40 results about "Learning Recognition" patented technology

The invention relates to the technical field of building disease detection, in particular to an urban building disease detection method and device, electronic equipment and a storage medium. Multi-modal image data formed by original visible light and thermal infrared image data is obtained, and an original thermal infrared image is subjected to geometric correction; calculating a mapping relation with an original visible light image so as to complete pixel-level registration, obtaining target multi-modal image data, inputting the target multi-modal image data into a hierarchical deep learning recognition model, recognizing building disease information, then performing three-dimensional space mapping, generating a building three-dimensional mesh model containing disease three-dimensional space setting coordinates, and finally performing three-dimensional mesh modeling. And then calculating a relationship between a model surface grid vertex and a disease point cloud density, generating a disease distribution thermodynamic diagram, analyzing disease aggregation characteristics in multiple dimensions according to the thermodynamic diagram, and quantitatively analyzing spatial correlation between the disease and a building construction node in combination with building component information. According to the invention, the urban building disease detection efficiency and precision are improved.

The invention relates to the technical field of geological informatization, in particular to an artificial intelligence modeling analysis method for a hydrate pilot production data set, which comprises the following steps of: acquiring logging data, lithology data, stratum physical property parameters and natural gas hydrate production dynamic data; screening, cleaning, complementing, de-noising and standardizing are carried out in sequence to obtain an artificial intelligence modeling data set; and establishing a stratum lithology machine learning recognition model, a stratum physical property machine learning recognition model and a natural gas hydrate artificial intelligence historical fitting model through a support vector machine SVM, a random forest RF and a neural network DNN. According to the method, a serial modeling architecture of lithology identification, physical property prediction and production history fitting is created, and the prediction output of the upstream model is used as the optimization input of the downstream model, so that the downstream production prediction model can learn physical property parameters which are recalculated based on machine learning and have higher precision; and the accuracy of final production prediction is improved from the data source.

The invention discloses a myopia image deep learning recognition model training method, particularly relates to the technical field of medical image processing and deep learning, and is used for solving the problem that an existing deep learning model lacks anatomical structure priori knowledge guidance in myopia eye bottom image analysis. The method comprises the following steps: acquiring a myopia eye bottom image and anatomical structure priori knowledge data, extracting a multi-scale feature map by using a deep learning model, analyzing the geometric morphology of a key anatomical component based on standard spatial relationship information, and generating a spatial constraint loss item; according to the method, key anatomical path topology coherence is evaluated based on topology connection information, topology constraint loss items are generated, a loss item fusion strategy is dynamically adjusted according to a training stage, finally, a model is iteratively trained to convergence through a gradient back propagation algorithm, and organic combination of medical priori knowledge and a deep learning model is realized. And the clinical rationality and reliability of model output are improved.

The application discloses a CSI-based position-independent human activity continuous learning recognition method, which comprises the following steps: 1, collecting CSI action sample data; 2, pre-processing the CSI action sample data; 3, constructing positive samples by randomly scaling the pre-processed samples in the time dimension; 4, constructing a multivariate time graph neural network and extracting CSI action sample features; 5, calculating the similarity between the sample feature values and the positive samples and the feature values of the remaining samples, obtaining a comparison loss, and optimizing the feature extraction network; 6, freezing the feature extraction network, sending the features obtained from the input samples into a classifier for training to obtain a classification model. When the application continuously learns new action categories, the user does not need to retrain the feature extraction network, and the new and old action recognition in any position in the room can be realized by providing limited position new category samples to train the classifier, and the practicability is relatively high.

The invention discloses a power grid fault data processing method and device and a medium, and the method comprises the steps: carrying out the multi-dimensional detection data collection of a target power grid, obtaining a power grid image and sensing parameters, and collecting the environment parameters in a target power grid environment; acquiring historical detection data of a target power grid, extracting a fault sample proportion coefficient, performing combination division on the historical detection data, and performing training of an integrated image fault identification branch and an integrated sensing fault identification branch; performing fault rate influence analysis and detection data influence analysis according to the environmental parameters to obtain a fault rate influence coefficient, an image influence coefficient and a sensing influence coefficient; according to the fault rate, the image and the sensing influence coefficient, the number of image recognition branches and the number of sensing recognition branches are calculated and obtained, fault recognition branch calling and fault recognition are carried out, a power grid fault recognition result is obtained, and the technical problem that the machine learning recognition fault perception rate and accuracy are low due to the fact that power grid line inspection fault data samples are few is solved.

