the field of biological manufacturing is promoted.

The invention relates to the technical field of data acquisition, in particular to an AOP-based user behavior data acquisition method. The method comprises the following steps: firstly, identifying user behavior related code features by statically analyzing Java source codes of an application program, and constructing a feature template library; during a compiling period, according to a template library identification target method, embedding a point burying instruction by utilizing a byte code enhancement technology, and constructing a monitoring section framework; and during running, dynamically adjusting an execution strategy according to the running state of the application program by virtue of a dynamic rule engine. Meanwhile, an annular buffer area is adopted to asynchronously process user behavior data streams, cache data are managed in a sub-generation mode, and asynchronous processing is triggered when a threshold value is exceeded. The method also constructs an adaptive thread pool model based on a hierarchical thread pool architecture, distributes tasks according to data value weights, monitors the task backlog rate of a high-priority sub-pool, adjusts the sampling frequency or acquisition dimension when the task backlog rate exceeds a threshold value, forms adaptive closed-loop control, and effectively acquires and processes user behavior data.

The invention discloses a patch position correction method and system, relates to the technical field of biomedical engineering and digital health, and adopts a hardware-level time synchronization and motion-artifact coupling modeling technology to effectively solve the problems of time delay drift and motion interference of signal acquisition in a dynamic environment and remarkably improve the basic quality of electrocardiosignals. Secondly, by establishing a body surface reference coordinate system and a pose compensation parameter generation algorithm, the micro-displacement state of the patch can be accurately recognized, and visual position adjustment guidance is generated, so that the patch is always kept at the optimal measurement position, and the phenomenon of signal attenuation caused by poor contact is fundamentally reduced; a signal quality multi-dimensional evaluation system adapts to quality discrimination requirements in different motion states through an intelligent weighted fusion mechanism, and combines an adaptive filtering and deep learning enhancement technology, so that the morphological integrity of a QRS waveform is kept, various interference components are effectively inhibited, and the system can still keep excellent electrocardiosignals in a strenuous motion state.

The invention relates to the technical field of speech enhancement, in particular to an audio optimization method and system applied to speech recognition, and the method comprises the steps: collecting audio data in real time, and uniformly dividing the audio data into audio frames; for each audio frame and each neighbor frame of the preset audio frame, evaluating whether a noise spectrum is re-selected when the audio data is enhanced by adopting a spectral subtraction method; if reselection is carried out, each mode of the audio frame is acquired; selecting a noise mode from the modes of the audio frame, and obtaining the number of lagging frames of other modes; obtaining a frequency spectrum change characteristic value and a voice characteristic value of each mode, and selecting a main voice mode; acquiring audio mode characteristic values of other modes except the main voice mode; and a noise frame and a new noise spectrum are obtained. The invention aims to improve the accuracy of voice recognition by improving the accuracy of noise frame selection and enhancing the audio enhancement effect of the spectral subtraction for the voice features of the audio frames.

The invention provides a vulnerability repair method based on a dynamic denoising and hybrid enhancement technology, which comprises the following steps: S1, performing dynamic denoising and hybrid enhancement on an original vulnerability repair model to obtain an optimized vulnerability repair model; s11, deriving approximate distribution of clean data and noise data from the training loss of the original vulnerability repair model by using an expectation maximization algorithm; s12, calculating the weight of the clean data and the weight of the noise data by using a confidence coefficient function of distribution perception, and re-weighting the approximate distribution by using the weight of the clean data and the weight of the noise data; s13, generating a mixed sample by using the interpolation coefficient and reconstructing a loss function to obtain a total loss value; s14, repeatedly training the vulnerability repair model by using the value of the total loss until the value of the current total loss is not reduced in continuous n rounds or reaches the total number of rounds of training, and outputting the optimized vulnerability repair model; and S2, performing vulnerability repair by using the optimized vulnerability repair model. The software development efficiency and quality are improved, and software vulnerabilities are reduced.

The invention belongs to the technical field of robot and human voice interaction, and particularly relates to a human voice interaction method and system. The method comprises the following steps: step 1, an intelligent agent enables radio duration to realize self-adaption according to environment loudness; 2, designing a multi-agent framework with a large model and a small model cooperating with each other; and step 3, based on the framework designed in the step 2, analyzing the radio in the step 1, and using a retrieval enhancement technology based on a tree-shaped document to improve the question and answer reliability of the intelligent agent and realize the voice interaction of the intelligent agent. The method is used for solving the problems that in the prior art, most voice interaction technologies can only achieve single-round interaction and cannot remember context content of interaction, and pure human voice is difficult to obtain when the traditional voice interaction technology is directly transplanted to a robot which can move and makes noise by itself.

