73results about How to "Fast fitting" patented technology

The invention aims to provide a method for inversing the sea wind direction through navigation radar images based on relative radiation correction. The method comprises navigation radar image sequence collection, navigation radar image preprocessing, navigation radar image relative radiation correction and sea wind direction reversing. According to the method, the navigation radar image relative radiation correction link is added to the wind direction inversing process through wind stripes, and influence of radial damping of the navigation radar echo intensity on wind direction inversing is eliminated effectively. A self-adaptive Lagrange least square segmentation fitting correction method is adopted for the navigation radar image relative radiation correction, so that it is guaranteed that sea clutter image features are not destroyed after correction is conducted, and applicability of engineering is improved as well.

The invention refers to a coat laminating method for spray adhesives and a laminating device thereof, in particular to a coat laminating method for spray adhesives and a mechanism thereof that protect tools from pollution. The laminating method comprises preparing feed, spraying adhesives on top surface of the coat, heating and softening, pressing the coat into a lower module, absorbing a molding part at an upper module, laminating, demoulding a finished product and other steps. A rubber coating surface of the coat is upwards to prevent the tools or the molding part from being polluted by the softened sticking agent. Furthermore, binding capacity and steadiness of the laminating can be greatly improved and the laminating velocity can also be improved. The method has the advantages of improving competitive strength and economic benefits of the product.

The invention provides a deep text matching method and device based on sorting learning, and the method specifically comprises the steps: firstly, obtaining sentence pairs composed of assumed sentences and reasoning sentences, the reasoning sentences comprising positive reasoning sentences and a plurality of negative reasoning sentences, and the assumed sentences being related to semantics of thepositive reasoning sentences and not related to semantics of the negative reasoning sentences; secondly, carrying out corresponding processing on sentences in the sentence pairs, forming sentence vectors, calculating loss values of a preset loss function according to matching degree values among the sentence vectors, and adjusting parameters of a depth matching model according to the loss values;and finally, carrying out text matching on the input statement by utilizing the finally obtained depth matching model through parameter adjustment. The input sentence pairs are expanded into the sentence sequences from the two sentence pairs and comprise the positive example type data and the negative example type data, due to the fact that the input number and types of the model are expanded, thefitting speed of the model is increased, and the matching precision of the model can be improved.

On the soft thin film and glass base plate, which constitute the touch screen layer, with the corresponding circuit diagrams being arranged in desired positions, the preparation for etching circuits are producing the tin indium oxide conducting layer are carried out. The location points are set up the four corners on the soft thin film and glass plate. The film is integrated to the glass plate closely by using the automatic lining machine. The laser cutter removes the unused soft thin film on the circumference of the circuit pattern. Then, the glass plate cutter cuts the glass plate to desired pieces. The invention increases the qualification rate and the production efficiency.

The invention discloses a 3D curved glass laminating device and method. The device comprises an upper pressing module, a lower pressing module and a heating module. A cell phone liquid crystal screenis fixed to the upper pressing module. Curved glass is fixed to the lower pressing module. The lower pressing module comprises a curved jig, an ejector pin device and an adjustable positioning device.The curved glass is placed on the surface of the curved jig. The ejector pin device is hidden under the surface of the curved jig in the laminating process and driven by a driving device to extend out of the surface of the curved jig so as to eject a laminated product. The adjustable positioning device is arranged on the surface of the curved jig to adapt to the appearance, size and structure ofthe to-be-laminated curved glass so that the to-be-laminated curved glass can be fixed to the proper position of the surface of the curved jig. The device can be flexibly suitable for laminating curved glass products of different curving forms and size structures.

The invention provides a differential evolution-based double-exponential thermal wave dynamic image sequence fitting reconstruction method. The method comprises the steps: designing an embedded defective test specimen and applying instantaneous thermal excitation to the specimen, and collecting temperature field change condition in real time by a thermal imager; selecting a part of typical pixel points to acquire a temperature-time change sequence and graphing; according to the shape of the sequence and fitting function, selecting a fitting model and fitting; evaluating through fitting evaluation parameter values, fitting all pixel points in an original image by adopting the model, and reconstructing the image according to the fitting coefficient. Compared with the prior art, the method has the beneficial effects that original data can achieve higher precision without pretreatment; the speed is faster, temperature sequence of single pixel point containing 100 data is fitted only for 10-150milliseconds; and under the condition of not influencing the effect, only four fitting coefficients are provided by the method, the image compression efficiency is higher, and the storage space can be further reduced.

