1269 results about "Mixture model" patented technology

The invention discloses a mechanical equipment state monitoring method and system based on multiple sensors, and the method comprises the five core steps: multi-modal data collection and preprocessing, dynamic feature fusion, adaptive threshold diagnosis, digital twin fault tracing and predictive maintenance decision. All-domain coverage of equipment is realized through a three-layer sensor network architecture, the problems of data synchronization and interference resistance are solved by utilizing a temperature and vibration integrated sensor, deep fusion and anomaly detection of multi-source data are realized in combination with an attention mechanism, a Gaussian mixture model, a three-dimensional convolutional neural network and the like, and finally a precise maintenance strategy is generated through digital twinning and reinforcement learning. The multi-sensor-based mechanical equipment state monitoring system comprises a sensor network layer, an edge computing layer, a cloud platform layer and a man-machine interaction layer, supports federated learning to protect data privacy, improves real-time diagnosis capability through edge-cloud collaboration, and enhances a reality interface to realize intelligent operation and maintenance interaction.

The invention discloses a grading, zoning and differentiating risk early warning method for mountain torrent disasters in medium and small watershed. Firstly, a drainage basin risk area is determined through mountain torrent disaster data analysis; the drainage basin delay time is calculated, the section water level and the drainage basin submerging range are simulated through the simplified hydrodynamic model, and quantitative analysis is conducted on the model; according to the method, a high-precision grid database is formed by combining unstructured grid division and risk source identification, and then clustering grading is performed on multi-dimensional characteristics of grids through a Gaussian mixture model (GMM) to obtain risk levels of the grids in a drainage basin. Finally, a differential early warning threshold value and a response strategy are formulated according to a grid partition grading result, and dynamic updating and linkage issuing are achieved on a GIS platform in combination with actual measurement data. According to the method, limited real conditions of small and medium-sized basin data and precision requirements of model simulation are taken into consideration, fine setting of the early warning threshold value is realized on the basis of grading and partitioning, and the accuracy of mountain torrent risk identification is also remarkably improved.

The invention belongs to the technical field of iron stick yam image recognition detection, and discloses an iron stick yam intelligent system and method based on image recognition, and the method comprises the steps: dynamically adjusting the polarization direction to generate a polarization suppression image, and generating a space mapping relation; based on the polarization suppression image, segmenting the main body area of the iron stick yam, filling and optimizing the hole edge, generating a high-precision contour mask, further extracting the texture difference between wrinkles and cracks, generating a direction sensitive characteristic pattern, and evaluating morphological defects; constructing a comprehensive feature vector, training a Gaussian mixture model to generate reference distribution, calculating a quality deviation index, and judging a risk level; constructing a double-branch feature vector, outputting a preliminary comprehensive quality score through a complementary aggregation model, introducing a planting density-curvature physical model, obtaining a final quality score, dynamically adjusting a quality judgment threshold, dividing quality grades, and triggering a response strategy according to confidence grade to form a closed loop; and comprehensiveness and accuracy of Chinese yam quality evaluation are improved.

The invention provides an electricity price-based permanent magnet motor energy storage and optimization operation method and system, and relates to the technical field of motor energy storage and optimization operation, and the method comprises the steps: predicting the power of a motor through a deep belief neural network, predicting the electricity price through a space-time diagram convolutional neural network, and quantifying a system error in combination with a Gaussian mixture model, based on a prediction result, performing global optimization by adopting a quantum genetic optimization algorithm to obtain an initial optimization strategy set, applying the initial optimization strategy set to a meta-heuristic hybrid optimization model, generating a time sequence association control strategy, performing system modeling through a hybrid Gaussian-hidden Markov model, and constructing a double-layer action evaluation control network to generate an operation control instruction; a distributed reinforcement learning framework is used for cooperative control, and an optimal control strategy is obtained based on adaptive dynamic programming and a meta-learning strategy migration mechanism.

