69results about How to "Increase differentiation" patented technology

The invention discloses a virtual reality interaction method for constructing multi-modal fusion based on application scene features, and relates to the technical field of virtual reality. In order to solve the technical problem that in the prior art, when an existing multi-modal fusion technology is used for processing fusion of different modal data, the characteristic difference between different modals is large, the technical scheme provided by the invention comprises the following steps: acquiring and preprocessing multi-modal data; constructing a cross-modal knowledge enhancement network, extracting initial features, and enhancing the consistency and difference between visual and auditory features; important context information in fusion features is optimized and enhanced based on an optimization network of frequency spectrum and space attention; constructing an edge guide learning network, and combining the edge guide learning network to the tactile data; the visual, auditory and tactile features are fused and applied to a virtual reality scene; and optimizing the model through structural similarity loss, mean square error loss, significance weight loss and edge loss. The method is suitable for cooperative work of multi-modal features in a complex environment.

The invention discloses a social robot detection method based on directed relation graph comparative learning, and the method comprises the steps: taking each user as a user node, obtaining the personal attribute information and the personal text information of each user node, splicing the personal attribute information and the personal text information, and converting the spliced information into a feature vector; constructing a user directed relation graph, learning features of adjacent users under different relations by using a directed relation graph convolutional neural network, and performing relation feature fusion based on an attention mechanism; a comparison view of the directed relation graph is constructed from the two aspects of attribute features and graph topological structures in combination with direction information, and then training and prediction are carried out. According to the method, the accuracy of social network platform robot account identification is improved through the directed relation graph attention mechanism and directed graph comparative learning, and the directed relation graph is combined with the attention mechanism to effectively utilize the direction information of the social network platform relationship and the importance information of the relationship between the users in real life.

The invention belongs to the field of multi-label image classification, and particularly discloses a neural collapse-based long-tail multi-label image classification method, which guides learning of different types of features by using label embedding of an equiangular tight frame, so that the different types of features meet the property of neural collapse, the difference between different types is maximized, and the neural collapse-based long-tail multi-label image classification efficiency is improved. The influence of the head class on the tail class in the same sample is reduced, and learning of tail class features is fully guaranteed. Besides, a comparative learning method is also adopted, so that each class feature is clustered and collapsed to a class prototype, and the model learns more robust features. And finally, features corresponding to different classes of samples extracted by the model are combined and input into a binary classifier meeting an equiangular tight frame, so that the sample features are fully utilized, the influence of an unbalanced data set on the training of the model is avoided, the generalization ability of the model is enhanced, and the training efficiency of the model is improved. And the prediction precision of multi-mark image classification is improved.

The invention discloses an industrial image defect detection method, which comprises the following steps that an industrial image defect detection model is constructed, and the industrial image defect detection model comprises a plurality of teacher networks, student networks and a bidirectional feature mapping module; generating a synthesis abnormal image set; training an industrial image defect detection model, constructing a contrast loss function, and optimizing parameters of the industrial image defect detection model; and inputting the to-be-detected image into the industrial image defect detection model to obtain a detection result. According to the invention, high-precision industrial image defect detection is realized, a plurality of teacher networks are utilized to train a student network, and a bidirectional feature mapping module is added between the teacher network and the student network to relieve the huge difference between model capabilities of the teacher network and the student network. And the difference between network output features of teachers and students is increased by comparing loss functions, so that the industrial image defect detection model is more accurate during detection and has stronger detection performance.

The overhead line fault detection method based on the multi-source three-mode data fusion network comprises the following steps: acquiring historical fault data of an overhead line, constructing a multi-source three-mode data set containing waves, images and texts, and performing space-time dual-scale data enhancement to obtain a data set for fault detection; building different fault feature extraction network models for waves, images and texts for feature coding of each modal data, performing semantic alignment on feature vectors of the output waves, images and texts, and performing feature fusion to obtain fusion features of the multi-source three-modal data; generating a graph structure based on the fusion features of the multi-source tri-modal data, and inputting the graph structure into a dynamic graph attention neural network for training; and three-mode fault data of faults are collected to form a multi-source three-mode data set, a graph structure is generated after feature fusion, the graph structure is input into the trained dynamic graph attention neural network, a corresponding fault type is obtained, and parts and parts where the faults occur are positioned.