The invention discloses an AI service intelligent fault tolerance and degradation method, device and equipment based on a Java bytecode enhancement technology, a medium and a product, and relates to the technical field of AI service integration. According to the method, in a class loading period of Java application starting, a loaded class is scanned to identify target methods marked with declarative fault-tolerant annotations, a bytecode operation framework is utilized to enhance the target methods so as to implant a unified fault-tolerant logic interception entry, and then in a running period, when the target methods are called, the target methods are subjected to fault-tolerant annotation processing. And triggering a fault-tolerant interceptor through the interception entrance to analyze the annotation to obtain a fault-tolerant strategy, allocating an independent fault-tolerant component instance to the fault-tolerant interceptor based on the unique identifier of the method, constructing and executing an enhanced call chain containing a fusing check and retry mechanism, and finally, if the execution of the call chain fails, executing the fault-tolerant interceptor. And if yes, a preset degradation method in the fault-tolerant strategy is automatically executed, so that complete non-intrusive decoupling of AI service fault-tolerant management and service logic can be realized, and the reliability of the system is remarkably improved.

Methods for reticle enhancement technology (RET) for use with variable shaped beam (VSB) lithography include inputting a substrate design, wherein a low-pass filter is applied to the substrate design to form a target wafer pattern. An initial mask pattern is determined from the target wafer pattern. An initial set of VSB shots is determined based on the initial mask pattern. A substrate pattern is calculated from a simulated mask pattern calculated with the initial set of VSB shots. The target wafer pattern is compared with the substrate pattern, and the initial set of VSB shots is adjusted until the substrate pattern and the target wafer pattern are within a predetermined tolerance. The adjusting of the initial set of VSB shots creates an adjusted set of VSB shots.

The invention discloses a crop planting scheme generation method and device based on a large language model, equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the natural language understanding of to-be-processed agricultural information, obtaining language understanding information, and converting the language understanding information into an information vector; constructing an agricultural knowledge graph based on the to-be-processed agricultural information; fusing the information vector and the agricultural knowledge graph to obtain a fusion result, and determining a retrieval result which corresponds to the agricultural knowledge graph and comprises a target entity and entity relationship path in the agricultural knowledge graph by utilizing a knowledge retrieval enhancement technology and based on the fusion result; determining a target context based on the retrieval result and an initial context corresponding to the large language model, and generating a crop planting scheme for the target context by using the large language model to obtain a target crop planting scheme; the target crop planting scheme comprises the explanation information corresponding to the planting suggestions, so that the efficiency of generating the crop planting scheme is improved.

The invention belongs to the technical field of RAG memory enhancement, and particularly relates to a self-adaptive RAG memory enhancement method, which comprises the following steps of: 1, determining an application scene and task requirements; 2, a backbone LLM is initialized and configured; step 3, attribute mining and identification; step 4, attribute labeling granularity selection and labeling; step 5, evaluating and sorting attribute priorities; step 6, constructing memory instance embedding representation; step 7, implementing a retrieval strategy based on attributes and embedding; step 8, integrating and optimizing retrieval results; and step 9, performing experimental evaluation and continuous improvement, namely determining the specific application scene of the self-adaptive RAG memory enhancement technology, retrieving appropriate text content for recommendation according to the theme and requirements input by the user, and determining the specific requirements of the task according to different scenes. According to the method, memory information is enriched by deeply mining and marking key attributes, and meanwhile, two efficient retrieval means are adopted to ensure quick positioning of related information.

The application belongs to the field of chemical mass and heat transfer enhancement technology, and discloses a macro-quantity structured micro-scale vortex packing structure, method and application. The packing structure is a packing with a multi-scale microstructure on the surface. The microstructure causes boundary layer separation and vortex shedding when fluid flows through, generates stable micro-scale vortexes at the microstructure, and forms a macro-distributed vortex group in the packing. The vortex can cut the boundary layer, stretch and fold the fluid micro-element, and promote the periodic update of the interface, thereby improving the gas-liquid mass and heat transfer efficiency. The application significantly reduces the pressure drop. Under the same operating conditions, the pressure drop of the packing structure of the application is reduced by more than 50% compared with the conventional packing, and the pressure drop increment caused by the vortex structure is less than 15%, realizing the synergistic optimization of mass transfer enhancement and low pressure drop.

