483results about How to "Rich in features" patented technology

A system and method are disclosed for transforming catalog data from multiple supplier sources to a standardized rich content catalog either by the suppliers themselves or by a third party using the system and method of the present invention. Incoming raw catalog data content is cleansed and normalized using an extensive knowledge base of patterns and incoming schemas are appended to the cleansed and normalized data. The resulting rich content catalogs are published for user browsing and data syndication. Users are administered to form groups for purposes of shopping, product pricing, and access authorization.

A distributed user interface (UI) system includes a client device configured to render a UI for a server-based application. The client device communicates with a UI server over a network such as the Internet. The UI server performs formatting for the UI, which preferably utilizes a number of native UI controls that are available locally at the client device. In this manner, the client device need only be responsible for the actual rendering of the UI. The source data items are downloaded from the UI server to the client device when necessary, and the client device populates the UI with the downloaded source data items. The client device employs a cache to store the source data items locally for easy retrieval.

The invention discloses an attention CNNs and CCR-based text sentiment analysis method and belongs to the field of natural language processing. The method comprises the following steps of 1, training a semantic word vector and a sentiment word vector by utilizing original text data and performing dictionary word vector establishment by utilizing a collected sentiment dictionary; 2, capturing context semantics of words by utilizing a long-short-term memory (LSTM) network to eliminate ambiguity; 3, extracting local features of a text in combination with convolution kernels with different filtering lengths by utilizing a convolutional neural network; 4, extracting global features by utilizing three different attention mechanisms; 5, performing artificial feature extraction on the original text data; 6, training a multimodal uniform regression target function by utilizing the local features, the global features and artificial features; and 7, performing sentiment polarity prediction by utilizing a multimodal uniform regression prediction method. Compared with a method adopting a single word vector, a method only extracting the local features of the text, or the like, the text sentiment analysis method can further improve the sentiment classification precision.

A distributed user interface (UI) system includes a client device configured to render a UI for a server-based application. The client device communicates with a UI server over a network such as the Internet. The UI server performs formatting for the UI, which preferably utilizes a number of native UI controls that are available locally at the client device. In this manner, the client device need only be responsible for the actual rendering of the UI. The source data items are downloaded from the UI server to the client device when necessary, and the client device populates the UI with the downloaded source data items. The client device employs a cache to store the source data items locally for easy retrieval.

A distributed user interface (UI) system includes a client device configured to render a UI for a server-based application. The client device communicates with a UI server over a network such as the Internet. The UI server performs formatting for the UI, which preferably utilizes a number of native UI controls that are available locally at the client device. In this manner, the client device need only be responsible for the actual rendering of the UI. The source data items are downloaded from the UI server to the client device when necessary, and the client device populates the UI with the downloaded source data items. The client device employs a cache to store the source data items locally for easy retrieval.

The invention discloses a method and a server for synchronous playing and a player. The method comprises the following steps: from a plurality of target multimedia apparatuses, a random apparatus is picked as a master control apparatus and the total number of the target multimedia apparatuses is configured on the master control apparatus; a player of each apparatus is connected to a synchronous playing server on the master control apparatus and sends a playing starting request message to the synchronous playing server upon a successful connection; and when the synchronous playing server receives the playing starting request messages from all the players, the synchronous playing server determines a current playing time curPlayTime according to a current playing system clock curSysTick maintained by itself and sends playing starting clock messages, carrying the curPlayTime, to all the players, which receive the playing starting clock messages and start to play multimedia files synchronously according to the curPlayTime in the messages. The method improves the precision of multi-player synchronous playing.

The invention provides a face detecting and tracking method and a device. The method comprises the steps of inputting a face image or a face video, preprocessing the face image or the face video in an illumination mode, detecting a face by usage of an Ada Boost algorithm, confirming an initial position of the face, and tracking the face by the usage of a Mean Shift algorithm. According to the face detecting and tracking method and the device, a self-adaptation local contrast enhancement method is provided to enhance image detail information in the period of image preprocessing, in order to increase robustness under different illumination conditions, face front samples under different illumination are added to training samples and accuracy of the face detection is increased by adoption of the Ada Boost algorithm in the period of face detection, in order to overcome the defect that using color of the Mean Shift algorithm is single, grads features and local binary pattern length between perpendiculars (LBP) vein features are integrated by adoption of the Mean Shift tracking algorithm in the period of face tracking, wherein the LBP vein features further considers using LBP local variance for expressing change of image contrast information, and accuracy of the face detection and the face tracking is improved.

