147 results about "Combination strategy" patented technology

The invention discloses a segmentation combination-based adaptive constant false alarm rate target detection method for an SAR (synthetic aperture radar) image, and belongs to the technical field of synthetic aperture radars. The method comprises the following steps of: dividing a reference window into four sub windows, extracting uniformity statistics of the four sub windows and judging whether the sub windows are uniform; obtaining parameters for estimating background clutter models by adopting corresponding sub window combination strategies according to the non-uniform number, and then obtaining a detection threshold value by using a false alarm probability and a relationship between the clutter models; and comparing the pixel value of a current detection unit with the detection threshold value, judging whether a target exists, detecting the whole SAR image to be detected by adopting a running water form, and performing target fusion operation on the detected SAR image. According to the scheme, the method has low calculation quantity and simple operation, solves the problems of low detection probability, high false alarm rate and the like when the environment is complex and changeable and multiple targets are adjacent under high-resolution large scenes in the prior art, obviously improves the detection effect, and can keep good detection performance under various complex detection environments.

A system and method for merging two rulesets provided in rule-based systems associated with originating applications executing at different locations, each ruleset comprising rules in potential conflict with each other, and each ruleset being in a different rule format. The rulesets to be merged are communicated to an assimilator service provided with a merge policy comprising a set of specifications of partially-ordered priorities and / or mutual-exclusion constraints. The rulesets are translated into a common core representation capable of being implemented in any logic program rule engine provided in a rule-based application at any location. The rulesets are assimilated to produce a new merged ruleset comprising logic required for resolving potential conflicts among rules in accordance with the merge policy that is implemented in any logic program rule engine provided at any location. The new merged ruleset is then translated into one of the originating application's rule format.

The invention discloses a Chinese-medicinal-material identification method based on deep neural networks. The method includes the following steps: using Chinese-medicinal-material pictures, which arecollected by a web crawler and artificial photographing, as input of a data set, and carrying out preprocessing; and adopting a Bagging method of ensemble learning for training and prediction processes, namely adopting a random sampling method to generate multiple sub-training-sets, utilizing classical convolutional neural network models and all the sub-training-sets to carry out fine-tuning training to generate multiple weak classifiers, wherein the adopted convolutional neural network models include AlexNet, SqueezeNet and GoogleNet, and finally cooperate with a Softmax classification algorithm, and using an ensemble-learning combination strategy to obtain a strong classifier to obtain a classification result, wherein a voting method is adopted for the ensemble-learning combination strategy. The method of the invention is used for auxiliary identification of Chinese medicinal materials, reduces amateur errors appearing in identification, and can analyze the Chinese medicinal materials in a manner of high accuracy, fast identification speed and stable performance.

A method for managing metadata file layout of a parallel file system belongs to the field of file system and is characterized in that the file system organizes the files by an extension block distribution way; after receiving a user request, a metadata server quickly carries out the operations of distribution, division, combination, release, and the like to the extension block in the corresponding file layout, thus achieving the requirements of the file layout on obtaining, submission or adjustment and realizing the optimum management to the file layout. The method adopts a principle of meeting the lowest requirement, responds to the request on the corresponding minimum extension block set, thereby meeting the simultaneous operation of multi-user on the same file to the maximum extent and improving the parallelism of the system; and simultaneously, the method adopts an extension block real-time combination strategy to ensure the maximum continuity of the file layout, optimizes the memory, reduces the seek time of the file layout and improves the performance of the file system.