The invention relates to a layout generation method and device, equipment, a storage medium and a program product. The layout generation method comprises the following steps: acquiring a target graph and target layout precision; determining a numerical value effective standard according to the target layout precision; and based on the numerical value effective standard and the geometric information of the target graph, generating a periodic arrangement graph of the target graph, and obtaining a target layout. According to the method and the device, the periodic arrangement graph with the target layout precision can be generated, so that advanced resolution enhancement technologies such as reverse photoetching technology correction and the like can be conveniently implemented on the full-chip layout.

The present invention discloses a method for constructing an EEG data learning model for de-individualizing differences. EEG data with corresponding calibration values ​​are obtained through preprocessing. Data enhancement technology is used to enhance the calibrated EEG data to prevent distortion while increasing the data volume, facilitating subsequent operations and improving the accuracy of subsequent model operations. A loss function is constructed to obtain a loss baseline value in the calibration process. After obtaining the loss gain coefficient, the EEG data is adjusted. The EEG data is continuously cleaned and adjusted according to the model to present differential integration. After the data is proportionally unified, a data learning model is constructed based on the data after differential integration, thereby removing the differences between the original data and being able to uniformly incorporate the EEG data, so that the error between the output result of the final data processing model and the actual output result is controlled within a small range.

The present invention discloses a testing method and a testing device, relating to the field of computer technology. A specific embodiment of the method includes: receiving a test request, obtaining a to-be-tested application and test cases according to the test request, and adding a Java Agent to the virtual machine parameters corresponding to the to-be-tested application; before loading the class files of the to-be-tested application, modifying the bytecode of the target classes of the to-be-tested application according to the Java Agent and based on the bytecode enhancement technology, wherein the target classes include: classes for remote calls and classes involved in persistent data; during the process of testing the to-be-tested application using the test cases, calling the target methods included in the target classes, and obtaining the running data corresponding to the target methods; determining the test data corresponding to the test cases according to the obtained running data, and storing the test data. This embodiment can conduct tests deep into the program, improve the testing efficiency, have a low testing cost, and is also applicable to the testing of complex scenarios.

This invention discloses a method, apparatus, device, and medium for implementing an intelligent quality valve in the requirement submission process, relating to the field of requirement process quality review in fintech. The method includes: acquiring a requirement description and uploading it to a created large-scale intelligent platform; expanding the document based on comparison with existing similar requirements and performing standardization processing to fill in missing elements, generating a requirement document; acquiring voice data from requirement review meetings and uploading it to the large-scale intelligent platform; analyzing the voice data to generate a requirement review meeting summary and submitting the summary to a DevOps platform; creating a requirement review model, which uses retrieval enhancement technology to acquire private domain data in real time to review the requirement document currently awaiting review. This application enables automated review operations, significantly reducing the burden of manual review and improving the accuracy and efficiency of the review process.

The invention relates to the technical field of PCB (Printed Circuit Board) detection, in particular to a method and a system for detecting the edge deviation of the appearance of a PCB based on a welding component. The polarization characteristic pattern is generated by acquiring the multi-spectral original image and the polarization state parameter of the to-be-detected area of the PCB, the reflection exposure abnormal area and the reflection exposure normal area can be effectively positioned, the fusion weight is automatically adjusted according to the spectral response characteristic of the signal area through the self-adaptive spectrum fusion enhancement technology, and the detection accuracy is improved. The problem that in a traditional method, image enhancement is not fine enough is solved, the adaptive fusion enhancement technology enables the multispectral image to better conform to the actual situation during enhancement, more detail information is reserved, key features can be accurately extracted by generating a semantic segmentation map and combining three-dimensional standard coordinate information, and the image enhancement accuracy is improved. The detection precision and reliability of the welding component and the appearance edge of the PCB are effectively improved, the deviation can be accurately positioned and quantified, and the quality control level in the production process of the PCB is improved.