The invention discloses a text data enhancement method based on government affair text language disease error correction, which comprises the following steps: constructing a proper noun and terminology dictionary, and recording the positions and types of proper noun and terminology; analyzing a syntactic structure and literary and physical characteristics according to a proper noun and terminology dictionary, and constructing a language disease generation rule; the method further comprises the following steps: preprocessing the to-be-processed text data to obtain a preprocessed text; the preprocessed text is translated back, wherein the preprocessed text is translated into a text of a target language, and then the text of the target language is translated back to an original language; comparing the texts before and after the back translation, and screening out new texts of which semanteme is not damaged and which have language diseases as enhanced data; generating a negative sample on the basis of the enhanced data by using a language disease generation rule, and expanding the negative sample as an unprocessed negative sample; the editing distance of the unprocessed negative samples is calculated, and texts with the editing distance smaller than a threshold value are reserved to serve as the finally generated negative samples.

The embodiments of the present application provide a robot scanning method, device, equipment and storage medium. In the robot scanning method, the map established by the robot in the target space at the current moment can be obtained, and the map can be compared with the map established at the historical moment to obtain the similarity comparison result between the two maps. If the similarity comparison result does not meet the set similarity condition, a distinguishing mark is added to the target space so that the similarity comparison result between the map of the target space after adding the distinguishing mark and the map established at the historical moment meets the similarity condition. Through this implementation method, the difference between the map established by the robot in the target space and the map established at the historical moment is improved, thereby reducing the possibility of positioning loss during the scanning process.

The invention provides a code generation method and device based on an up-multiplexing mixed diversity expert large model, and belongs to the field of artificial intelligence, and the method comprises the steps: obtaining code description information; inputting the code description information into a trained code large model, and obtaining a code output by the code large model; the code large model is a large model based on an upper multiplexing mixed diversity expert; the large code model is constructed by adding a shared expert and a plurality of common experts on the basis of the dense model; sharing parameters of an FFN module of the expert multiplexing dense model; the shared experts are used for extracting general knowledge of the lexical elements; a common expert adopts a random initialization mode to initialize parameters; and common experts are used for extracting special knowledge of lexical elements. According to the method, a shared expert and a plurality of common experts are added on the basis of a dense model to construct a code large model, and a diversity multiplexing mechanism is adopted, so that the accuracy of generated codes is improved.

The invention discloses a K-means-based disaster recovery power distribution network dynamic regulation and control method and device, and a medium. The method comprises the steps of obtaining a data set of a disaster recovery power distribution network; selecting a plurality of data points as clustering centers according to the data set to obtain a first clustering center set; distributing the data set according to the distance between the data set and the first clustering center set to obtain a clustering result; and generating a disaster recovery strategy according to a clustering result to regulate and control the disaster recovery power distribution network. According to the K-means-based disaster recovery power distribution network dynamic regulation and control method and device and the medium provided by the invention, a high-dimensional power grid data set is clustered into a plurality of groups, the complexity of the problem can be reduced, and clustering analysis can help identify potential risk points in a power grid, so that prevention measures are formulated in advance, the probability of fault occurrence is reduced, and the reliability of the power grid is improved. The problem that it is difficult to effectively cluster data to generate a disaster recovery strategy to deal with disasters can be solved.

The present invention relates to a use of a composition including cyclo(hispro) (CHP) and parathyroid hormone (PTH) for preventing, improving or treating a bone loss-related disease.

A method for enhancing steel surface images based on a regional autonomous strategy belongs to the field of computer vision and image processing technology. The steps are as follows: 1. Image preprocessing: Acquire the original image of the steel surface and convert the input RGB color image to a grayscale image; 2. Defect region detection: Use the Canny edge detection algorithm to perform edge detection on the grayscale image and extract potential defect boundary information; 3. Regional CLAHE enhancement: Based on the region mask generated in step 2, perform contrast-limited adaptive histogram equalization on the defect region and the background region respectively, and set different enhancement parameters; 4. Median filtering; 5. Weighted image fusion: Perform weighted fusion processing on the enhanced image obtained in step 3 and the filtered image obtained in step 4 to obtain the final output image. This invention can improve the overall quality and defect identifiability of steel surface images through the above method.