The invention relates to the technical field of geological disaster intelligent monitoring, in particular to a geological disaster intelligent monitoring system and method based on multiple sensors. The method comprises the following steps: covering a monitoring area through multiple sensors, and collecting and summarizing environment data; separating effective signals and noise by adopting a space-time adaptive decomposition method, completing data denoising by combining a signal-to-noise ratio dynamic threshold and a local weighted regression smoothing trend residual error, then aligning time and dimensions through linear interpolation and standardization, and integrating multi-sensor data based on an adaptive weighted fusion method; the fused data is subjected to generation and review of a to-be-reviewed mark, so that the data integrity is ensured; key features are extracted, a mixed model based on LSTM and a random forest is constructed to evaluate disaster risks, and a risk grade division rule is set; and finally, outputting fusion data and an analysis result, and generating an early warning signal and a decision suggestion for potential geological disasters. According to the invention, accurate monitoring and risk assessment of geological environment changes are realized.

The invention provides a Beidou electric power data intelligent acquisition and dynamic communication scheduling method and system. The method comprises the steps that a master station constructs a database and trains a data priority classification model, a channel quality prediction model and an intelligent compression model; a substation collects power data through an edge calculation unit, constructs a skyline candidate set to perform data screening, and performs intelligent classification marking by using a data priority classification model; performing reliability evaluation on the data stream by using a Gaussian mixture model, selecting a compression strategy according to a reliability score, a data type and a priority, and packaging into a data frame; determining a transmission strategy in combination with a channel quality prediction result and a context-aware intelligent switching protocol, and sending a data frame; the master station receives the data frame, performs integrity verification, decompresses and reconstructs the data frame, and feeds back a communication state for model updating; and monitoring the operation state, performing early warning based on the anomaly detection model, and triggering a self-healing strategy. According to the invention, the Beidou communication resource utilization rate and the system reliability are improved.

Provided are a method and an apparatus for constructing a road congestion prediction model, a device, a medium, and a product. A road traffic network is defined as a directed weighted graph. Historical dynamic traffic features of each road segment in the road traffic network are obtained as sample data, including recent dynamic traffic features and periodic dynamic traffic features. The sample data is input into a mixture of adaptive graph learners (MAGL) model for learning, and a probability prediction vector is output. The sample data is input into a trend expert model, and a trend distribution vector of a predicted probability of future traffic conditions is output. The periodic dynamic traffic features are fused to determine a periodicity prediction vector. An aggregated logit vector is obtained. An objective function is determined based on the aggregated logit vector. Congestion prediction training is performed to obtain a road congestion prediction model.

The invention discloses an intelligent fault diagnosis method fusing state monitoring and a multi-modal large model, and the method specifically comprises the steps: synchronously collecting time sequence data and a space image through a heterogeneous sensor group and monitoring equipment disposed in power grid equipment, and forming original data; based on the original data, a physical constraint feature vector is generated in combination with an equipment thermodynamic equation and a material deformation rule; performing health index prediction through the lightweight LSTM network based on the physical constraint feature vector; when detecting that the health indexes continuously decrease, clustering an HI time sequence curve by adopting a Gaussian mixture model, judging a degradation stage according to a clustering center distance, and obtaining a stage recognition result; and based on finite element simulation parameters, introducing a reinforcement learning model, optimizing the simulation parameters by taking maintenance cost minimization as a target, and outputting a predictive maintenance work order. According to the invention, intelligent fault diagnosis and accurate maintenance of the power grid equipment are realized, the fault processing efficiency and accuracy are improved, and the power failure loss is reduced.

The invention discloses a large model reasoning method and system based on tree diagram and knowledge graph retrieval enhancement, and the method comprises the steps: segmenting a document into text blocks, recognizing the entities, relationships and factual descriptions of the text blocks, and forming a knowledge graph of the text blocks; constructing a factual summary of the documents based on the knowledge graph of the text blocks, and performing clustering recursion on all the documents by using a Gaussian mixture model to generate a similarity tree structure graph; the method comprises the following steps: obtaining a user problem, selecting a reasoning path from a similarity tree structure diagram through a large language model debate iteration method, obtaining a corresponding document cluster, pruning the cluster based on the user problem and a confidence score of a document, selecting a plurality of text blocks from the pruned document, and finally obtaining a knowledge graph based on the selected text blocks. And generating a reasoning result. The invention relates to the technical field of natural language processing, can realize retrieval enhancement based on a tree diagram and a knowledge graph, and accurately queries related information when a large model answers a complex multi-hop question, so that the reasoning accuracy is effectively ensured.

