33results about How to "Add valid information" patented technology

The invention discloses an instant messaging information processing method and device. The method comprises the following steps: obtaining the communication intimacy of an object with each member in the target group; setting the display priority of group information in the target group according to the communication intimacy, wherein the display priority of the group information sent by a group member with high communication intimacy is higher than the display priority of the group information sent by the group member with low communication intimacy; and displaying the group information on the basis of the display priority. Received information is ranked according to the intimacy of the object with other members, so as to prevent interference information and further improve user experience.

The invention provides a geographical location information-based method and a geographical location information-based system for outputting an advertisement reminding message, belonging to the technical field of communication. The geographical location information-based method comprises the following steps of: obtaining current geographical location information of a terminal in which a first user and / or a second user are or is when the condition an instant messaging message is sent to the second user by the first user is collected; obtaining advertisement data matched with the geographical location information according to the current geographical location information of the terminal in which the first user and / or second user are or is, and generating an advertisement reminding message according to the matched advertisement data; and pushing the matched advertisement reminding message of the terminal in which the second user is to a client end of the first user for outputting; and / or pushing the matched advertisement reminding message of the terminal in which the first user is to a client end of the second user, and outputting reminding information as the instant messaging message sent to the second user by the first user. By using the invention, the advertisement information and the geographical location of the users are combined, and thus the attraction of the advertisement is enhanced to ensure that better advertisement effect is achieved.

The invention discloses a tricholoma matsutake fast nondestructive testing system and method based on a convolutional neural network. The system comprises a deep learning convolutional neural networkmodel, a control end and a consumer terminal; the deep learning convolutional neural network model comprises sample collection, data acquisition, and deep learning convolutional neural network modeling and optimization; in the sample collection, a sample set is established by finishing sample screening of a detection object, and the sample set is divided into a training set, a verification set anda test set; the data acquisition comprises sample chemical content measurement and spectral data acquisition; in the deep learning convolutional neural network modeling and optimization, preprocessedspectral data and corresponding chemical content are modeled through the deep learning convolutional neural network model and pooling processing; a detection result of the tricholoma matsutake by thedeep learning convolutional neural network model is stored in the control end; and the consumer terminal can obtain detection data of the tricholoma matsutake by accessing the control end. Accordingto the tricholoma matsutake fast nondestructive testing system and method, the detection cost can be effectively reduced, and the market supervision of the supervision department can be facilitated.

The invention discloses a rehabilitation training motion state monitoring method and system fusing electrocardiogram and myoelectricity characteristics. The method comprises the following steps: obtaining an electrocardiogram signal and a myoelectricity signal and performing characteristic extraction respectively; performing characteristic fusion based on a random fusion coefficient vector to obtain a fusion characteristic vector; learning a multi-classification motion state recognition model based on a support vector machine by adopting the fusion characteristic vector; performing iterative optimization on the fusion coefficient vector by adopting a particle swarm algorithm to obtain an optimal fusion coefficient vector and a multi-classification motion state recognition offline model; and performing motion state recognition based on the fusion characteristic vector of the electrocardiogram signal and the myoelectricty signal collected in real time. According to the invention, electrocardiogram and myoelectricity signals are fused for motion state detection, so accuracy and reliability are greatly improved, and real-time monitoring of the motion state of rehabilitation training ofa patient is achieved; the motion state can be used as the basis for real-time adjustment of rehabilitation training task intensity and a control strategy, the rehabilitation training effect is enhanced, and secondary injuries caused by excessive training are avoided.

The invention discloses a scanning picture matching method and a scanning picture matching device, which belongs to the field of image processing and are used for solving the problems of poor accuracy and low efficiency of scanning picture matching in the prior art. The scanning picture matching method comprises the following steps of: performing binarization processing on a scanning picture according to color information of the scanning picture; dividing the scanning picture subjected to binarization processing into a plurality of pixel blocks, and performing OR operation on pixel values corresponding to all pixels in each pixel block to obtain a lower sampling picture of the scanning picture, wherein the sampling picture is composed of OR operation results of each pixel block; dividing the obtained lower sampling picture into a plurality of pixel regions, and summing the OR operation results of all pixel points in each pixel region to obtain characteristic information of each pixel region forming the scanning picture; and further calculating similarity of the scanning picture and a picture to be matched, and judging whether the scanning picture is matched with the picture to be matched or not.

