66results about How to "Solve few problems" patented technology

While deep learning is used for classification, multi-modal information with rich samples and classification contribution variability of each modality are considered, and the problem of insufficient samples is solved by using a semi supervision method. Data of different modalities of a hyperspectral image is sent into a deep neural network, the semi supervision method is used and a large number ofunlabeled samples are utilized, and the deep neural network based on self-encoding is used for feature learning. All labeled and unlabeled data are sent into the self-encoding deep neural network tocarry out learning, similar networks are designed for different modalities, a respective initialization parameter is obtained through self-encoding reconstruction, and hidden attributive classification of labeled samples is obtained through a clustering method. For the unlabeled data, a deep characteristic is calculated through a multi-target deep network, then a similar marked sample is searchedbased on a clustering label, and finally, labels of the unlabeled samples are predicted according to the label information of the labeled samples.

The invention discloses a generative adversarial network technology-based medical text generation method. The method comprises the following steps of downloading multiple medical documents of a fieldwhich a department belongs to; representing words in the medical documents by utilizing word vectors; performing text classification on the medical documents, and reserving the required medical documents; obtaining optimal output sequence tags of the required medical documents; obtaining indexes of keyword sets of the required medical documents, and randomly disordering the indexes to obtain new medical documents and corresponding sequence tags; performing training by adopting a generative adversarial network-based generative model to generate medical texts; outputting the generated medical texts; obtaining BLEU parameters of the generated medical texts; and assessing the generated medical texts to finally obtain target medical texts. By disordering data to generate the new medical texts,the problems of patient privacy security and relatively few medical texts are solved.

The invention discloses an infrared image splicing method based on coarse-to-fine registration, and the method comprises the steps: carrying out the coarse registration of an infrared image, carryingout the fine registration of the infrared image, improving the image registration precision, and improving the image splicing effect, wherein in the coarse registration process, feature points are extracted by utilizing every two adjacent images to construct a homography matrix, and each image can use an untransformed original image as an extraction object of the feature points in the coarse registration process, so that the extraction quantity of the feature points is ensured. The fine registration process is based on a fine registration sequence for sequential registration; deformation errors existing in the coarse registration process are adjusted, the feature point extraction process is obtained on the basis of the rule that the percentage of the area of the minimum circumscribed convex polygon covered by all feature point pairs consistent with the feature point pairs and the area of an original image is in direct proportion to the image registration accuracy, and the precision offine registration is guaranteed.

The invention provides an online assembly defect identification system and method. the method comprises: through establishing a B/S architecture, establishing a real-time monitoring management systemcomprising a product assembly line, a lower computer image processing module and an upper computer server module; typical assembly defects of each assembly station on a product assembly line are systematically analyzed, and a target area is set for a product image of each assembly station on the basis of a set of fixed image acquisition system, so that the calculation loss of image processing is greatly reduced, and meanwhile, based on a PSO-SVM algorithm, establishing a product manufacturing defect identification model. Through training and learning a training sample library image, the product manufacturing defect identification model can identify the assembly defects on the product assembly line in real time. and the upper computer server module is used for controlling the PLC to carry out real-time maintenance of the assembly defects, so that real-time acquisition, analysis, identification and processing of the assembly information of the intelligent manufacturing workshop are realized, and the problem that the product assembly defects of the intelligent manufacturing workshop are difficult to identify in time is solved.

The invention discloses a five-axis motion control card, and relates to the technical field of numerical control. The control card comprises a singlechip module, an FPGA module, a system auxiliary module, an expansion interface module, a dual-port communication module, a digital quantity IO expansion module, a DA module, an AD module and an input/output interface module, wherein the system auxiliary module is connected with a singlechip module and the FPGA module; the singlechip module is connected with the expansion interface module; the dual-port communication module is connected with the FPGA module and an upper computer; the DA module is connected with the FPGA module; the AD module is connected with the FPGA module; the digital quantity IO expansion module is connected with the FPGA module; the DA module is connected with the input/output interface module; the AD module is connected with the input/output interface module; and the digital quantity IO expansion module is connected with the input/output interface module. The singlechip and the FPGA are adopted, so the control card has the advantages of improving the interpolation efficiency, reducing external circuit modules, enhancing the stability, reducing the area of the control card and having low cost.