The invention discloses an invading object detection method and device applied to a perimeter security system, and relates to the technical field of invading object detection.According to the invading object detection method and device, a background model is dynamically updated by combining adaptive background modeling with time sequence analysis, background stability is kept in dynamic environments such as illumination variation and leaf shaking, and the invading object detection efficiency is improved. Different types of background changes are distinguished through a Gaussian mixture model, and the coexistence condition of long-term static background and short-time dynamic change can be adapted; in the target detection stage, a front-and-back frame difference method and a background subtraction method are combined, a moving target is extracted by utilizing complementary characteristics of the two methods, in the target classification and behavior analysis stage, spatial morphological characteristics of the target are extracted, the motion trend of the target is analyzed and predicted in combination with a time sequence, and track prediction is performed by adopting Kalman filtering. Even if the target is shielded for a short time or moves slowly, the tracking stability can still be kept.

The invention relates to the technical field of gesture recognition, in particular to an anti-interference optimized gesture recognition method. Comprising the following steps: acquiring a gesture video stream through a camera, constructing a dynamic background model by using a frame difference method and a Gaussian mixture model, eliminating a static background and interference, and extracting a target area image; performing local brightness histogram analysis on the target region image, and optimizing the image quality by adopting a region adaptive compensation algorithm and a multi-scale edge enhancement technology; positioning a gesture area in real time by using a color histogram and a feature matching algorithm, and dynamically updating a gesture track in combination with Kalman filtering; and extracting gesture shapes, tracks and dynamic mode features through deep learning, comparing the features with a standard model library, and outputting gesture categories and corresponding function instructions. According to the method, a multi-level optimization strategy is adopted for a complex background, a dynamic target and a changeable illumination environment, so that the anti-interference capability and the recognition precision of gesture recognition are improved.

The invention discloses a soil cadmium pollution risk identification method based on a spatial information enhancement algorithm, and relates to the field of soil cadmium pollution risk prediction and risk identification, and the method comprises the steps: predicting a constructed soil cadmium pollution data set through employing a machine learning algorithm, and comparing the prediction precision of the machine learning algorithm, selecting an optimal model as a soil cadmium concentration prediction model; dividing a target area into grid units, utilizing a soil cadmium concentration prediction model to perform prediction, outputting soil cadmium concentration spatial distribution data, utilizing a Bayesian information criterion BIC to determine the optimal clustering number of soil cadmium pollution, adopting a Gaussian mixture model to perform clustering on the soil cadmium concentration, obtaining soil cadmium clustering features, and determining the soil cadmium pollution. Identifying and evaluating a soil cadmium pollution risk area; and carrying out overlay analysis on the risk receptor and the soil cadmium pollution risk area, and verifying the accuracy of identifying the risk area by the soil cadmium concentration prediction model. According to the method, accurate identification and reliable evaluation of the soil cadmium pollution risk are realized.

The invention relates to the technical field of mine network safety operation, in particular to a mine network safety operation system, which comprises a data acquisition and preprocessing module, the data acquisition and preprocessing module acquires various equipment and sensor data of a mine network, and the data quality is improved through denoising of an auto-encoder and processing of a Gaussian mixture model; the threat detection and intelligent analysis module adopts a DCNN and RL combined model and can accurately monitor attack behaviors; the defense strategy generation and response module generates and adjusts a defense strategy in real time according to an attack type and a network state by means of a hybrid reinforcement learning and genetic algorithm; the quantum encryption and secure communication module encrypts transmission data by using a QKD technology to ensure the security of the transmission data; and the event response and recovery module quickly responds to the security event and recovers the system through a self-adaptive algorithm. According to the invention, mine network security challenges are effectively handled, security is comprehensively improved, mine production continuity is ensured, and reliable guarantee is provided for mine network security operation.