The invention discloses an operating method of an intelligent financing investment adviser robot system and the intelligent financing investment adviser robot system. The operating method of the intelligent financing investment adviser robot system includes the steps: A, acquiring and processing user data; B generating an investment strategy; C, matching the investment strategy with the user; D implementing a one-key ordering purchase function; and E, maintaining and adjusting the investment strategy after investment. The operating method of the intelligent financing investment adviser robot system utilizes the mobile internet technology to provide companion type financing adviser service to the user, maximumly knows about the demand of the user, most reasonably and most efficiently recommends the most suitable financing product to the user, continuously tracks the user for knowing about growth in the financing field and changes of the investment combination after the user purchases the product, continuously provides adjustment service for the investment combination strategy so as to enable the strategy to be adapted to different growth stages of the user and different market environments, and at the same time provides a performance analysis tracking analysis tool for investment combination to enable the user to check the investment performance conveniently and momentarily.

The embodiment of the invention discloses an order task processing method and device, a server and a medium, and the method comprises the steps: responding to an order task, and obtaining a pluralityof order tasks meeting a preset order attribute; The first workstation is used for dividing the plurality of order tasks into at least one wave task according to the task combination strategy, and calculating and judging the first workstation which corresponds to the specified wave task and meets the requirement on the matching degree between the specified wave task and the workstation by taking the wave as a unit and according to the attribute of an order article in the specified wave task; And allocating the picking tasks corresponding to all orders in the specified wave task to the first workstation. According to the embodiment of the invention, the problem of low order task processing efficiency in the prior art is solved, the order task processing efficiency is improved, and the working efficiency of a warehouse is improved.

The invention discloses a cellular atypia automatic grading method based on deep learning and a combination strategy, comprising: first, under different resolutions, utilizing a deep learning method to identify the grade of a pathological tissue image; then, under each resolution, utilizing a trained depth model to process the large image under a current resolution combined with a sliding window method; utilizing one of combination strategies - an absolute majority vote method to determine the grade of the large image under a current resolution, thereby obtaining the grade of the large image under each current resolution; and finally utilizing a relative majority vote method to determine the final grade of the image out of grades of a plurality of resolutions. The cellular atypia automatic grading method employs big slice images as research objects, utilizes a deep learning and sliding window method and a combination strategy to accurately evaluate image cell atypia grades, helps doctors to estimate pathological tissue image cancer grades, and accurately and rapidly performs clinical diagnosis.

The invention relates to an electromyographic signal classification method based on a multi-kernel support vector machine. For a sample with complex distribution, based on the classification performance of a single-kernel support vector machine, the classification accuracy and the quantity of support vectors are easily influenced. The method combines a multi-kernel support vector machine method with a binary tree combination strategy and comprises the following specific steps of: collecting electromyographic signals of the lower limbs of a human body through an electromyographic signal acquisition instrument; denoising the electromyographic signals containing interference noise by using a wavelet coefficient inter-scale correlation denoising method; extracting the features of the denoised electromyographic signals to obtain the features of the electromyographic signals by using denoised wavelet coefficients; and classifying on the basis of the multi-kernel support vector machine. The method can well meet the multi-classification requirement of lower extremity prosthesis control, and takes into account both accuracy and instantaneity, and has broad application prospects in the multi-movement mode recognition of intelligent prosthesis control.

The invention discloses a neural network-based boundary combination named entity recognition method, which comprises the following steps of 1, extracting the entity boundary information based on a neural network model, and constructing a boundary recognition model; 2, implementing a boundary combination strategy, and combining the entity boundaries to obtain a candidate entity set; and step 3, constructing a neural network classifier, and screening the candidate entity set. According to the method disclosed by the present invention, by employing the boundary combination strategies and introducing the neural network techniques, the characteristic that the neural network automatically extracts the high-dimensional abstract features in a layered manner is fully exerted; by dividing the entityrecognition into three steps of boundary recognition, boundary combination and candidate entity recognition, the defects of a traditional sequence model are overcome, and the problem of feature sparseness generated by a traditional machine learning method is avoided to a certain extent, so that the performance of the nested named entity recognition is improved, and a very good effect is achieved.