The invention relates to an image display method and an image display system based on streetscape attribute information. The image display method comprises the following steps: step a, increasing a streetscape object labeling point in a streetscape image, establishing a streetscape space-attribute database with the streetscape image taking space longitude and latitude and image row number as labels; step b, storing a vector diagram corresponding to the streetscape target into an attribute value, in the streetscape space-attribute database, of the vector diagram, associating the vector diagram of the streetscape target with the space longitude and latitude and the image row number of a labeling point pixel of the streetscape target; step c, selecting the vector diagram in the streetscape space-attribute database, and drawing the vector diagram of the streetscape object on a corresponding streetscape object labeling point position. The image display method is beneficial to improving the navigation ability of the streetscape and improving the experience effect.

The invention discloses a 3D face identification method under local shielding conditions based on radial lines. The method comprises steps that (1), the radial lines emitted from a nose tip point areextracted, and the part of the radial lines above the nose tip point is uniformly sampled; (2), local areas among sampling points are extracted, and a local characteristic descriptor is further constructed; (3), a cost function of the characteristic descriptor is constructed, and similarity vectors having local matching characteristics and corresponding to the radial lines is generated; (4) the similarity vectors corresponding to the radial lines are matched to acquire similarity values, sum of the similarity values of all the radial lines is a similarity value of two 3D faces, so face identification is accomplished. The method is advantaged in that through intercepting and sampling the radial lines, the local areas among the sampling points are extracted, the local characteristics are further extracted, the effective information of the radial lines is enhanced, and thereby accuracy of 3D face identification under the local shielding conditions is improved.

A normalized radar screen monitoring method includes the steps that S1, the model of a to-be-monitored radar and icon information corresponding to a PPI of the model are determined; S2, according to the determination condition, corresponding radar model information, the icon information and radar image and character data information are extracted from a radar data storage unit; S3, by means of anerected external video data acquisition unit, PPI image videos and PPI display data of the to-be-monitored radar are collected in real time; S4, the collected PPI image videos are subjected to noise removal and de-jittering, image information of a non-PPI part is deleted, target monitoring and target tracking are carried out, and to-be-monitored target data is output; OCR is carried out on the collected PPI display data, and recognition data is output; S5, data fusion is carried out on the target data and the recognition data output in S4 and the information extracted in S2, and a virtual radar image and character report is output and displayed. The technical problem about unified platform display of various kinds of radar system information data is solved, and more visual radar information data is provided for a command and control center.

The invention belongs to the technical field of electronic maps, and particularly relates to a positioning image display method and system based on street view attribute information. The positioning image display method based on street view attribute information comprises the following steps: a, adding street view target marking points into a street view image, and establishing a street view space-attribute database taking the spatial longitude and latitude and the row number of the image as marks for the street view image; b, drawing an attribute image of a relevant street view target on each marking point of the street view image, and storing the attribute images and parameters thereof in a relevant record of the street view space-attribute database; c, inquiring through spatial positions or attribute values of the street view space-attribute database under various conditions, determining a street view object marking point, and selecting a relevant attribute image in the street view space-attribute database for displaying. By adopting the positioning image display method, the street view navigation capability is favorably enhanced, the experience effect is improved, and high-added-value applications such as inquiry, search, navigation and advertisement can be performed.

The invention provides a formula identification method and device. The formula identification method comprises the following steps: performing distortion correction on image layout; segmenting basic elements in the corrected image layout, and determining a region where the basic elements are positioned as a formula region according to the features of the basic elements in the image layout; performing formula identification on the formula region according to a formula symbol. By adopting the formula identification method and device, a plurality of local features in a short text image can be realized, a formula in the short text image is detected and identified, valid information in the short text image can be effectively enhanced, random codes are reduced, and the accuracy of a whole answering system can be further increased.

