51results about How to "Ensure robustness" patented technology

The invention discloses a hyperspectral image classification method based on frequency domain denoising and element gradient correction, and the method comprises the following steps: carrying out the preprocessing of all hyperspectral image data, and dividing an overall training sample set formed by the processed hyperspectral images into a training set and a verification set; constructing a sample weighting model based on frequency domain denoising and element gradient correction; in the training process, a parameterization frequency spectrum gating sensing transformation module is utilized to map features to a frequency domain through discrete Fourier transform, a learnable frequency spectrum response function is utilized to adaptively suppress spectrum jitter noise, and finally pure features are reconstructed. And automatically constructing a high-confidence pseudo-clean verification set based on a Gaussian mixture model and time domain consistency. According to the method, a time domain momentum updating mechanism is introduced, the variance of statistical estimation is effectively smoothed, random interference caused by training fluctuation is resisted, and the accuracy of pseudo clean set construction and the convergence stability of overall model training are further improved.

The application provides a campus second-hand commodity transaction credible traceability evaluation method and system, relates to the electric digital data processing technical field, and fundamentally improves the accuracy and foresight of the credibility evaluation through the closed-loop linkage control system constructed from multi-dimensional feature deep quantification to individual-group correlation comparison, and then to adaptive strategy dynamic decision-making; The beneficial effects brought about include: 1) the initial trust evaluation ability of new users without historical data is enhanced; 2) the identification accuracy of hidden fraud behaviors with "mode anomaly" characteristics using high credit identity as a cover is improved; 3) the dynamic self-adaptation of risk intervention strategy is realized, which can intelligently adjust the control strength according to the overall risk situation of the platform, and achieves a better balance between ensuring platform safety and optimizing user experience.

The invention relates to a tread pattern depth estimation method based on a physical information fusion network, and belongs to the technical field of intelligent traffic systems and vehicle active safety. The method comprises the following steps: firstly, acquiring three-axis acceleration data under different working conditions through a tire built-in sensor; zero phase shift filtering, phase alignment and resampling preprocessing are carried out on the signals to form standard waveform data; then constructing a spatial-temporal feature extraction network, extracting time sequence features through a convolutional layer and a circulating layer, and fusing normalized load and speed working condition information; a physical constraint module based on a tire rigidity mechanism is introduced, the pattern depth and the vertical deformation amount are predicted by using learnable parameters, and a theoretical load is reversely deduced to establish a mechanical equilibrium constraint; and finally, network parameters and physical parameters are synchronously optimized by adopting a mixed loss function, and wear state grades and confidence coefficients are output through a probability statistical method. According to the method, the estimation precision and the physical interpretability of the model under the complex variable load working condition are effectively improved.

The invention relates to the technical field of light field regulation and control, and discloses a static phase distortion compensation system of a spatial light modulator, which comprises a polarization splitting prism arranged in the light emitting direction of a laser and used for adjusting a light beam emitted by the laser into linearly polarized light in the horizontal direction, the first quarter-wave plate adjusts the linearly polarized light into circularly polarized light and irradiates the circularly polarized light to the spatial light modulator; a second quarter-wave plate arranged in the light emitting direction of the spatial light modulator adjusts an emergent light beam of the spatial light modulator into linearly polarized light, and a linear polarizer performs polarization analysis on the emergent light beam of the second quarter-wave plate; the static phase error of the spatial light modulator is determined through the polarization direction included angle of the incident beam of the first quarter-wave plate and the emergent beam of the second quarter-wave plate, and the polarization direction included angle is determined through the Malus law relation of the incident light intensity and the transmission light intensity of the linear polarizer. The static phase distortion of the spatial light modulator is compensated by using the phase error, and the static phase distortion compensation cost is reduced.

This invention belongs to the field of computer vision technology and discloses a method and system for video depth understanding and generation based on a constraint-guided model. The method includes: acquiring original reference images and extracting image sequences; extracting original feature information based on the image sequences, receiving user-inputted custom constraints, and correcting the original feature information based on the custom constraints; extracting target images from a preset database based on the corrected original feature information and the custom constraints; generating a preset number of candidate videos based on the target images, and determining whether to output the current candidate video as the final generated video; if not outputting the current candidate video as the final generated video, generating guidance information. This invention overcomes the dilemma of existing technologies, achieving a precise balance between controllability and innovation. Furthermore, this invention optimizes generation efficiency and quality while reducing computational costs.