The invention is suitable for the technical field of computers, and provides a large language model training data generation method based on flow playback and implicit feedback, and the method comprises the steps: obtaining interaction session data, including user questions, model answers and user subsequent behaviors, of a user and a large language model in online service; based on the interactive session data, identifying a user preference signal through a hierarchical implicit feedback judgment algorithm; the hierarchical implicit feedback judgment algorithm preferentially processes high-confidence implicit signals, including collaborative learning recognition based on user editing behaviors, recognition based on user query reconstruction and automatic judgment based on confidence scoring; a structured preference training data set is generated according to the user preference signal and comprises a plurality of preference pairs, and each preference pair comprises user questions, correct answers and wrong answers; the authenticity and quality of the data are effectively improved, the problems of reward model cracking and model catering are relieved, and continuous optimization and rapid iteration of model performance are achieved.

The invention relates to a panda behavior recognition method based on deep learning and time sequence prediction, and the method comprises the steps: S1, constructing a panda behavior deep learning recognition model which comprises a region division module, a multi-view motion analysis module, a multi-scale posture estimation module and a posture guide fusion module; s2, carrying out region division and part coding on the input video, and generating a part coding video comprising three regions, namely a panda head region, a panda whole region and a background region; s3, performing multi-view motion analysis on the part coding video, and extracting multi-view motion features of the panda; s4, extracting multi-scale features of the joint attitude information of the panda in the part coding video, and performing feature fusion to generate an attitude thermodynamic diagram; and S5, taking the posture features in the posture thermodynamic diagram as guide signals to be fused with the multi-view action features, and outputting the giant panda behavior category. According to the method, the accuracy and robustness of giant panda behavior monitoring can be improved.

The invention discloses an insulator degradation analysis method, system and device and a medium, and the method comprises the steps: inputting the real-time inspection data of an insulator into a multi-task deep learning identification model which is obtained through introducing an insulator feature extraction branch and a cross-modal feature attention fusion branch into a backbone network of a YOLOv5 model, and carrying out the degradation identification; obtaining degradation state information, and importing the degradation state information into a power grid geographic information comprehensive management system; according to historical inspection comprehensive data and real-time climate data of the insulator, an inspection polymorphic layer of the insulator is constructed through the system, a multi-dimensional algorithm is adopted to analyze the degradation rule of the insulator, and a degradation risk analysis result is obtained; on the basis of a power grid geographic information integrated management system, insulator degradation identification, map positioning and multi-dimensional algorithm fault law analysis are combined, so that the utilization rate of insulator degradation state information is improved, and solid data support and scientific basis are provided for formulating a power grid operation and maintenance strategy.

This invention discloses a method, device, and medium for intelligent identification of wire harness defects based on deep learning, relating to the field of industrial defect identification technology. The method includes: using a deep learning recognition network to perform joint deep learning representation and defect candidate identification on topology-aligned image sequences, outputting an initial list of defect candidates; verifying the structural relationships of the initial list of defect candidates based on the wire harness topology prior graph, and incorporating the consistency verification conclusion into the defect candidates as a judgment criterion, while correcting the target region identifier, generating a structural constraint defect judgment list; and performing online inference calibration on the structural constraint defect judgment list and re-judging boundary samples to generate a wire harness defect judgment conclusion. This invention uses a deep learning recognition network to perform joint deep learning representation and defect candidate identification on topology-aligned image sequences, enabling defect candidate identification to simultaneously consider local texture anomalies and global morphological relationship information, and improving the distinguishability of defects in complex critical connection areas.

The invention discloses a pathogenic microorganism infection immune feature recognition system and method, belongs to the technical field of immune feature recognition, and aims to solve the problems of incomplete feature extraction, weak model generalization ability, poor adaptability to novel pathogenic microorganisms and the like due to the fact that a traditional immune feature recognition technology mostly adopts single-dimensional features or a traditional machine learning model. In order to solve the problems of low recognition accuracy, high false positive rate and difficulty in meeting actual requirements of clinical diagnosis and epidemic situation monitoring in the prior art, the pathogenic microorganism infection immune feature recognition system comprises a data acquisition module, a data preprocessing module, a multi-dimensional immune feature extraction module, a fusion deep learning recognition model module and a result output and verification module, according to the method, through a deep learning architecture fused by Transform, CNN and LSTM, the long-distance dependency relationship, local features and time sequence features among the features are captured at the same time, optimization strategies such as transfer learning and regularization are combined, and the recognition accuracy and generalization ability of the model can be significantly improved.

