217results about How to "Keep more" patented technology

The invention discloses a text classification method based on a cyclic convolution network. The text classification method comprises the following steps: step 1, representing context vectors of all words by using a bidirectional cyclic network; step 2, combining the context vectors and a word vector of a current word into the representation of the current word; step 3, extracting the most important context information by using a maximum pool technology to obtain text representation; step 4, carrying out text classification by using the text representation. According to the method disclosed by the invention, more word order information in a text can be kept and a long-distance text dependence relation is captured; semantics of the words can be accurately described and the words and phrases, which have the greatest influences on the text classification, are found by the maximum pool technology, so that the text classification accuracy rate is effectively improved. A test shows that the efficiency of the method is averagely improved by 1% on the aspect of a plurality of text classification data sets.

The invention discloses a convolutional neural network medical CT image denoising method based on residual error learning. The method comprises the following specific steps: 1) constructing a medicalCT image model; 2) constructing a neural network model; 3) training the network; 4) updating parameters; and 5) denoising the medical CT image: inputting the medical CT image containing the noise intothe constructed network model, and outputting the medical CT image without the noise through the network. The method has the advantages that medical CT image denoising is carried out by combining knowledge in the aspect of the convolutional neural network in deep learning. Noise in the image is approximately learned in a residual error learning mode, good pertinence is achieved, and meanwhile thetraining efficiency of the neural network is improved. The convolutional neural network and the residual error learning method are adopted, so that the feature information in the image can be betterlearned, and more local image information can be reserved in the image denoising process. And meanwhile, the image denoising capability is also improved.

The invention relates to a medical image fusion method and an image detection method based on fusion medical image learning, which relate to an image detection technology based on fusion medical imagelearning. The invention solves the technical problems that the medical image is polluted by noise, the signal-to-noise ratio is low, the gray level difference between different tissues is small, theapplication of the medical image is affected, and the single mode image cannot provide richer information of the pathological tissue from different angles, and the like. The method comprises the stepsof: reading the two kinds of modal images and preprocessing the two kinds of modal images respectively to obtain denoised images; using improved shearing wave transform for multi-scale image segmentation. According to the fusion rules, two kinds of modal images are fused to get fused images. All the fused images are combined into a fused image data set. The improved YOLO v2 depth learning algorithm is used to train the dataset, and the training network is generated. The method performs detection with a trained network. Different modalities of medical images are fused together to provide moreinformation from different perspectives.

The invention relates to a preparation method for a high-temperature-resistant ultra-thin antenna window and belongs to the technical field of ceramic-based composite materials. According to the method, the high-temperature-resistant ultra-thin antenna window is prepared by a series of processes of preparation of a perform, repeated dipping of high-purity silica sol, improvement of rigidity of the composite material, machining, washing, pickling, carbon heat removal, secondary repeated dipping of the high-purity silica sol and ceramic thermal treatment. The ultra-thin antenna window prepared by the method disclosed by the invention can meet a requirement on high-temperature resistance and is also high in rigidity and high in machinable property.

The invention relates to extraction of active ingredients of Chinese medicines, in particular to a comprehensive extraction process for auricularia auricula polysaccharide. In the comprehensive extraction process for the auricularia auricula polysaccharide, the auricularia auricula polysaccharide is extracted by cooperating a super micron-milling technology, ultrasonic waves and complex enzyme, the dissolution rate of the auricularia auricula polysaccharide can be greatly improved by combining specific extraction process conditions, and the comprehensive extraction process has the advantages of short reaction time, mild reaction conditions, thorough breaking of cell walls, and high dissolution rate of the polysaccharide. Through comprehensive research, the comprehensive extraction process is easy to operate, and ensures that ingredients are retained comprehensively and the extraction yield is greatly improved.