The invention discloses an unmanned aerial vehicle game decision-making system based on anomaly detection. The unmanned aerial vehicle game decision-making system comprises an unmanned aerial vehicle sensor database, an upper computer and a sensor real-time data receiver, the unmanned aerial vehicle sensor database and the sensor real-time data receiver are respectively connected with the upper computer; the upper computer comprises a data preprocessing module, a pre-training module and an anomaly detection module. And the unmanned aerial vehicle sensor database stores the unmanned aerial vehicle flight state data accumulated in the historical flight process. The sensor real-time data receiver collects real-time state data of the unmanned aerial vehicle in the flight process and transmits the real-time state data to the unmanned aerial vehicle anomaly detection module of the upper computer for anomaly detection. The upper computer comprises a data preprocessing module, a pre-training module and an anomaly detection module, data transmitted by the database and the receiver are preprocessed, adversarial pre-training is carried out by using normal data in the database, and a trained model is used for carrying out anomaly detection on real-time data of the receiver. According to the method, the antagonism training framework anomaly detection model is innovatively introduced to fully learn the intrinsic characteristics of normal data, the difference between the normal data and the abnormal data is enhanced, the problem that a traditional anomaly detection system cannot completely recognize tiny anomalies is solved, and the anomaly detection precision of the unmanned aerial vehicle is remarkably improved.

The invention belongs to the technical field of damage detection, and particularly relates to a structural damage detection method based on voiceprint recognition, which comprises the following steps: step 1, a detection system is arranged and applied to a server, and the detection system is composed of a knock detection module, an audio anomaly detection module, a damage classification module, a damage rechecking module and a recognition interface module; 2, knocking a to-be-detected structure on site through a knocking detection module; step 3, segmenting all the audio files, and extracting key segments of the audio under the maximum knocking force from the audio files; and 4, training the generative adversarial network model based on optimal architecture search. According to the method, the generative adversarial network algorithm based on optimal architecture search is adopted for data expansion, the generative adversarial network architecture most suitable for voiceprint data characteristics can be automatically searched, the performance of the generative model is optimized, the learning and reproduction capabilities of voiceprint characteristics are improved, and recognition of structural damage is facilitated.

The application discloses a portable multifunctional energy storage and power supply equipment which is expandable and detachable, and belongs to the field of new energy storage and application. The equipment comprises more than one energy storage and power supply unit, and adjacent energy storage and power supply units are arranged in a stacked mode and are electrically connected with each other. The energy storage unit comprises a unit main body, the unit main body comprises more than one module expansion bin, a functional module is arranged in the module expansion bin, the unit main body is provided with a unit control panel at a position corresponding to the module expansion bin, a standard interface terminal is arranged on the unit control panel, a module control panel is arranged in the functional module, a standard interface which is matched with the standard interface terminal is arranged on the module control panel, and the standard interface terminal is electrically connected with the standard interface. The energy storage and power supply equipment has expandability, can integrate different functional modules through a unified standard interface, greatly widens the application range of energy storage products, expands application scenarios, and makes the energy storage and power supply equipment more portable.

The invention discloses an external traditional Chinese medicine composition for treating migraine. The composition is a medicament prepared from the following raw material medicines in parts by weight: 16-30 parts of fresh radish root, 8-15 parts of fresh ginger, 1-3 parts of fresh mint leaf and 0.3-2 parts of clove powder. The pharmaceutical composition is a pure traditional Chinese medicine preparation, aims at the basic pathogenesis that wind-cold is mainly used and orifice clearing is unfavorable when headache attacks, and achieves the purpose of rapidly relieving pain by purging evil to dredge orifices and promoting qi circulation. In the formula, the four medicines of radish root, clove, ginger and mint are adopted, and through the application of a fresh medicine preparation and aromatic resuscitation-inducing medicines and the compatibility of medicines with pungent and warm medicines as main medicines and medicines with pungent and cool medicines as assistant and guide medicines, the aromatic and pungent effects of the medicines can be furthest taken, so that the effects of inducing resuscitation and relieving pain are achieved. Meanwhile, the four medicines in the pharmaceutical composition are all medicinal and edible traditional Chinese medicines, all have application experience of ancient doctors for nasal administration, and the pharmaceutical composition has relatively good safety while being low in price.

