57results about How to "Solve the inconvenience of application" patented technology

The invention provides a personalized recommendation method and system based on collaborative filtering and deep learning, and the method comprises the steps: obtaining historical behavior feature data of commodities purchased by a user, carrying out the preprocessing, and sorting the purchasing behaviors of the user according to the time, wherein the sorted data is called a behavior feature sequence of the user; performing the personalized recommendation system modeling, which comprises the steps of obtaining input vectors of a user and a commodity from an interaction matrix, then respectively generating embedded vectors of the user and the commodity, weighting the embedded vectors through an attention neural network, and performing linear and nonlinear interaction on the weighted embedded vectors, thereby obtaining the explicit and implicit relationship between the user and the commodity; finally, estimating the click rate of the user to the commodity; and training and testing the model by using the user behavior characteristic sequence. According to the method, the collaborative signals of the users and the commodities are fully mined, a basis is provided for capturing personalized demands of the users, and the accuracy and interpretability of a recommendation system can be improved.

The invention relates to a blood purification adsorbent for uremia and a preparation method of the adsorbent. The blood purification adsorbent is carbon nanotube/nano silica/polystyrene resin composite microspheres, carbon nanotubes and nano silica are dispersed in polystyrene resin, and the polystyrene resin adopts a porous crosslinked structure. The preparation method comprises the steps of suspension polymerization, a chloromethylation reaction, post-crosslinking and surface modification. The provided and prepared blood purification adsorbent has high strength, adsorptive property and adsorption efficiency, has better adsorptive performance for special toxins and is particularly applicable to blood perfusion for uremia.

The invention provides a makeup style migration method and system based on regional style consistency, and the method comprises the steps: inputting a plain face image and a reference makeup face image, and deformation images to obtain a pseudo makeup face image; based on the number of preset sub-makeup styles, performing makeup segmentation on the pseudo makeup face image through an SOM network by using the color texture information of the makeup, decomposing the makeup style into the sub-makeup styles, and dividing the overall makeup into different sub-makeup style areas to obtain an approximate result of a makeup segmentation result of the target makeup face image; inputting the plain face image and the reference makeup face image into an adversarial generative network, adjusting a generative network module in the adversarial generative network through regional adaptive normalization, and applying the same normalization mode to the same sub-style region to enhance the consistency oflong-distance similar makeup regions; and enabling the generation network module to output the target makeup face image, so that a better makeup style migration image effect can be realized.

The invention provides an air conditioner on-hook spatial distribution automatic identification method and system based on a streetscape picture, which realizes air conditioner on-hook spatial distribution identification by taking the streetscape picture provided by a network platform as a data source, and comprises the following steps: capturing the streetscape picture through a crawler; and pre-screening the crawled street scene pictures based on the air conditioner hanging scene characteristics, labeling a window entity and an air conditioner hanging target entity, inputting the window entity and the air conditioner hanging target entity into a YOLOX network for air conditioner hanging entity model training and window entity model training, and after training is completed, predicting each street scene picture by using a corresponding model. Outputting the number of air conditioner outdoor units and window entities in each picture; and calculating the ratio of the number of air conditioner outdoor units to the number of windows, performing space projection in combination with the coordinate information of each street view picture, and outputting the corresponding air conditioner coverage rate of the region. By mining the network streetscape picture information, identifying the ratio of the air conditioner hanging unit to the window and extracting the air conditioner space distribution information, the labor cost can be greatly saved.

The invention provides a close-range photogrammetry target positioning method and a close-range photogrammetry target positioning system with a known photographing position. The method comprises the following steps: acquiring a three-dimensional coordinate of a camera station position, namely measuring and acquiring the three-dimensional coordinate of the camera station during image shooting; measuring two-dimensional coordinates of the shot target point image on the image; introducing a unit quaternion to construct a rotation matrix from an image space to an object space; obtaining an initial value of an object space three-dimensional coordinate in a relative orientation manner, carrying out differential calculation on a rotation matrix constructed by the unit quaternion, and establishing an error equation about the unit quaternion; and establishing a unit quaternion constraint condition equation and block adjustment with limited conditions, including performing iterative calculation by using an indirect adjustment method with limited conditions, and solving three-dimensional coordinates of a target point. According to the method, image control points do not need to be arranged, solving of unknown numbers of photography bases during block adjustment is reduced, the method is simple and easy to implement, the measurement speed is high, the operation time is shortened, and the measurement operation efficiency is greatly improved.