The invention provides a multi-attention feature fusion speaker recognition method, which comprises the following steps of: constructing a deep speaker characterization model which comprises a feature extractor and a speaker classifier, taking Fbank features as input of the deep speaker characterization model, extracting the Fbank features as speaker characterization through the feature extractor, in a training stage, using a speaker classifier to map the speaker representation to a speaker tag, and constructing a loss function to optimize a deep speaker representation model; and in a test stage, comparing similarity between speaker representations by adopting a cosine distance, and judging whether the speakers are the same speakers or not according to a threshold value. According to the method, weighted fusion is performed on the features of different branches through multi-attention feature fusion, and the multi-attention feature fusion comprises a space attention mechanism and a channel attention mechanism, so that effective information in each branch is enhanced, and speaker recognition performance with higher robustness is obtained.

The invention discloses a small sample remote sensing image scene classification method based on iterative feature distribution learning. The method comprises the following steps: step 1, setting an iterative feature distribution learning network framework; step 2, generating a similarity matrix between the image samples; step 3, training a small sample classifier by utilizing guide knowledge based on classification prediction probability and a similarity matrix; and step 4, correcting the feature distribution by using the prediction probability distribution matrix and the attention weight, and then iteratively updating the whole network. The method is simple in structure and reasonable in design, firstly, the similarity matrix is generated to obtain feature association between samples, secondly, the classifier can be trained under the guidance of historical classification prediction, finally, the attention mechanism is adopted to correct the features, and the corrected features are input into the network again to realize iterative updating of the whole network. According to the method, through iterative feature distribution learning, the characterization capability of the category can be further improved, so that the small sample classification accuracy is improved.

The invention discloses a mode wavefront restoration method based on a neural network, and the method comprises the steps: extracting the mass center offset and second-order moment information of a light spot, and fitting a nonlinear relation between the mass center offset and second-order moment of the light spot and a Zernike coefficient through the neural network; and finally, directly predicting a Zernike coefficient corresponding to the wavefront to be measured from the light spot centroid offset and the second-order moment information through a neural network. Compared with a traditional method, the method has the advantages that the second-order moment information of the light spots is extracted while the mass center offset of the sub-light spots is extracted, effective information in the sub-apertures is increased, the corresponding relation between the light spot information and the Zernike coefficient is fitted through the neural network, the detection precision of the Shack-Hartmann wavefront sensor is improved, and under the same recovery precision, the detection precision of the Shack-Hartmann wavefront sensor is improved. The requirement of a Shack-Hartmann sensor for the number of sub-apertures is reduced, and the method is expected to be used in the fields of dark and weak beacons, high-resolution wavefront measurement and the like.

The invention discloses an enhanced influence diffusion social recommendation method based on GCN, which simulates the process of social recursive dynamic propagation influence based on transitivity and dynamism characteristics of trust, constructs an enhanced influence diffusion model, fuses the trust based on historical interaction records of users and articles into recursive social dynamic modeling, and improves the social recommendation efficiency. The problem that different users recommend different articles due to trust relations is solved. Meanwhile, an attention mechanism is introduced, and different importance is given to users in the social graph, so that the same-order domain weight distribution problem is solved; when the long-distance social relation is calculated in a recursive manner, a residual error connection mode is proposed to reduce the influence of noise; and finally, predicting future behaviors and preferences of the user. According to the method, the user sparsity is reduced by fully utilizing the preferences of neighbors around the user, and a model based on the neural network is designed to simulate the social influence recursive propagation process, so that the recommendation performance is improved.