The invention provides a submarine cable defect detection method based on time-space texture cooperative enhancement, and aims to solve the problems of insufficient defect feature extraction in a low signal-to-noise ratio environment, high false detection rate caused by dynamic background interference and lack of interpretable basis for detection results in the existing submarine cable detection technology. In order to solve the problems, a space-time texture collaborative enhancement mechanism is introduced, and defect detection in a complex environment is realized through multi-modal feature fusion. Specifically, in order to solve the problems of dynamic background interference and weak target features in submarine cable detection, the invention provides a space-time difference enhancement technology, designs a composite texture analysis architecture, extracts linear structure features and local statistical distribution features of a cable, and constructs a dual verification mechanism of defect characterization; a dynamic feature weight distribution strategy is further provided, and the fusion proportion of the spatial-temporal features and the texture features is adjusted in a self-adaptive mode according to the environment complexity. Meanwhile, the method provides reliable technical support for underwater operation and maintenance decisions.

The present application provides a method and apparatus, storage medium, and equipment for training a speech enhancement model, relating to the field of speech enhancement technology. The method comprises: obtaining N groups of training samples, wherein the i-th group of training samples comprises: i-th training data and i-th target data; training a speech enhancement model using the N groups of training samples; obtaining the i-th group of training samples comprises: obtaining the i-th room impulse response and i-th clean speech data, processing the i-th room impulse response and the i-th clean speech data to obtain the i-th training data in the i-th group of training samples; determining the i-th control curve based on the i-th room impulse response, multiplying the i-th control curve with the i-th room impulse response to obtain the i'th room impulse response; and convolving the i-th clean speech data with the i'th room impulse response to obtain the i-th target data in the i-th group of training samples. The present application can reduce distortion after signal processing and solve the problem of aligning training data and target data.

The present disclosure relates to the field of microneedle transdermal enhancement technology, in particular to a micro-projection massage head and a massager, wherein the micro-projection massage head includes a massage nanochip, the micro-projections are provided on the massage nanochip, and the height of the micro-projection does not exceed 60 micrometers; the massager includes a vibrating mechanism and a micro-projection massage head; during operation, the vibration mechanism can drive the micro-projection massage head to vibrate, and the working vibration frequency of the micro-projection massage head is not less than 1000 times per minute. According to the present disclosure, the technical problems such as discomfort and easily causing damage to subcutaneous tissue during the application of the existing microneedle transdermal enhancement technology can be solved, and the improvement effect of the effective ingredients on the skin can be revealed in time.

The application relates to the technical field of aero-engines and discloses an automatic design method for a turbine blade of an aero-engine, which comprises the following steps: analyzing design requirements through a language large model, combining retrieval enhancement technology with a turbine blade expert knowledge base, utilizing step-by-step reasoning templates to perform cross-disciplinary coupling verification, generating and optimizing a component design scheme and parameters, subsequently calling a parameterized modeling and an AI simulation agent model to perform multi-physical field prediction, comparing the result with requirements, outputting a design if the requirements are met, and otherwise automatically iteratively adjusting parameters until the requirements are met or the maximum number of iterations is reached. The application can realize end-to-end intelligent closed-loop design from design requirements to design results, reduce manual intervention, quickly generate a design scheme meeting performance requirements, improve design efficiency and reduce design cost, and improve design quality.

This invention discloses a reinforcement learning-based underwater image enhancement training optimization method, belonging to the field of underwater image enhancement technology. Addressing the problems existing in current technologies, this method treats low-frequency color correction and high-frequency texture reconstruction in underwater image enhancement as two enhancement objectives with different optimization focuses. It utilizes a reinforcement learning mechanism to dynamically adjust the training strategies for both, solving the problem that existing fixed training strategies are difficult to adapt to changes in training states. A frequency-domain perceptual state representation is constructed by combining low-pass filtering and residual decomposition to explicitly describe the relative optimization state between low-frequency color restoration and high-frequency texture restoration during the current training process, providing a decision-making basis for the policy controller. An adaptive training optimization mechanism based on frequency-domain state feedback is built, automatically completing policy scheduling according to the actual restoration state during training, reducing the reliance on manually designed training strategies and improving the adaptability, generalization, and scalability of the method in complex underwater degradation scenarios.

