38results about How to "Retain original features" patented technology

The present invention provides a Radviz-based fuzzy clustering result visualization method, comprising the steps of 1) performing data pre-processing on a result of a fuzzy clustering algorithm; 2) designing a reasonable clustering cluster dimension anchor point layout for a RadViz circumference; 3) projecting data samples inside the Radviz in a round dot mode or a pie chart mode; 4) extending dimension anchor points to dimension circular rings, and achieving visual encoding of the Radviz circular rings; 5) fusing membership degree distribution information to a Radviz front view; and 6) extracting coexistence and association relations among clusters, and mapping the coexistence relations by using chord lines. A fuzzy membership degree matrix obtained by the fuzzy clustering algorithm is displayed based on Radviz, as much fuzzy clustering information as possible is provided for users, researchers are allowed to freely explore information such as data attributes, membership degree matrixes and clustering clusters, and therefore the users can make a decision more rapidly, visually and accurately.

The invention provides a graph neural network construction method and an abnormal flow detection method based on a graph neural network. The graph neural network construction method for abnormal flowdetection comprises the steps of S10, obtaining features with correlation and timing in original flow data; S20, converting the features in the step S10 into graph structure data; and S30, constructing a depth map neural network model. The features with correlation and time sequence in the network flow are deeply mined, the graph neural network model is constructed, the related features between the internal fields of the network flow message segment are mined on the basis, and the model with a good classification effect on the flow data is finally obtained through pre-training. For unknown network traffic to be identified, the unknown network traffic to be identified can be quickly analyzed and judged through the trained model only by performing simple session restoration, and possible anomalies in traffic data can be quickly found.

The invention relates to a method and system for calculating magnetic anomaly signal frequency characteristics, and the method comprises the steps: obtaining a proportional relation between an upper limit frequency point of the frequency domain energy of a magnetic anomaly signal and the movement speed of a ferromagnetic target object,an inverse proportion between the upper limit frequency point and the nearest contact path between the ferromagnetic target object and the sensor, and an non-linear relationship between theupper limit frequency point and the magnetic moment direction and the sensitive direction of the sensor through a magnetic dipole magnetic anomaly signal analysis model, a frequency spectrum formula of the magnetic anomaly signal and a control variable method; and accordingto the calculation formula of the upper limit frequency point of the frequency domain energy of the magnetic anomaly signal, quickly estimating the upper limit cut-off frequency of the magnetic anomaly signal in a certain detection scene.

The invention discloses a self-adaptive denoising method for partial discharge signals on the basis of spectral kurtosis and S transformation. The self-adaptive denoising method comprises the following steps of: through a discrete signal rapid spectral-kurtosis diagram and S transformation time-frequency spectrum calculation, rapidly acquiring time-frequency characteristics of the partial discharge signals; by combination of the time-frequency characteristics of the partial discharge signals, optimizing value ranges of all parameters in a partial-discharge-signal matching atom dictionary in a self-adaptive manner; then finding atoms with maximum inner product with noised partial discharge signals as optimum matching atoms from the optimized results of the partial-discharge-signal matching atom dictionary by an MP (Matching Pursuit) algorithm; finally, on the basis of the optimum matching atoms obtained by all iterations, carrying out sparse representation on the original partial discharge pulse signals so as to achieve the purpose of denoising the partial discharge signals. The self-adaptive denoising method disclosed by the invention has the advantages that the accuracy of the denoising result is high, no waveform distortion is caused, no limitation on the number of measured local discharge pulses and the positions of the pulses is formed, and the application is wide.

The invention provides a content management system recognition method and a device, and a storage medium, and aims to improve the effectiveness of a content management system recognition result. The content management system identification method comprises the steps of acquiring webpage feature information from a predicted webpage, and the webpage feature information comprises at least one of webpage response header and response body information, static resource information, webpage directory structure information, webpage association information, webpage error page information, webpage footertechnical support information and a reading guide file; extracting a preset keyword from the webpage feature information corresponding to the predicted webpage and converting the preset keyword intoa corresponding feature vector; and based on the feature vector, determining a webpage type corresponding to the predicted webpage by using a CMS clustering model, the CMS clustering model being obtained by performing clustering iteration by using webpage feature information extracted from a webpage sample.