This invention relates to the field of foreign object disturbance identification technology, and discloses a method, system, device, and medium for online foreign object disturbance identification in a weighing rain and snow gauge based on analog data-driven simulation. The method includes: outputting a formula for calculating the weight increase based on the rain-collecting area of ​​the weighing rain and snow gauge; generating a basic precipitation signal with noise and instantaneous disturbances based on the weight increase calculation formula and disturbance rules; generating a disturbance label array based on the basic precipitation signal; training a learner using window data filtered from the disturbance label array; using the trained classifier as a machine learning recognition model for online disturbance identification; deploying the machine learning recognition model into the main controller of the weighing rain and snow gauge; and using a two-stage pipeline architecture to achieve online identification of foreign object disturbances in the weighing rain and snow gauge. This invention employs a two-stage pipeline architecture for anomaly-triggered classification, ensuring high-precision real-time identification while maintaining low power consumption.

The application discloses a Parkinson disease severity recognition method based on tensor ring decomposition and region segmentation, relates to the technical field of machine learning recognition, and comprises the following steps: collecting VGRF signals, accurately dividing the VGRF signals according to target personnel indexes, gait window indexes, foot indexes, set region indexes and sampling point time indexes, constructing a five-order time domain tensor, and improving the learning ability for large-dimension tensors; converting the five-order time domain tensor into a frequency domain five-order tensor, extracting low-rank structures in the frequency domain five-order tensor as core tensors, reducing the data dimension of calculation, and improving the learning power of a classification model; extracting statistical features of patients in gait, feet and each set region in the core tensors, further maintaining the correlation features between each dimension of data on the basis of reducing the data dimension of calculation, and then accurately distinguishing the severity of Parkinson disease of the patients and improving the accuracy of the classification model in recognizing the disease severity of the patients.

This invention discloses a photovoltaic junction box defect detection method based on region-constrained two-stage visual recognition. The method includes: inputting the original image into a one-stage target detection model to obtain candidate junction box targets; dividing the image into left, center, and right regions according to the width of the original image; cropping the selected junction box targets to obtain multiple local images; inputting each local image into a two-stage instance segmentation model to obtain segmentation results for lead wires, soldering, and residual adhesive, respectively; and finally, performing an inverse mapping fusion algorithm according to the local cropping offset, superimposing the recognition results and quantification indicators onto the original image for output. This scheme combines deep learning recognition with spatial prior constraints, candidate scoring algorithms, mask deduplication algorithms, workstation assignment algorithms, parameterized defect rules, and result inverse mapping fusion, which can improve the stability, interpretability, and engineering applicability of photovoltaic junction box detection.

This invention discloses a method, system, device, and storage medium for hail cloud feature analysis and identification based on multi-source data. The method includes: obtaining satellite feature products, derived feature products, and radar feature products strongly correlated with the hail cloud formation and dissipation process; performing spatial registration processing on the satellite feature products, derived feature products, and radar feature products to obtain spatiotemporally aligned satellite feature sets and radar feature sets; constructing a dataset based on the spatiotemporally aligned satellite feature sets and radar feature sets; constructing a GSCF deep learning recognition model; and training and testing the model. This application effectively solves the problem of heterogeneous fusion of multi-source data, fully leverages the advantages of macroscopic meteorological information from satellite data and fine structural information from radar data, improves the quality and representational ability of fused features, provides reliable data support for accurate hail cloud identification, and is of great significance for improving disaster early warning efficiency.