The invention discloses a multi-source heterogeneous agricultural data element fusion system and method, and relates to the technical field of data fusion processing, and the method comprises the steps: carrying out the data screening of cleaned data; the screened data is preprocessed by the constructed agricultural field knowledge base, the logic consistency of the data is checked through a rule engine, and a corresponding coping strategy is adopted according to a check result; identifying abnormal points in the multi-source data by adopting a Gaussian mixture model and a constructed logic rule combination, and correcting or complementing the detected abnormal points in combination with a specific scene; key features are selected through information gain, multi-source fusion is carried out on the key features and original multi-source data, an abnormal mode library is established after abnormal data modes are classified, and source categories of the abnormal data are matched from a fault matching rule library according to abnormal features of the abnormal data; and the fault matching rule base is established, so that the source category of the abnormal data can be accurately identified, and a targeted solution is provided for processing the abnormal data.

The invention discloses a large language model end-cloud collaborative inference system based on low-rank fine tuning, and belongs to the technical field of inference optimization of end-side cloud computing. Establishing an end-cloud collaborative reasoning architecture, and in an offline stage, performing parameter fine tuning on a large language model by a cloud side based on training data of different downstream tasks; in the online stage, user requests are classified through'variational auto-encoder-Gaussian mixture model 'clustering, whether a low-rank adapter matched with a current task exists in an end side cache is judged, and if yes, reasoning is executed on the end side; and otherwise, forwarding the task to the cloud side. After a plurality of user requests are processed by the architecture, historical user requests and cache states are analyzed based on a Mama model, and an end-side low-rank adapter library is dynamically updated. And monitoring end cloud load and reasoning delay in real time, and issuing the new adapter to the end side according to the task repetition rate increment. According to the method, dynamic balance of the system is realized, and high efficiency and adaptability of the system are ensured while calculation and storage overhead are reduced.

The invention provides a micro-grid robust optimization scheduling method considering a segmented multi-interval uncertainty set, and the method comprises the steps: carrying out the segmentation fitting of the source load prediction power and error data of wind, light and loads based on a Gaussian mixture model, and combining with a box-type interval uncertainty set to build a multi-interval uncertainty set of power segmentation segments; the uncertainties are used for describing wind, light and load uncertainties; constructing a microgrid robust optimization scheduling model considering the segmented multi-interval uncertainty set and the carbon emission by taking the minimization of the operating cost and the carbon emission cost of the microgrid as a target; and solving the micro-grid robust optimization scheduling model by adopting a column sum constraint generation algorithm. According to the method, the problems of uncertainty modeling and carbon emission of the micro-grid can be solved, the conservative property of a robust optimization scheduling strategy is reduced, and the economical efficiency of micro-grid scheduling is improved.

The invention discloses a network space security risk intelligent identification method and system. The method comprises the following steps: deploying traffic collection equipment at key network nodes and boundary equipment, and collecting network traffic data and log data; extracting network behavior features and log key fields, and constructing a data feature model; performing vectorization processing based on the data feature model, performing traffic classification by using a deep neural network, and identifying a device type and a device identifier; carrying out noise reduction, redundant information removal and standardization processing on flow data in the data feature model to generate standardized data and storing the standardized data in a vector database; entity extraction is carried out on the standardized data, a knowledge graph is constructed, data distribution is analyzed through a Gaussian mixture model, the probability of abnormal behaviors is calculated in combination with historical data, and attack risks are calculated based on Bayesian decision; and generating an alarm according to the calculated attack risk, tracing the historical record of the attack risk in the knowledge graph, sending a blocking instruction, and identifying the security risk of the cyberspace.