The invention provides a robot navigation method and device based on edge and cloud cooperative computing. The method comprises the following steps: acquiring real-time environment sensing data through a body sensor of the robot; executing real-time positioning and map construction, and generating a local obstacle avoidance instruction; generating a task unloading decision according to the task division strategy, the current network state and the server resource state; and fusing the optimized global map, the long-term trajectory prediction result and the local obstacle avoidance instruction to generate a global comprehensive navigation strategy of the robot. According to the method, robot navigation tasks are intelligently divided and dynamically unloaded through an innovative edge and cloud cooperative computing framework, the low delay of the edge side and the powerful computing power of the cloud are fully utilized, and the core bottleneck that the computing power of a robot body is limited can be effectively solved. The overall delay of the navigation system is obviously reduced, and the navigation efficiency, safety and intelligent level in a large complex scene are improved.

The utility model relates to the technical field of forearm positioning tools, in particular to a forearm neutral position fixing device which comprises a supporting part, a holding assembly is arranged at one end of the supporting part in a sliding mode, a bandage is connected to the top of the supporting part, the supporting part comprises a bottom plate, arc-shaped plates are installed on the two sides of the bottom plate, and an inserting hole is formed in one end of the bottom plate. The holding assembly comprises a sliding plate, the sliding plate is matched with the inserting hole in an inserted mode and locked and fixed through a locking bolt, and a handheld part is installed on the top of the sliding plate. According to the forearm neutral position fixing device, due to the sliding fit design of the supporting component and the holding assembly, a user can flexibly adjust the holding position according to self requirements, stable fixing is achieved through the locking bolt, and it is ensured that the forearm is always kept in the neutral position in the moving process. In addition, the bottom plate ensures that the device is firm and durable; and the pad effectively relieves skin discomfort and pressure pain caused by long-time wearing.

The application discloses an interactive form design method and system based on a language large model, and belongs to the technical field of human-computer interaction and form automatic generation. In view of the problems that a traditional form design tool is tedious to operate, highly depends on manual experience, and an existing method based on a large model has high training cost, weak interaction ability and difficulty in real-time collaboration, the following scheme is used for improvement: defining a JSON Schema specification of a form and constructing a visual form designer, integrating a language large model interaction entrance in an interface, dynamically assembling a request containing user instructions, a JSON Schema specification and existing form data according to whether the form is empty, and submitting the request to the language large model; receiving a response in a streaming manner, and real-time analyzing and progressively updating a form designer preview interface, so that a user can directly generate or modify a form through natural language instructions, and multi-round interaction and manual subsequent adjustment are supported. The method is mainly used for reducing a form design threshold, improving design efficiency and intelligent collaboration level.

This invention discloses a method for diagnosing self-discharge anomalies in energy storage battery packs based on short-term charging data, belonging to the field of battery safety monitoring technology. Addressing the shortcomings of existing methods that rely on complete charge-discharge cycles and are difficult to adapt to multi-stage varying operating conditions, this invention extracts features from multi-stage charging data, combines an autoencoder to achieve adaptive extraction of features under multiple operating conditions, and constructs a data-driven model fused with GCN-BiLSTM (Graph Convolutional Network-Bidirectional Long Short-Term Memory) to accurately estimate the RCC (Remaining Charging Capacity) of individual battery cells. Finally, based on the size, distribution, and changes of the RCC of each individual cell in the battery pack, we can diagnose overall battery pack inconsistencies, individual cell SOC inconsistencies, and self-discharge faults. This provides a guarantee for the safe operation of energy storage systems.