The application discloses a curtain wall system connecting node stress state recognition method based on deep learning, and the method comprises the following steps: collecting original stress time series data and synchronous environment temperature data of a curtain wall connecting node to form a sample set; based on dynamic time clustering, the original stress time series data of each sample is segmented and divided, and an adaptive feature mapping function combining segmented information and a wavelet base function is used for feature mapping, and then dimension reduction is performed through principal component analysis to obtain a dimension reduction feature vector; a deep learning recognition network is constructed, and a double supervision loss function containing a weighted time series focal loss and an attention consistency regularization term is used to train the deep learning recognition network; and the obtained enhanced feature sequence and external physical features are input into the trained deep learning recognition network to output a stress state category of a node to be recognized. The application realizes automatic and engineering deployable transformation from original multi-source monitoring data to a clear state grade.

The invention discloses a bone nail recognition system and method based on machine recognition and artificial intelligence, and relates to the technical field of bone nail intelligent recognition. Bone nail tray partition images are collected through an image collection module, and the image quality of a bone nail tray is optimized by automatically adjusting shooting parameters; a camera of the image acquisition module moves in a partition above the tray through a three-axis movement mechanism, and then an image splicing and fusion unit of the image processing and recognition module splices and fuses partition images of the bone nail tray to obtain a complete image of the bone nail tray; and the deep learning recognition unit inputs the complete image of the bone nail tray into a pre-trained deep learning model, and outputs to obtain a bone nail information recognition result, so that the accuracy and efficiency of bone nail recognition can be improved.

The invention discloses an automatic step searching and processing method for grid-connected test data of a photovoltaic inverter, and belongs to the technical field of photovoltaic power generation system testing, and the method comprises the steps: collecting and preprocessing time sequence data through a power grid simulator and a power analyzer; inputting the preprocessed data into a pre-trained convolutional neural network model, and automatically outputting coordinates of a starting point and an ending point of step change; obtaining a trigger signal of the power grid simulator, and performing time compensation alignment on the trigger signal and the multi-channel data; based on the aligned data and step point coordinates, extracting a steady-state interval of each step platform, and calculating a platform value, response time and overshoot; and finally, filling the result into a standard test report template, and automatically generating a test report. According to the method, the problems of low efficiency, high subjectivity and high error rate caused by dependence on manual analysis in the prior art are solved, and the accuracy, consistency and efficiency of testing are remarkably improved through full-automatic data processing and machine learning recognition.

The invention belongs to the technical field of underwater detection and identification, and relates to an underwater target multi-mode identification method and system based on a cascade mode. The method comprises the steps that an optical image and an acoustic image of a target in the same underwater area are collected and preprocessed; inputting the preprocessed optical image into a pre-trained first deep learning recognition model to obtain a first recognition result based on the optical image and a first confidence coefficient corresponding to the first recognition result; and performing cascade decision judgment based on the first confidence coefficient. The underwater optical image recognition result and the underwater acoustic image recognition result are deeply fused through an ordered conditional judgment type cascade structure, high-quality optical information is preferentially utilized, and when the optical information is unreliable, the optical information is intelligently switched to dominant acoustic information or finer fusion judgment is started. Therefore, high-precision and high-robustness recognition of the target is realized in various complex underwater environments.

The invention belongs to the field of machine visual defect detection, and particularly relates to a thermal battery die-casting die visual cleanliness detection method and device, and the method comprises the following steps: obtaining the product size information of a to-be-detected die, and configuring the visual field parameters of an industrial camera according to the product size information; moving a tray bearing the mold to the position under the camera; jacking the tray, and adjusting the longitudinal position of the camera to a preset shooting height; turning on a double-layer concentric annular dome light source, and shooting a front image of the mold; carrying out segmentation processing on the image, extracting a key feature region, and obtaining a to-be-detected image; and analyzing the to-be-detected image by adopting a pre-trained deep learning recognition model, and judging the cleanliness state of the mold. According to the invention, the automatic and high-precision visual inspection of the internal cleanliness of the concave cylindrical mold is realized, the low-efficiency and subjective manual visual inspection is effectively replaced, and the detection efficiency and consistency are improved.

The application discloses an intelligent terminal identity authentication and permission management method based on machine learning recognition. The method uses the user operation habits, physiological characteristics and environment interaction data collected by a terminal device, establishes a user recognition model through a machine learning method, forms a continuous confidence evaluation of the user identity, divides the user identity into multiple confidence levels according to the confidence evaluation, and opens the function permissions of the terminal device according to the confidence levels, so that the high confidence level corresponds to the high permission opening, and the low confidence level corresponds to the low permission restriction, thereby realizing the security management. The application realizes the change from the traditional discrete authentication to the continuous cognitive authentication, and improves the use convenience while ensuring the safety.