The invention discloses a distributed photovoltaic power prediction method, system and device based on a Gaussian mixture model and a medium, and belongs to the technical field of photovoltaic power prediction.The distributed photovoltaic power prediction method comprises the steps that a time sequence vector is collected, principal component analysis is carried out on the time sequence vector, low-dimensional feature representation is obtained, and a power feature vector of each photovoltaic power station is formed; performing clustering analysis based on a probability model on the power feature vector to generate a plurality of photovoltaic power station clusters; for each cluster, acquiring meteorological input data through a set data source priority rule and a completion mechanism; and constructing a neural network power prediction model based on the accumulated power data in the cluster and the corresponding meteorological features, and outputting a future power generation power prediction value of the photovoltaic power station in the corresponding cluster. According to the invention, N photovoltaic power stations in a region are divided into M clusters through a GMM clustering method, so that the design is simplified; and the power prediction of the whole area is realized.

The invention provides a photovoltaic cleaning robot sensor data fusion and obstacle avoidance method and system, and relates to the technical field of photovoltaic cleaning robots, and the method comprises the steps: obtaining multi-source sensor data, carrying out the probability distribution modeling through employing a Gaussian mixture model, and achieving the data fusion through employing a Kalman filtering algorithm of an adaptive covariance matrix, and constructing a three-dimensional semantic map, marking obstacle information, and dynamically adjusting a motion track in combination with global path planning and a fuzzy logic controller of a local obstacle avoidance layer to realize smooth obstacle avoidance of the photovoltaic cleaning robot. The environment perception capability and the obstacle avoidance efficiency of the photovoltaic cleaning robot are improved, the collision risk is reduced, and the safety and the efficiency of photovoltaic panel cleaning operation are guaranteed.

The invention relates to a user psychological state monitoring method and system based on voice and semantic recognition, and the method comprises the steps: firstly obtaining voice data and corresponding text data of a user, and respectively extracting a voice spectrum feature vector and a text feature vector; then, inputting the two features into a cross-modal joint coding network model, and aligning a speech spectrum with a text embedding space by using a multi-head attention mechanism to obtain a joint feature; thirdly, confidence coefficients of the joint features are extracted through a Bayesian network model, weighted features are obtained, a Gaussian mixture model is used for fitting an emotion fluctuation trend, and an emotion distribution probability is calculated; the confidence coefficient weight is dynamically adjusted according to the emotion distribution probability, new weighted features are obtained, and after time sequence alignment processing, local information is extracted by using a convolutional neural network and the features are fused; and finally, calculating a classification probability through a softmax function, and determining a psychological state classification result of the user.

The invention relates to the technical field of mechanical equipment state monitoring, in particular to a method and sensor for detecting temperature and vibration data of dynamic equipment, and the method comprises the steps: collecting a temperature signal and a vibration signal of the dynamic equipment, and carrying out the fusion processing, thereby obtaining a fusion feature set; establishing a feature distribution baseline based on a Gaussian mixture model, calculating a relative entropy of the fusion feature set and the feature distribution baseline, and generating a dynamic threshold sequence; constructing a detection model according to the fusion feature set and the dynamic threshold sequence, and outputting to obtain an abnormal mode label; performing time serialization processing on the abnormal mode label, predicting a fault probability in a future time period according to the abnormal mode label subjected to time serialization, and generating a fault prediction result; and performing priority ranking on the fault types in the fault prediction result, and generating a maintenance report according to a priority ranking result. The reliability and practicability of temperature vibration data detection of the dynamic equipment are comprehensively improved, and an efficient solution is provided for health management of the dynamic equipment.

The invention provides a bionic robot control method and system based on AI learning, and the method comprises the steps: obtaining dynamic environment data including an obstacle position, a target state and topographic features, and collecting and transmitting real-time data to a data processing module through a sensor network; for the dynamic environment data, noise and redundant information are removed, key features are extracted, and a structured environment feature data set is generated; based on the structured environment feature data set, using a Gaussian mixture model to construct an environment model, and if the environment change exceeds a preset threshold, updating parameters of the environment model to reflect the latest environment state; according to the environment model, predicting a future environment change trend by adopting a time sequence prediction algorithm, and if the deviation between a prediction result and a current environment state exceeds a preset range, triggering a model retraining mechanism; and on the basis of the prediction result, in combination with the current state and the target task of the bionic robot, an optimal decision strategy is generated by using a reinforcement learning algorithm.