The invention discloses a non-invasive blood glucose monitoring method, system and equipment based on PPG (photoplethysmography) signals and a medium. The non-invasive blood glucose monitoring methodcomprises the following steps: acquiring to-be-monitored PPG signals; carrying out filtering processing on the to-be-monitored PPG signals, and removing high-frequency components of the to-be-monitored PPG signals; removing baseline drift of the to-be-monitored PPG signals subjected to filtering processing; carrying out segmental processing on the to-be-monitored PPG signals of which the baselinedrift is removed, and dividing the to-be-monitored PPG signals into a plurality of single-period to-be-monitored PPG signals; carrying out characteristic extraction on each of the single-period to-be-monitored PPG signals, and extracting a plurality of signal characteristics; and inputting the all extracted signal characteristics into a pre-trained classifier, carrying out classification and identification, and identifying blood glucose concentration labels corresponding to the to-be-monitored PPG signals.

Disclosed is a method for detecting an abnormal message, comprising: dividing a text of a detected message into a plurality of text segments; obtaining one or more account attributes of each text segment, and determining a publication proportion parameter corresponding to the account attributes of each text segment; determining a first factor corresponding to the account attributes of each text segment according to the publication proportion parameter; determining a second factor of the detected message according to the first factor corresponding to the account attributes of each text segment; and determining according to the second factor of the detected message whether the detected message is an abnormal message. Through the combination of publication account attributes of messages with undifferentiated text segmentation and the use of Bayesian algorithm, batches of junk messages of a microblog account are effectively limited, and the flexibility of junk message processing is improved.

The invention provides a time sequence related network intrusion detection method based on an attention mechanism. The method comprises the the steps of training an SSAE network by using an intrusiondetection data set to obtain trained network flow data to be classified, and performing one-hot coding and standardization processing on the network flow data to be classified to obtain preprocessed network flow data; and inputting the preprocessed network flow data into a classifier based on an attention mechanism of the trained laminated sparse auto-encoder SSAE network, and carrying out classification processing on the network flow data to be classified by the classifier to obtain a network intrusion detection result of the network flow data to be classified. According to the method, a double-layer Bi-GRU network structure added with an attention mechanism is designed as a classifier, so that the method has relatively high classification accuracy and relatively low false alarm rate, andmeanwhile, the training and testing time of the model is greatly shortened.

The invention discloses a signature method suitable for an Android intelligent POS machine application. The method comprises the following steps of receiving a shelving request of any application, calling a signature dynamic library for the application on the premise of not changing a native signature, performing secondary signature reinforcement on the application by adopting a signature certificate of a CA signature, writing an applicable related terminal model into signature data, encrypting the signature data, and shelving the application in an application store, after the terminal downloads the application through the application store, sequentially verifying the public key certificate, the signature data, the terminal model and the native signature, and after all verification processes are passed, installing the application in the terminal. According to the method, the security of the terminal application is improved through secondary signature reinforcement, and meanwhile, the application is signed for the second time in combination with the matched related terminal model, so that the terminal can judge the compatibility of the application and the terminal at the initial stage of signature verification, and the user experience is further improved.

The invention belongs to the technical field of optical materials, and particularly relates to a luminescent material, a preparation method thereof and a multiband optical temperature sensor. The invention provides a luminescent material which comprises an inner core and a shell, and the shell wraps the periphery of the inner core; the inner core comprises a first rare earth ion doped MNbO4 nanocrystal (in the formula, M represents any one element selected from lanthanide elements); the shell comprises a nanocrystal of second rare earth ion doped MNbO4 (in the formula, M represents any one element selected from lanthanide elements); the first rare earth ions and the second rare earth ions are different rare earth ions respectively. According to the luminescent material, the preparation method thereof and the multi-band optical temperature sensor provided by the invention, the technical blank that a luminescent material with multiple spectrums and multi-band sensitive temperature performance does not exist at present can be effectively filled up.

The invention provides a cartoon animation generation method with the combination of a hand drawing style. The method includes following steps: (1) firstly constructing three-dimensional grids and a skeleton subspace model of a role by employing a hand drawing; (2) then creating a group of viewpoint relevant models matched with multi-viewpoint hand drawings with the combination of posture reconstruction and a differential domain grid deformation algorithm; and (3) finally combining the hand drawing style with an input motion sequence through motion annotation and a heuristic generalization algorithm to enable the generated animation to maintain original characteristics of the motion sequence. According to the animation generated by employing the method, the original characteristics of themotion sequence are maintained, the drawing style of hand drawing is reflected, and the method can effectively generate cartoon role animation representing the hand drawing style.