The invention discloses a method for embedding and extracting a strong robustness hologram watermark on the basis of DWT-SVD (Discrete Wavelet Transform-Singular Value Decomposition). The method comprises the following steps: 1) carrying out transform on an original watermark image to generate a Fourier transform encryption hologram digital watermark, and carrying out singular value decomposition to obtain a singular value matrix; 2) carrying out channel separation on an RGB (Red Green Blue) colorful carrier image, carrying out second-level wavelet decomposition on a B component, and carrying out the singular value decomposition on a low-frequency coefficient matrix to obtain the singular value matrix; 3) embedding, multiplying the singular value matrix generated in the step 1) by an embedding coefficient K, embedding the obtained value into the singular value matrix generated in the step 2) through a gray level adding method, carrying out reduction and wavelet inverse transform on the generated singular value matrix, and combining channels to obtain the RGB colorful carrier image which contains the digital watermark; and 4) carrying out extraction, carrying out channel separation on the RGB colorful carrier image which contains the digital watermark, carrying out the second-level wavelet decomposition and the singular value decomposition to obtain the singular value matrix, subtracting the singular value matrix generated in the step 2) from the singular value matrix obtained in the step 4), dividing the obtained value by the embedding coefficient K, and carrying out the inverse transform to obtain original watermark information.

The invention discloses an SAR image noise suppression method based on a direction wave domain mixture Gaussian model, belonging to the field of image processing, mainly solving the problems of fuzzydistortion of an SAR image caused by the current denoising method and improvement of excessive zero values added by coset decomposition of direction wave, and comprising the following steps: (1) carrying out cartoon texture model decomposition on the SAR image; (2) carrying out improved direction wave transformation on the texture part; (3) carrying out fitting of the mixture Gaussian model on thehigh frequency subband coefficient of the coset; (4) estimating noise variance; (5) estimating noise-free coefficient by Bayesian Minimum Mean-Square Error (MMSE); (6) carrying out improved directionwave inverse transformation on the noise-free coefficient to obtain the denoised texture part; and (7) adding the cartoon part and the denoised texture part to obtain a denoised SAR image. The SAR image noise suppression method has the advantages of improved definition, better maintenance of detail information and smoother uniform region of the denoised SAR image and can be used in filtering speckle noise in the SAR image.

The inulin extracting process includes the following steps: washing, crushing, water leaching, filtering in a primary filter, a tubular filter and an ultrafiltering and decolorizing machine successively, nanometer filtering to concentrate, decolorizing with active carbon, secondary filtering, electrodialysis, evaporating concentration and spray drying. The process has high inulin product purity and yield, short extracting period, simple operation and low production cost.

The invention provides a self-adaptive filtering method for data of a pipeline magnetic flux leakage inner detector, and relates to the technical field of fault diagnosis techniques of pipelines. The method comprises the following steps that magnetic flux leakage signals are acquired by the pipeline inner detector, are amplified and stored, and are transformed into the actual digital voltage signals, the digital voltage signals are subjected to baseline correction, and then the abnormal data are searched, eliminated and compensated; the single-channel signal classification are conducted, and different types of signals are subjected to line filtering through corresponding single-channel self-adaptive filtering algorithms; the influence degree of noises between the adjacent channels is compared and analyzed to implement multi-channel equalization on the magnetic flux leakage signals; and the multi-channel signal area classification is conducted, and the different signal areas are subjected to surface filtering through corresponding multi-channel self-adaptive filtering algorithms. According to the method provided by the invention, the predication of the service life of the pipeline is more accurate, the feature is much easier to extract in the fault diagnosis of the pipeline, the filtering effect is more obvious, and the function of rapidly and accurately processing big data is achieved.

The invention relates to a soft and insulation high wet-absorption fabric, which is prepared by postprocessing including boiling gray fabric, wherein the gray fabric is prepared by interweaving warp yarns and weft yarns; the warp yarns and / or the weft yarns adopt covering yarns with yarn cores being modified polyester fiber; the modified polyester fiber is dissolved during the boiling process to enable the warp yarns and / or the weft yarns adopting the covering yarns to have a hollow structure finally; the modified polyester fiber is modified polyester filament yarns dissolved under the alkaline condition. According to the fabric provided by the invention, the conduction of heat can be reduced, and more heat can be reserved, and a heat insulation function is realized. According to the hollow structure, the comfort in wearing and use is improved, meanwhile, the water absorbing capacity and the quick dry capacity of the fabric are improved, and better comfortable and good feeling is brought.