The present application relates to the technical field of robots, and in particular to a motion generation method, system, device and storage medium for a humanoid robot, comprising: physical simulation rendering generates a visual video containing visual representation of joint motion state of the robot; analyzing the visual video to obtain frame images, natural language instructions and motion sequences, decoding output trajectory data; and outputting control instructions through a hierarchical architecture of high-level motion planning and bottom-level dynamics tracking to drive the robot to perform actions. The present application implicitly encodes the joint state of the humanoid robot in the visual features, improves the inference efficiency and generalization ability of the model, and the hierarchical decoupling architecture of the large model high-level trajectory planning and the reinforcement learning bottom-level dynamics tracking controller pays a small performance cost, which can compress the inference time delay to within 50ms and the parameter quantity to below 2 billion, and balances the high generalization of the stateless VLA and the small model, few-step inference, edge low latency, and low power requirement.

The invention relates to the technical field of oil quality analysis, in particular to an aviation oil property analysis method and system based on spectral oblique projection. According to the method, when the oblique projection operator is constructed, the limitation of a traditional oblique projection operator is broken through through dynamic multi-component cooperation and self-adaptive residual error separation, key information which is in complex association with properties in the aviation oil spectrum is extracted more accurately, the pertinence and robustness of feature extraction are improved, and the accuracy of feature extraction is improved. Higher-quality input variables are provided for an aviation oil property prediction model, and the method is a core technical support for realizing high-precision aviation oil property analysis.

The application discloses a real-time fine mapping method suitable for an underwater robot, which comprises the following steps: firstly, a 3D Gaussian map is constructed by using a first frame; then, key frame determination is performed on the following image frames; and finally, an incremental map is constructed by using the key frames; the process effectively overcomes the optical interference specific to the underwater environment, filters out the geometric noise caused by suspended particulate matters and the like, and significantly improves the geometric consistency and color authenticity of the map; all frames are tracked in real time, accurate poses are obtained and output, the process effectively overcomes the interference of underwater optical degradation on pose estimation, and ensures the robustness and accuracy of the tracking process.

The invention relates to the technical field of financial risk control, in particular to a financial transaction abnormal data real-time monitoring method and system based on deep learning. Comprising the steps of collecting external macroscopic variables in real time; identifying the macroscopic state of the current market; generating a macroscopic state control signal; transmitting the macroscopic state control signal to a preset dynamic baseline management unit; retrieving a situation baseline model matched with the macroscopic state from a preset situation baseline model library; calling a situation baseline parameter corresponding to the situation baseline model; coding the financial transaction flow data by using a deep neural network; generating a microcosmic behavior feature vector; calculating a semantic distance between the microscopic behavior feature vector and a normal feature space; comparing the semantic distance with a preset discrimination threshold value; and generating an abnormal risk alarm in response to the semantic distance being greater than the discrimination threshold. According to the invention, the false alarm rate when the market fluctuates violently and the missing report rate under the stable market are obviously reduced, and the monitoring accuracy is improved.

The invention is suitable for the technical field of communication, provides a serial port communication data processing method and a related device, and ensures the accuracy of receiving and executing a control command of serial port communication in an electromagnetic interference environment. The method mainly comprises the following steps: when a first module receives a data stream of a control command sent by a second module, searching a frame header of a data frame from the data stream; when the frame header of the data frame is found, frame information behind the frame header in the data stream is received, and the frame information comprises a frame command, a frame data length, effective data, a checksum value and a frame tail of the data frame; when it is determined that the number of data bytes after the frame command is equal to the frame data length, receiving of the data frame is ended, and a target data frame is obtained; when the frame tail of the target data frame is verified to be correct, performing checksum calculation on data from a frame header to valid data in the target data frame to obtain a target checksum value; and only when the target checksum value is matched with the checksum value, the frame command is executed according to the requirement of the valid data.

The application relates to the field of automatic control of heating, ventilation and air conditioning, and discloses a control method for improving the cooperative operation of an air conditioning system. The method first constructs a virtual flow monitoring model of terminal equipment and heat exchange efficiency gradient calculation logic; secondly, a flow velocity estimation model is established based on the principle of pipe network fluid dynamics to calculate the cold energy transport lag time required for the temperature change generated on the cold source side to be transferred to each thermal saturation terminal equipment; finally, a source network load cooperative regulation strategy is executed to lock the upper limit of the opening degree of the terminal regulating valve and simultaneously reduce the pressure difference set value of the supply and return water main on the transmission and distribution side. Through the time and space synchronous regulation strategy, the application effectively solves the large flow and small temperature difference problem of the air conditioning water system under the premise of guaranteeing the comfort of users, and improves the comprehensive operation energy efficiency of the system.