This invention discloses an AI-based method and system for identifying the authenticity of traditional Chinese medicine (TCM) powder based on microscopic images, belonging to the field of TCM identification technology. The method includes: performing microscopic imaging of the TCM to be tested and performing depth-of-field fusion to obtain a microscopic image; inputting the microscopic image into a trained detection model for TCM identification; constructing a detection model based on a deep learning recognition model with a Transformer architecture; the training process of the detection model is as follows: selecting a subset of standard sample microscopic images and manually annotating the microscopic identification features to obtain manually annotated images; using the manually annotated images to train the detection model to obtain a lightweight model; using the lightweight model to identify and annotate the microscopic identification features of all standard sample microscopic images, and then manually reviewing and correcting them to obtain corrected standard sample microscopic images; training the detection model using the corrected standard sample microscopic images to obtain the trained detection model.

The invention provides a game coin recognition learning system capable of recognizing game coins betted by players with high precision. The game coin recognition learning system (10) is provided with a game recording device (11), which uses a camera (212) to record the state of game coins (W) stacked on a gaming table (4) as an image; a game currency determination device (12) having an artificial intelligence device (12a) for performing image analysis on the recorded state image of the game currency (W) to determine the number and type of game currency (W) betted by the player (C); and a teacher device (13) that, when it is determined that a determination result of the coin determination device (12) is suspected to have an error, inputs an image used in the determination by the coin determination device (12) and the number and type of correct coins W for the error into the artificial intelligence device (12a) as teacher data, and causes the artificial intelligence device (12a) to learn.

This invention relates to a method and system for classifying and recognizing human-machine safety status in oil and gas operation scenarios based on deep learning. The method includes the following steps: S1, acquiring video or image data of the oil and gas operation scenario; S2, preprocessing the acquired video or image data; S3, extracting features from personnel and equipment targets in the oil and gas operation scenario using the backbone network and feature fusion structure in the deep learning feature extraction network; S4, classifying and recognizing unsafe human-machine status, based on the extracted feature information, using a deep learning classification and recognition model to classify and recognize the operator's violations and the unsafe operating status of the equipment, obtaining the corresponding unsafe human-machine status recognition results; S5, outputting the results. This invention, through the cooperation of data preprocessing and a deep learning recognition module, improves the system's recognition accuracy and robustness under complex lighting, occlusion interference, and dynamic operation conditions.

This invention discloses a simulation system and method for pattern recognition in underwater vortex optical communication, belonging to the field of underwater optical communication technology. The system includes: a beam generation module for generating conjugate superimposed vortex beams with different topological charges and illuminating them onto an underwater disturbance module; an underwater disturbance module for setting up an underwater disturbance environment; a data acquisition module for capturing beam intensity images of the conjugate superimposed vortex beams after transmission through the underwater disturbance module; and a data processing module including a deep learning recognition model and a display control device. The deep learning recognition model is used to recognize the beam intensity images obtained by the data acquisition module; the display control device is used to display the recognition results, the parameters for training the deep learning recognition model, and to adjust the topological charge of the conjugate superimposed vortex beams in the beam generation module. The method involves using the above system for recognition. This invention is adaptable to complex underwater environments and exhibits high robustness in the recognition simulation system.

The invention provides a steam generator heat transfer tube expanded tube contour automatic analysis eddy current detection method and system. The detection method comprises the steps that S1, eddy current signals are used for conducting data collection on changes of an expanded tube contour; s2, extracting a signal feature region as a deep learning recognition target and a signal feature region; s3, establishing a deep learning model of expanded tube contour measurement; s4, training the deep learning model; and S5, the trained deep learning model is used for measuring the contour of the expanded tube. The application of the deep learning model enables the whole detection process to realize high automation and intelligence, and through the training and optimization of the deep learning model, the rapid and accurate identification of tiny defects in the expanded tube contour of the heat transfer tube can be realized, the timely discovery of potential safety hazards is facilitated, and the normal operation of a steam generator is ensured.