The invention discloses a method for inhibiting lepidium meyenii myrosinase activity. The method comprises the step that maca is subjected to short-time heat treatment by using stem, wherein fresh lepidium meyenii is blanched in high-temperature steam, such that activity of myrosinase in lepidium meyenii is inhibited or deactivated. The method has high lepidium meyenii myrosinase deactivation efficiency. With the method, myrosinase activity is low, such that lepidium meyenii storage life or shelf life can be prolonged. Also, with the method provided by the invention, glucosinolate content in prepared lepidium meyenii is maintained high, such that lepidium meyenii bioactivity is well maintained, and original nutrients and appearance quality in lepidium meyenii are well preserved. Also, the method is simple and feasible, and energy-saving. The method has the advantages of low cost, good working environment, good sanitary environment, and no pollution.

The invention discloses a polarimetric SAR terrain classification method based on a denoising convolutional neural network. The method comprises the following steps: (1) generating a feature vector ofeach pixel point; (2) generating a training sample set and a test sample set; (3) generating a mean value graph; (4) constructing a denoising convolutional neural network; (5) training a denoising convolutional neural network; and (6) inputting the test sample into the trained denoising convolutional neural network to obtain a classification result. The polarimetric SAR terrain classification method based on the denoising convolutional neural network is adopted to classify the images, the loss of terrain information in the denoising process is reduced, more polarimetric scattering informationis reserved, and finally the classification precision is improved.

The invention belongs to the technical field of food processing, and particularly relates to a low-oil content crispy surimi-based product and a production method thereof. The finished product with an oil content of less than 13% is prepared through the following steps of: adopting fresh surimi or frozen surimi as the main raw material; chopping for 1-2 minutes; adding starch and water and continuously chopping for 1-2 minutes, then adding gingerade, powdered sugar, gourmet powder and onion sections and chopping for 1-2 minutes, and finally adding salt and chopping for 3-5 minutes; shaping, heating for curing and freezing; and then vacuum frying, deoiling and seasoning the product after freezing. Through the invention, the problems of single variety, poor convenience and edibility and the like of deeply-processed aquatic products in the prior art are solved, and a vacuum frying technology is applied to the development of low-oil content crispy surimi-based products. The crispy surimi-based product obtained by the invention is crispy in taste, nutritional, healthy, convenient to eat and easy to store and is suitable for industrial production.

The invention discloses a high-dimensional image de-noising method based on a tensor dictionary and total variation. Based on the study of high-dimensional image processing, the tensor dictionary learning combines a total variation regular term, a high-dimensional image de-noising model combing the tensor dictionary learning with a TV regular term is provided, and then the model is solved by usingan alternate iterative method to obtain a reconstructed MSI image after iteratively updating. The high-dimensional image de-noising method based on the tensor dictionary and the total variation has the advantages that a high-dimensional image is regarded as a tensor entirety, stereochemical structure information cannot lose, the correlation among wave bands is also considered, and the accuracy ofan algorithm is improved through a tensor dictionary learning mode; under the premise without losing a high-dimensional image space structure, a high-order TV regular term is adopted, more complete edge information is well preserved, and a good reconstruction effect is obtained; the experimental result achieves a better effect in both subjective vision and objective evaluation indexes, and more texture information and contour information can be retained.

The invention relates to the field of food processing, in particular to probiotic grain powder and a production method and application thereof. The probiotic grain powder comprises 105-1,012 cfu / g of probiotic, 10-300 mg / 100 g of GABA and 5-50 parts of peptide or free amino acid, and enzyme activity is larger than or equal to 500 u / g. The grain powder is prepared from 100 parts of grain, 0.5-2 parts of compound bacteria, 0.5-2 parts of compound enzyme, 1-20 parts of white granulated sugar, 0.1-0.5 part of amino acid, 1-10 parts of taurine, 1-10 parts of egg white protein powder, 50-200 parts of water, 0.1-5 parts of glycerinum, 1-6 parts of trehalose and 0.5-2 parts of monoglyceride, wherein compound bacteria, compound enzyme, white granulated sugar, amino acid, taurine and egg white protein powder serve as auxiliary materials, and water, glycerinum, trehalose and monoglyceride serve as protective agents. The method for preparing the probiotic grain powder comprises the steps of main material soaking and grinding into thick liquid, heating and curing are conducted, the auxiliary materials are added for fermentation, and finally drying is conducted. The technologies of sprouting, enzymolysis, nutrition compound and ultra-low-temperature freezing are adopted, and the nutrient contents and viable count are kept to the maximum extent.