The invention discloses a method of detecting flood disaster changes through object-level high-resolution SAR (synthetic aperture radar) images and aims to solve the problem that a false target possibly existing in an SAR image and similar to a water region in geometrical characteristics and spectral characteristics causes great 'pseudo-changes' and causes difficulty and disturbances to flood disaster change detection. Each time-phase image is subjected to contourlet transform; image noise is suppressed at the premise of keeping image edge characteristics, an optimal decomposition scale is selected through simple sample training, and the position of a mark point in a possible area of a water body is quickly acquired through block histogram statistics, on the optimal decomposition scale. Contour information of the water body is acquired through a mark-point-based watershed segmentation and region merging strategy, and interference of the false target is further eliminated through discrimination rules based on multiple features. Finally, water contours extracted from multiple time-phase images are directly compared according to registration results, and a region having water body changes is obtained.

The invention discloses a massive fragment data aggregation system for two kinds of massive fragment data. The system comprises a fragment uploading module of a client, and a cache module, a verification library module, a document storage module, a fragment receiving module, a fragment processing module, a fragment combination module and a fragment aggregation module of a server. Correspondingly, the invention also discloses a massive fragment data aggregation method. By the system and the method, a plurality of technologies such as message queuing services, memory banks, document databases, daemon processes and the like are adopted, so that the massive fragment data can be highly efficiently processed and combined; and a strict combination strategy is adopted, so that combination accuracy is ensured.

The invention discloses a video segmentation method based on depth recovery and motion estimation. The video segmentation method based on the depth recovery and motion estimation comprises the following steps of: (1) working out a background subtraction measure by using homography matrix estimation or using a camera motion and dense depth map recovered by a video sequence consistent depth recovery method; (2) performing dense motion estimation, and estimating dense motion fields d and occlusion maps o of continuous two frames of images; (3) calculating a video segmentation result according to an interactively generated combination strategy of multiple measures; and (4) repeating the step (3) for at least two times, and then, ending. Firstly, according to the video segmentation method based on the depth recovery and motion estimation disclosed by the invention, videos can be segmented by iterative optimization of motion, depth and segmentation information. Secondly, according to the video segmentation method based on the depth recovery and motion estimation disclosed by the invention videos of which the backgrounds do a planar motion can be segmented without estimating camera parameters and the depth information. Finally, the video segmentation method based on the depth recovery and motion estimation disclosed by the invention is a video segmentation method of combining multiple measures, the accuracy of various measures can be measured, and reliable measures are screened out to involve in video segmentation calculation.

The invention relates to an intelligent control method based on adaptive planning of a virtual ship for an under-drive unmanned ship formation. The method comprises the following steps of 1, setting aformation and initializing parameters; 2, collecting a position coordinate (xL, yL) and a heading angle psiL of a leader ship, conducting wave filtering, and transmitting the position coordinate andthe heading angle to a following ship; 3, according to the formation, the position coordinate and the heading angle information of the leader ship, obtaining a reference position (xr, yr) and a reference motion posture psir of the following ship in the formation in real time; 4, introducing the virtual ship and conducting real-time adaptive planning to obtain a reference track of the following ship; 5, using a combination strategy of RBF neural networks and a minimum parameter learning algorithm to train learning parameters online to generate intelligent formation control signals, wherein theintelligent formation control signals include the rotating speed nF of a mainframe of the following ship and a rudder angle command signal deltaF. Compared with the prior art, the method has the advantages that the method adapts to curved path tasks, overhead is avoided, leader ship speed information is not needed, and the method is simple, convenient and excellent in real-time performance.

The invention discloses a complaint short text classification method based on deep integrated learning, which comprises the following steps: preprocessing a client complaint text set to obtain a preprocessed complaint text set; Designing complaint classification labels according to the theme classification of the preset complaint text, and marking corresponding complaint classification labels on the preprocessed complaint text set to obtain a training sample set; Performing text feature extraction on the training sample set by adopting a BTM topic model to obtain text feature vectors; Carryingout text feature extraction on the training sample set by adopting a convolutional neural network to obtain a convolutional semantic feature vector; Performing normalization and fusion on the text feature vector and the convolutional semantic feature vector by adopting a normalization combination strategy to obtain a combined text feature vector; And inputting the combined text feature vectors into a random forest model for training, combining classification results of a plurality of decision trees by adopting a weighting method according to the difference of different decision trees, and obtaining the category with the maximum probability as a text classification result of the training sample set.

