151results about How to "Implement expansion" patented technology

The invention relates to a method and a device for manufacturing a live-action map. The method comprises the following steps: acquiring a plurality of video images acquired by a plurality of cameras; synthetizing and processing the plurality of video images to form a full-view video image; identifying a target object in the full-view video image and geographic information of the target object; marking the target object to acquire the marked target object; renewedly mapping the marked target object to the full-view video image; and performing map drawing processing on the full-view video image according to the geographic information of the target object and generating the live-action map. The method and the device have the beneficial effects that real-time manufacturing of the live-action map can be realized by a user, and a traditional mode that the live-action map needs to be manufactured by a special team and the user can use the map by only payment or internet is broken through; and by adopting the live-action map manufactured by the method and the device, the live-action map can be used free, and the dependence of internet loading on signals and geographical conditions can also be avoided.

The invention relates to a dynamic traffic simulation platform which comprises a perception layer, a fusion layer, an application layer and a presentation layer. The perception layer is connected with a dynamic traffic data collecting device, multivariate heterogeneous traffic flow data are obtained and transmitted to the fusion layer in real time, and the fusion layer is connected with the application layer. The multivariate heterogeneous traffic flow data and GIS data are received for processing, the application layer obtains road network dynamic traffic flow data, and a traffic simulation model is formed. Through the traffic simulation model and a traffic OD matrix, dynamic traffic simulation is performed, and the presentation layer receives the traffic simulation result of the dynamic traffic simulation in the application layer, and provides service application and presentation to a terminal client. The problem that in the prior art, a large number of real-time dynamic traffic data can not be rapidly matched with a road section well is solved. Therefore, the multifunctional dynamic traffic simulation platform is capable of performing OD back stepping and real-time dynamic traffic early warning on dynamic traffic flows on the basis of a GIS grid rapid matching technology.

The invention discloses a method for realizing dynamic function online interception expansion of a system through multilingual cloud compilation. By intercepting a tag, compilation and execution of a business logic interception code are realized, and then the compilation and the execution are carried out. The compilation running process specifically comprises the steps of detecting whether expansion interception exists or not, and if the expansion interception exists, determining the business logic interception code; searching for a corresponding instance from a memory, and calling and running the corresponding instance by utilizing a multilingual dynamic script execution engine; and if the corresponding instance does not exist, performing compilation by utilizing a corresponding script compilation engine according to a language type of the business logic interception code to generate the corresponding instance. According to the method, the business logic interception code can be directly introduced and runs after being compiled through multilingual cloud compilation in system running, so that a previous code is not influenced, a dynamic script engine system is realized, the dynamic function interception expansion of the system is realized, and the flexibility, openness, maintainability and expandability of system function expansion are improved.

The invention provides a relation extension method based on entity links. The method is characterized in that the technological means of entity links are used, text in natural language is linked to the corresponding entities in knowledge bases, unlabeled entities in message box attributes are identified, using the labeling systems of encyclopedic entries to serve as the types of the entities according to the matching degree between candidate entity types and entity types at two ends of a relation; meanwhile, a model fusion method is used to sort the candidate entities, and the advantages of a nonlinear boosting integrated model are fused into a linear model so as to increase model performance and guarantee relation accuracy. By the method, unlinked entities in the message box can be identified effectively, the problems of entities with identical names and entities with aliases are solved, a relation with the target entity is built, and relation extension is achieved.

The invention provides a method for detecting the change of urban surface features in remote sensing image based on Siamese convolution network, the Siamese convolution network is a twin convolution neural network SCNN, Based on the data enhancement technique, the initial sample set is selected from the registered two-phase city images, the twin convolution neural network (SCNN) is set up, the twin convolution neural network (SCNN) is trained based on the initial sample set, and the initial sample set is expanded by the data enhancement technique. The twin convolution neural network (SCNN) istrained based on the expanded sample set, and the trained SCNN model is obtained to detect the change of urban terrain. The invention realizes the expansion of the sample through the data enhancementtechnology, and designs a Siamese convolution neural network, avoids the tedious steps of artificially designed features in the traditional change detection method, and realizes the end-to-end operation; Fully considering the spatial attributes of the image, the accuracy and reliability of the change detection are improved.

The invention discloses a distributed cache method, which is applied in a system with one machine frame and a plurality of single boards, the method comprises: a cache board in the system records a non-cache board which provides the memory of self part as the supplementary cache; the cache stores the data to be deleted due to full local caching in the supplementary cache provided by the non-cache board and deletes the data from the cache board; the cache board responds to the external data request according to the data in the local caching and supplementary cache. The invention also provides another distributed cache method, a distributed caching system and buffers and non buffers in the system. By adopting the invention, the expansion of capacity of caching under the condition of not adding extra storage media can be realized.

