66results about How to "Overcoming the problem of low precision" patented technology

The invention discloses a method and system for automatic focusing of a laser projector. The method includes S1, the laser projector projecting at least one light beam into space; S2, a collecting module collecting light spot images in real time; S3, a recognizing module recognizing the laser light spots according to second derivative of distribution of the light spot strength; S4, a computation module computing parameter values of definition of the light spot images; S5, a focusing module focusing relative relation of a light source constantly and repeating the step 1-4; S5, the focusing module comparing the parameter values acquired in real time to acquire the optimum parameter value, and focusing the light source to the optimum position according to the optimum parameter value. With the method and the system, automatic focusing of the laser projector can be realized, the problem of low precision brought by manual focusing in the prior art is solved, and meanwhile, focusing efficiency is greatly improved.

The invention discloses a model training method, device, system and storage medium. The method is applied to a model optimization node. The method includes: calling an intelligent contract stored in ablock chain to obtain a first block chain address of a data analysis model, wherein, the first block chain address is obtained and written into the intelligent contract by a model owning node after the data analysis model is written into the block chain; Querying the block chain according to the first block chain address to obtain a data analysis model; training data analysis model using preset first training data set to obtain model optimization information; Writing the model optimization information into the block chain to obtain a second block chain address; the second block chain addressis written into the intelligent contract, and the second block chain address is used for the model owning node to obtain the model optimization information. By adopting the technical proposal of the disclosed embodiment, user privacy can be ensured, and the problem of low precision of a model obtained by training due to insufficient number of training data sets in the related art can be overcome.

The embodiment of the invention relates to the field of communication transmission, and particularly discloses a method for acquiring synchronous time, a device and a system thereof. The method comprises the following steps: receiving pseusynchronous time information, processing the pseusynchronous time information and obtaining pseusynchronous time T1; acquiring synchronous time information from a synchronous time source, processing the synchronous time source, obtaining synchronous time T2, calculating a time difference value Delta T of the synchronous time T2 and the pseusynchronous time T1 and storing the difference value Delta T; and when failing to acquire synchronous time information from the synchronous time source, correcting the pseusynchronous time T1 with the stored time difference value Delta T, obtaining corrected time T3 and selecting, and outputting the corrected time T3 as synchronous time of equipment needing to synchronize time. The invention solves the problem that in the prior art, the transmission delays of the two-way transmission channel of the transmission network are different so as to cause low accuracy and the like of the synchronous time obtained by operating the network time protocol.

The invention discloses a regional traffic safety evaluation method based on a connectivity-based spatial weight matrix, and the regional traffic safety evaluation method belongs to the technical field of traffic safety. The regional traffic safety evaluation method comprises the steps of: firstly, acquiring relevant data of regional traffic safety evaluation through data investigation; secondly, establishing a Bayesian conditional autoregressive model, which regards a traffic analysis zone as a basic unit, as a regional accident risk prediction model, and establishing the spatial weight matrix which marks network connectivity by means of the number of paths contained in a region; and finally, carrying out model parameter estimation. The regional traffic safety evaluation method overcomes the problem that the model precision is not high because a traditional accident risk prediction model does not consider spatial relativity, the connectivity-based spatial weight matrix takes the influence of geographic dimensionality and actual traffic links between regions into account, the regional accident risk prediction model has good data fitting degree and data adaptability, and more significant factors associated with accident occurrence can be dug out, thereby effectively guiding practical application.

The invention discloses a wifi and geomagnetism fused indoor positioning system and method based on an HMM (Hidden Markov Model). The method comprises the following steps: matching a wifi signal acquired by a mobile terminal in real time with a pre-established wifi fingerprint library to obtain a local position area; initializing the position in the area to obtain an initial probability distribution matrix; obtaining a state transition probability matrix by using a position relationship of label points in the local position area; and fusing a geomagnetic signal with a pre-established geomagnetic fingerprint library to obtain an observation probability matrix, and estimating the position of a user according to a forward algorithm. When the user is positioned in the local position area determined by the wifi signal, the current real-time geomagnetic signal is used to update the observation probability matrix, and the position of the user is estimated by the observation probability matrixand the state transition matrix; otherwise, the wifi signal is introduced to re-establish the local position area, and the initial probability distribution matrix, the state transition probability matrix and the observation probability matrix are updated according to the preceding method. According to the wifi and geomagnetism fused indoor positioning method based on the provided by the invention, various information is effectively fused, and the positioning accuracy of the system is improved.