The invention relates to a method for extraction of Auricularia auricula polysaccharide from Auricularia auricula. The method includes: selecting dry Auricularia auricula as raw materials, conducting soaking, washing and mashing, then adding cellulase, controlling enzymolysis conditions and performing enzymolysis to break cell walls; subjecting the enzymolysis generated Auricularia auricula residue to hot water extraction; combining an Auricularia auricula enzymolysis liquid and an Auricularia auricula extract liquid, conducting filtering, then adding a tannic acid solution to remove protein from the filtered liquid, subjecting a polysaccharide solution collected by centrifugation to alcohol precipitation by an ethanol solution; and finally, carrying out vacuum drying to obtain a polysaccharide product. The extraction method provided by the invention has the advantages of reasonable process, simple operation and comprehensive component retaining, the polysaccharide yield and purity are high, and the product safety is good.

The invention discloses an engine surge fault prediction system and method based on a fusion neural network model, and belongs to the technical field of time series data prediction, and the system comprises a prediction module which is used for generating a prediction time series with a specified length through three-dimensional structure time series data of an engine; a feature extraction modulewhich is used for extracting local features of the prediction time sequence, semantic relationships among data and overall sequence trend features; and a classification module which is used for judging whether a surge fault exists or not according to the local features of the prediction time sequence, the semantic relationship among the data and the trend characteristics of the whole sequence. According to the method, the prediction time sequence with the specified length is generated firstly, that is, prediction of the working state data of the engine in a future period of time is achieved, and then whether the working state data of the engine in the future period of time include the surge fault data or not is judged so that the surge fault of the engine is predicted more accurately and quickly in advance.

The invention discloses a degreased trotter processing method, including the following materials and steps: a, baking trotter through an alcohol biow lamp until the surface of the trotter turns cream-colored, washing through clear water, and blanching through steam for 15-20mins; b, putting one portion of seasoned soup and 2.5 portions of trotter into a stewing cooker, adding new seasoning pouches and condiments, heating through steam for 2.5-3.5h, then scooping out the lard at the upper layer, adding appropriate amount of balm, mixing and stirring for two times,and getting out the mixture from the cooker; c, airing and cooling, then weighting and packing the trotter, vacuumizing and sealing the pouch, sterilizing, packing with an external package and sealing the package. The trotter pretreatment method is sanitary, nontoxic and convenient to operate, and can make the trotter totally degreased during the stewing process and make the trotter easy to be colorated. Since most of the grease is removed during the stewing process, the degreased trotter is not greasy to eat and can be eaten upon opening the package. The sealed and packed trotter can be preserved for 240-300 days at room temperature.

The invention discloses a non-supervision change detection method based on fuse change detection operators and dimension driving. The non-supervision change detection method mainly solves the problems that due to the fact that image noise is eliminated through filter in the prior art, a large number of image details are lost, and the change detection difficulty is increased. The non-supervision change detection method is achieved through the steps of firstly, generating a logarithm ratio image and a mean value ratio image through image data for change detection; secondly, fusing the logarithm ratio image and the mean value ratio image to generate a primary difference image; thirdly, judging the reliable dimension of each pixel point on the primary difference image, and fusing the reliable dimensions of all the pixel points on the primary difference image to generate a change detection difference image; fourthly, conducting clustering on the change detection difference image to generate a change detection result image. More image details are reserved while noise is effectively restrained, the accuracy of the change detection result is improved, and the non-supervision change detection method can be used for change detection of an SAR image.

The invention relates to a processing method for a dried chive product. The processing method comprises the following steps of: picking, cleaning and sterilizing raw materials; cutting the raw materials into segments; drying the raw materials in an oven; sorting the raw materials; sampling the raw materials for inspection; freezing and packaging the raw materials; and putting the product in a warehouse. When the processing method provided by the invention is adopted, the defect that fresh chive cannot be stored for a long time is greatly overcome, and the condition of incapability of using the chive due to season, place of origin and the like can be effectively adjusted. The chive can be recovered when being immersed in fresh water; and original color, nutrition and special flavor of the chive are reserved.