The invention discloses a radiotherapy system and a radiotherapy method based on deep learning. The radiotherapy system comprises a medical case database module, an image data processing module, a radiotherapy processing module, a deep convolutional neural network module and a radiotherapy scheme generating module. According to the radiotherapy system and the radiotherapy method provided by the invention, in the combination of image data and some radiotherapy data of patients, radiotherapy plans of existing patients are learned, so that a radiotherapy scheme can be directly generated for a newpatient, so that doctor's time in repeatedly debugging radiotherapy plans is greatly saved, labor intensity and a requirement on the professional level of operating personnel are reduced, and a moreproper radiotherapy plan is provided to the patient; therefore, the effect of radiotherapy can be enhanced.

The invention relates to an intelligent terminal application system based on network structure, which at least comprises one local network with intelligent terminal, access network, and fireproof wall; the first server area with several different server groups while each group has at least one server with different functions, and the first server area has authorization domain analysis server; network load balancer connected to the fireproof wall and the first server area, to select one of minimum-load server to connect the intelligent terminal, according to the requested server group and the load of server.

The invention discloses a double-loop-line intersected belt sorting machine, an intersected belt sorting system and a sorting method thereof. The double-loop-line intersected belt sorting machine comprises at least one annular rail, two sets of intersected belt cars for forming two conveying loop lines circularly rotated on the annular rails side by side, and a power device for driving the two sets of intersected belt cars to move along the annular rails. The double-loop-line intersected belt sorting machine is delicate in design and simple in structure; as two sets of cars are used in one annular rail, the number of the intersected belt cars is increased by using internal space of the loop lines under the precondition of not increasing the floor area to multiply the sorting capacity, theuse efficiency of the sorting machine is improved, and the equipment cost is reduced; and meanwhile, parcels can be transferred between two conveying annular intersected belt cars, so that the idlingrate of the cars can be reduced, meanwhile, the parcel loading flexibility is improved, and the arrangement requirements and the occupied space of a particle loading line are reduced.

The invention discloses a method and system for achieving system function dynamic extension and replacement through multi-language cloud compilation. According to interception labels, service logic interception codes corresponding to to-be-extended system functions are stored in a script storage library; interception device codes are compiled at corresponding position adding places according to the interception labels during local compiling, and corresponding service logic interception codes are compiled and run through cloud compilation according to the interception device codes. The corresponding service logic interception codes can be added online in real time according to service requirements, and therefore the service logic interception codes can be directly introduced and run after being compiled through cloud compilation during system running, previous codes are not influenced at all, a dynamic script engine system is obtained, system function dynamic interception extension is then achieved, and the flexibility, openness, maintainability and scalability of system function extension are improved.

The embodiment of the invention discloses a data enhancement method and device, a data processing method, electronic equipment and a computer storage medium. The method includes: selecting T image transformation manners from N image transformation manners, wherein N is an integer which is greater than 1, T is an integer which is greater than 0, and T<=N; and sequentially carrying out transformation processing, which corresponds to the T image transformation manners, on a current sample image to obtain a new sample image. According to the embodiment of the invention, image samples can be expanded, and a recognition accuracy degree of a network model obtained by training can be increased.

The invention discloses a multilingual cloud compiling method and system for achieving dynamic interception and extension of system functions. According to business requirements, a method needing to be extended is marked with an interception label; before an item is compiled, target interception labels corresponding to the system functions needing to be extended are screened; then, for the method marked with the target interception labels, during local compiling, a compiling interception device code is added at the corresponding position according to the target interception labels, and then according to the interception device code, a corresponding business logic interception code is compiled through cloud compiling and runs. By means of the multilingual cloud compiling method and system, the business logic interception code can be dynamically and remotely added in real time through the interception labels according to the business requirements, and therefore during system running, the business logic interception code is compiled through cloud compiling and then is directly introduced and runs, previous codes are not influenced, a dynamic scrip engine system is achieved, then, dynamic interception and extension of the system functions are achieved, and flexibility, openess, maintainability and scalability of extension of the system functions are improved.

The invention discloses a visual malicious software detection device and method based on a deep neural network, and the method comprises the steps: converting an executable file sample into a bytes file and an asm file through employing a disassembly technology, combining a normal software data set collected and marked by a user with a famous BIG 2015 malicious software data set, and obtaining a balance experiment data set; in order to effectively extract high-dimensional features in a data sample, converting the sample further into an RGB three-channel image by using a visualization technology combined with data enhancement. The invention also provides unique deep neural network classification architecture, which is used for improving the performance of the detection method. The method disclosed by the invention is explained from other numerous neural network model methods; the superiority of the RGB three-channel image in the aspect of malicious software detection compared with a gray level image is verified through experiments, and the data enhancement technology is beneficial to visualization of malicious software detection. And a novel thought and a novel method are provided for other researchers to carry out malicious software detection experiments.