The invention provides an automatic focusing method and system for a light source projector. The method comprises the steps that S1 a light source projector projects at least one light beam to the space; S2 an acquisition module acquires the light spot image of the light beam; S3 according to the acquired light spot image, a calculation module calculates definition parameter values; and S4 a focusing module compares the parameter values acquired in real time to acquire the optimum parameter value, and adjusts a light source to the optimum position according to the optimum parameter value. According to the method and the system, automatic focusing of the light source projector is realized; the problem of poor precision caused by manual focusing in the prior art is overcome; and the focusing efficiency is greatly improved.

The invention provides an array antenna electromagnetic characteristic solving method based on finite element region decomposition, and aims to give consideration to the calculation precision, solvingcapability and efficiency of solving the electromagnetic characteristics of a large array antenna at the same time, and the method comprises the following specific steps: (1) creating a unit geometric model; (2) setting materials and boundary conditions; (3) generating unit grids; (4) translating and copying the unit grids; (5) replacing the material and boundary attributes of the non-structuralunit; (6) solving by adopting a finite element region decomposition method; and (7) calculating electromagnetic characteristics. According to the invention, a finite element region decomposition method is applied to electromagnetic characteristic solution of the array antenna, so that the difficulty in geometric modeling and mesh generation of the large array antenna is reduced, and both the solution precision and the solution efficiency are considered.

Provided is a meteorological data complement method based on automatic convolutional encoding and decoding algorithms. The method comprises the steps that meteorological data is preprocessed; the number of iterations is set; automatic convolutional encoding is conducted on a four-dimensional matrix; automatic convolutional encoding is conducted on a feature matching matrix; automatic convolutional encoding is conducted on the feature matching matrix; automatic convolutional decoding is conducted on the feature matching matrix; automatic convolutional decoding is conducted on the feature matching matrix; automatic convolutional decoding is conducted on the feature matching matrix; a loss function is calculated; a convolution kernel and an offset weight are updated; a meteorological data attribute value is updated; meteorological data complement is completed. According to the meteorological data complement method based on the automatic convolutional encoding and decoding algorithms, the problem that the precision of data complement is low is solved, and the method has the advantages of being high in robustness and missing value filling accuracy.

The invention discloses a polarized light orientation method of an unmanned plane under the cloudy conditions. The method comprises the following steps of extracting color features with the highest distinguishing degree from all pixel points in a collected image; performing super pixel division on the images; obtaining a plurality of super pixel blocks; distinguishing the sky regions and cloud layers for polarization orientation from the divided images according to the color features through clustering; by using an expansion method, processing the images; reducing the extracted sky region; eliminating the cloud layer edge interference; according to the Rayleigh scattering model and images, building a relationship model of a sun direction vector and polarization angle and the solar zenith angle under the unmanned vector coordinate system; according to the azimuth angle and sun meridian line direction and relationship model, obtaining the course angle of the unmanned plane vector in thegeographic coordinate system. The problem of poor adaptability in the prior art is solved; the positioning precision of the unmanned plane under the cloudy conditions can be improved.

The invention discloses a high-score image change detection method based on color space smoothing and an improved frequency tuning salient model, comprising the following steps: obtaining an image difference map and a logarithmic ratio map, decomposing the difference map and the logarithmic ratio map, and then fusing the decomposed image to obtain a fused difference image; on the basis of the fusion difference image, the color space smoothing based on Mean Shift being applied to the fusion difference image to remove the disturbance region and retain the change region in the fusion difference image. Aiming at the smoothed fusion difference image, the improved frequency tuning saliency model is used to protect the edge of the fusion difference image, and the difference image with large variation separability is obtained. Based on the two-dimensional histogram threshold algorithm, the change information of the difference image with large change separability is extracted, and the change result is post-processed by the morphological algorithm to obtain the final change result map. The invention has high detection precision and guarantees the integrity of the detection result.