The invention discloses an intelligent question-answering method based on a machine reading understanding and common question answering model, and the method comprises the steps: carrying out the collection of question-answering pairs, generating the question-answering pairs based on a system document, and carrying out the arrangement and filtering, thereby forming a database; model pre-training: training an MRC model and an FAQ model by using the database in the step 1; verifying user login; user question information collection: obtaining question information input by a user from a front end and inputting the question information into the pre-trained model; and answer generation: adopting an FAQ model to generate an answer based on semantic similarity matching, and if a result is not ideal, inputting a user question text into an MRC model to obtain a question answer and returning the question answer to a front end for display. The method supports multi-mode retrieval, can intelligently and deeply understand documents to provide answers under the condition that a common question answering model is not ideal in effect, so that the generated answers are more targeted, more references are provided for solving questions of a user, meanwhile, supplementation can be provided for a common question answering data set, and a question answering cold start mechanism is realized.

The invention provides a remote sensing image solid waste identification method and system based on multi-strategy enhancement, and the method comprises the steps: cutting an original remote sensing image according to the input original remote sensing image and a preset overlapping ratio, and storing the coordinate position information of the left upper corner point of each cut remote sensing image block in the whole remote sensing image; putting the remote sensing image blocks into a deep learning target detection network for remote sensing image solid waste recognition to be recognized, and recognizing a frame of the solid waste, wherein the deep learning target detection network for remote sensing image solid waste recognition adopts a multi-strategy enhanced key point target recognition network guided by position information; screening the detected frame to obtain a final frame identification result, and displaying the final frame identification result in the original remote sensing image. The scheme provided by the invention can be effectively applied to the high-resolution remote sensing image, so that the solid waste in the remote sensing image can be quickly detected.

The invention provides a multi-terminal distributed uniform equipower wireless power supply system, which comprises more than one power transmitting unit and more than two relay units, a plurality of loads are respectively connected to each relay unit, and wireless power transmission is established between the power transmitting units and the relay units by adopting a magnetic coupling resonant power transmission mode. Each power transmitting unit comprises a power supply module, a compensating circuit and a transmitting coil which are connected in sequence; the power supply module is formed by connecting a direct-current power supply, a driving circuit and a high-frequency full-bridge inverter; each relay unit comprises a relay coil, a compensating circuit, a compensating coil, a rectifying and voltage stabilizing circuit and a load; and in each relay unit, a relay coil provides a receiving coil function of the unit, and a compensation coil provides a transmitting coil function of the unit. The method can be applied to long-distance and multi-load application, and it is ensured that each load in a distributed system can effectively receive enough power.

The invention provides a positioning method and system based on signal quality control and feature fingerprint enhancement, and the method comprises the steps: carrying out the feature extraction based on signal quality control: carrying out the quality control from a signal level, and extracting frequency domain CSI information and time domain PDP information; establishing fingerprints based on time domain and frequency domain combined features, including combining frequency domain CSI information and time domain PDP information, and constructing a multi-dimensional feature offline position fingerprint database; carrying out robust fingerprint positioning based on distance measurement enhancement, comprising the steps of constructing online test position fingerprint information by adopting the same combined feature mode, calculating a fingerprint positioning result by utilizing a multi-dimensional feature offline position fingerprint database, constraining the fingerprint positioning result by using a first path energy distance measurement result, and finally obtaining a robust fingerprint positioning result. According to the positioning technology solution provided by the invention, the accuracy and the stability of an indoor positioning result are improved.

The invention discloses a user service access processing method, a piece of access equipment and an access terminal. The access equipment is arranged to be connected to an authenticated service channel and a non-authenticated service channel of the access terminal. The access equipment carries out user data access through the authenticated service channel when access of the access terminal is successfully authenticated, and the access equipment carries out user data access through the non-authenticated service channel when access of the access terminal is not authenticated or fails to be authenticated. The scheme can be adopted to solve the problem of application inconvenience caused by the fact that a user cannot access some controlled websites when the user does not initiate authentication or fails to be authenticated.

The invention provides a GPS elevation fitting method considering gravity terrain correction, which is used for mountain terrain and comprises the following steps of: calculating terrain correction and interlayer correction of Bouguer gravity anomaly at grid points according to a digital elevation model of a measurement area; adding the terrain correction and interlayer correction of the Bouguer gravity anomaly, and performing converting to obtain the influence of terrain factors in the elevation anomaly at the grid points; carrying out gravity terrain correction to deduct the influence of terrain factors in elevation anomaly at a GPS benchmark; performing numerical fitting on the elevation anomaly of the GPS benchmark after the terrain correction to obtain a local quasi-geoid model after the terrain correction; and combining the local quasi-geoid model after the terrain correction with the terrain influence in the elevation anomaly calculated at the corresponding position to obtain a real local quasi-geoid, and taking the real local quasi-geoid as a final GPS elevation fitting result considering gravity terrain correction. The data acquisition cost is lower than that of a gravity measurement method, and the GPS elevation fitting precision of mountainous areas can be effectively improved.