The invention discloses a fused attention model-based Chinese text classification method. The method comprises the following steps of: respectively segmenting a text into a corresponding word set anda corresponding character set through word segmentation preprocessing and character segmentation preprocessing, and training a word vector and a character vector corresponding to the text by adoptionof a feature embedding method according to the obtained word set and character set; respectively carrying out semantic encoding on the word vector and the character vector by taking a bidirectional gate circulation unit neural network as an encoder, and obtaining a word attention vector and a character attention vector in the text by adoption of a word vector attention mechanism and a character vector attention mechanism; obtaining a fused attention vector; and predicting a category of the text through a softmax classifier. The method is capable of solving the problem that more redundant features exist in the classification process as existing Chinese text classification methods neglects character feature information of texts, the extracted texts are single in features, all the pieces of semantic information of the texts are difficult to cover and features having obvious contribution to the classification are not focused.

The invention provides an automatic text abstraction method based on a pre-training language model. According to the method, an ultra-large-scale unsupervised Chinese corpus is used for training a complex deep language model; the low-layer network structure of the model can extract and retain the grammar and structure information of a text, and the high-layer network structure can extract and retain the semantics and context information of the text, so that the richer text features and semantic information are provided for an automatic text summary task; a pre-training language model and an Encoder are combined to realize, the text features and the semantic information in the pre-training language model are fully utilized, so that a better semantic compression effect is provided, and the performance of an automatic text abstract is improved; a pre-training language model and a decoder are combined, not only the semantics in an original text are considered in the text generation process, but also the semantic information of vocabularies is also considered, so that the readability of the generated text and relevance with the original text are improved, and the performance of an automatic text abstract is improved.

The invention designs a driving path planning method based on the state grid method. Firstly, according to the external environment information and the initial global reference trajectory of the self-driving vehicle, the vehicle motion state and the surrounding environment information of the vehicle are detected by the vehicle-mounted equipment, and the surrounding environment information of the vehicle is updated in real time. Then, the relevant indexes reflecting the driver's personalized trajectory are extracted through the driving simulator experiment, such as the driver's trajectory preview time, the lateral distance compensation value of the lane centerline and the minimum lateral distance value of the distance obstacle when the vehicle is traveling, and the path search strategy is determined according to the above related indexes. Finally, the preview point state of the local trajectory is determined and the local reference trajectory is generated based on the state grid method.The driving path planning method of the invention comprehensively considers the personalized differences of different drivers in the driving process, and the factors considered are more comprehensive, and can provide support for the personalized development of the automatic driving vehicle.

A system used for sending out information of advertisement comprises obtaining unit of input information, user library, information library, advertisement library, character-processing unit and information-sending unit. The method for sending out information of advertisement is also disclosed.

The invention discloses a product packaging surface defect detection and classification method based on machine vision. The method comprises the steps of 1, acquiring a high-definition color image of defect-free product packaging, making the high-definition color image into a standard image, conducting real-time shooting with a camera, conducting online acquisition of a high-definition color image of product packaging to be detected, and making the high-definition color image as an image to be detected; 2, conducting image matching on the image to be detected and the standard image based on SURF algorithm; 3, conducting difference image operation on the two images matched in step 2 to obtain a defective image; 4, conducting feature extraction on the defective image to obtain the geometrical features and color features of the defective image; 5, classifying product packaging surface defects by means of RBF neural network algorithm. Automatic defect detection and classification are conducted by means of a machine vision system, human factor interference can be avoided, labor cost is reduced greatly, and then huge hidden cost caused by training and management when artificial detection is adopted is avoided.

The invention discloses a cellulose composite microsphere, a preparation method thereof and application. The particle size of the cellulose composite microsphere is 1mum-1200mum, the specific surface area of the cellulose composite microsphere is 100m<2> / g-500m<2> / g, and the aperture of the cellulose composite microsphere is 200nm-900nm. The preparation method comprises the following steps of: preparing a mixture solution of cellulose and a composite material by using the cellulose as a matrix, a polymer material as composite material and a water solution of alkali / urea or alkali / thiourea as a solvent, and obtaining the cellulose composite microsphere through sol-gel phase transition and cross-linking agent cross linkage. The advantages of cellulose microsphere are reserved by the cellulose composite microsphere, and due to the adding of the composite material, new functional groups are simultaneously added for the surface of the microsphere and further decoration and modification are facilitated. An organic solvent used in the preparation method disclosed by the invention can be reused, the whole preparation technology is simple, low in time consuming, undemanding for equipment and convenient for industrial production, and the prepared cellulose composite microsphere has a good flow property and mechanical property and a wide application.