The invention relates to an attaching method for a back-adhesive skin and an attaching device thereof, in particular to an attaching method for the back-adhesive skin internally arranged in a vehicle door panel for avoiding smearing equipment. The attaching method comprises the steps of preparing materials, pressing the skin in a lower die, adsorbing a molding component to an upper die, pressing in, die unloading of finished products, etc. The coating of the vehicle door panel skin is arranged to face the molding component, so that the pollution of the equipment or the molding component caused by the adhesive can be avoided. Furthermore, the method and the device have the advantages of greatly enhancing the attaching binding force and stability, improving the attaching speed and strengthening the competition and economic benefits of the product.

The invention relates to an automatic film laminating machine, which comprises a laminating machine support, a workbench disposed on the laminating machine support, and a machine shell distributed on one side of the workbench. The laminating machine is characterized in that: a turntable is embedded on the workbench, a first to-be-laminated workpiece working position and a second to-be-laminated workpiece working position are correspondingly distributed on the turntable, on which observation holes are distributed at the same time; the first to-be-laminated workpiece working position and the second to-be-laminated workpiece working position are both provided with an independent upward pushing positioning assembly below them; the upward pushing positioning assembly is equipped with an image positioning device; a pressure head assembly is arranged above the working end of the turntable; separating rolling wheels are distributed at the input end of the pressure head assembly, and a downward departing material receiving disc as well as an initial Mylar disc are distributed at one lateral end of the separating rolling wheels; an upward departing material receiving disc is distributed at the output end of the pressure head assemble; and the first to-be-laminated workpiece working position and the second to-be-laminated workpiece working position are connected to an air pressure pump. Thus, image recognition can be used to adjust a Mylar position, and finally the laminating pressure head is controlled for intelligent and precise lamination. Therefore, the invention is able to adapt to fast processing production on an assembly line.

The invention discloses a wind power prediction method based on a morphological filter and a similar section search algorithm, comprising the following steps: (1) designing a high-frequency filter; (2) for an average trend component m(t) which is a low-frequency component, replacing global prediction with a local prediction method, searching for similar sections based on Euclidean distance, and using the similar sections to train a least squares support vector machine model so as to perform prediction; (3) using a local prediction method (namely, according to certain principles) to select part of data from historical data for modeling and prediction, and adopting a similar section search algorithm (namely, a 'secondary similarity' algorithm); and (4) synthesizing the prediction results of two independent components to get a final result of wind power prediction. The method has the advantages of clear physical significance of wind plant prediction power prediction, stable prediction result, and high prediction precision.

The invention discloses a CRTSIII track plate-based shape size quick calculation method. The method comprises the steps of 1) importing a standard three-dimensional design model of a CRTSIII track plate in point cloud processing software; 2) by utilizing the point cloud processing software, establishing a direct plane feature, a direct plane circle feature and a direct boundary comparison point ofthe three-dimensional model of the track plate; 3) by utilizing the point cloud processing software and the established direct plane feature, direct plane circle feature and direct boundary comparison point, establishing features such as an auxiliary point feature, an auxiliary linear feature, an auxiliary plane feature and the like of a CRTSIII track plate shape size detection item; 4) by utilizing the point cloud processing software and the established point feature, plane circle feature, linear feature and plane feature, establishing a direct distance feature and a direct angle feature ofthe CRTSIII track plate shape size detection item; and 5) by utilizing the point cloud processing software, cutting the direct plane feature of the three-dimensional model of the track plate.