The embodiment of the invention discloses a data uploading method, a data uploading system and a data uploading device. The method includes the following steps that: a cloud server receives slice data transmitted by a first content distribution network server, wherein the slice data are obtained through performing slicing processing on data to be uploaded by using a client end of the data to be uploaded, and the first content distribution network server is a content distribution server which is determined by the client end according to the internet protocol proximity and is nearest to a base station to which the client end belongs; and the cloud server merges the received slice data into the data to be uploaded according to preset combination strategies. With the data uploading method, the data uploading system and the data uploading device of the invention adopted, the success rate of data uploading can be improved.

The invention designs a password strength evaluation model based on multi-model ensemble learning. Firstly, a real password training set is used to train several existing password evaluation models assub-models. Secondly, several trained sub-models are used as base learners for ensemble learning, and the advantage integration of each sub-model is realized by the combination strategy of partial weak-term voting. Finally, a universal password evaluation model is implemented based on high accuracy. Password strength evaluation based on a multi-model ensemble learning model for passwords with different levels of complexity is performed, the evaluation results are high in accuracy and versatility, the multi-model ensemble learning has good applicability in password evaluation.

This invention discloses one match information process method and its device to improve platform set flexibility, which comprises the following steps: receiving user send strategy label to express user require match information; according to the strategy label to get source information and integration strategy information; finding the source information relative source; according to the integration strategy information to integrate the match source into one set source. This invention also provides one match information process device and platform system.

The invention discloses a game skill release area method of multiple skill disks. The method comprises the steps that a skill operation area is divided into a main skill disk and a subsidiary skill disk, skills of the two skill disks can be alternately released, skill icons of the main skill disk are clicked to release main skills, the main skills enter cooling time after being released, and the skill icons are replaced with subsidiary skill icons at the positions at this time; the subsidiary skill icons can be clicked to release the subsidiary skills, the subsidiary skills enter the cooling time after being released, when the cooling time of the main skills is ended, the icons at the positions are changed into the main skill icons, and the main skill icons can be clicked to release the main skills at this time; the difficulty of master operation is reduced; the number of the skills capable of being released is increased, and the diversification of a battle is enriched; more main and subsidiary skill combination strategies are brought, and the interestingness of the battle is improved.

A method and apparatus for cache management of distributed objects have been disclosed. A query strategy, based on a fetch strategy and a merge strategy, may be used to query a database and reconcile data in a local cache. In one approach, a merge action depends upon whether an entity was added, deleted, or changed relative to the original version of the entity as it was last fetched.

The invention relates to a recognition method for components of Chinese character pictures and belongs to the field of image processing and pattern recognition. The method includes firstly pre-processing Chinese character component pictures with specific fonts, performing skeleton extracting, forming initial sections according to endpoints and intersections which are detected out, and merging some sections in an artificial interactive mode; next, performing statistical modeling on the marked sections and describing directional information of the sections by using a four-dimensional vector; then constructing section neighbor relations to complete the component modeling process according to the maximum spinning tree principle; for an input Chinese character, firstly decomposing the character into sections, then generating a group of section sets with the maximum similarity by contrasting each component in a library, finally obtaining recognition results of components of the input Chinese character through the optimum selection strategy, namely, obtaining the optimum solution through solution of the variety knapsack problem. Local features are introduced, the optimum combination strategy is adopted, the complete recognition method for the components of the Chinese character pictures is formed, and the recognition rate can be improved effectively.