The invention relates to a monocrystal structure of a luminescent material of a cadmium coordination compound with blue light performance, blue light performance and a preparation method thereof. The molecular formula of a product provided the invention is CdC10H5N3O4. The molecular weight of the product is 343.57, and the product belongs to a single orthorhombic system. An orange bulk crystal is synthesized by a hydrothermal method. The preparation method provided by the invention is as follows: respectively dissolving cadmium chloride and 2-pyridyl-1H- imidazole-4, 5-dicarboxylic acid in an aqueous solution, and then, adding a small amount of pyrazole, adjusting the pH value through an HCl solution, putting the solution in a stainless steel kettle sleeve, and burning the stainless steel kettle sleeve in an oven of 150 DEG C for three and a half days to obtain an aurantiacus bulk crystal. The product prepared by the method provided by the invention is a monocrystal with an intact crystal form; and the prepared compound is clear in composition and structure, as shown in a luminous performance test, the monocrystal structure provided by the invention exhibits blue fluorescence and has high thermal stability. The yield of the product prepared by the method provided by the invention is high and can reach 45%.

The invention discloses an automatic identification method for inhaul cable jacket diseases based on transfer learning and inhaul cable robot, and the method comprises the steps: preparing a concretecrack picture, building a ResNet network model in deep learning, and carrying out training, to obtain a training model; obtaining a disease picture of a stay cable PE jacket, dividing the data set into a training set, a verification set and a test set, establishing a transfer learning model of the ResNet convolutional neural network, and training the model; and controlling the stay cable robot tocarry out image acquisition, and carrying out stay cable PE jacket disease automatic identification by using the trained ResNet model. The detection precision is high, the stay cable robot can replacemanual work to perform automatic crawling detection, the automation degree is high, real-time recognition can be achieved, efficiency is high, cost is low, and compared with a traditional manual detection method for stay cable PE jacket diseases, the method has obvious advantages.

The invention discloses an electrocardiogram anomaly recognition method and system based on combination of AlexNet and transfer learning, and belongs to the technical field of feature extraction classification prediction. The technical problem to be solved by the invention is how to accurately and efficiently complete electrocardiogram anomaly recognition in combination with an AlexNet deep convolutional neural network and transfer learning, and how to get rid of dependence on sample data capacity and automatically learn features of data samples. According to the technical scheme, the method specifically comprises the following steps: preprocessing: converting each electrocardiogram signal in a data set into an electrocardiogram image, and clipping an electrocardiogram of an abnormal typein different directions; feature extraction: putting the data-enhanced image into a pre-trained model for training, taking the pre-trained electrocardiogram image as the input of an AlexNet model, automatically extracting features, and carrying out transfer learning by using the AlexNet model pre-trained by an ImageNet data set; and classification prediction: putting high features obtained by a pre-trained AlexNet deep convolutional neural network model into a support vector machine for electrocardiogram classification.

A water-soluble thermosensitive high-molecular material is a copolymer prepared from N-isopropyl acrylamide, 1-vinyl-2-pyrrolidone, and acrylic acid. It features its thermal reversible water solubility, that is, when the temp of its aqueous solution is raised to LCST, deposit occurs and when the temp is lowered, the deposit is dissolved, changing the ratio of its components can regulate the deposition temp.

The invention relates to a construction method of a semi-supervised neural machine translation model based on word-to-word translation, and belongs to the technical field of natural language processing. The method comprises the steps that firstly, monolingual corpora of a source language and a target language and parallel corpora of the source language and the target language are acquired; training a cross-language language model by using the monolingual corpus; initializing an encoder-decoder of the translation model by using the trained language model; acquiring bilingual dictionaries of thetwo languages; constructing a dictionary prefix tree in the translation model according to the bilingual dictionary; training an auto-encoder in the translation model; training a translation model byusing the parallel corpora of the source language and the target language; training the translation model and training the translation model at the same time; and training the translation model by using the parallel corpora of the source language and the target language through the auto-encoder, and fusing the translation model to obtain a final translation model. The method is simple and effective, the model can be translated normally, and the translation performance of the model is greatly improved.