The invention provides a screen protective layer mounting device and a using method of the device. The screen protective layer mounting device is convenient to operate, excellent in pasting quality and high in working efficiency. The screen protective layer mounting device comprises a base plate (1), wherein the upper plane and the lower plane of the base plate (1) intersect to form an included angle, at least two positioning protruding blocks (2) are arranged on the base plate (1), an area defined by the at least two positioning protruding blocks (2) is the same as equipment with a protective layer to be pasted in shape and right clamps the equipment with the protective layer to be pasted, a boss (3) is arranged on the inner edge of each positioning protruding block (2), the height of each boss (3) is consistent with the thickness of the protective layer to be pasted, the size of an area defined by the bosses (3) is the same as the screen size of the equipment with the protective layer to be pasted, and a tight absorbing layer is arranged in the area defined by the positioning protruding blocks (2). The method is easy to operate, mounting is rapid, and mounting efficiency and mounting precision can be effectively improved. The screen protective layer mounting device and method can be applied to the mounting field of protective layers.

The invention provides a method for manufacturing a circuit board with a superouter fitting, which comprises the following steps of: providing a mother board of a circuit board and manufacturing and forming a plurality of circuit board units on the mother board of the circuit board; forming a plurality of slots penetrating through the mother board of the circuit board on the mother board of the circuit board, wherein each circuit board unit is adjacent to at least one slot; providing a fitting mother board for forming the superouter fitting and adhering the fitting mother board to at least one part of area of the circuit board units and covering each slot; and simultaneously cutting the mother board of the circuit board and the fitting mother board to obtain a plurality of mutually-separated circuit boards.

A Doppler parameter quadratic fitting method based on time division multiplexing comprises the steps of (101) establishing a quadratic fitting unit which comprise a multiplying unit a2, a multiplying unit b2, a multiplying unit c2, a multiplying unit d2, a multiplying unit e2, a divider a2, a summator a2, a mean value module a2, a mean value module b2 and a mean value module c2; (102) calculating fitting parameters m, m21, m22, m23, m24 and m25 by means of the quadratic fitting unit with the abscissa xi of a scatter signal as the input end; (103) calculating the fitting coefficient summation by means of the quadratic fitting unit based on the fitting parameters m, m21, m22, m23, m24 and m25; (104) achieving quadratic curve fitting of the scatter signal based on the fitting coefficient summation. Compared with traditional quadratic fitting, the Doppler parameter quadratic fitting method based on time division multiplexing has the advantage that the number of hardware resources needed is reduced greatly by means of time division multiplexing of hardware resources in the quadratic fitting unit.

The invention discloses a rapid fitting method of a human body model. The method comprises the following steps: establishing a standard human body model in an initial posture; obtaining posture and body shape parameters of a target human body model, wherein the three-dimensional human body posture and body shape parameters correspond to skeletons and a plurality of base parameters of a three-dimensional standard human body model; selecting an initial posture base of a standard three-dimensional human body model according to the obtained posture parameters of the target human body model; according to the obtained bases and skeleton parameters of the plurality of groups of target human body models, starting fitting from the selected initial posture base; obtaining a three-dimensional human body model with the same body shape as the target human body; and obtaining a three-dimensional target human body model grid with the same posture as the target human body two-dimensional image. The initial posture bases of the three-dimensional standard human body model can be a plurality of preset initial posture bases including T-pose. According to the method, the fitting process of the standard human body model and the target human body model is optimized, a plurality of initial posture bases are used for replacing a traditional single T-pose initial posture, the process of driving the skeleton to move is optimized through the variable speed frame interpolation method, and the whole fitting process achieves good balance in time and effect.

The invention discloses a financial big data system based on online deep learning and market operation value quantification. The system includes a person flow stay perception module, a shopping mall BIM space module, a shopping mall value super index evaluation module, a shopping mall value super index evaluation data set storage and instant training module, and a financial big data and shopping mall BIM visualization module. A subjective operation value quantized value is fitted through passenger flow data and business volume personalized data, an optimized convolutional neural network is trained in real time for processing a large amount of data sampled during operation, and real-time quantification of the operation value of a shopping mall is realized. According to the invention, fairness from objective data to subjective rating is ensured, data desensitization is realized, network training is convenient, and real-time monitoring and evaluation of managers are facilitated.