The invention relates to a generative adversarial network image defogging method fusing multi-level features. The method is based on an end-to-end defogging algorithm of deep learning, during down-sampling, feature extraction is carried out on foggy images with different resolutions, the extracted different feature maps are learned through an SE-ResNet module, information between channels is better fitted, and performance degradation caused by network deepening is prevented. And splicing the learned multi-level feature maps, and fusing more image features. And then, during up-sampling, putting a feature map after down-sampling fusion into an SE module for learning so as to better distribute channel weights and enhance useful features. And splicing the learned feature map with the up-sampling feature map, and fusing more image information. And finally, adding the residual image of network learning and an input foggy image to obtain a final image defogging result. Experimental results show that the image defogging method provided by the invention is better in defogging performance.

The invention relates to an aspect-level sentiment analysis method and system based on gated cavity convolution and graph convolution, and the method comprises the steps: representing context and aspect words in sentences of a to-be-analyzed text through word vectors, and obtaining the embedding of the context and the embedding of the aspect words; inputting the obtained context and aspect word embedding into a gated dilated convolutional neural network to obtain context representation integrated with aspect information; and inputting the context representation integrated with the aspect information into a graph convolutional neural network to obtain the predicted sentiment polarity of the text to be analyzed. According to the method, the gated cavity convolutional network and the graph convolutional neural network are combined, the text language information is effectively utilized, meanwhile, the importance of aspect information in aspect-level sentiment analysis is emphasized, and the accuracy of aspect-level sentiment analysis tasks is improved.

The embodiment of the invention relates to the technical field of big data, and discloses a sensitive data identification method. The method comprises the steps of obtaining information of a data table of a big data platform; pre-judging whether sensitive data exists in the data table or not according to the information of the data table; obtaining operation information of the data table; and verifying the pre-judgment according to the operation information of the data table so as to determine whether sensitive data exists in the data table or not. The invention further provides a sensitive data identification device and a computer readable storage medium. According to the invention, sensitive data existing in a big data platform can be quickly identified.

The present invention provides a Brillouin scattering signal denoising method based on spectrum center alignment. The steps are as follows: 1) Convert the Brillouin spectrum signal collected by the Brillouin optical time domain sensing system into a two-dimensional grayscale image ; 2) Smooth filtering of the two-dimensional grayscale image; 3) Use edge detection to obtain the position where the Brillouin frequency shift changes sharply, and divide the image into different areas along the fiber axis; 4) Obtain the Brillouin frequency of each area Frequency shift position; translate each area to align the Brillouin spectrum centers of different areas to obtain a new image; 5) filter the new image, and move each area in the opposite direction after denoising is completed; 6) fit each position The Brillouin frequency spectrum on the fiber can be used to obtain the Brillouin frequency shift information along the entire sensing fiber, and the temperature and strain distribution information on the fiber can be obtained. The denoising method of the present invention can avoid boundary blurring, enhance the denoising effect, reduce the size of the filter window, and reduce the computational complexity.

The invention discloses a 3D target detection method based on context perception feature aggregation. The method comprises the following steps: S1, extracting a local feature point set of an input point cloud and predicting coordinates and feature offset; s2, generating an optimized search radius according to a feature cluster formed by voting operation; s3, generating semantic features according to the optimized search radius in the step S2; and S4, performing detection frame classification and regression task of the 3D target according to the semantic features in the step S3 to obtain a final output result with a 3D detection frame. Compared with the prior art, the target detection method provided by the invention has the advantages that the precision of a target detection task is improved while the lightweight class of the network is maintained, and an ideal target detection effect is achieved.

The invention discloses a method and device for improving the definition of an ultra-high-definition 4K-resolution VR video, and the method comprises the steps: setting a hotspot spherical surface anda non-hotspot spherical surface on a spherical VR video, enabling the spherical VR video to be unfolded into a rectangular VR video with the resolution of M * H, and enabling the hotspot spherical surface and the non-hotspot spherical surface to be respectively mapped to a hotspot region and a non-hotspot region of the rectangular VR video; processing the resolution of the picture of the non-hotspot area to obtain a rectangular VR video with the resolution of N * H; and reversely carrying out the above steps to recover and obtain the spherical VR video. According to the method, the space withthe ultra-high-definition resolution is fully utilized, a spherical VR video with the equator resolution higher than 3840 can be supported, and the spherical effective information is increased; the method can be seamlessly compatible with equipment with various standard ultra-high-definition resolutions; only the width of the non-hotspot spherical surface is slightly scaled, so the extra distortion and the image quality reduction degree caused by a full-image stretching or compressing mode are remarkably reduced, and the performance consumption of image rendering is remarkably reduced; and the coherence of the original spherical picture is not damaged.