The invention provides a bacterial colony image collecting system. The bacterial colony image collecting system comprises a box, a base, an image collecting part and one or more sets of lateral lightsource assemblies, wherein the base is arranged at the bottom of the box and used for placing a culture dish; the image collecting part is arranged at the top of the box, and a lens of the image collecting part faces the base to collect an image of a bacterial colony in the culture dish; the lateral light source assemblies are arranged on the side wall of the box to provide lateral light or fluorescent light exciting light for the culture dish, and are provided with light dispersing parts, and the light dispersing parts are located on light outlet paths of light sources of the lateral light source assemblies so that light passing through the light dispersing parts can uniformly cover the culture dish on the base. Through the bacterial colony image collecting system, the problems are solvedthat in the prior art, a bacterial colony image collected by the bacterial colony image collecting system is not clear, and a few fluorescent light exciting light sources exist.

The invention discloses an agricultural crop irrigation device. The device includes a main support plate, wherein a lower surface of each end of the main support plate is equipped with a mounting plate, the middle of the main support plate is equipped with a connection water pipe, one end of the connection water pipe is equipped with a three-way control valve, and the other side of a hollow disk is uniformly provided with water holes. The device is advantaged in that the structure is scientific and reasonable, the device is safe and convenient to use, the sprayed water is blocked by adjustingblades, so the spraying distance of the water can be shortened, a problem that the water sprays too far to water crops is solved, through rotation of a water outlet sleeve, the water can flow out froma second water pipe and the water outlet sleeve, a problem that the waver volume is small when an irrigation device is relatively close to a nozzle is solved, the water of the hollow disk flows out from the lower water holes, better penetration of the water into roots of the crops is facilitated, the water can be effectively sprayed to leaves and branches of the crops through the upper water holes, and growth of the crops is better promoted.

The invention provides an entity disambiguation method and device based on a UCL knowledge space, and the method comprises the steps: firstly constructing a basic knowledge base, and completing the construction of the UCL knowledge space; then obtaining a candidate entity set related to the entity to be disambiguated from the UCL knowledge space, and generating an embedded representation of the candidate entity and the entity to be disambiguated by using a word vector representation method; then extracting concept features of an entity to be disambiguated and contexts thereof, and extracting features of contexts of candidate entities; finally, the four vector representations generated previously are used as input, a self-attention matching network based on a deep structured semantic matching model DSSM is adopted, and the matching degree is obtained; and obtaining a final disambiguation result according to the sorting of matching results, and completing entity linking between the entities in the text and the entities in the UCL knowledge space. According to the method and the device, the problem of less entity related information in the short text can be solved, and the entity disambiguation accuracy can be improved.

The embodiment of the invention provides a method and device for a decision engine, a machine readable storage medium and a processor. The method comprises the following steps: determining an external service scene and sample data selected based on an artificial intelligence AI model node in a decision engine; searching a first Docker mirror image corresponding to the service scene from a preset scene mirror image library; controlling the first Docker mirror image to model according to the sample data to obtain a model file; and controlling the first Docker mirror image to deploy the model file. AI model nodes are arranged in the decision engine, modeling and deployment are carried out in a Docker mirror image, a user can carry out operation on the AI model nodes, any AI model can be modeled and deployed through the Docker mirror image, the limitation that a traditional decision engine only supports an AI model in a PMML format is broken through, and the flexibility of the decision engine is improved.

The invention discloses an Android App historical screen recovery and evidence collection system based on memory analysis, and relates to an Android App historical screen recovery and evidence collection system. The objective of the invention is to solve the problems that an existing Android mobile phone ROOT permission and evidence obtaining system does not have universality, some evidence obtaining systems have less memory extraction information, and an evidence obtaining method for a specific App does not have universality. The system comprises an Android memory extraction subsystem, a memory evidence storage subsystem and an Android App historical screen recovery and evidence collection subsystem. The Android memory extraction subsystem is used for completely extracting the memory space of the target App; the memory evidence storage subsystem is used for storing the file memory evidence extracted by the Android memory extraction subsystem; and the Android App historical screen recovery and evidence collection subsystem is used for realizing App historical screen recovery and evidence collection by utilizing a model in the MVC architecture. The system is applied to the field ofandroid App historical screen recovery and evidence obtaining.