The invention discloses a recommendation method integrated with a curiosity mechanism of a user, which comprises the following steps of: calculating novelty and conflict degree according to historicalrecords, then obtaining stimulation degree through weighted summation, and training respective wundt curve of each user according to a stimulation degree list of the historical records; performing Learning by using an accuracy-based recommendation method to obtain a relevancy list; Calculating curiosity of the article to the user according to the von special curve, and performing sorting to obtain a curiosity list of each user; And finally, sorting the two lists by using a borda count method.

The invention discloses a fast hearing aid fitting device comprising a hearing aid, a processor, a DSP chip, a digital analog converter, a sound converter and a sound output device. The hearing aid fitting method comprises the following steps: S1, inputting a listening test chart; S2, selecting a listening program according to the listening test chart; S3, enhancing or attenuating the sound processed by the processor according to the feedback of a patient; S4, acquiring the listening program conforming to the patient and a corresponding enhancement or attenuation value; and S5, configuring the hearing aid with corresponding data according to the parameters in step S4. The patient can be quickly matched with a hearing loss type in the corresponding DSP chip to achieve a fast matching effect, and the fast hearing aid fitting device can greatly shorten the fitting and hearing compensation experience time of the hearing aid.

The invention discloses a cleaning and laminating all-in-one machine for a ceramic disc and a wafer. The all-in-one machine comprises a cleaning device, at least one group of laminating devices and afeeding device positioned on one side of the laminating devices, wherein the cleaning device comprises a feeding station, a cleaning station, a discharging station and a first material moving assembly; the laminating device is located on one side of the cleaning device and is in butt joint with the cleaning device through a material taking assembly; the laminating device comprises a preheating station, a laminating station, a cooling station, a discharging station and a second material moving assembly; feeding devices and the laminating devices are consistent in number and correspond in a one-to-one mode; and each feeding device comprises a storage station, a washing station, a wax throwing station, a baking station, an overturning station and a mechanical arm structure. The cleaning operation of the ceramic disc and the processing operation of a sapphire wafer can be carried out at the same time, then the processed sapphire wafer is laminated to the surface of the ceramic disc, and therefore the laminating efficiency of the sapphire wafer is greatly improved.

The invention relates to a teaching evaluation sentiment analysis method fusing cognitive migration, and belongs to the field of education informatization. The method comprises the following steps: cleaning and preprocessing collected data; performing word embedding on an unstructured teaching evaluation text by using a BERT language model, inputting the unstructured teaching evaluation text into a bidirectional GRU network for learning, and outputting context hidden memory information of the text; the method comprises the following steps: modeling historical practice records of students by using a neurocognitive diagnosis model, and obtaining personalized cognitive competence vectors of the students as prior knowledge characterization; the extracted text hidden memory information and the prior knowledge information are fused into a cognitive migration matrix, and the prior knowledge is migrated into the text memory through an attention mechanism, so that the real emotional characteristics contained in the text are enhanced. According to the method, sentiment classification can be accurately carried out on the teaching evaluation text carrying the conflicting sentiment features, and the problem of sentiment misclassification caused by existence of the conflicting sentiment features in teaching evaluation is solved.

The invention discloses a long-acting anti-fouling fabric and a preparation method thereof, and relates to the technical field of fabrics. The long-acting anti-fouling fabric prepared by the invention sequentially comprises the Lycra cloth, the foamed rubber and the anti-fouling film from inside to outside; the anti-fouling film is prepared by reacting hyperbranched polyglycidyl ether with modified cyclodextrin; the modified cyclodextrin is prepared by grafting a hydroxylamine polymer to sulfydryl-carboxymethyl-cyclodextrin, so that the stability is enhanced, a large number of hydroxyl groups are introduced, the hydrophilicity is enhanced, the fabric has stain resistance, the density of a stain-resistant film is also enhanced, and the tear resistance of the fabric is further enhanced; the foaming rubber is prepared by blending chloroprene rubber, polyurethane rubber and fluoropolyether nitrile, vulcanizing and foaming; after the polyurethane rubber is added and blended, the vulcanization speed is increased, the tear resistance of the foaming rubber is improved, the Lycra cloth, the foaming rubber and the anti-fouling film are tightly connected, the peel strength of the fabric is enhanced, and the service life of the fabric is prolonged.