The invention relates to a high-precision ultralow frequency six-dimensional force micro-vibration measuring system, and the system belongs to a field of high-precision measuring system technology. The problem that the measuring system in prior art has low measuring precision and inflexible installation interface is solved. The measuring system comprises a box body, a load disc, four vertical piezoelectric sensors, six horizontal piezoelectric sensors, six lateral positioning plates, a signal adjustment device and a data acquisition and processing system; wherein the box body is formed by a shell body and a base, the load disc is formed by a bottom disc and a mounting column, the vertical piezoelectric sensors are impacted by a top surface in the shell body and a top surface of the bottom disc, the horizontal piezoelectric sensors are respectively impacted by the side surfaces of the bottom disc and the side surfaces of the lateral positioning plates, charge signals measured by ten piezoelectric sensors are converted into voltage signals through the signal adjustment device, the voltage signals are analyzed and calculated by the data acquisition and processing system, so that three dynamic forces and three dynamic bending moments of a disturbance source are obtained. The measuring system can precisely measure the disturbance force of the disturbance source, the installation interface is flexible and easy to operate, and the system has high measurement reliability.

The invention discloses a photoetching hot area detection method based on federal personalized learning. The method comprises the following steps that global model parameters returned by each node by a central server are aggregated, common characteristics of each node are fused, the global model parameters are updated, and the latest global model parameters to each node are fed back; each node downloads a global model parameter from the central server, and then trains a local model parameter by using local data to find the optimal local model parameter under the current global model parameter so as to overcome model heterogeneity and data heterogeneity of different nodes; and after the local model parameters are finely adjusted, the nodes train all the parameters by using local data to find the optimal current parameters for searching common features of different nodes. According to the method, the problem of model overfitting caused by too little local data is solved; data between chip design manufacturers is protected, and privacy protection is achieved; and the stability and the overall precision of the federal personalized learning model in the heterogeneous environment are improved.

An SAR image segmentation method of high-order multi-scale CRF semi-supervision is proposed. The method comprises the following steps of (1) inputting an SAR image; (2) carrying out the wavelet transform on the image; (3) obtaining the eigenvectors of a high-order multi-scale CRF; (4) obtaining the local conditional probabilities of the high-order multi-scale CRF; (5) carrying out the initial segmentation of the SAR images; (6) calculating the edge probability of each pixel point; (7) calculating the joint posterior edge probability of each pixel point; (8) calculating a posterior edge probability of each pixel point; (9) segmenting the SAR images; (10) ending segmentation. By taking into account the high-order potential energy and the inter-scale potential energy of each pixel point whencalculating the posterior edge probability, the method of the invention fully utilizes the spatial context structure information and improves the accuracy of the segmentation result.

The invention discloses a user load classification method based on an optimal clustering number and an apparatus thereof. The method comprises the following steps of acquiring time sequence curve data of a user load and carrying out normalization processing; extracting a user electricity usage behavior characteristic; using an Renyi entropy gradient descent method to carry out optimal clustering number training; and using the optimal clustering number to carry out hierarchical clustering. In the invention, a similarity and relevance among different users are comprehensively considered; a problem that precision of a traditional Euclidean distance similarity during high dimension data processing is not high is overcome; the different clustering numbers of Renyi entropy values are calculated so that the optimal clustering number can be found; a disadvantage that the clustering number needs to be set in advance in a traditional algorithm is overcome; operation efficiency is increased and operation cost is reduced.

The invention provides a remote sensing image rotation target detection method based on multistage fusion and angular point offset. The method is used for solving the technical problems that in the prior art, the detection accuracy of targets with different scales is low, and the running speed of the detection process is low. The method comprises the following implementation steps of: 1, acquiring a minimum enclosing rectangle of a rotary marking box of each target; 2, generating a training set; 3, constructing a deep full convolutional neural network; 4, training the deep full convolutional neural network; 5, detecting a rotating target in the image; 6, post-processing the box of the rotating target; and 7, drawing the final rotation detection boxes of all targets at corresponding positions in the image to obtain a detection result graph. According to the invention, targets with different scales in the image can be better distinguished, false detection is reduced, and the precision of the target box after image target detection is improved.