A computer implemented method of detecting a fault in a system comprises the steps of executing at least three virtual machines, each virtual machine executing a same application software, in separated and isolated memory segments and in a dedicated core of a multi-core processor; the virtual machines being synchronized and concurrently executed by a common hypervisor; wherein non-faulty virtual machines provide an identical output message within a predefined time-interval; detecting a fault in an output of a virtual machine, the fault corresponding to a different output message of the faulty virtual machine. Developments include a distributed vote mechanism, pull / push mechanisms, association of output vote messages with a safety extension comprising identification information, virtual machine recovery using data context.

A distributed user interface (UI) system includes a client device configured to render a UI for a server-based application. The client device communicates with a UI server over a network such as the Internet. The UI server performs formatting for the UI, which preferably utilizes a number of native UI controls that are available locally at the client device. In this manner, the client device need only be responsible for the actual rendering of the UI. The source data items are downloaded from the UI server to the client device when necessary, and the client device populates the UI with the downloaded source data items. The client device employs a cache to store the source data items locally for easy retrieval.

A distributed user interface (UI) system includes a client device configured to render a UI for a server-based application. The client device communicates with a UI server over a network such as the Internet. The UI server performs formatting for the UI, which preferably utilizes a number of native UI controls that are available locally at the client device. In this manner, the client device need only be responsible for the actual rendering of the UI. The source data items are downloaded from the UI server to the client device when necessary, and the client device populates the UI with the downloaded source data items. The client device employs a cache to store the source data items locally for easy retrieval.

The invention provides a face verification anti-counterfeit recognition method and a system thereof based on an interactive action. The method comprises a step of carrying out the initial recording of the information of an register static face image and the information multiple register face action images, a step of waiting the reading of a static face image to be read, matching a character and a stored character when the shooting of static face image to be detected obtained by a user to be verified is detected, and if a matching degree reaches the storage characteristic of a preset threshold value, conforming to a verification requirement, and a step of randomly selecting and prompting the user to be verified to complete a corresponding face action according to the recorded face action, extracting the characteristic of the action image to be detected of the user be verified, matching the characteristic with the historical verification feature information of a corresponding face action, completing face identity verification if a matching rate reaches a preset threshold value, adding the matched image into the historical verification feature information, returning to select a next face action to continue matching if the matching is not approved or does not reaches a desired effect, treating the verification as a failure if the action exceeds a preset number of times, and ending the verification.

The invention discloses a fault classification method based on a one-dimensional multi-path convolutional neural network, and belongs to the technical field of industrial process monitoring. Accordingto the method, a traditional two-dimensional convolutional neural network is improved; multiple paths of parallel one-dimensional convolutional neural networks are applied in the variable direction;the cross correlation among variables is deconstructed, the variables are independently subjected to convolution and pooling to extract time sequence feature information, the extracted features are more diversified, the robustness is higher, the defect that a traditional two-dimensional convolutional neural network is sensitive to the prior arrangement order of the variables in input data is overcome, and the method is more suitable for complex and high-order industrial process data; experiments show that the fault classification model obtained based on the one-dimensional multi-path convolutional neural network training provided by the invention can effectively perform fault classification of industrial process data, and has higher generalization capability compared with a common model.

The invention discloses a device for automatically carefully sorting and grading shrimps. The device comprises a feeding system, a sorting channel, an image collecting system, a grading system and an image processing system; the feeding system is used for outputting to-be-sorted shrimps to the sorting channel in a single row; the sorting channel consists of a buffering channel and a sampling channel which are sequentially connected, the sampling channel is divided into a plurality of single-row channels, the buffering channel is internally provided with a plurality of direction adjusters, and used for separating raw material shrimps, output by the feeding system, into singles, and adjusting the posture of each raw material shrimp entering single-row channel; the image collecting system is used for collecting images of the raw material shrimps; the grading system comprises spray nozzles corresponding to single-row channels, and the spray nozzles are controlled by the image processing system and used for blowing the shrimps to enter different receiving tanks; the image processing system is used for analyzing the images, grading the image of each shrimp, and emitting the signal for controlling the grading system according to the graded results. The invention further discloses a method for automatically carefully sorting and grading shrimps.