The invention provides an unplanned return ICU risk prediction method, system and device, and the method comprises the steps: fusing an MIMIC-IV public data set and local private data, and carrying out the standardization preprocessing of multi-source medical data through a rule module and a clinical knowledge base; constructing a partitioned feature system comprising basic features, laboratory indexes, hemodynamic parameters and respiratory metabolism dimensions, and optimizing feature expression by adopting a sliding window and a feature interaction method; performing medical field fine tuning based on a large language model, and converting the structured features into natural language description in combination with cue words; carrying out model training by adopting optimization strategies such as label smoothing and gradient cutting, and evaluating model performance in a mode of combining internal verification and external verification; and realizing interpretability analysis of model prediction by using an SHAP framework. According to the scheme, the accuracy and generalization ability of ICU return risk prediction and the applicability of an actual scene are remarkably improved.

The application discloses a kind of embroidery textile production quality detection methods based on data analysis, it is related to intelligent textile technical field, including, real-time acquisition embroidery production line multi-source sensor data, time stamp synchronization and normalization processing are carried out through Internet of Things edge node, output structured process flow data;Structured process flow data is dynamically constructed process causal diagram model based on it, embroider line batch quality, needle speed parameter and environmental temperature and humidity parameter are as node, bayesian conditional probability is as edge weight, the causal effect value of needle speed parameter to broken needle defect is calculated through do-calculus;According to dynamic parameter adjustment instruction, needle speed parameter and embroider line tension value are real-time regulated, and simultaneously, through fuse signal, fault embroidery machine table coordinate is synchronized to operation and maintenance terminal.The application is accurately quantified the causal influence intensity of needle speed parameter and environmental temperature and humidity interaction through bayesian conditional probability and do-calculus framework.

The invention discloses an intelligent garage management method and system for collaborative dynamic guidance, and aims to improve the parking resource utilization rate and the guidance efficiency. According to the method, a parking request containing user preferences and vehicle attributes is received, and a target parking lot is recommended and reserved. After the vehicle enters the parking lot, the system collects real-time data through the Internet of Things, and generates dynamic situation information in the parking lot by combining historical data and utilizing a space-time fusion prediction model. Based on this, the system executes multi-target dynamic optimal parking space distribution, and plans a driving path with the lowest comprehensive passing cost. In the guiding process, the system carries out continuous monitoring, and parking space and path re-planning is triggered immediately in case of congestion or parking space occupation and other abnormalities. The corresponding system comprises a request processing module, a parking lot sensing module, a data fusion and situation prediction module, a dynamic parking space distribution module, a real-time path planning module, a state monitoring and re-planning module, a billing updating module and the like, automatic and intelligent management from reservation, guiding to parking lot leaving billing is achieved through the cloud, and the parking efficiency and the user experience are remarkably improved.

The utility model discloses a connector for a rod tip on a fishing line set. The connector aims to solve the problems that a traditional fishing line is easy to wind and not attractive due to the fact that a knot is knotted and exposed. The connecting machine comprises an integrally-formed L-shaped main body, the main body is composed of a fishing line knotting containing part and a buckle containing part, and through holes are formed in the two containing parts in a penetrating mode; the fishing line knotting containing part is in a big-end-up cone shape, a top end opening is larger than a bottom opening, and a cavity used for completely hiding a fishing line knot is formed. The buckle containing part is used for containing a cylindrical buckle with a penetrating through hole. In the use process, a fishing line knot is stuffed into the knotting containing part cavity to be hidden, a fishing line sequentially penetrates through the knotting containing part bottom through hole, the buckle through hole and the buckle containing part through hole, and finally the buckle is fixed in the buckle containing part to complete connection. According to the utility model, thread knots are effectively prevented from winding, and the aesthetic degree and connection stability are improved.