Embodiments of the invention provide a video click rate rating prediction method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining information of ato-be-rated video; wherein the information of the to-be-rated video comprises video content quality information used for describing the content quality of the video; inputting the information of the to-be-rated video into the video click rate rating model, and predicting the click rate rating of the to-be-rated video in a preset future time period; wherein the click rate rating reflects the clickrate; wherein the video click rate rating model is obtained by training according to the click rate rating of the first video and the information of the first video, the first video is a video with aknown click rate rating, and the information of the first video comprises video content quality information. According to the video click rate rating prediction method and device, the electronic device and the storage medium provided by the embodiment of the invention, the video click rate can be accurately predicted, and advertisers can be helped to improve the advertisement putting precision.

The invention discloses a multi-time multispectral image feature level information fusion method based on iterative implicit regularization. The method comprises the following steps: (1) inputting twomulti-time multispectral images; (2) realizing super pixel extraction; (3) performing unsupervised clustering on the superpixels to generate grouping information; (4) realizing feature extraction; (5) performing operation on the extracted features to generate a pseudo tag; (6) forming a training sample set according to the sample data, the pseudo label data and the group information; (7) selecting and initializing a classifier; (8) judging whether a termination condition is met or not, if yes, executing (10), and if not, executing (9); (9) training a classifier by using a method based on iteration implicit regularization, and returning to the step (8) after the classifier training is finished; and (10) classifying all the features by using a classifier to obtain a final fusion result image. According to the method, the change information of the two hyperspectral images can be distinguished through fusion, and different change types can be distinguished.

The invention discloses a locomotive running gear fault detection method and device based on time sequence images and a storage medium, and the method comprises the steps: obtaining a locomotive running gear time sequence image data set, and unifying the size; selecting a plurality of normal fault-free sequence images, and performing gray scale difference of two adjacent images in a time sequenceto obtain a time sequence residual data set; calculating an image gray average value in the time sequence residual data set to obtain a standard template; extracting a plurality of images from the time sequence residual data set, and calculating the structural similarity between the images and a standard template based on an image structural similarity algorithm to obtain a similarity matrix; calculating parameters [theta] and [epsilon] based on the similarity matrix; for the to-be-detected image M, calculating the gray difference M' between the to-be-detected image M and the image adjacent tothe to-be-detected image M in time sequence; and calculating the structural similarity MS between M'and the standard module, if MS theta is less than epsilon, determining that M is a normal image, and otherwise, determining that M is a fault image. The method can solve a problem that a single standard template cannot effectively detect a fault when the shape of a to-be-detected object changes with time.

The invention relates to an optical scanner controller support comprising: a lifting device comprising a base and slide bars mounted in parallel on both sides of the base, wherein a lead screw is arranged in parallel to the slide bar at one side and is connected with a lifting assembly, and two parallel lifting plates are arranged on the slide bar at one side and the slide bar and lead screw on the other side; a pitching device including a bracket connected to the lifting plates on both sides through shafts, wherein a pitching assembly is installed on a lifting plate on one side, the lead screw drives the lifting plates and the bracket connected to the lifting plates to slide up and down along the slide bars, and the pitching assembly is connected to a gear structure by the bracket and the rotation of the bracket is driven by rotating the pitching assembly; and a counterweight device including a counterweight steel rope connected with the lifting plates on both sides, wherein the other end of the counterweight steel rope is connected with a balancing weight as heavy as a lifted object. The optical scanner controller support can realize the change of the height of an optical scanner to a work surface and the emergent light angle of the optical scanner by the movement of the mechanical structure.

The invention discloses a method for determining recommended information and a related device. The method comprises the following steps: querying a target input method portrait, at least comprising a first tag of a target input method, of a target user; searching a first target information flow label corresponding to the first label of the target input method through a corresponding relation between an input method label system and an information flow label system; and based on the first target information flow label and the multiple pieces of candidate recommendation information, determining target recommendation information from the multiple pieces of candidate recommendation information by using an information click probability prediction model. The target input method label of the target user can be converted into the target information flow label representing the target information flow portrait of the target user by utilizing the corresponding relationship between the input method label system and the information flow label system. The invention solves the problem that the accuracy of recommended information is low due to the fact that a target user who does not have any click behavior on information in an information flow product cannot obtain an information flow portrait.