The invention relates to a ship detection method and device. The method comprises the steps of obtaining an accurate positioning rotating rectangular frame of a ship in a remote sensing image based onship features of the remote sensing image; extracting a slice image of the ship in the remote sensing image based on the accurate positioning rotating rectangular frame of the ship in the remote sensing image; inputting a slice image of a ship in the remote sensing image into a pre-established classification model, and obtaining a ship model of the ship output by the pre-established classification model; according to the ship detection method provided by the invention, the classification difficulty is low, the positioning result is more accurate, adjacent ship targets can be distinguished, target confusion is avoided, the influence of insufficient training samples and non-uniform categories is reduced, the classification performance of the classifier is improved, and the identification ofspecific models of the ship targets is realized.

The embodiment of the invention discloses a display cabinet interaction method, a display cabinet and a computer readable storage medium. The method comprises the steps that a preset image is played through an audio-video device arranged on the display cabinet; the content of the preset image is acquired, and a target displayed commodity corresponding to the content of the current played image isdetermined; all displayed commodities displayed on the display cabinet and display positions corresponding to the displayed commodities are obtained; a target display position is acquired according tothe target displayed commodity, and the target display commodity is highlighted and displayed. The problems that an existing display cabinet is monotonous in function and display information of the displayed commodities is little are solved; the purposes that the displayed commodities are highlighted and users can conveniently know the corresponding relation between the display content and the displayed commodities are achieved. The invention further discloses the display cabinet and the computer readable storage medium. By means of the scheme, functions of the display cabinet are expanded, meanwhile, the experience of users is enhanced, and the purposes of displaying the commodities and selling the displayed commodities can be achieved by the display cabinet.

The invention relates to a 3D printing simulation lumbar vertebra model for training a low-age orthopedist to find an optimal screw feeding point of a lumbar pedicle screw and a manufacturing method;the model comprises skin, muscles and a lumbar vertebra skeleton, and the lumbar vertebra skeleton is provided with an optimal channel, an optimal screw feeding point, alternative channels and alternative screw feeding points in a pedicle projection area; the optimal screw feeding point is arranged at the intersection of the optimal channel and the zygopophysis cortex on the vertebral body and isconnected with the optimal channel; the alternative channels are arranged around the optimal channel in parallel and cover the vertebral pedicle projection area; and the alternative screw feeding points are arranged at the intersections of the alternative channels and the vertebral zygopophysis, vertebral lamina or transverse process cortex, and are connected with the corresponding alternative channels. According to the invention, the culture curve of a low-age orthopedist is shortened, the culture efficiency is improved; meanwhile, the model is high in simulation degree and strong in repeatable operability.

The invention provides a force feedback technology based surgical robot master-slave control experiment system, and relates to the field of surgical robots. The surgical robot master-slave control experiment system comprises a master operation hand, a system workstation, a display, a mapping simulation module, a slave hand simulation module and a human body structure simulation module, wherein themaster operation hand sends coordinate information to the system workstation, the system workstation achieves coordinate mapping of the master operation hand and the slave hand simulation module through the mapping simulation module, and the slave hand simulation module and the human body structure simulation module send image information to the display for displaying. Based on force feedback andcoordinate mapping technologies, a mechanical arm model can be simulated in a one-to-one mode, a motion simulation process is completed, and a touch sense is brought into an actual operation processof a surgical robot, so that the sense of reality of a simulation experiment process is increased.

The invention provides an eremosparton songoricum DREB transcription factor EsDREB gene and an application thereof; the protein has an amino acid sequence as shown in SEQ ID No 3; the length of the gene encoding region is 624 bp, and 207 amino acids are encoded; a subcellular localization experiment shows that the gene is localized in a nucleus; and a yeast one-hybrid experiment demonstrates that the gene as a transcription factor has the activity of transcription activation. Meanwhile, the yeast transforming the gene has higher survival rates than a control of non-transgenic yeast in drought, cold, heat stress, which indicates that the DREB transcription factor EsDREB of a desert plant of eremosparton songoricum has abiotic stress resistance. The gene of the invention is applicable to research of crop abiotic stress transgenic breeding.