30results about How to "Control computation" patented technology

The invention belongs to the field of text data mining and provides a web text-based acquiring and screening method of seismic macroscopic anomaly information, applied to collecting and screening seismic macroscopic anomaly text information from the internet. According to the method, based on a Heritrix frame, by the use of a seismic macroscopic anomaly subject descriptor group, a crawling strategy from seismic macroscopic anomaly subject relevancy judging and link ordering to information extraction is customized for three information sources, namely common webs, post bars and social networks, and subject related webs crawled are subjected to information screening mainly from three aspects, namely subjective sentence judging, text subjectivity judging and seismic macroscopic anomaly matching. The method has the advantages that online collection of seismic macroscopic anomaly information is provided with a scientific, efficient and accurate technical means and information acquisition efficiency is greatly improved.

The invention provides a line parameter detection and data non synchronous verification-based double-end ranging method. The method includes the following steps that: fault staring points corresponding to sampling data at two ends of a fault line are detected, the initial value of a non-synchronous angle between the fault staring points of the data at the two ends is calculated according to the electrical quantity of the fault line in normal operation, the initial value of a fault distance is calculated; and iteration is carried out to realize correction: the non-synchronous angle of the two ends is corrected according to the fault distance, online parameters of the line are calculated, ranging results are optimized, and when allowable error is satisfied, ranging is terminated. According to the method of the invention, the change of the parameters of the line and synchronization of the data at the two ends of the line are fully considered; a ranging result optimization method is utilized to well eliminate the influence of fault phasor measurement deviation on the ranging results, and therefore, fault location accuracy of the power transmission line can be effectively improved.

The invention discloses an indoor scene three-dimensional point cloud reconstruction method based on depth information fusion. The method comprises the following steps: collecting a color image and a depth image of a scene; fusing the color image and the depth image to perform feature extraction and matching; determining the pose of the depth camera and screening out a key frame; generating a three-dimensional point cloud of each key frame according to the key frames; and splicing the three-dimensional point cloud of each key frame frame by frame according to the pose of the depth camera to obtain a globally consistent scene three-dimensional point cloud model. According to the method, whether a point is a feature point is judged by using image brightness information and depth information at the same time in feature extraction, so that the extracted feature points have high representativeness. Meanwhile, homogenization operation is performed on the extracted feature points, it is ensured that the number of the feature points is within a reasonable range, the calculated amount of a system is effectively controlled. The system reliability is improved. The accuracy of the front end is improved, the robustness of the front end is enhanced, and the result of back-end optimization is more accurate or the number of iterations of back-end optimization is reduced.

The invention provides an aero-engine high pressure turbine disc optimization design system based on particle swarm and method, and relates to a design technology field of the aero-engine. The system comprises a definition module of the optimization task of the aero-engine high pressure turbine disc, a primary particle swarm / file and a grid generation module, a global optimum selection module based on a gridding neighborhood, a particle update module, a file update module and a file size control module based on the gridding neighborhood, first the gridding jam index precision is improved by using active volume of file in gridding, then the particle jam distance is calculated in the finite region by using the gridding neighborhood message, the confidence level operand is provided for selecting the global optimum and deleting the file particle, the obtained ultimately file is the result of a plurality of the target optimization. During designing a plurality of the target optimization for the aero-engine high pressure turbine disc, starting from increasing precision of the global optimum in particle swarm optimization and of the non-particle deleting, the system and method provides the high-quality result of a plurality of the target optimization.

The invention provides a license plate recognition method and device, electronic equipment and a storage medium, and relates to the field of license plate recognition. The license plate recognition method provided by the invention comprises the following steps: firstly, inputting a license plate picture into a feature extraction network to obtain a first feature map of the license plate picture; inputting the first feature map into a spatial transformation network to obtain a second feature map formed by rotating the first feature map; and finally, inputting the second feature map into a number recognition network to determine a recognition result of the license plate photo. According to the method provided by the invention, the spatial transformation network is added between the feature extraction network and the number recognition network; therefore, before the number recognition network performs license plate number recognition according to the feature map, the feature map is rotated by the spatial transformation network, so that the feature map input into the number recognition network is rotated, and the accuracy of recognizing a picture with an inclined license plate is improved.

The invention provides a high-dimensional data clustering method, electronic device and system. The method comprises the following steps: obtaining the high-dimensional data to be clustered after dimension reduction and encryption from a plurality of clients; obtaining the corresponding base clustering by calculating the high-dimensional data to be clustered after dimension reduction and encryption; calculating the similarity between any two base clusters in the base clustering set; initializing the preset reference clustering; using the iterative algorithm to obtain candidate reference clusters successively based on the preset reference clusters and the tag sets corresponding to the base clusters. The preset reference clusters are updated iteratively until the preset reference clusters are no longer changed, if the preset reference clusters and candidate reference clusters have higher average similarity with the other clusters in the base cluster set, then the preset reference clusters are used as the final clustering results. The invention improves the accuracy of the high-dimensional data clustering result and controls the calculation amount.