The invention provides a troposphere delay and multi-path error separation method and system of a large-height-difference baseline. The method comprises the steps of: firstly recovering a double-difference residual error of an independent baseline of a measurement region into a non-difference residual error of each satellite of each measurement station, namely, solving a group of independent baselines of the measurement region, and converting the double-difference residual error of each baseline into an inter-station single-difference residual error through adding a gravity center reference for the first time; adding the gravity center reference to the single-difference residual error of each baseline again, and recovering the inter-station single-difference residual error to the non-difference residual error of each observation station; then, constructing a non-difference zenith troposphere delay error model of each observation station in combination with a wet delay projection function of a satellite oblique path signal; and finally, subtracting a model calculation value from the non-difference residual error to obtain a multi-path error, thereby realizing separation of the two types of errors. The troposphere delay and multi-path error separation method and the system of the invention support high-precision GNSS short-distance relative positioning under narrow geographical conditions of high mountains.

The invention provides a 5G base station time synchronization error elimination method based on a known coordinate reference station, and the method comprises the steps: adding one or more reference observation stations with known coordinates to receive a 5G system positioning reference signal, and transmitting the obtained TDOA value information to a system service platform in real time through a 5G communication link to participate in positioning calculation; the relative time synchronization error between the base stations is obtained; and the system service platform broadcasts the calculated relative time error parameters between the base stations as differential information to a user terminal to be positioned through a 5G communication link, the user terminal receives TDOA values and base station coordinates from the base stations and the relative time error parameters between the base stations broadcasted by the system service platform, and calculates the accurate position coordinates of a user to realize positioning. According to the invention, only one low-cost reference station receiving terminal is additionally arranged to eliminate the time synchronization error of the base station, and high-precision 5G positioning can be satisfied without using high-cost time synchronization such as an optical fiber.

The invention provides a seismic source rapid determination method and system based on a single rectangular model, seismic data and high-frequency GNSS data are combined to realize rapid determination of a seismic source, and the method comprises the following steps: a strong seismograph and a seismograph collect original motion information of a seismic station, a GNSS receiver collects basic navigation positioning information of a GNSS observation station, and the basic navigation positioning information is sent to a ground computing center; the ground computing center constructs a seismic station characteristic function time sequence, judges whether the seismic station is triggered by an earthquake or not, and accurately picks up the initial movement moment of the seismic station according to the seismic station characteristic function time sequence when the earthquake occurs; the ground computing center determines the epicenter position, and generates GNSS observation station motion information according to the basic navigation positioning information of the GNSS observation station; and the ground computing center estimates the point source magnitude, estimates the initial fault size of the single rectangular model according to the point source magnitude, inverts the parameters of the single rectangular model, and sends earthquake early warning information to an earthquake early warning center.

The invention provides a multi-fractal quantification method and system for terrain complexity in a three-dimensional scene. The method comprises the steps: obtaining slope data: building scatter points on the surface of a three-dimensional terrain of a selected region, obtaining elevation values, and obtaining the slope value of each scatter point according to the relation between the scatter points and the elevation values of adjacent points; carrying out multi-fractal calculation, namely respectively and correspondingly adopting a box-dimension multi-fractal calculation method for a plurality of order moment q values, dividing the established scatter point region into sub-regions with different scales, calculating gradient distribution probability measurement and score distribution functions in each sub-region, checking whether the multi-fractal characteristics of the scatter area are obvious or not by judging whether the logarithm of the fractal function and the logarithm of the scale are linearly related or not, and extracting and reserving corresponding singular indexes and fractal dimensions if the multi-fractal characteristics are obvious; drawing a multi-fractal spectrum according to the obtained multiple groups of singular indexes and fractal dimensions, extracting multi-fractal characteristic parameters through a multi-fractal spectrum image, and supporting quantitative expression of terrain complexity.