The invention aims to provide a time-frequency spectrum generation method and device applied to securities data quantitative trading. After low-frequency information is separated out through a recursive low-pass filter, in a time-frequency spectrum, a frequency is limited in a medium and high-frequency limited range, so that the problem of data low-frequency processing is solved and medium and high-frequency information required for generating the time-frequency spectrum can be obtained. The time-frequency spectrum is established by adopting a recursive wavelet algorithm, so that the problem of securities market data timeliness is solved. In addition, by performing differential processing on a dynamic time-frequency spectrum, the problem of low time-frequency spectrum precision caused by Heisenberg uncertainty factors is solved.

The invention provides a method for extracting an offshore culture area in a satellite image based on scale space transformation, which is characterized by comprising the following steps of 1) extracting a water body area in a remote sensing image by utilizing an NDWI index on the basis of remote sensing image segmentation; 2) respectively carrying out large-scale and small-scale space transformation on each waveband of the remote sensing image of the water body area, and carrying out subtraction operation on a large-scale space transformation result and a small-scale space transformation result to obtain a multi-waveband local enhancement image; 3) selecting a sample set in the target area, and performing vector projection transformation on the multi-band local enhanced image to obtain abreeding area feature map; and 4) observing the target area, determining a threshold value, and then extracting the breeding area from the breeding area feature map. According to the offshore culturearea extraction method, a cross-scale operation mechanism in an image scale space is applied, and the high-precision extraction of multiple types of offshore culture areas in the satellite remote sensing image is achieved through a double-threshold method on the basis of full-band vector projection transformation.

The invention provides an image correction method and device and a storage medium in the technical field of image processing. The method comprises the following steps of performing rectangular boundary extraction in images and image correction, corresponding linear equations to parameter space; applying constraint conditions to three parameters in spherical space; utilizing average value of parameters in linear equations; determining the three parameters corresponding to the linear equations of boundaries; obtaining linear equations of boundaries of the rectangular graphs. The method is simple, high in extraction efficiency and precision and wide in application range. Therefore, image boundary extraction efficiency is greatly improved. Constraint conditions of conversion points in a parameter space on a spherical surface are continuous. The problem that a traditional discrete image boundary extraction method based on Hough transform is low in precision is solved. Meanwhile, the defect that a method for extracting boundaries of images based on Hough transform cannot extract the boundary with the inclination angle of pi / 2 relative to a horizontal coordinate axis is overcome, so that obtained images are corrected based on a two-dimensional spatial interpolation function.

The invention discloses a pedestrian pose resolving method for a vehicle. The method comprises the following steps: a plurality of vehicles which can be networked performing real-time perception on pedestrians in a visual field range by using all-round cameras to obtain pedestrian three-dimensional point clouds of a single-vehicle visual field; filtering invalid vehicles according to the driving speed return value of the vehicle-mounted system to form a local area Internet of Vehicles capable of sharing valid data; synthesizing and splicing pedestrian three-dimensional point clouds of multiplevehicle views based on the local area Internet of Vehicles to obtain global three-dimensional point clouds; obtaining a global pedestrian three-dimensional pose by utilizing a position relationship among all points in the point cloud; sharing the global pedestrian three-dimensional pose to all vehicles in the local Internet of Vehicles; and based on the bicycle safety range, analyzing the globalpedestrian three-dimensional pose to obtain pedestrian early warning information in the bicycle safety range. By utilizing the method, the perception range of the vehicle can be expanded in scenes such as parking lots and intersections, and meanwhile, the vehicle without perception capability can perceive pedestrian information in a safety range.

The invention discloses a parameter lambda introduced bridge technology state degradation evaluation method. A bridge technology state degradation process model, which can be used for describing construction and operation period of a bridge, is established through adopting three parameters including a technology state score when the bridge is constructed, a statistical service life of bridges with the same type and operation and utilization time of the bridge in combination with consideration of environment, load and material property changes. When the bridge technology state degradation process model provided by the invention is used for carrying out bridge technology state evaluation and predication, the bridge can be pertinently detected, cured, maintained and reinforced and manpower and material recourses can be used purposely; the rationality of maintenance, reinforcement and modification scales is guaranteed, the bridge keeps a good technology state and the service life of the bridge is effectively prolonged; the parameter lambda introduced bridge technology state degradation evaluation method has important practical meanings and realization meanings on bridge structure safety, sustainable operation and social and economic development.