The invention relates to the technical field of heart sound signal processing, and discloses a heart sound feature recognition method based on wavelet spectral analysis. The method comprises the following steps: collecting heart sound signal data and carrying out signal filtering and segmentation processing to obtain segmented cardiac cycle signals; performing n-level discrete wavelet transform decomposition to obtain an approximation coefficient and n detail coefficients; performing multi-dimensional statistical feature calculation to obtain a multi-dimensional feature vector; performing principal component analysis dimension reduction processing to obtain a dimension reduction feature vector; fitting feature distribution by using a Gaussian mixture model and introducing a dynamic adjustment factor to obtain an adaptive classification threshold; the hidden Markov model is constructed through the dimension reduction feature vectors, the likelihood probability is calculated through a forward algorithm, the likelihood probability is compared with a self-adaptive classification threshold value, and the heart sound feature recognition result is output. The time-frequency features of the heart sound signals are fully utilized, the feature distinction degree and the recognition accuracy are improved, and the recognition accuracy is improved. And meanwhile, the method has good robustness and generalization ability.

The invention provides a data correction method and system combined with lattice structure additive manufacturing process characteristics, and the method comprises the steps: firstly employing a Newton iteration method, taking the radius of a pillar as a variable, optimizing the relative density of an iteration target and generating a lattice structure geometric model under the condition that a process constraint condition is satisfied, and extracting actual geometric parameters after forming through CT scanning, the elastic modulus is corrected through the pillar diameter deviation and the defect volume fraction, a compression failure mechanism is combined, a density-related failure criterion is introduced, a nonlinear relation with the relative density is constructed through specific energy absorption data, material plasticity parameters are inversely optimized, and a defect coupling evaluation model is constructed according to the surface powder sticking rate and the pillar diameter deviation. And establishing a Gaussian mixture model based on a stress-strain curve to screen out abnormal data, and finally fusing the parameters to construct a data correction model. The dot matrix structure design precision can be improved, and then a high-quality data basis is provided for performance prediction-structure design two-way feedback.

The invention belongs to the field of data storage, and discloses a frequency access data storage migration method and device for big data, and the method comprises the steps: obtaining a data access log, carrying out the preprocessing of the data access log, and obtaining a time window statistical matrix; carrying out timeliness attenuation weighting calculation on a plurality of data access times in the time window statistical matrix to obtain a weighted access frequency vector; obtaining a multi-dimensional feature matrix based on the weighted access frequency vector; clustering the feature vector of each time window slice in the multi-dimensional feature matrix through a Gaussian mixture model to obtain a clustering tag vector; inputting the clustering label vector into an access frequency prediction model to obtain an access frequency prediction value of the data in a future preset time step; the migration decision matrix is determined based on the access frequency predicted value and the data storage cost parameter, data migration is carried out based on the migration decision matrix, the data migration complexity can be reduced, and the access speed and the storage cost are balanced.

The invention provides a short-term photovoltaic power prediction method and system, computer equipment and a medium, and belongs to the field of photovoltaic power generation output power prediction.The method comprises the steps that short-term photovoltaic power and meteorological working condition data samples are obtained, and a Gaussian mixture model is used for conducting multi-modal clustering processing on the meteorological working condition data samples to obtain membership probability embedded vectors; through a Pearson's correlation coefficient weighting and sliding window mechanism, extracting features from the data sample and the membership probability embedding vector, and constructing a multi-modal time sequence feature tensor; a multi-head attention mechanism in a traditional Transform network is replaced with a class domain fusion self-attention mechanism, and a class domain fusion attention model is formed; inputting a multi-modal time sequence feature tensor to train a class domain fusion attention model to obtain an initial prediction value; and residual error estimation is carried out on the initial prediction value by using a residual error learning error compensation strategy, a short-term photovoltaic power prediction result is output, and the accuracy and robustness of the model are improved.