This invention discloses a method for predicting protein structure based on phased multi-strategy copy exchange. In the frame of differential evolution (DE), generate initial conformation populations with a variety of folding types in each temperature layer; to each temperature layer, divide a conformation search into two phases according to iterations; randomly select a conformation as a target unity from the populations in the first phase; divide the populations into two parts according to energy in the second phase; randomly select one unity as the target unity from first 50% of populations with low energy; randomly select three conformation unities different from the target unity to generate a testing unity through variation, cross and section combination strategy; judge whether to accept the testing unity according to the energy of the conformation so as to perform copy exchange to corresponding unities of adjacent temperature layers; and, under the guidance of the phased strategy and the supplement of copy exchange strategy, acquire a series of metastable conformation through updating continuously.

The invention discloses a large-scaled wireless sensor network synchronizer based on frequency offset bidding and dynamic topology. After a network region clustering mechanism is adopted for reducing a node synchronous hop number, the node selects a parent node with stable clock frequency and few synchronous hops for synchronization, and effects of restraining ambulation of multiple hop errors can be achieved; a fast and slow synchronization combination strategy is adopted for meeting requirements on two performance indexes: synchronization precision and convergence time in the network; the node uses the hop number information to dynamically select the parent node, time synchronization requirements for a dynamic topological network can be met, and anti-destruction and expansibility are good; and synchronization among nodes in a single hop range in the network adopts a TPSN bidirectional synchronization similar idea, transmission delay during the synchronization process is eliminated, and high synchronization precision can be achieved.

The invention provides an improved hierarchical cascade support vector machine parallelization method. Optimization is obtained through an improved Cascade SVM algorithm. Firstly, a c is introduced in the improved algorithm to be used to measure a change situation of the number of model support vectors obtained in training of each layer in a hierarchical training process; and then through adjusting a merging strategy and a hierarchical structure in the model training process, a merging mode of the support vectors obtained by training of each layer is optimized and adjusted from merging of every two to average segmentation after merging of all, and thus a defect, existing in the merging mode of every two, that filtering non-boundary samples is insufficient is avoided. Under the premise of ensuring that the classification accuracy is not reduced, the method can effectively shorten the model training time and improve the training and classification efficiency of a model by virtue of a current-mainstream parallel framework of Spark.

The invention discloses a data processing device for asset allocation. The device includes: a data acquisition module, which obtains unstructured data through a human-computer interaction port, and obtains structured data through a financial information database; a data processing module, which processes the structured data and the unstructured data to obtain a preliminary Analyzing data; the digital configuration module, based on the preliminary analysis data, and solving based on the Markowitz configuration model to obtain the digital configuration results of large-category assets; the product adaptation module, according to the digital configuration results of the large-category assets and the product collection of large-category assets, Optimizing the products in the product set, so as to obtain the optimal product investment portfolio. The invention also discloses a data processing method for asset allocation. The invention realizes automatic provision of individualized asset combination strategy by means of technical means, and solves the problems of high labor cost, narrow audience and large service difference in traditional investment consultants.

The invention discloses a method for preparing a cationic unordered rock salt structure anode material through high-valent-state transition metal ion replacement combination strategies. A lithium saltis mixed with oxides of high-valent-state transition metal elements M (one of more of Ti, V2, Nb, Mo and Zr), oxides of M' (one or more of Fe, Ni and Mn) and villiaumite through a solid-phase ball milling method, and then through high-temperature treatment, the cationic unordered rock salt structure anode material is obtained. The prepared anode material has the high charge and discharge specificcapacity and cyclic stability, and the preparation method is simple in process and easy to implement, and application and popularization are facilitated.