The invention discloses a degradation assessment method for a technical state of a bridge based on an initial state of the finished bridge. By adopting four parameters including a technical state score during bridge finishing, non-degradation time of the technical state of the bridge, statistical service life of a same-type bridge and bridge operation time, a model capable of describing a bridge technical state degradation process when the bridge is continuously changed with environments, loads, material properties and the like after being finished to enter an operation period is built. The bridge technical state degradation model is used to perform bridge technical state assessment and prediction, so that the bridge can be pointedly detected, maintained, repaired and reinforced, the reasonability of repair, reinforcement and reconstruction scales is ensured, the bridge is enabled to keep good technical state, and the service life of the bridge is effectively prolonged.

The invention provides a text similarity measuring system based on multi-feature fusion and relates to the field of intelligent information processing. According to the system, the text similarity is measured by fusing multiple features based on word frequencies, word vectors and Wikipedia labels. The invention aims to solve the problem of semantic loss caused by non-considering of contexts in a conventional text similarity measuring system and the problem of low similarity result accuracy caused by larger text length difference. The text similarity measuring system is implemented by the following steps: carrying out preprocessing such as word segmentation and stop word removal on a training text; training corpora of the processed training text as a word vector model; measuring the similarity based on the word frequencies, the similarity based on the word vectors and the similarity based on the Wikipedia labels between input text pairs to be computed, and carrying out weighted summation to obtain a final text semantic similarity measuring result. According to the system, the measurement accuracy of the text similarities can be improved, so that the requirement on intelligent information processing is met.

The invention provides a method and a device for acquiring cell information. The method comprises the steps of receiving a request message, which is used for acquiring cell description information of a target cell configured in a target base station, of a current base station; and sending indication information used for indicating a transmission mode of the target base station for transmitting the cell description information to the target base station after the request message is received so as to enable the target base station to send the cell description information of the target cell according to the transmission mode. According to the invention, a problem that cell information acquisition accuracy is low because cell information acquired through a single transmission mode is one-sided in related technologies is solved, and thus an effect of improving the cell information acquisition accuracy is achieved.

The invention discloses a high-order multi-scale condition random field CRF unsupervised synthetic aperture radar SAR image segmentation method. The method comprises the following steps: performing wavelet transform on the input SAR image; computing histogram features; computing semivariance feature; forming eigenvector; calculating the conditional probability of the local class; performing initial segmenting of synthetic aperture radar sar image; calculating an edge probability of each pixel point; calculating a joint posterior edge probability of each pixel point; calculating a posterior edge probability of each pixel point; estimating parameters; segmenting SAR image segmentation. The invention solves the model parameters through the iteration of the relevant parameters and the iteration of the characteristic parameters, fully utilizes the characteristics of the image itself, and greatly reduces the demand for the quantity of the training data.

The invention discloses a wifi and geomagnetism fused indoor positioning system and method based on an HMM (Hidden Markov Model). The method comprises the following steps: matching a wifi signal acquired by a mobile terminal in real time with a pre-established wifi fingerprint library to obtain a local position area; initializing the position in the area to obtain an initial probability distribution matrix; obtaining a state transition probability matrix by using a position relationship of label points in the local position area; and fusing a geomagnetic signal with a pre-established geomagnetic fingerprint library to obtain an observation probability matrix, and estimating the position of a user according to a forward algorithm. When the user is positioned in the local position area determined by the wifi signal, the current real-time geomagnetic signal is used to update the observation probability matrix, and the position of the user is estimated by the observation probability matrixand the state transition matrix; otherwise, the wifi signal is introduced to re-establish the local position area, and the initial probability distribution matrix, the state transition probability matrix and the observation probability matrix are updated according to the preceding method. According to the wifi and geomagnetism fused indoor positioning method based on the provided by the invention, various information is effectively fused, and the positioning accuracy of the system is improved.