The invention discloses a preparation process of shell powder coating. The preparation process of the shell powder coating comprises the following steps: selecting shell raw materials; (2) acid-washing the shells; (3) alkali-washing the shells; (4) calcining the shells: putting the aired shells into a calcining furnace, wherein the calcining temperature is controlled to 300 to 350 DEG C and the calcining time is 8 hours; (5) pulverizing the shells; (6) putting the pulverized shell powder into a dry powder stirring machine, adding sticky rice glue powder and stirring for 30 minutes to obtain the shell powder coating; (7) packaging the shell powder coating, to obtain a finished product. Offshore, middle sea and deep sea shells are selected as the raw materials, and more chitin in the shellscan be retained; the purely natural sticky rice pure glue powder is added, so that the shell powder coating is green.

The invention belongs to the technical field of foodstuff processing, and concretely relates to a cereal instant tea and a production method thereof. The instant tea provided by the invention comprises the following raw materials: ultrafine grinded cereal materials, ingredients, soluble starch, NaHCO3 and water, wherein the cereal is selected from rice, broken rice, buckwheat, millet, corn or barley and combinations thereof, and the ingredients are red date, matrimony vine, radix codonopsis, safflower, etc. The cereal and the ingredient are grinded into ultrafine particles with a size of less than 100 [mu]m. The production method comprises immersing the cereal with water, draining water, roasting, ultrafinely grinding the roasted cereal and the ingredients, mixing the grinded cereal and the ingredients with the NaHCO3, adding a binder, uniformly mixing, molding, drying and cooling to obtain the product. The product provided by the invention is quick in dissolving speed, and high in digestion and absorption rate, and has a unique cereal fragrance.

The invention provides an extraction method for a web title. The method comprises the following steps of: determining the position of a search term in the web title; judging whether a length from a first character of the web title to the last character of the search term is less than or equal to a preset title display length or not, searching for a segmentation character if the length from the first character of the web title to the last character of the search term is not less than or equal to the preset title display length, and returning corresponding characters as results when the character length of a certain segment of characters in the web title is less than or equal to the preset title display length and the characters comprise the integral search term and the found segmentation character. The invention also provides an extraction device for the web title. The displayed web title is high in readability, comprises much reserved core information, and has red marks, so that retrieved contents can be obtained quickly and conveniently by a retriever.

The invention belongs to the field of biological medicines, and in particular relates to a spina date seed extract and brain-strengthening and sleeping-promoting oral liquid containing the spina date seed extract. A preparation method of the spina date seed extract comprises the following steps: crushing spina date seeds, performing water extraction on active substances, performing semi-bionic extraction, concentrating, and removing impurities by using an alcohol precipitation method to obtain the spina date seed extract. The brain-strengthening and sleeping-promoting oral liquid containing the spina date seed extract is further prepared by taking the spina date seed extract as a raw material and blending with other components such as honey. The oral liquid contains sleeping-promoting and nerves-calming effective components of the spina date seeds, including spina date seed flavonoid substances, spina date seed alkaloid and jujuboside D and E, and has the effects of calming the nerves, strengthening the brain, promoting sleeping and benefiting intelligence.

The invention provides a preparation method of a high-temperature proton exchange membrane. The preparation method comprises the following steps of 1) preparing a bis-hydroxyethyl sulfone-4, 5-dicarboxyl imidazole polycondensate; 2) performing modification on the polycondensate; 3) carrying out casting and film forming; and 4) performing phosphoric acid doping. The invention also discloses the high-temperature proton exchange membrane prepared by the preparation method, and a proton exchange membrane fuel cell which takes the high-temperature proton exchange membrane as a polyelectrolyte membrane. Compared with the high-temperature proton exchange membrane in the prior art, the high-temperature proton exchange membrane prepared in the invention has higher proton conductivity, lower cost, lower methyl alcohol permeability, higher mechanical property, higher stability and high temperature resistance performance, and higher safety and environmental protection property in use.