The invention relates to the technical field of wind power plant power control, in particular to a power control method and system based on wind power plant wake flow collaboration. Comprising the following steps: constructing a power cooperation model and a plurality of operation cycles; acquiring an environment monitoring packet of the current operation cycle according to a preset monitoring time node, and setting an operation control strategy according to the environment monitoring packet and the power cooperation model; obtaining an operation record packet according to a preset correction time node, and judging whether a correction instruction of the power collaboration model is generated or not according to the operation record packet; various operation wind fields are established according to historical environment parameters, all wind turbine generators in the wind power plant are aggregated according to operation characteristics of different wind fields, multiple groups of corresponding equipment domains are generated, wake flow disturbance is accurately analyzed, and robustness and continuity of a cooperative control effect are ensured by dynamically adjusting operation control parameters. The overall operation efficiency of the wind power plant is improved.

This invention relates to the field of combustion monitoring and fault diagnosis technology for thermal power boilers, and in particular to a method for condition monitoring and fault diagnosis of coal-fired boilers based on multimodal data fusion. The method includes the following steps: S1. Asynchronous data alignment; S2. Differentiated feature extraction; S3. Cross-modal collaborative attention interaction; S4. Feature adaptive integration; S5. Fault diagnosis. This invention breaks through the perception limits of a single sensor, significantly improving diagnostic accuracy; it innovates an asynchronous alignment mechanism for heterogeneous data, reducing hardware deployment costs; it possesses powerful capabilities for identifying unseen operating conditions, greatly enhancing system security; and it has high industrial applicability and multi-task expansion value.

The invention discloses a generated picture blind box detection method and system based on a diffusion model, and relates to the technical field of safety detection, and the method comprises the steps: taking a pre-training diffusion model as a reference model, constructing an ordinary differential equation integral pipeline, and generating a denoising velocity field; obtaining an identification curvature field through an approximate curvature calculation formula and a pseudo-Gaussian compression strategy; obtaining a diagonal high-frequency wavelet component by using first-order discrete wavelet transform; curvature features and diagonal high-frequency wavelet features are extracted by using a dual ResNet-50 model; and splicing to form a joint representation, inputting the joint representation into a multi-layer perceptron, and judging whether the picture is generated by a diffusion model or not according to output. According to the method, the pre-trained diffusion model is used as a reference model for generating the probability flow trajectory, the curvature features of the probability flow trajectory are used as main discrimination signals, and the high-frequency diagonal wavelet signals are assisted to jointly form joint representation for detecting abnormal information, so that a picture source or a model type does not need to be identified in advance; and blind box detection can be carried out on images generated by various diffusion models.

The invention relates to the technical field of autonomous planning navigation, in particular to an anti-interference multi-route autonomous planning navigation method in a polar region environment, which comprises the following steps: environment perception and data acquisition: collecting polar region environment data by using a millimeter wave radar, a laser radar, a meteorological sensor, an infrared night vision device and a high-definition camera. According to the method, a dynamically updated environment situation map is constructed through a multi-source data fusion technology, an accurate and reliable basis is provided for planning, a multi-route parallel generation and multi-target optimization mechanism is adopted, the diversity and robustness of a scheme set are ensured, the anti-interference capability of a decision-making system is enhanced, and the decision-making efficiency is improved. And a whole-course linkage mechanism from real-time interference evaluation to dynamic rolling optimization to closed-loop verification is also established, and self-adjustment and continuous optimization can be carried out according to environmental feedback, so that the autonomy, safety and adaptability of polar navigation are remarkably improved.

The invention relates to a target detection model hiding function embedding method and system based on output hiding, and the method comprises the steps: taking a conventional target detection model as a carrier, constructing a mixed data set on the premise of not changing the output dimension of the conventional target detection model, introducing a composite loss function comprising initial task loss and modulation loss, and carrying out the combined fine tuning. And hiding and embedding the predetermined target recognition result into the model output representation. And the embedded implicit information can be extracted through a decoding network initialized by a preset key, so that the hidden identification of the existence state of the predetermined target is realized. Compared with the prior art, the method has the advantages that original function precision of the carrier model is kept, non-inductive integration of special functions is achieved, and good concealment, effectiveness, safety and the like are achieved.