This invention provides a turnout design method and system, relating to the field of key characteristic enhancement technology for urban rail transit turnouts. The method includes: obtaining the target train speed, existing turnout dimensions, turnout frog angle, and the range of values ​​for the turnout guide curve radius; calculating the design target speed based on the target speed; back-calculating the lateral speed based on the existing turnout dimensions and iteratively lengthening and adjusting to obtain a first parameter; optimizing the radius based on the range of values ​​for the turnout guide curve radius to obtain the optimal guide curve radius; determining the half-section and switch rail impact angle based on the optimal guide curve radius, the turnout front length, and the turnout rear length to obtain the optimal parameter combination; and verifying the optimal parameter combination to obtain the final turnout alignment design scheme. The method and system described in this invention improve the lateral throughput speed of turnouts, meet the turnout speed requirements of signaling professionals, and enhance the turnout performance of existing turnouts.

Methods for reticle enhancement technology include inputting a target wafer pattern, the target wafer pattern spanning an entire design area, and iterating a proposed mask for the entire design area until the proposed mask meets criteria towards producing the target wafer pattern. Each iteration includes calculating a predicted wafer pattern from the proposed mask. The calculating comprises calculating a cost and derivative data, the cost and the derivative data being based on comparing the predicted wafer pattern to the target wafer pattern. The cost further comprises specifications for mask manufacturability.

This invention discloses a device for enhancing the stability of a power line joint, relating to the field of power line joint enhancement technology. It includes a base, a connecting rod, and a terminal block. The connecting rod, measuring 200mm x 26mm x 10mm, is located on one side of the base. The terminal block is located on the side of the connecting rod away from the base. The device is gently slipped onto the cable, and then the cable core is carefully inserted into the cable tray, ensuring complete alignment for optimal electrical connection. Next, the terminal block is accurately placed above the core, and the M8 nut and M8 locking bolt are tightened to securely hold the core in place, preventing loosening or detachment. The device limits significant cable sway at the joint, reducing stress concentration and preventing breakage, strand breakage, and unraveling due to stress concentration.

This application discloses an automated detection method for pin defects based on 3D imaging and deep learning, including the following steps: (1) performing structured light 3D adaptive anti-overexposure imaging on the pin to obtain effective point cloud at the micron level of the pin tip; (2) performing local repair on the pin tip defect cloud obtained in step (1), calculating the normal vector and curvature features of the repaired point cloud, fusing 3D point cloud features and 2D texture features to obtain multimodal fusion features of the pin tip; (3) inputting the multimodal fusion features obtained in step (2) into a lightweight twin feature matching network based on few-sample deep learning, combining meta-learning transfer enhancement technology to complete the identification and classification of pin tip micro-breakage and micro-bending defects, and outputting defect location and defect degree parameters; (4) based on normal vector constraints and pseudo-defect feature comparison, performing double filtering on the suspected defects detected in step (3), removing pseudo-defects to obtain the real defect detection results.

The invention relates to the technical field of construction of detection models, in particular to a method for constructing a light-weight small target detection model, which comprises the following steps of: constructing an HS decoupling head structure by using a double-branch collaborative mechanism aiming at a detection head part of the light-weight small target detection model, and extracting a deep channel attention mechanism and multi-scale texture features to obtain a light-weight small target detection model; optimizing spatial precision according to a prediction result, and determining a lightweight small target detection model based on a depth lightweight reconstruction mechanism; according to the invention, by adopting the HS decoupling head technology based on the double-branch cooperation mechanism, the technical effect of significantly improving the small target detection precision is realized; by adopting a depth lightweight trunk reconstruction technology based on MobileNetV3, the technical effect of maintaining strong feature extraction capability under extremely low resource consumption is realized; by adopting the neck network depth separable convolution comprehensive replacement and GSConv enhancement technology, the technical effects of greatly reducing the model calculation complexity and keeping the feature fusion capability are realized.

Systems for reticle enhancement technology (RET) for use with variable shaped beam (VSB) lithography include a device configured to determine an initial mask pattern from a desired pattern for a substrate; a device configured to calculate a first substrate pattern from the initial mask pattern; a device configured to determine an initial set of VSB shots based on the initial mask pattern; a device configured to calculate a second substrate pattern from a simulated mask pattern calculated with the initial set of VSB shots; a device configured to compare the first substrate pattern with the second substrate pattern; and a device configured to adjust the initial set of VSB shots until the second substrate pattern and the first substrate pattern are within a predetermined tolerance, creating an adjusted set of VSB shots.