The invention discloses an aero-engine group health evaluation method based on multi-working-condition dynamic clustering, and belongs to the field of aero-engine health state evaluation. The method comprises the following steps: firstly, carrying out clustering analysis on set parameters in engine operation data, and carrying out merging processing on small-scale abnormal clusters to obtain a working condition category division result; secondly, constructing a health baseline data set, carrying out standardized preprocessing on sample data in a working condition category division result, and carrying out nonlinear dimensionality reduction to obtain a low-dimensional feature data set; thirdly, performing clustering analysis on the low-dimensional feature data set by adopting a Gaussian mixture model, and calculating an average mahalanobis distance between a sample of each clustering category and a health reference center to obtain multi-level health levels corresponding to different clustering categories; and finally, through fusing the membership soft probability and the sample individual mahalanobis distance, constructing a continuous health score and obtaining a health grade determination interval. According to the method, health state characteristics under different working conditions can be effectively identified, individual difference modeling and group transverse comparison evaluation are supported, and the accuracy is improved.

The invention provides an under-mine video stream dynamic denoising method based on multi-modal fusion, which comprises the following steps: constructing a time sequence synchronous fusion mechanism of visible light, infrared and laser radar data, and realizing time-space alignment of multi-source heterogeneous data; a dynamic noise model is established by introducing a fractional calculus optical flow field concept and combining a Gaussian mixture model, so that a dynamic noise region is accurately identified; an improved self-adaptive wavelet threshold function is constructed, a function threshold parameter can be linked with a dust concentration sensor in real time, and the de-noising intensity is dynamically adjusted according to the actual dust concentration; designing a dual-path feature enhancement neural network to effectively separate and enhance structural features and texture features in the video image; a cascaded detection decision system is created, a lightweight network is used as a primary detector, a high-confidence detection result is directly output, and a low-confidence detection result is input into a Transform correction module for secondary reasoning. According to the invention, dynamic denoising, feature enhancement and target intelligent monitoring of the video stream under the mine can be realized.

The invention relates to the technical field of optical fiber sensing and fault diagnosis, in particular to an optical fiber gyroscope fault monitoring method and system. The method comprises the following steps: extracting four-dimensional characteristics of phase difference, light intensity fluctuation, polarization state drift and temperature drift from an output signal of an optical fiber gyroscope, and constructing a high-dimensional characteristic matrix after unifying a time reference; establishing a dynamic prediction model based on historical data to generate a residual matrix, and orthogonally separating the residual into an internal structure abnormal signal and an environment disturbance signal through covariance characteristic decomposition; mapping the structure residual error sequence into a topological graph, calculating an evolution index in real time, and capturing a fault gradient trend in combination with sliding window gradient analysis; a multi-dimensional vector is constructed by fusing topological features, an adaptive classifier of an online Gaussian mixture model is adopted to identify a fault mode and quantify a health index, and meanwhile, a health measurement result is fed back to a prediction model and topological analysis parameters to realize closed-loop optimization. According to the invention, full-process adaptive monitoring from anomaly detection to health measurement is realized.

The invention discloses a method and a system for analyzing a pollution source of a rainwater pipe network. The method comprises the following steps of: 1, obtaining a first key fluorescence characteristic parameter combination according to basic physicochemical indexes by adopting a fluorescence characteristic parameter prediction model; 2, three-dimensional fluorescence characteristic parameters of the pollution source are obtained, and a pollution source three-dimensional fluorescence characteristic data set is obtained; screening by adopting a multivariate statistical method to obtain a second key fluorescence characteristic parameter combination; 3, determining a fluorescence characteristic parameter combination according to the first key fluorescence characteristic parameter combination and the second key fluorescence characteristic parameter combination; 4, inputting the fluorescence characteristic parameter combination into the Bayesian mixture model to obtain the contribution rate of the pollution source to the receptor water quality, and completing the analysis of the pollution source of the rainwater pipe network; according to the method, the DOM fluorescent fingerprint information is predicted by adopting basic data; and based on the three-dimensional fluorescence characteristic parameters and in combination with a Bayesian mixture model, precise quantitative analysis of the pipe network pollution source is realized.