The invention belongs to the field of medical image segmentation, and provides a retinal vascular image segmentation method and system based on differential attention. The method comprises the following steps: acquiring a retinal vascular image; and obtaining a retinal fundus vascular image segmentation result based on the retinal vascular image and a differential attention-based multi-scale residual network, wherein the differential attention-based multi-scale residual network comprises a multi-scale input module, an encoder module, a differential amplification module and a decoder module; the multi-scale input module is used for extracting multi-scale information of the retinal vascular image; the encoder module is used for encoding the multi-scale information; the differential amplification module is used for extracting low-frequency information and high-frequency information of the encoded multi-scale information and then extracting features of the low-frequency information and the high-frequency information; and an attention mechanism is introduced into the decoder module, so that an attention degree of an area needing to be paid attention to is improved, irrelevant areas are inhibited, and the extracted low-frequency and high-frequency features are finally restored to original resolutions.

The invention relates to the technical field of antibacterial coatings, and discloses a sterilization method of a cold-sprayed electric heating coating. The sterilization method comprises the following steps: spraying metal powder or ceramic reinforced metal composite powder on the surface of a matrix by utilizing a cold spraying process to obtain an electric heating coating with the thickness of200-2000 microns; and applying voltage of 8V or below to the outside of the electric heating coating, and electrifying for 1-5 minutes to realize sterilization. According to the sterilization method,metal particles are quickly deposited on the surface of the matrix through the cold spraying process to form a compact coating with high conductivity, an antibacterial agent does not need to be added,quick and efficient sterilization can be realized under the condition of low-voltage electrification, and the pollution of the traditional coating containing heavy metals to the natural environment is effectively avoided; the sterilization method can be applied to the surface of a medical operating table, the surface of an infusion table and the surface protection of ocean detection equipment.

The invention discloses a ball bearing fault determination method and device and a storage medium, and relates to the technical field of rolling bearings. The fault determination method of the ball bearing comprises the following steps: acquiring a vibration signal of the ball bearing, and performing de-noising processing on each time period signal in the vibration signal based on a wavelet packet decomposition method to obtain each time period signal after de-noising processing; performing time-frequency analysis on the signal of each time period by using short-time Fourier transform to obtain the time period when the ball bearing has a fault and the fault frequency corresponding to the time period; based on a time-frequency analysis method of Choi-Williams distribution, carrying out refined analysis on the signal of the fault occurrence time period of the ball bearing, and when the fault frequency comprises the fault characteristic frequency of the ball bearing, determining the fault occurrence duration corresponding to the fault characteristic frequency according to the fault characteristic frequency of the ball bearing; and determining the fault type and the fault severity of the ball bearing according to the fault characteristic frequency of the ball bearing and the fault occurrence duration corresponding to the fault characteristic frequency.

The invention is suitable for the technical field of quick-freezing and fresh-keeping of fruits and vegetables, and provides a pure physical fresh-keeping method for quick-freezing and fresh-keeping of fruits and vegetables. The method comprises the following steps of: taking natural 90%-ripe fruits and vegetables as raw materials; dehydrating the raw materials by adopting a cold-wet air convection method; then performing standing to reach water balance; then carrying out adhesion-free independent quick freezing; and finally placing and storing the frozen fruits and vegetables in a freezing chamber at -18 DEG C or below. Therefore, the quick freezing process does not damage the cell structure, the nutritional ingredients, the structure, the original characteristics and the sensory characteristics of the thawed product are kept at the level of fresh fruits, no chemical additive is used, the discharge amount of waste residues and carbon dioxide is less, the freezing time is as long as three years, and the material supply can be provided for shortage seasons or time periods.

The invention discloses a batch memory scheduling method based on Bank division. According to the method, first, according to sources of memory requests, the memory requests are divided into CPU memory requests and GPU memory requests, and the CPU memory requests and the GPU memory requests are synthesized into batch buffer requests respectively; and second, batch memory requests needing to be processed are selected from a CPU batch buffer and a GPU batch buffer, wherein if the batch GPU memory requests are selected, the memory requests to be processed next time are selected by use of the scheduling strategy of first-ready first-come-first-serve (FR-FCFS) of row buffer hitting priorities; and if the batch CPU memory requests are selected, a Bank needs to be divided, so that memory access requests of different kernels are mapped to different Banks, and the memory access requests of multiple CPU applications are isolated. Through the technical scheme, mutual interference of multiple CPUmemory access requests and multiple GPU memory access requests is eliminated, the characteristics of the memory access request of each kernel are reserved to the maximum extent, the hit rate is increased, and therefore the purposes of lowering memory power consumption and improving system performance are achieved.