The invention discloses a method, a device and a system for processing video images. The method includes that a transmitting end divides an original video image into a plurality of sub-images according to a dividing strategy corresponding to an aspect ratio of a display screen of a receiving end; the divided multiple sub-images and combination strategy information corresponding to the aspect ratio of the display screen of the receiving end are transmitted to the receiving end; the receiving end receives the multiple sub-images transmitted by the receiving end and the combination strategy information corresponding to the aspect ratio of the display screen of the receiving end; the multiple sub-images are combined into a recombined video image with an aspect ratio identical to the aspect ratio of the display screen of the receiving end according to a combination strategy represented by the combination strategy information; and the recombined video image is displayed on the display screen. According to the scheme provided in an embodiment of the invention, the method, the device and the system have the advantage that the video image displayed at the receiving end is matched with the display screen when an aspect ratio of the transmitting end of the video image is inconsistent with the aspect ratio of the display screen of the receiving end.

The invention relates to a method for combining HARQ with power control based on IEEE802.16e. By introducing the parameter of 'the proportion of repeat request users', the method realizes the compromise of power and delay in a multi-carrier system, by fully utilizing the repeat request combination gain of an HARQ mechanism during the power control process, the method effectively reduces the total transmission power and meets the certain requirements of signal to interference and noise ratio and delay and respectively provides simple and easy combination strategies aiming at two sub-channel dividing mechanisms in a protocol. As for a non-real-time system, if the objective frame error rate and the total transmission power restriction of each user are given, the strategy can meet the frame error rate and the power restriction required by the system by the selection of a proper repeat request user proportion and at the cost of certain delay; or when the objective frame error rate and the tolerable maximum delay of the system of each user are given, the strategy can select the corresponding repeat request user proportion to realize the minimum total transmission power.

The application discloses a processing method for virtual resources, a server, and a client. The method comprises: a fist virtual resource, a second virtual resource, and a combination strategy are configured, the first virtual resource is returned to a first client based on a first virtual resource request message received from the first client, and the second virtual resource is returned back to at least one second client according to a second virtual resource request message received from at least one second client; when an article exchange request message is received from the first client, the first virtual resource and at least one second virtual resource of at least one second client are obtained; and when the first virtual resource and the at least one second virtual resource are determined to be able to be combined based on the combination strategy, combined usage of the first virtual resource and the at least one second virtual resource is determined and an article exchange confirmation message is returned back to the first client. According to the technical scheme, the resource utilization rate of the client during virtual resource usage can be improved.

The invention discloses a face detection method based on hierarchical network and cluster merging. The face detection method based on hierarchical network and cluster merging divides the convolutionalneural network into two-level of networks. In the first-level network training process, by obtaining heat maps of the original input picture at the seven-level resolution, by obtaining the initial candidate face area at the seven-level resolution according to the local hottest area on the seven heat maps, loss of face information on the sample can be reduced. In the second-level network trainingprocess, the specific facial feature area is added in the training positive sample, so that the convolutional neural network can extract the features of the facial feature area in a targeted manner, and at the end of the network, a picture containing a face detection frame is obtained through a candidate box merging strategy based on cluster and facial features. The face detection method based onhierarchical network and cluster merging overcomes the problems that a current method is complicated in network and large in calculated amount, effectively processes the face which changes the postureor is shielded in the picture, and improves the face detection accuracy based on the convolutional neural network.

The invention discloses a remote sensing image segmentation method based on hard boundary constraint and two-stage combination. The method mainly includes the steps that first watershed segmentation of edge constraint and edge distribution are conducted to obtain sub-elements; and then layered sub-elements of the edge constraint are combined, the sub-elements can conduct extreme growth under the edge control, and a so-called initial element set is obtained; and on the basis of the two steps, the edge constraint is abandoned, secondary element layer combination is conducted to obtain a final segmentation result. In the two combining processes, an element combining strategy capable of combining elements repeatedly is designed so as to quicken the combining processes. Compared with a Fractal Net Evolution Approach (FNEA) method of eCognition software, the remote sensing image segmentation method has the advantage that edge position precision of the elements is high; mistaken segmentation rate is slightly small, over-segmentation rate is remarkably reduced; and dependency degree on scale parameters is low.