The invention provides a welding image defect recognition method and device, a storage medium and equipment. The method comprises the steps of acquiring a to-be-recognized welding image; inputting theto-be-recognized welding image into a defect recognition model, recognizing the welding image through the defect recognition model to obtain a welding image type, and judging whether the welding image has defects or not and the defect type when the defects exist or not; wherein training processing is carried out on the initial defect recognition model as follows: forming a defect recognition model by combining a convolution basis of a pre-training model with a full-connection classifier; carrying out migration training on the defect recognition model; wherein migration training refers to freezing a convolution basis, training a full-connection classifier, and then finely adjusting a defect recognition model. Based on the combination of deep learning and transfer learning technologies, efficient training of the defect recognition model can be realized under the condition of limited sample number, and the accuracy of welding image defect recognition is improved.

The invention relates to a database data encryption and decryption method and device. The database data encryption and decryption method comprises the steps that a cryptographic algorithm and a secret key are acquired, encryption and decryption are conducted on an encryption and decryption object through the cryptographic algorithm and the secret key, and the encryption and decryption object is a database table, or a line in the database table, or a view or a line in the view. The database data encryption and decryption device comprises an acquiring module and an encryption and decryption module, wherein the acquiring module is used for acquiring the cryptographic algorithm and the secret key, and the encryption and decryption module is used for conducting encryption and decryption on the encryption and decryption object through the cryptographic algorithm and the secret key which are acquired by the acquiring module. The encryption and decryption object is the database table, or the line in the database table, or the view or the line in the view. The database data encryption and decryption method and the database data encryption and decryption device achieve expansion of an encryption and decryption mode of a database, and satisfy requirements for efficient data encryption and decryption of a mobile computing environment and network application.

The embodiment of the invention provides an error correction method and device and relates to the technical field of computers. The method comprises the steps that a target word to be subjected to error correction is determined; an error correction pair matched with the target word is looked up in a target error correction dictionary and used as a candidate error correction pair, wherein the target error correction dictionary is a pre-established error correction dictionary, each error correction pair stored in the target error correction dictionary is obtained after a sample word is subjectedto error correction through an error correction model, and the error correction model is obtained after a first preset model is trained by using the error correction pair in a known error correctiondictionary; according to the found candidate error correction pair, an error correction result of the target word is obtained. Compared with the prior art, the error correction method has advantages that when the target word to be subjected to error correction is subjected to error correction, an error correction word of the target word can be found in the target error correction dictionary with high probability, and therefore the failure probability of error correction on the target word is lowered.

The invention discloses a method for fast switching an application program based on an input method. The method comprises the following steps that the preset input method is selected, and reference content is input into a current edit box through the preset input method; the intention of a user is determined according to the reference content; the corresponding application program is called according to the intention of the user, and is provided for the user for selection; the application program selected by the user is fast entered into. Compared with the prior art, according to the method for fast switching the application program based on the input method, after the reference content is input into the current edit box through the preset input method, the reference content is not adopted as input content in the prior art by a mobile terminal, the reference content is analyzed and judged, then the intention of the user is determined, the corresponding application program is called to the user for selection according to the intention of the user, and therefore the application program can be switched fast, the user operation is simplified, and user experience is improved. The invention discloses the mobile terminal for fast switching the application program based on the input method at the same time.

The invention provides a three-dimensional human body posture estimation method and a computer readable storage medium. The method comprises the following steps: acquiring a single-person image from an original image by adopting a human body detection network and carrying out standardization processing on the single-person image; predicting two-dimensional coordinates of the key points from the single-person image by using a two-dimensional attitude estimation method; generating a three-dimensional coordinate from the two-dimensional coordinate, including: predicting a first three-dimensional coordinate of the key point by using a three-dimensional attitude generator; performing symmetric processing on the two-dimensional coordinates according to a symmetric structure of a human body joint, and predicting second three-dimensional coordinates of the key points by using a three-dimensional attitude generator; enabling the first three-dimensional coordinate and the second three-dimensional coordinate to respectively calculate difference values with the corresponding labels, the sum of the results bing used for back propagation, and obtaining three-dimensional human body posture estimation. A connection relation and a symmetric relation between key points of a human body are fully utilized, and the purpose of optimizing a prediction result can be achieved; and meanwhile, on the basis of an original data set, the expansion of training data is realized, and the robustness of the model is enhanced.