The invention relates to a front multi-vehicle tracking method integrating a millimeter radar and depth learning vision. The millimeter radar is used for obtaining front data information, according tothe echo reflection strength and width information, invalid information is removed, and only front vehicle information remains. According to the method integrating the millimeter radar and a camera,by filtering radar information and generating a motion trajectory through an online tracking model, and trajectory correlation is performed. The front vehicles related to the trajectory are recorded and numbered. According to the generated trajectory and the numbered front vehicles, the steps just need to be repeatedly executed on data of the next period, consistency checking is performed, and thedata is added into the numbered trajectory. For newly appearing vehicles, trajectory generation, trajectory collection and numbering are performed according to the beginning steps. By combining advantages of the millimeter radar and the vision depth learning, and the target tracking accuracy and robustness for front multiple vehicles can be effectively improved.

The invention provides an open quantitative analysis method based on a medical image. The method comprises the steps that 1, medical image data is acquired, focus tissue in the medical images is segmented, automatic or auxiliary positioning and tumor extraction are conducted on the focus tissue, and segmentation of the diseased part is achieved; 2, feature extraction is conducted on the segmented focus tissue, image features are mined, and a focus tissue image feature database is built; 3, a model for assisting in diagnosis, prognosis and prediction precision is built on the basis of the focus tissue image feature database by combining clinical information and focus image features of a patient through a computer analysis algorithm. According to the analysis method, by applying a big data analysis method to image omics, diagnosis of a doctor is better assisted.

The invention discloses a network attack detection method, device and equipment and a storage medium, and belongs to the technical field of networks. The invention provides a scheme for detecting a network attack based on multi-dimensional features. An integrated learning mode is adopted. Through the multiple classification models, whether the data flow is abnormal or not is judged according to the features of the multiple dimensions, then the classification results of the multiple dimensions are summarized to comprehensively judge whether the network attack occurs or not, the considered features are more diversified, the method can adapt to diversified service forms of an existing network, and the accuracy of detecting the network attack is improved.

The invention provides an aerially-photographed vehicle real-time detection method based on deep learning and mainly aims to solve the problem that in the prior art, it is difficult to perform precisedetection on an aerially-photographed vehicle target under a complicated scene on the basis of guaranteeing instantaneity. The method comprises the implementation steps that 1, an aerially-photographed vehicle dataset is constructed; 2, a multi-scale feature fusion module is designed, and a RefineDet real-time target detection network based on deep learning is optimized in combination with the module, so that an aerially-photographed vehicle real-time detection network is obtained; 3, a cross entropy loss function and a focus loss function are utilized to train the aerially-photographed vehicle real-time detection network in sequence; and 4, a trained detection model is used to detect a vehicle in a to-be-detected aerially-photographed vehicle video. According to the method, the designedmulti-scale feature fusion module can effectively increase the information utilization rate of the aerially-photographed vehicle target, meanwhile, the aerially-photographed vehicle dataset can be trained more sufficiently by use of the two loss functions, and therefore the detection accuracy of the aerially-photographed vehicle target under the complicated scene is improved.

The invention discloses a living body face detection method based on WLD-TOP (Weber Local Descriptor-Three Orthogonal Planes). The method comprises the following steps: (1) training stage: reading a training set video, performing face region detection on each frame, converting the frames into a gray level facial image frame sequence to construct a three-dimensional image matrix, constructing a filtering template, calculating WLD features, generating a WLD-TOP feature vector, and inputting the feature vector into an SVM classifier for training to establish an SVM model; and (2) testing stage: for an image sequence under test, performing face detection on each frame, converting the frames into a gray level facial image sequence, constructing a three-dimensional image matrix and a filtering template, calculating WLD features, generating a WLD-TO feature vector, and feeding the WLD-TOP feature vector into a trained SVM model to obtain a living body face detection result. The Weber law is adopted on the basis of the LBP-TOP, so that the size relationships between neighborhood pixels and a center pixel are reflected, and the differences between the neighborhood pixels and the center pixel are quantified. Thus, the features of a descriptor are more complete.