The invention discloses a binder-free preparation method for preparing a composite carbon-based fertilizer by using biogas residues, which comprises the steps of conducting pretreatment, specifically, collecting biogas residues after solid-liquid separation, air-drying and crushing to obtain dried biogas residue powder; performing high-temperature pyrolysis, specifically, transferring the pretreated dried biogas residue powder into a muffle furnace, and performing pyrolysis for a period of time under a pyrolysis temperature condition to obtain a biogas residue biochar material; preparing a biogas residue carbon-based compound fertilizer; and carrying out granulation preparation, specifically, placing the prepared composite biogas residue biochar base fertilizer in an extrusion die, and performing granulation under certain compression speed and pressure conditions to obtain the granular composite biogas residue biochar base fertilizer. The biogas residues are utilized for reproduction, and the environmental and economic burden is relieved; the process is simple, and the cost is low; a binder is not added, so that the method is environment-friendly; powder is granulated, so that the transportation and application difficulties are solved; and meanwhile, the granules are dried at high temperature after being formed, nutrients are tightly combined with charcoal functional groups, the nutrient release time is prolonged, and the fertilizer quality is improved.

The invention provides a method and system for quickly estimating surface albedo, and the method comprises a channel conversion model buildingprocess, including inputting an old channel surface albedo database, a surface albedo inversion result of a new satellite with a consistent space-time range and a corresponding solar zenith angle, converting the surface albedo, forming a new and old channel surface albedo mapping table based on a space-time nearest neighbor principle, constructing a multi-layer artificial neural network, and performing training according to the new and old channel surface albedo mapping table; a new channel surface albedo estimation process, including extracting the longitude and latitude of an observation target, the observation moment and the solar zenith angle during observation when atmospheric remote sensing is carried out based on new channel observation, searching an old channel database forthe surface albedo and the solar zenith angle of an old channel at the nearest distance and the nearest moment , and taking the surface albedo and the solar zenith angle as the input of an artificial neural network, obtaining the surface albedo corresponding to the new channel and according to the hypothesis of the surface reflection model, converting to the surface albedo corresponding to the solar zenith angle at the observation moment.

The invention provides a building inclination model layered household display method fusing a vector house type image, and the method comprises the steps: collecting the vector house type image data corresponding to a building inclination model, with the vector house type image data comprising the vector contour data of a building room and the corresponding attribute information, and the attribute information comprising the height information; importing the vector house type image data into a spatial database, and organizing the vector house type image data by adopting the spatial database; recording longitude, latitude and height information of a position point needing to be inquired in the three-dimensional geographic information system, conducting space matching on the longitude, latitude and height information, vector contour data stored in a space database and the corresponding height information to identify a corresponding unique building room; and extracting corresponding vector contour data and corresponding attribute information, drawing a three-dimensional space contour, and displaying a corresponding planar graph and the attribute information. According to the invention, data fusion application of different data sources is realized, and the problem of layered household display of a building inclination model in a three-dimensional geographic information system is solved; and the integrity and continuity of the building inclination model data are not influenced.

The invention provides an emergency crowd emergency evacuation simulation method based on a time-space interaction process, and the method comprises the steps: constructing a space-time interaction mechanism between an emergency and an individual and between individuals based on a unified space-time framework; based on time-space interaction mechanism analysis, constructing an individual time-space dynamic evacuation behavior, simulating a population evacuation process driven by an emergency by using a solid force model, and carrying out evacuation evaluation index analysis. According to the scheme, implementation is easy and convenient, practicability is high, the crowd evacuation behavior under the complex geographical scene is simulated, and the simulation result can provide reasonable and effective decision reference for emergency evacuation under emergencies; the problems of low practicability and inconvenience in practical application in related technologies are solved, the user experience can be improved, the great loss of manpower and material resources is avoided, and the method has important market value.

The invention provides a semantic point cloud acquisition and semantic octree map construction method based on division-to-division, which comprises the following steps of: performing semantic acquisition on an acquired image by using a first image semantic segmentation network to obtain a primary segmentation result; sequentially connecting an image restoration network and a second image semantic segmentation network in series behind the first image semantic segmentation network to form a sis processing framework; determining a processing area of the image restoration network, outputting a corresponding result to a second image semantic segmentation network by the image restoration network, and outputting a second segmentation result by the second image semantic segmentation network; according to a priority mechanism, optimizing and complementing to obtain a final optimized semantic segmentation result; images and depth information are obtained through a camera, and dense point clouds are obtained in combination with SLAM system pose information; in combination with depth information, an optimized image semantic segmentation result is mapped to dense point clouds to obtain dense point clouds with semantics, and an octree semantic map with more complete and more stable semantic information can be obtained.