The invention relates to a grid unit particle number optimization method in rarefied airflow numerical simulation. The method comprises the following steps of: judging whether the number of simulationparticles in any grid cell is within a set range or not in the numerical simulation of the thin air flow by adopting a DSMC, and if the number of the simulation particles is within the set range, notoptimizing the number of the particles in the grid cell; if the number of the simulation particles exceeds the upper limit value of the set range, repeatedly executing particle fusion operation according to a particle fusion criterion so as to enable the number of the simulation particles in each grid unit to be within the set range; and if the number of the simulation particles is lower than thelower limit value of the set range, repeatedly executing particle separation operation according to the particle separation criterion, so that the number of the simulation particles in each grid unitis within the set range; and controlling the particle number of each grid unit in the simulation domain, so that the simulation particle number in each grid unit does not change greatly along with the unit density and the unit size, and the distribution of the simulation particles on different grid units is balanced.

The invention relates to the field of information processing, in particular to a pooling processing method and system and a storage medium. The pooling processing method comprises the steps of obtaining a pooling weight graph corresponding to a feature graph of a to-be-processed image through a preset weight calculation model; and obtaining a pooling result of the to-be-processed image according to the feature map and the pooling weight map. By adopting the embodiment of the invention, the characteristics of the to-be-processed image can be reserved for dynamic pooling, so that the pooling processing performance is improved.

The invention discloses a method for constructing a product industry parallel data set by using an incremental singular value decomposition method. The method for constructing the data set is based oninfluence factor data and yield and quality data of a certain product production process. Parallel sample data similar to the target data are obtained from historical data samples, and process designand data analysis is performed by utilizing the parallel sample data. According to the method, an incremental singular value decomposition method is utilized, data dimension reduction and principal component analysis can be conducted on mass data in a library, data disturbance caused by random factors during data collection can be filtered out through the analysis, and the dimension and the sample number of imported data are controlled through a left singular matrix U and a right singular matrix. Meanwhile, the algorithm is based on a batch updating mode, and the complexity of the whole calculation is low.

The invention discloses a C language oriented source code cloning detection method. According to the technical scheme, adopting a context five-sense-organ grammar to define a C language grammar and analyzing a source program, generating an analysis tree of the source program, then converting the whole analysis tree to obtain a converted analysis tree, and then restoring the converted analysis treeinto a source code in a text form. In addition, formatting normalization processing is carried out on the obtained source code in the text form, and cloning detection is performed on the obtained C function through an LCS algorithm to obtain a cloning function detection result of the current to-be-detected function, and during cloning detection, only the code sequence length of the function fallsinto the allowable length range of the cloning function to serve as a cloning comparison object of the current to-be-detected function. The Type 3 clone can be detected, and the calculated amount ofdetection is controlled to a certain extent.

The invention discloses a method, device and equipment for annotating an image and acquiring an image annotation model, and belongs to the technical field of image processing. The method comprises the steps that a first image is acquired, and at least one to-be-labeled target object exists in the first image; obtaining feature maps of a plurality of sizes of the first image, wherein the feature map of each size in the feature maps of the plurality of sizes comprises a plurality of first feature vectors; selecting a first feature vector meeting a screening condition from the first feature vectors of the feature maps of the multiple sizes to obtain a first feature vector corresponding to the at least one target object; and obtaining correlation data between the first feature vectors corresponding to the at least one target object, and labeling each target object in the at least one target object based on the correlation data. By selecting the first feature vector meeting the screening condition, the calculation amount when the target object is labeled is controlled, and the labeling efficiency is improved. And the target object is labeled based on the correlation data between the first feature vectors, so that the labeling accuracy is improved.

The invention discloses a control method and system of an underwater rocker arm system based on an ADP in the technical field of automatic control. The control method of the underwater rocker arm system comprises the steps that an underwater rocker arm system model considering underwater resistance is established; establishing an energy optimal performance index function of the underwater rocker arm system model; constructing a neural network model of the underwater rocker arm system based on the energy optimal performance index function; and solving the neural network model of the underwater rocker arm system based on an ADP algorithm to obtain an energy optimal control strategy of the underwater rocker arm system. According to the invention, less energy is consumed when the underwater rocker arm moves and resets within a small range.