The invention provides an organoid cell microcarrier detection method and a model building method thereof, and belongs to the field of organoid cell detection based on computer vision. The method comprises the following steps: deploying a high-precision camera in a micro-fluidic chip sorting area, collecting an organoid cell microcarrier image, and carrying out a series of preprocessing to obtain a standard data set; and constructing a cell microcarrier detection model comprising a multi-layer feature extraction module, a feature fusion module and a target detection module. After model training, channel pruning and lightweight operation are further adopted to optimize the model so as to adapt to embedded equipment with a limited memory. According to the method, automatic identification of the cell microcarrier in the chip is realized, and the detection efficiency and the intelligent level are remarkably improved; meanwhile, a multi-level feature extraction and fusion module is utilized, so that the model can still keep high accuracy and stability under the condition of coping with dense or sparse cell distribution; in addition, the lightweight design ensures that the model can be smoothly deployed on microfluidic equipment with limited computing power.

The application discloses a dynamic residual error mixed expert low-illumination image enhancement method and device and a storage medium, and relates to the technical field of image processing. The method separates and models image structure information and texture information in a latent representation space. Secondly, a mixed expert network and a multi-scale hierarchical denoising network structure are used to realize fine denoising and reconstruction. Finally, a unified model framework is constructed to restore key structures and detail features in low-light images. The application can be widely applied to image enhancement processing in the fields of night monitoring, automatic driving, mobile photography, remote sensing detection and the like.

The invention discloses an abnormal transaction identification method and device based on a multi-model fusion strategy and a readable medium, and the method comprises the steps: obtaining transaction data of different abnormal transaction types of a to-be-analyzed user and a selection sequence of a base model, calculating the weight of each abnormal transaction type by adopting an analytic hierarchy process according to the selection sequence of the to-be-analyzed user to the base model; inputting transaction data into a trained abnormal transaction recognition model, obtaining a first abnormal transaction number corresponding to each abnormal transaction type through a base model of each abnormal transaction type, multiplying the first abnormal transaction number corresponding to each abnormal transaction type by a weight, forming an input sequence, and inputting the input sequence into a meta-model; obtaining a second abnormal transaction number corresponding to each abnormal transaction type; and according to the number of the second abnormal transactions corresponding to each abnormal transaction type and the weight, calculating to obtain the abnormal proportion of each abnormal transaction type. The abnormal transactions of various types can be identified, and the prediction precision can be improved.

The invention discloses a substance identification device, a database calling method, equipment, a medium and a product, the substance identification device is provided with a detection module, a storage module and a data processing module, and the data processing module is configured to perform first similarity comparison on a protein fingerprint spectrum of a to-be-detected object and a standard protein fingerprint spectrum; obtaining a first matching set; verifying a second interference feature in the protein fingerprint spectrum of the to-be-detected object based on the first matching set; if the verification is passed, performing second similarity comparison on the protein fingerprint spectrum of the to-be-detected object and the standard protein fingerprint spectrum to obtain a second matching set; and determining an identification result of the to-be-detected object based on the first matching set and the second matching set. Similarity comparison is directly performed on the protein fingerprints of the to-be-detected object based on the predetermined related characteristics of the protein fingerprints, so that identification complexity is reduced while identification time consumption is shortened, and substance identification efficiency is improved.