The specification discloses an image processing method, device, storage medium and electronic equipment, wherein the method comprises: inputting acquired living body sample images and attack sample images into an initial image detection model for model training, controlling the initial image detection model to adopt a first feature processing network to perform first feature processing on the living body sample images and a second feature processing network to perform second feature processing on the attack sample images, and adjusting model parameters of the initial image detection model based on asymmetric angle interval losses and network self-supervision losses corresponding to the first feature processing network and the second feature processing network, so as to obtain an image detection model corresponding to the initial image detection model.

The invention provides an adaptive SSMVEP-MI fusion brain-computer interface method and system based on eye movement tracking. The method comprises the following steps: determining a current effective fixation area of a user in real time through eye movement tracking; based on the region, only activating a corresponding local SSMVEP visual stimulation target, and generating associated specific motor imagery prompt information; electroencephalogram signals are synchronously collected, and SSMVEP features and MI features are extracted; self-adaptively selecting a decoding path according to whether the accumulated number of the MI effective samples with the labels reaches a threshold value or not: if not, independently decoding by using the SSMVEP features; if yes, fusion decoding based on signal quality evaluation is carried out on the bimodal features; and finally outputting a control signal to drive external equipment or update an interactive interface. According to the method, the visual interference is remarkably reduced, the recognition accuracy and the training efficiency at the initial stage of motor imagery are improved, and the robustness and the availability of the system are ensured through a dual adaptive mechanism of the sample number and the signal quality.

The application discloses a converter dual-mode switching control method based on online parameter identification performance evaluation, and specifically comprises the following steps: step 1, establishing a state space model of a two-phase interleaved parallel Boost converter; step 2, designing an extended Kalman observer; step 3, judging the failure condition of the extended Kalman observer and generating a quantized performance index; step 4, designing a model predictive control and using the extended Kalman observer to provide a real-time parameter dynamic updating model; step 5, designing a sliding mode controller; step 6, dynamically adjusting the output weight of the model predictive controller and the sliding mode controller by evaluating the performance of the extended Kalman observer in real time; and step 7, outputting a control signal to control the on-off of a switch tube and then controlling the output of the converter. The application solves the problem of performance decline of the model predictive control when the model parameter estimation is inaccurate or fails.

This application relates to a method, system, device, and medium for intelligent screening of clothing images. The method includes: acquiring a dataset of clothing images to be processed; performing basic quality filtering to obtain a screening set; performing image enhancement on the screened images to generate a standardized set; extracting the visual feature vector and quality score of each image; identifying the main body region based on the features using a clothing region segmentation model to generate a mask; combining the mask and features to extract semantic labels of design elements to form a structured design point set; matching the design point set with a pre-built clothing trend database to generate a trend matching score; simultaneously calculating a design innovation score based on its historical frequency of occurrence; dynamically weighting the quality score, trend score, and innovation score to generate a comprehensive score; and finally, performing a hierarchical decision based on the comprehensive score and quality score, outputting the images to be added to the database and their corresponding level labels. This application realizes intelligent evaluation and screening of clothing designs.

The invention provides a coarse and fine switching method and device in satellite top-crossing tracking and pointing, and the method comprises the steps: carrying out the state estimation and uncertainty quantification based on the instantaneous pointing error of a satellite top-crossing tracking and pointing system at a satellite top-crossing stage, generating and dynamically updating the two-dimensional probability density distribution describing the position of a light spot, and carrying out the dynamic updating of the two-dimensional probability density distribution; fusing with a light intensity distribution diagram influenced by atmospheric turbulence to obtain an effective receiving index, and determining a region of interest for fine aiming detection from probability distribution; performing spatio-temporal clustering analysis on multiple frames of detection data continuously collected by the fine aiming equipment, and calculating the coordinates of a fine aiming tracking point and the identification confidence of the fine aiming tracking point; and predicting the tracking interruption probability based on the atmospheric turbulence prediction data and the probability distribution, performing judgment in combination with the confidence coefficient, and if the judgment is passed, switching the control system to a fine aiming closed-loop tracking mode. According to the invention, rapid, reliable and anti-interference automatic switching from coarse aiming to fine aiming can be realized in a high dynamic scene in which a satellite rapidly passes through the top, so that rapid establishment and stable maintenance of a free space optical communication link are ensured.