This invention provides a machine vision-based method for detecting surface defects in automotive die-cast parts, belonging to the field of automotive die-cast part inspection technology. The method includes: acquiring optical images, three-dimensional morphology data, and micro-area temperature distribution of the die-cast part surface; real-time correlation with the pressure-temperature-time curve of the die-casting process; generating a fused image through pixel-level feature overlay; delineating high-risk areas; establishing a dynamic rule base and a self-learning recognition mechanism; automatically generating a new defect rule base after identifying new defects; and mapping the defect detection results to a three-dimensional digital model of the casting in real time. This invention constructs a full dataset by fusing multi-source data and generating a high-precision fused image, accurately delineating high-risk areas. Through a dynamic rule base and self-learning mechanism, it achieves efficient and accurate identification of typical and new defects, effectively improving the accuracy and efficiency of surface defect detection in die-cast parts, enhancing the adaptability to various defect identification methods, and improving product quality and production stability.

The application discloses a kind of daily ceramic multi-defect identification and grading method, belong to daily ceramic automation detection technical field.The method is aimed at the problem of few defect detection types, low accuracy, insufficient automation of prior art, its gist is that: first, multiple detection stations and the special lighting scheme matched with it are used to collect multiple images of tableware surface;Then the multiple images are input into the pre-trained deep learning recognition model for collaborative identification, and the defect type result is output;Finally, according to the recognition result and the preset rule, the quality grade is determined, and automatically sorted into the corresponding output channel.The application is mainly used to realize the comprehensive, accurate and efficient automation detection and grading of multi-class defects of daily ceramic tableware.

This invention discloses an image recognition system for artificial intelligence based on reinforcement learning. The invention relates to the field of artificial intelligence image recognition technology and includes a multi-device image input module, an optimal transmission color alignment preprocessing module, a Q-learning convolutional kernel dynamic configuration feature extraction module, a reinforcement learning recognition decision module, a performance monitoring module, an adaptive adjustment module, and a self-optimization module. The multi-device image input module receives image data from multiple heterogeneous devices and detects and identifies the device type and color space type corresponding to the image data. The advantages of this invention are: it adopts optimal transmission color alignment preprocessing technology, fundamentally solving the problem of cross-device color distribution mismatch; it is compatible with image input from multiple heterogeneous devices such as surveillance cameras, industrial inspection cameras, consumer mobile phones, and remote sensing satellites; the cross-device recognition accuracy is improved compared to existing normalization schemes; and the system's versatility is significantly improved.

The invention discloses a cage-rearing laying hen abnormal sound monitoring method and system. The system comprises a pickup sensor, a deep learning model, a deep learning processor and a cloud server. Aiming at the problems of complex sound wave reflection and large equipment noise interference in a laying hen cage culture environment, an array microphone is used for collecting henhouse audio, band-pass filtering and multi-window spectrum subtraction are adopted for denoising, and effective sound production segments are cut by combining energy entropy proportion method endpoint detection; constructing a ResNet deep learning recognition model based on the MFCC features, and carrying out the training of the ResNet deep learning recognition model; the MFCC feature extraction module is deployed on an edge device to extract MFCC features and first-order and second-order difference combinations thereof in real time as three-dimensional input, and the three-dimensional input module inputs an optimal deep learning model for classification and recognition and then uploads the three-dimensional input to a cloud server; preferably, the server generates graded early warning according to the frequency of the abnormal sound, and the graded early warning is visually displayed in the front-end applet. Experiments show that compared with traditional manual inspection, the efficiency is greatly improved, and early warning of laying hen respiratory diseases can be achieved.

The application provides a cerebellar Purkinje neuron rapid identification system and method based on multi-modal deep learning, relates to the cross technical field of biomedical engineering and computer vision, and fuses specific fluorescence features and multi-dimensional morphological features, constructs a special multi-modal deep learning model, solves the interference problem of other neurons in a mixed culture system, and greatly reduces the false positive rate; a small sample learning architecture solves the industry pain point of a small amount of biological sample labeled data, improves the model generalization capability, and adapts to multi-scene sample identification. The identification system fuses Purkinje neuron specific fluorescence labeling features and cell morphological features, constructs a multi-modal deep learning recognition model based on small sample learning, is matched with a full-automatic microscopic imaging and analysis system, realizes rapid and high-accuracy identification of Purkinje neurons under a mixed culture system, does not need a complex cell purification step, greatly shortens the experimental period, and reduces the technical threshold.