The invention provides an industrial automatic quality inspection device based on image recognition and a control method, the industrial automatic quality inspection device comprises a microhardness tester and a control platform, the control platform controls a hardness pressure head on a turntable to rotate to a position right above an object to be tested, controls the hardness pressure head to move downwards, and loads a set test force to the object to be tested, so that a square indentation is generated on the object to be tested; two diagonal lines of the indentation are exposure areas; an objective lens on the turntable is controlled to be arranged right above the indentation, and at the moment, an auxiliary cross line is overlapped with a diagonal line of the indentation and is located at the central position of a diagonal line exposure area; the color of the auxiliary cross line is different from the imaging color of the to-be-detected object and the indentation; according to the RGB value of each pixel point in the collected image, the number of auxiliary pixel points of auxiliary intersection lines located in the diagonal exposure area in the two diagonal directions is counted; determining the length of the two diagonal lines according to the counted number of the auxiliary pixel points; and determining the hardness of the to-be-measured object according to the lengths of the two diagonal lines. The hardness detection efficiency and accuracy are high.

The application provides a test case automatic generation system based on mutation, which analyzes initial test cases, obtains corresponding component structures and field type information according to case subnode data obtained through analysis, and generates new test cases by applying mutation operations with correct semantics according to the type of corresponding subnode data, thus solving the problem that mutation operations damage the format of test cases and cannot generate effective test cases. Meanwhile, the application uses randomization and semantic-based mutation, adopts corresponding semantic mutation strategies according to the type of data, enhances the pertinence and effectiveness of generated test cases, enhances the difference and richness of generated test cases through randomization mutation strategies, and adopts a plurality of mutation strategies designed according to the characteristics of XML and JSON formats, so as to eliminate the difference of XML and JSON format test case generation tasks from different sources to the greatest extent and enhance the universality of the test case generation method.

The present invention provides a low-rank multi-scale multi-modal pedestrian re-identification method, electronic device and storage medium, belonging to the field of deep learning technology; including obtaining pedestrian images containing three modes of RGB, near infrared and thermal infrared, and inputting the pedestrian images of the three modes into a three-branch feature extraction network, respectively obtaining RGB modal features extracted from an RGB camera, NI modal features extracted from near infrared and TI modal features extracted from thermal infrared; using a low-rank multi-scale multi-modal network model to calculate the three modal features separately to obtain multi-modal features; using the multi-modal features to perform pedestrian re-identification on pedestrians, and obtaining pedestrian re-identification results of the pedestrians. The present invention absorbs useful information in other modalities through a multi-scale multi-modal interaction module; and fuses information from the other two modalities through a low-rank multi-modal fusion module, so that the model can more efficiently fuse multiple modal features while maintaining low model complexity.

The invention discloses an import and export enterprise risk assessment method based on machine learning, and the method comprises the following steps: constructing an enterprise risk general situation database according to historical monitoring data, and carrying out the statistics of the total number of all items; generating a time weight and a risk weight of each type of items; calculating the risk value of each risk item for the overall item of the enterprise; dividing risk levels of the enterprises according to the corresponding risk values; calculating the membership degree of the enterprise risk in each risk level; and judging whether the enterprise is in a qualified range according to the membership matrix and feeding back. According to the invention, through historical monitoring data, an enterprise risk general situation database, a time risk double attention module, an enterprise risk assessment module, a risk grade division module, a first-level fuzzy comprehensive evaluation model and a feedback module, risk analysis of a target enterprise is realized; therefore, whether the enterprise has a risk can be judged according to the risk value of the target enterprise.

The utility model discloses a liftable highway traffic incident early warning device and early warning system, and relates to the technical field of road traffic management, the early warning device comprises a telescopic housing, the housing is provided with a control module, a display early warning module, a sound early warning module, a lifting module and a communication module, the control module can display information obtained by the communication module through the display early warning module and the sound early warning module, and controls the lifting module to enable the inner shell and the outer shell to slide in the direction close to each other or away from each other. According to the early warning device, through the control module, the shell is adjusted to ascend and descend, information obtained by the communication module is displayed through the display early warning module and the sound early warning module, the structure is simple, and effective early warning can be achieved. The lifting type road traffic incident early warning system comprises a plurality of early warning devices and a service platform used for sending sound, character and identification early warning information to the early warning devices. The early warning range is reasonable and controllable, and the early warning efficiency is high.