The invention discloses a human body model driving method, and the method comprises the steps: obtaining the posture and body shape parameters of a target human body model, and enabling the three-dimensional human body posture and body shape parameters to correspond to the skeleton and a plurality of basic parameters of a three-dimensional standard human body model; establishing a standard human body model in an initial posture; inputting the obtained multiple groups of basis and skeleton parameters into a standard three-dimensional human body model for fitting; obtaining a three-dimensional human body model with the same body shape as the target human body; driving the skeleton to move from the initial posture to the target posture; and obtaining a three-dimensional target human body model grid with the same posture as the target human body two-dimensional image. According to the method, fitting of the standard human body model and the target human body model is optimized, an animation series from the initial posture to the target posture is formed, and the process of driving the skeleton to move is optimized by using the variable-speed frame interpolation method, so that the whole fitting process achieves a good balance in time and effect.

The invention discloses a sparse SAR target classification method and device based on transfer learning. The method comprises the following steps: (1) reconstructing a sparse SAR image based on a matched filtering SAR image by using a BiIST algorithm; (2) constructing convolutional neural networks with the same structure in a source domain and a target domain, and pre-training a neural network in the source domain by using the simulation data set; (3) migrating partial parameters in the neural network pre-trained in the step (2) to a network of a target domain, and randomly initializing other parameters; and (4) finely adjusting the network, taking the sparse SAR image obtained in the step (1) as input data for training, and outputting an obtained target classification result and accuracy. The sparse reconstruction algorithm adopted by the invention can effectively suppress the side lobes and clutters of the SAR image, improve the quality of the image, and provide guarantee for the training of a subsequent classification network; according to the sparse SAR target classification method based on transfer learning, the speed of network training convergence can be increased, and the target classification precision is further improved.

The invention discloses a mobile terminal replacement prediction method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining business data, constructing RandomForst and XGBOOST models, converting leaf node positions of a sample in the models into one-hot coding features, and generating a first feature; performing feature combination on the first features to generate second features, and combining the first features and the second features; selecting the merged features by adopting an integrated filtering method, and removing the features with poorcharacterization capability; constructing a sample weight function, and optimizing a model loss function; and constructing a logistic regression random forest based on a bagging framework, and performing mobile terminal replacement prediction through a model. According to the technical scheme, the method has the advantages of being high in representation capacity, low in cost, accurate in prediction and high in robustness, operators can conveniently and accurately market mobile phones to users, and high commercial value is achieved.

The invention relates to a wireless portable surgery force line track measuring system in unicondylar knee replacement, which is mainly composed of two parts, one part is a data acquisition, processing and wireless sending module, and the other part is a data receiving and real-time display module. The data acquisition, processing and wireless sending module mainly comprises a customized pressure sensor array, an analog-to-digital converter, a multi-channel analog switch, an algorithm hardware implementation circuit and a system-in-package chip made based on a special control logic and interface chip and a radio frequency transmitting chip. A force line track measuring algorithm comprises a pressure maximum point coordinate determination algorithm based on least square fitting and trajectory fitting. The force line track generated by the distal end of the femoral prosthesis on the prosthesis gasket shell in the unicondylar knee replacement surgery can be measured in real time, so that whether the prosthesis position is proper or not is judged.

A Doppler parameter quadratic fitting method based on time division multiplexing comprises the steps of (101) establishing a quadratic fitting unit which comprise a multiplying unit a2, a multiplying unit b2, a multiplying unit c2, a multiplying unit d2, a multiplying unit e2, a divider a2, a summator a2, a mean value module a2, a mean value module b2 and a mean value module c2; (102) calculating fitting parameters m, m21, m22, m23, m24 and m25 by means of the quadratic fitting unit with the abscissa xi of a scatter signal as the input end; (103) calculating the fitting coefficient summation by means of the quadratic fitting unit based on the fitting parameters m, m21, m22, m23, m24 and m25; (104) achieving quadratic curve fitting of the scatter signal based on the fitting coefficient summation. Compared with traditional quadratic fitting, the Doppler parameter quadratic fitting method based on time division multiplexing has the advantage that the number of hardware resources needed is reduced greatly by means of time division multiplexing of hardware resources in the quadratic fitting unit.