The invention provides a CT image segmentation system based on an attention convolutional neural network, and the system comprises a feature coding module which uses a parallel convolutional neural network to gradually reduce the size of a feature map of an input image, and achieves the simultaneous extraction of image semantic information and spatial information through the multiplexing of a network layer and the interception and fusion of features of all layers; the semantic information extraction attention module which is used for generating attention features by pooling and further refining the semantic information features extracted by the feature coding module; the feature fusion pooling attention module which is used for fusing the refined semantic information features with the semantic information and spatial information features spliced by the feature coding module to form an attention feature map by using parallel connection of maximum pooling and average pooling; and the feature map decoding module which is used for gradually and finely restoring the attention feature map into the size of the input image by using a convolution module and an up-sampling module. Accordingto the invention, by fusing the attention module, efficient and accurate image segmentation is realized.

The invention discloses a sentence semantic similarity calculation method which includes the steps: extracting characteristics of a first sentence and a second sentence by a deep learning model to obtain a first sentence global semantic vector and a second sentence global semantic vector; extracting characteristics of words of the first sentence and the second sentence through characteristic engineering to obtain a first sentence local semantic vector and a second sentence local semantic vector; splicing the first sentence global semantic vector and the second sentence global semantic vector with the first sentence local semantic vector and the second sentence local semantic vector to obtain a first sentence one-dimensional characteristic vector and a second sentence one-dimensional characteristic vector; calculating the vector distance between the first sentence one-dimensional characteristic vector and the second sentence one-dimensional characteristic vector to obtain similarity between the first sentence and the second sentence. Sentence characteristics extracted by the method are more comprehensive and deeper and have certain pertinence, and calculated similarity is higher in accuracy.

The invention discloses a prosodic hierarchy labeling method. The method comprises the steps as follows: acquiring to-be-labeled text data and audio data which have a corresponding relation; extracting a to-be-labeled text characteristic set of each word according to the to-be-labeled text data; extracting an acoustic characteristic set of each word according to the audio data; acquiring a prosodic hierarchy structure by a prosodic hierarchy labeling model according to a word identification of each word, the to-be-labeled text characteristic set of each word and the acoustic characteristic setof each word. The invention also discloses a model training method, a prosodic hierarchy labeling device and a model training device. The prosodic hierarchy labeling model is established by combination of text characteristics and acoustic characteristics, richer characteristics can be provided for prosodic hierarchy labeling, the prosodic hierarchy labeling accuracy can be improved, and the voicesynthesis effect can be enhanced.

The invention discloses a lithium ion battery health state estimation method based on a codec model, which comprises the following steps of: (1) acquiring battery charging and discharging period dataincluding terminal voltage and current data and the maximum discharging capacity in each charging and discharging period; and (2) constructing a codec model of an attention mechanism according to thecharacteristics of the acquired data, wherein the codec model comprises an encoder and a decoder, voltage and current values are used as encoder input, an SOH estimated value of the battery is obtained and used as decoder output, and the number of nodes of each layer is determined; (3) preprocessing and normalizing the data acquired in the step (1), inputting the data into the codec model with randomly initialized weight, and minimizing the output error of the codec model through Adam algorithm training; and (4) inputting a new test sample into the codec model trained in the step (3), and calculating a prediction error so as to evaluate the accuracy of model prediction.

The invention provides a compressed multi-scale feature fusion network-based image super-resolution reconstruction method. The invention aims to solve a technical problem that a reconstructed high resolution image has a low peak signal to noise ratio and low structural similarity in the prior art. The implementation process of the invention includes the following steps that: a training sample setcomposed of high- and low-resolution image pairs is obtained; a multi-scale feature fusion network is constructed; the multi-scale feature fusion network is trained; a compressed multi-scale feature fusion network is obtained; and the compressed multi-scale feature fusion network is adopted to perform super-resolution reconstruction on an RGB image to be reconstructed. According to the compressedmulti-scale feature fusion network-based image super-resolution reconstruction method of the invention, a plurality of multi-scale feature fusion layers which are connected with one another sequentially in a stacked manner in the multi-scale feature fusion network are adopted to extract the multi-scale features of low-resolution images, and nonlinear mapping is performed on the multi-scale features of the low-resolution images; and therefore, the improvement of the low peak signal to noise ratio and low structural similarity of the reconstructed high-resolution image can be benefitted. The method can be applied to fields such as remote sensing imaging, public safety, medical diagnosis.