This invention discloses a method and system for multimodal emotion recognition of speech and facial expressions in smart buildings. The method includes: preprocessing collected speech information and facial expression videos, including: extracting MFCC features from the speech information; segmenting the facial expression video into frames and extracting facial expression regions from the facial expression images; inputting the MFCC features into a residual network model to extract high-order emotion features from the speech information; inputting the segmented facial expression images and the extracted facial expression regions into a 3D-CNN model to extract high-order emotion features from the facial expression images; and using an MLP model to perform multimodal fusion of the high-order emotion features of the speech information and facial expression images to identify the emotional state of the person. During the fusion process, a pre-designed loss function is used to complete end-to-end training of the multimodal feature fusion of speech and facial expression images, and to increase the difference between the high-order emotion features of the speech information and facial expression images. This invention can improve the recognition rate of multimodal emotion recognition of speech and facial expression images.

The present invention relates to a use of a composition including cyclo(hispro) (CHP) and parathyroid hormone (PTH) for preventing, improving or treating a bone loss-related disease.

The present invention relates to a power load clustering method, comprising: obtaining power load data to be clustered, preprocessing the obtained power load data; initialization processing; calculating an initial cluster center vector matrix according to the preprocessed data set; allocating the remaining load vectors to the clusters where the corresponding cluster centers are located according to an improved similarity measurement standard to obtain multiple cluster clusters; calculating the average value of the intra-cluster vector of each cluster cluster as the current cluster center vector of the cluster cluster; judging whether the set clustering standard is met according to the current cluster center vector and the corresponding previous cluster center vector, if the judgment is yes, outputting the current clustering result and ending the clustering, otherwise further judging whether the set maximum number of iterations is reached, if the judgment is yes, outputting the current clustering result and ending the clustering, otherwise returning to continue the iterative calculation. Compared with the prior art, the present invention can effectively improve the accuracy of power load curve clustering.

The invention relates to the related field of carbon emission, in particular to a carbon emission influence factor acquisition method and system, and the method comprises the steps: carrying out the reduction of original carbon emission influence factor data through employing a neighborhood rough set attribute reduction algorithm, and obtaining a primary target influence factor; screening the primary target influence factors by adopting an improved principal component analysis method to obtain target influence factors; wherein the improved principal component analysis method comprises a correlation analysis step and a principal component analysis step, the correlation analysis step is used for performing correlation analysis on the primary target influence factors to obtain a target correlation matrix, and the principal component analysis step is used for performing correlation analysis on the primary target influence factors; and the principal component analysis step is used for performing principal component analysis on the target correlation matrix to obtain target influence factors. The method can solve the problems that in the prior art, carbon emission influence factors are large in similarity and redundancy, potential key factors are ignored, and recognition is inaccurate.

The invention relates to a welcome lamp system of a vehicle and a control method thereof. The welcome lamp system of the vehicle comprises a welcome lamp, a control unit and a touch switch, wherein the welcome lamp is arranged on the vehicle and is provided with a driving circuit; the control unit is electrically connected to a driving circuit of the welcome lamp; the touch switch is arranged on a vehicle body and electrically connected to the control unit, and the touch switch is configured to generate an electric signal when being touched and output the generated electric signal to the control unit; the control unit is configured to control the light emitting mode of the welcome lamp according to the electric signal received from the touch switch.

Embodiments of the present disclosure disclose an information interaction method and device, electronic equipment, medium and product. The method comprises: obtaining accessory identification information corresponding to a to-be-identified accessory when the to-be-identified accessory is attached to a heating smoking set; determining an accessory verification result corresponding to the to-be-identified accessory according to the accessory identification information and a preset verification mode; and in the case that the accessory verification result is verified, retrieving interaction information corresponding to the to-be-identified accessory according to the accessory identification information; wherein the interaction information is used to present the interaction feedback effect of the to-be-identified accessory in the working state. The technical scheme realizes the digital and intelligent linkage between the heating smoking set and the accessory, upgrades the traditional accessory with only physical function to the intelligent accessory that can provide exclusive interaction feedback, and effectively improves the differentiated and immersive use experience of the user for the heating smoking set.