The invention discloses a child massage healthcare device which comprises a scanning device and an information processing and display device. The scanning device can scan information of dermis tissue, and transmit the information to the information processing and display device. The state of an illness of a child is collected through the scanning device, the information and original set information are subjected to error amplification comparison, a disease database is arranged in a microprocessor, the result of the error amplification comparison is compared with data in a database, and therefore the preliminary characteristic of the state of the illness of the patient is obtained.

The embodiment of the invention discloses a key information extraction and live broadcast content processing method, device and equipment. The scheme comprises the steps of obtaining an original text of to-be-extracted key information; determining an attention weight vector for the original text, and correspondingly generating a first target word according to the attention weight vector; judging whether the first target word is included in the original text or not; if yes, according to the attention weight vector, obtaining the attention weight of the first target word when the first target word is generated, according to the attention weight of the first target word, generating penalty data for the first target word, and according to the penalty data, continuing to generate a second target word, the penalty data is used for reducing the possibility that the continuously generated second target word is the same as the first target word; and generating key information according to the first target word and the second target word.

The invention discloses a fast non-destructive detection system and method for matsutake based on convolutional neural network, including a deep learning convolutional neural network model, a control terminal and a consumer terminal; the deep learning convolutional neural network model includes sample collection, data collection, Deep learning convolutional neural network modeling and optimization; sample collection completes sample screening of detection objects to establish a sample set, and divides the sample set into training set, verification set and test set; data collection includes sample chemical content measurement and spectral data collection ; Deep learning convolutional neural network modeling and optimization Utilize deep learning convolutional neural network model and pooling processing to model preprocessed spectral data and corresponding chemical content; deep learning convolutional neural network model for matsutake The detection results of the matsutake are stored in the control terminal; the consumer terminal can obtain the detection data of the matsutake by accessing the control terminal. The invention can effectively reduce the detection cost, and is beneficial to the supervision department to supervise the market.

The invention provides a geographical location information-based method and a geographical location information-based system for outputting an advertisement reminding message, belonging to the technical field of communication. The geographical location information-based method comprises the following steps of: obtaining current geographical location information of a terminal in which a first user and / or a second user are or is when the condition an instant messaging message is sent to the second user by the first user is collected; obtaining advertisement data matched with the geographical location information according to the current geographical location information of the terminal in which the first user and / or second user are or is, and generating an advertisement reminding message according to the matched advertisement data; and pushing the matched advertisement reminding message of the terminal in which the second user is to a client end of the first user for outputting; and / or pushing the matched advertisement reminding message of the terminal in which the first user is to a client end of the second user, and outputting reminding information as the instant messaging message sent to the second user by the first user. By using the invention, the advertisement information and the geographical location of the users are combined, and thus the attraction of the advertisement is enhanced to ensure that better advertisement effect is achieved.

A normalized radar screen monitoring method includes the steps that S1, the model of a to-be-monitored radar and icon information corresponding to a PPI of the model are determined; S2, according to the determination condition, corresponding radar model information, the icon information and radar image and character data information are extracted from a radar data storage unit; S3, by means of anerected external video data acquisition unit, PPI image videos and PPI display data of the to-be-monitored radar are collected in real time; S4, the collected PPI image videos are subjected to noise removal and de-jittering, image information of a non-PPI part is deleted, target monitoring and target tracking are carried out, and to-be-monitored target data is output; OCR is carried out on the collected PPI display data, and recognition data is output; S5, data fusion is carried out on the target data and the recognition data output in S4 and the information extracted in S2, and a virtual radar image and character report is output and displayed. The technical problem about unified platform display of various kinds of radar system information data is solved, and more visual radar information data is provided for a command and control center.