The invention discloses an oil-immersed insulation paper board partial discharge failure probability characterization method based on a double-risk competition model. The method comprises the steps: carrying out partial discharge processing on a plurality of oil-immersed insulation paper board samples under a certain external voltage amplitude, and recording the time from the start of the partialdischarge of each sample to the breakdown of the paper board and the number of partial discharge times in the time; respectively fitting the two variables by adopting three-parameter Weibull distribution; and finally, calculating an oil-immersed insulation paperboard double-risk competition insulation failure probability model considering time and discharge times by adopting the double-risk competition model, and carrying out visual representation. According to the method, the insulation failure time and the insulation failure discharge frequency are comprehensively considered, and the insulation failure probability of the oil-immersed insulation paper board under the partial discharge effect is predicted from the perspective of double-risk competition, so that the insulation failure probability of the oil-immersed insulation paper board under the partial discharge effect can be effectively and comprehensively calculated, and the one-sidedness of a univariate model is avoided.

The invention belongs to the field of artificial intelligence, and provides a user classification method and device based on a random forest, equipment and a storage medium. The method comprises the following steps: acquiring a training sample set, wherein training samples in the training sample set comprise user features and user classifications; clustering the training samples according to user features in the training sample set, and determining a sample subset according to a clustering result; respectively constructing corresponding classification trees according to the determined sample subsets; and determining a random forest according to the classification tree, and classifying to-be-classified users according to the random forest. Through the recommendation method, the construction efficiency of the random forest can be effectively improved, the similarity problem of classification trees can be effectively reduced, generation of a sample subset missing training samples is avoided, and the classification precision of the random forest is improved, so that user classification can be more accurately performed, and the efficiency of obtaining required products by users is improved.

The invention relates to a subway passenger analysis method and system based on data mining. According to the method, source data is obtained through passenger travel transaction records, processing and multi-dimensional analysis of the source data are carried out, the more accurate and effective classification of subway passenger travel is achieved, meanwhile the travel track of a passenger is provided, and a reliable data basis is provided for planning of rail transit on the basis of ensuring network large passenger flow operation safety. According to the technical scheme, the final clustering number in the clustering process is selected by calculating a deviation coefficient, the proper clustering number can be quickly selected according to the data characteristics, a clustering result more conforming to passenger distribution rules is obtained, the travel track of the passenger is analyzed by using a space-time analysis method, the address and the working place of the passenger and the approximate distance between the address and the working place are calculated, the travel condition of the subway passenger is analyzed more deeply, and a reliable data basis is provided for planning of rail transit.

The invention discloses a touch screen laminating method and a glue coating device thereof. The method comprises the following steps: (1) coating glue to the whole surface of a cover plate by using a coating knife, the glue being planar; (2) irradiating the cover plate coated with the glue by using a UV lamp, activating the glue, and curing the glue until the glue is in a jelly shape and is not easy to flow; and (3) in a vacuum environment, attaching the display screen to the cover plate, wherein no bubble is achieved. The glue on the surface of the cover plate is planar, the thickness of the glue is uniform, the laminating speed of the cover plate and the display screen is high, and bubbles are thoroughly removed; and after the glue on the cover plate is in the jelly shape, the cover plate is attached to the display screen, so that the glue is poor in flowability and does not overflow, and the product quality is high.

The invention discloses a short-term load prediction method based on a long short-term memory network, and the method comprises the steps: collecting short-term load data, taking the collected short-term load data as a training sample, and carrying out the VMD decomposition, and obtaining n random component IMFs; for the n decomposed components, constructing a corresponding WOA-LSTM model for each component, optimizing LSTM parameters in the models by using WOA, and finding out the most suitable parameters; training the component data IMF by using the LSTM with the optimal parameter to obtain a prediction result of each component; based on the LSTM models constructed after respective optimization, the LSTM models are used to measure the short-term load data, and each LSTM model outputs a corresponding prediction result; and adding all prediction results to obtain a final load prediction result.