The invention discloses a gastric cancer pathological image canceration region segmentation method based on deep learning, which realizes complete segmentation of a gastric cancer canceration region and can greatly reduce the labeling time of professional doctors. Firstly, a data set is effectively expanded by utilizing a data enhancement technology; then adopting the semantic segmentation networkfor semantic segmentation of encoding-decoding, and adopting a new loss calculation mode, thus the problem that the network cannot converge when partial annotation data is used for training the network is solved, and accurate segmentation of all canceration areas is achieved in combination with an iterative training strategy.

The invention discloses a model training method and device, a prediction method and device, a data processing device and a medium. The network model training method comprises the following steps: acquiring original data information of a historical sample; carrying out feature extraction on the original data information to obtain historical features, and carrying out transformation processing on the historical features to obtain historical feature vectors; utilizing the historical feature vector to generate a simulation feature; combining the simulation features with historical features to serve as a training feature set, and training a deep neural network model after feature transformation; simulation data is generated through the deep learning model by utilizing real historical data, theproblem of lack of samples is solved, the deep neural network model is trained by combining the historical data and the simulation data so as to predict the data, and the accuracy of data prediction can be effectively improved.

The invention provides an identified association graph self-optimization mechanism based on deep learning, and the mechanism combines a knowledge graph technology and a deep learning technology, carries out the identification processing of continuously collected data through a deep learning model, and then adds the data to an association graph. Data processing is carried out through a plurality ofdistributed data storage nodes and distributed computing nodes in the association graph, and a series of global iterations are carried out through three parts of local computing, communication unitsand fence synchronization on the basis of a block synchronization parallel computing model. Self-adaptive dynamic optimal allocation of computing resources is realized according to the resource utilization rate, the processing performance and the locality of the data of the system. And carrying out continuous disambiguation analysis and clustering calculation on the data added into the associationgraph to carry out continuous simplification and correction so as to realize continuous self-optimization of the association graph.

The invention discloses a method for dynamically intercepting and extending functions of systems by means of multi-language cloud compiling. Business interception codes can be compiled and executed by the aid of interception tags. The method particularly includes detecting whether extension and interception are available or not and determining codes of interception devices if the extension and interception are available; searching corresponding examples from memories by the aid of the codes of the interception devices and invoking and operating the corresponding examples by the aid of multi-language dynamic script execution engines; generating corresponding examples by means of compiling by the aid of corresponding script compiling engines according to language types of the codes of the interception devices if the corresponding examples are unavailable. The method has the advantages that the codes of the interception devices can be compiled by means of multi-language cloud compiling when the systems are operated, compiled codes can be directly introduced and operated without influence on the previous codes, accordingly, dynamic script engine systems can be implemented, the functions of the systems can be dynamically intercepted and extended, and the functional extension flexibility, openness, maintainability and scalability of the systems can be improved.

The invention discloses a corpus expansion and speech synthesis system construction method and device based on artificial intelligence. The method comprises the following steps of according to a corpus in a small sample sound library, training and acquiring a WaveNet model; using the WaveNet model to generate a speech waveform corresponding to a given text; supplementing the corpus corresponding to the generated speech waveform to the small sample sound library and acquiring a large sample sound library; and using the corpus in the large sample sound library to construct a statistical parameter speech synthesis system. In the scheme, a speech synthesis effect can be increased, and manpower, material resources, time cost and the like are saved.

The embodiment of the invention provides a lane control method, device and system and a processing server. The method includes the steps that lanes with one-way congestion in two-way lanes are determined; for non-one-way congestion lanes, the number of remaining lanes, except the lanes which meet the current traffic flow of the non-one-way congestion lanes, in the number of all the lanes is determined; if at least one lane remains, traffic signal lamps corresponding to the non-one-way congestion lanes are controlled so that the non-one-way congestion lane regions can be cleared; when the non-one-way congestion lane regions are cleared, navigation indication devices on lane boundaries of the lanes corresponding to the remaining lanes in number are controlled to work so that the lanes corresponding to the remaining lanes in number can become lanes corresponding to the one-way congestion lanes in the passing direction, wherein the navigation indication devices are arranged inside the non-one-way congestion lanes and are adjacent to the one-way congestion lanes. The lane control method, device and system and the processing server effectively solve the problem of one-way congestion.