The invention discloses a three-dimensional face recognition method under partial occlusion based on radial lines. The realization steps of the method are as follows: (1) extract the radial lines emitted from the nose point on the human face, and compare the radial line nose point The above parts are uniformly sampled; (2) extract the local area between the sampling points, and construct the local feature descriptor; (3) construct the cost function of the feature descriptor, and generate the similarity with local matching characteristics corresponding to the radial line (4) Match the similarity vectors corresponding to the radial lines to obtain similarity values, and the sum of the similarity values ​​of all radial lines is the similarity value of two three-dimensional faces, thereby completing face recognition. The present invention intercepts and samples radial lines, extracts local regions between sampling points and extracts local features, increases effective information of radial lines, and improves three-dimensional face recognition accuracy under partial occlusion.

The invention discloses a scanning picture matching method and a scanning picture matching device, which belongs to the field of image processing and are used for solving the problems of poor accuracy and low efficiency of scanning picture matching in the prior art. The scanning picture matching method comprises the following steps of: performing binarization processing on a scanning picture according to color information of the scanning picture; dividing the scanning picture subjected to binarization processing into a plurality of pixel blocks, and performing OR operation on pixel values corresponding to all pixels in each pixel block to obtain a lower sampling picture of the scanning picture, wherein the sampling picture is composed of OR operation results of each pixel block; dividing the obtained lower sampling picture into a plurality of pixel regions, and summing the OR operation results of all pixel points in each pixel region to obtain characteristic information of each pixel region forming the scanning picture; and further calculating similarity of the scanning picture and a picture to be matched, and judging whether the scanning picture is matched with the picture to be matched or not.

The invention discloses a FGR-AM method for remote sensing scene recognition, comprising: performing effective information enhancement processing and invalid information suppression processing on the image features extracted by the third and fifth bottleneck convolution modules; From the image features extracted by the first bottleneck convolution module, the contour information contained in the remote sensing image and the more interesting features in vision are extracted at the same time, and the detailed features contained in the remote sensing image are extracted from the image features extracted by the fifth bottleneck convolution module; Integrate channel attention and spatial attention enhancement features; map multi-dimensional features to orthogonal k-dimensional features to identify and classify remote sensing images. The present invention takes into account the main features and detailed features of the image, extracts and fuses the interested information and detailed information, so that the recognition accuracy of the network is improved, and the network can accurately identify the scene in complex scenes and similar scenes. identification.

The invention discloses a method and device for extracting near-infrared spectrum features. The method comprises: acquiring N samples to be tested; using a spectrometer to acquire near-infrared spectrum data of N samples to be tested; performing preprocessing on the near-infrared spectrum data to obtain Two-dimensional near-infrared spectral smoothing data; two-dimensional near-infrared spectral smoothing data is arranged and converted to obtain four-dimensional spectral data; feature extraction is performed on four-dimensional spectral data; feature extraction is performed on four-dimensional spectral data to obtain two-dimensional Characteristic data; the advantages of the present invention are: the integrity of the data can be guaranteed, feature extraction can be performed in the full spectrum range, and information will not be lost.

The invention discloses an image processing method and device, a storage medium and a terminal. The method comprises the following steps: acquiring an original image; carrying out image blurring processing on a to-be-processed image multiple times to obtain multiple blurred images, wherein the blurring scales of the multiple blurred images are different, and the to-be-processed image is the original image or an image obtained after image denoising processing is carried out on the original image; performing difference processing for multiple times to obtain a plurality of difference images, and selecting two blurred images from the plurality of blurred images for difference calculation in each time of difference processing; performing image fusion processing on the plurality of differential images to obtain an effective information image; and performing enhancement processing on the to-be-processed image according to the effective information image to obtain a processed image. Through the scheme of the invention, effective information in the original image can be enhanced, so that the image quality is improved.