The invention provides a solar radio interference filtering method, a readable storage medium and electronic equipment. The method comprises the following steps: collecting intensity data in a set time period before and after a solar burst event, and screening out frequency channels where eligible radio signals are located according to radio flux value data of all frequency channels at a certain moment in a non-burst time period; preprocessing the acquired eligible radio station signal data, and establishing a data set; establishing a recurrent neural network, training the recurrent neural network by using the preprocessed radio station signal data, and predicting the signal value of a radio station in a solar burst area by using the trained recurrent neural network; and according to the linear additive property of the signal, subtracting the predicted radio station value from the original radio flux value of the burst area at the corresponding moment to obtain a value of a clean puresolar burst event. Radio station interference signals can be predicted more accurately, and more solar burst information can be reserved. The final solar burst observation effect of the method is obviously superior to that of a common image filtering method.

The invention discloses a processing method of lentinan. The processing method includes the steps of firstly, coarsely grinding; secondly, twin-screw extruding; thirdly, semi-bionic extracting; fourthly, centrifugalizing and concentrating; and fifthly, ethanol precipitating and drying. Since an ultrasonic method is cooperated with the semi-bionic method to extract the lentinan, more effective components can be extracted and reserved, production cycle can be shortened, and cost can be lowered.

The invention belongs to a food-processing technology, in particular to a production method of natural yellow croaker meal. The method comprises the following steps: (1) removing the viscera of a yellow croaker, cleaning the yellow croaker and pulping with a certain proportion of water; (2) pumping the yellow croaker pulp into a multifunctional reaction kettle, using NaOH to adjust the pH value of the reaction liquid to be 6.5-7.5 at a temperature of 30-60 DEG C and zymolyting the reaction liquid for 2-6 hours; (3) screening fish bones, adding accessories to fish meat so that the contents of solid substances achieve 30 percent; and (4) homogenating, spraying, drying, collecting, screening and packaging to obtain natural yellow croaker meal. The invention has the advantages that the method can keep rich nutrient compositions of the yellow croaker meal which has true flavor, good palatability and little nutrition loss, and the technology can solve the pollution in the processing process of marine food products and the source loss of raw materials.

The present invention discloses an electroencephalogram classification method based on an attention mechanism and a convolutional neural network. Collected electroencephalograms of n of channels are set as Xn (t) and a segment of the electroencephalograms intercepted every 2 s are used as a classification sample Xn (t) for the electroencephalograms in a normal state and in an abnormal state; a short-time Fourier transform is performed on each channel in the classification sample Xn (t), and rows represent time domain and columns represent frequency domain to obtain n of time-frequency domain matrixes; elements in the time-frequency domain matrixes are complex numbers, each element in the time-frequency domain matrixes is subjected to module value taking to obtain a real-number domain STFTmatrix; the real-number domain STFT matrix is transformed into pictures with colors presenting amplitude sizes; and the transformed pictures are input into the ResNet-50 convolutional neural network for feature extraction to realize electroencephalogram classification. The method can greatly improve the classification accuracy of the electroencephalograms in the normal state and abnormal state ofepilepsy patients.

The invention relates to a processing method of a cordyceps sinensis product which can be taken both as a medicine and food, belonging to the field of Chinese herbal medicines. A grinding method comprises the following steps of: rapidly freezing, coarse crushing, freezing for the second time, adding dry ice and saccharose, crushing by a low-temperature ultrafine grinder, and thus ultrafine powder is collected. The cordyceps sinensis ultrafine powder can be prepared into instant tea bags, powder, capsula, tablets, wine, oral liquid and buccal tablets. According to the invention, content of effective components of cordyceps sinensis can be reserved to the maximum, absorption of a human body to the effective components of cordyceps sinensis can be enhanced, the absorptivity can be improved by 3-5 times, and saccharose is added before superfine grinding, so that the situation that the rate of absorption and utilization of cordyceps sinensis is reduced by an embedding function of macromolecule substances added in a preparation process of a cordyceps sinensis product can be avoided, moreover, the grinding method of cordyceps sinensis is simple and practicable, industrial production is convenient to realize, the lost is low, the popularization is easy, large economic benefit can be generated, and limited cordyceps sinensis resources can be protected.