The invention provides a sea ice seawater segmentation method for a synthetic aperture radar (SAR) image. The sea ice seawater segmentation method comprises: preprocessing cross and co-polarized remote sensing data of a synthetic aperture radar; synthesizing the preprocessed image data into a three-channel remote sensing image by utilizing three groups of data including HV polarization, the difference between HV polarization and HH polarization and the ratio between HV polarization and HH polarization; selecting three-channel remote sensing images of different seasons and different marine environment conditions to perform manual marking of sea ice and seawater ranges, and performing enhancement operation on the marked images to generate a training sample set; taking the training sample data generated in the above steps as input, completing the training of a deep learning model, and obtaining a sea ice and seawater automatic segmentation model; and carrying out a step of subsequently obtaining a preliminary result of sea ice and seawater identification and finally obtaining a sea ice and seawater segmentation image. According to the invention, the response difference of sea water and sea ice in different SAR polarization images is fully utilized, and sea ice and sea water segmentation can be fully automatically realized.

The invention discloses a depth estimation method based on a light field EPI image, and the method comprises the steps: 1, carrying out the refocusing of light field data, and obtaining the light field data under different focusing parameters; 2, extracting sub-aperture images of a horizontal view angle and a vertical view angle from the refocused light field data; 3, extracting light field EPI images in the horizontal and vertical directions from the sub-aperture images; 4, building a double-branch depth estimation model based on light field EPI image correlation reasoning, and using the extracted EPI images in the horizontal and vertical directions for training; and 5, carrying out depth estimation on the to-be-processed light field data by utilizing the trained depth estimation model. According to the method, the relevance between the central pixel and the neighborhood of the EPI image can be fully utilized, and the data enhancement can be realized by utilizing the light field refocusing principle, so that the accuracy of light field EPI image depth estimation can be effectively improved.

The invention discloses a sample expansion method and system based on foreground and background feature fusion. The method comprises the steps: dividing a remote sensing ground feature classificationdata set into a source data set and a target data set based on ground feature categories; constructing a small sample source ground object classification task based on the source data set, and training a feature extractor, a hybrid model and a classifier based on the small sample source ground object classification task; constructing a small sample target ground object classification task based onthe target data set; performing sample expansion by using the trained feature extractor and the hybrid model based on the target classification task; wherein each task comprises a first task and a second task; the mixed feature is a feature synthesized by a foreground feature and a background feature by using a mixed model; according to the method, the hybrid model is trained based on the classification task, additional manual annotation is not added to expand the training sample, the training cost is reduced, the trained feature extractor and the hybrid model are utilized to expand the target data set, the classifier is trained, and the sample expansion method is realized.

An X-ray inspection machine and a method for tire expansion and spin driving. A vertical inspection machine expands the tire outwards from the bead part for positioning, centering, and spin driving, so as to improve the stability of the tire during high-speed spin, thereby improving imaging quality in X-ray inspection, improving X-ray inspection quality and enhancing the efficiency. The X-ray inspection machine comprises: a lead chamber; an X-ray running device; an imaging system running device; a tire-expanding and spin-driving device, having a rack and at least one set of guide rods, each side of each guide rod being mounted on the rack by a mounting bracket; guide bushings, co-axially sleeved on the guide rods by a bearing to bear the tire, so as to drive the tire to spin by pressing the tire bead; a connection rod, the center of which is fixedly and axially provided on the rack, one end of which is connected to a hinge pin for axially arranging the mounting bracket, and the other end of which is connected to a slide rack; the slide rack, moving along the straight line back and forth by a first driving device on the rack; and a second driving device, for being connected to the guide brushings and driving the guide brushings to move axially to drive the tire to spin.

The invention discloses a sample classification method, system and medium based on a multi-sample reasoning neural network. The method comprises the following steps: (1) establishing a multi-sample reasoning neural network MSIN; (2) taking several training samples of different sample fields as input values, inputting them into the multi-sample reasoning neural network MSIN, and training the multi-sample reasoning neural network MSIN for a specified number of rounds; after each round of training, inputting the verification samples to the MSIN for testing, and saving the MSIN which minimizes thetotal loss function of the MSIN as the final network. Step (3): taking a plurality of test samples from different sample domains as input values of the multi-sample reasoning neural network MSIN, inputting the input values into the trained multi-sample reasoning neural network, and outputting the sample categories corresponding to the test samples or the sample domains in which the test samples are located.

The invention relates to an image data enhancement method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring an original image, and performing mask fusion processing on the original image to obtain a detection target image; determining a perspective transformation base point of the detection target image; performing perspective transformation on the detection target image according to the perspective transformation base point and a preset perspective transformation condition to obtain a perspective view angle image; and fusing the perspective view angle image with a preset background image to obtain an enhanced image, wherein the enhanced image data comprises the enhanced image. By adopting the method, the image data can be expanded, and the image data volume is increased.