72results about How to "Avoid computational overhead" patented technology

A computer architecture and programming model for high speed processing over broadband networks are provided. The architecture employs a consistent modular structure, a common computing module and uniform software cells. The common computing module includes a control processor, a plurality of processing units, a plurality of local memories from which the processing units process programs, a direct memory access controller and a shared main memory. A synchronized system and method for the coordinated reading and writing of data to and from the shared main memory by the processing units also are provided. A processing system for processing tasks is also provided. The processing system includes processing devices and an absolute timer. The absolute timer defines a time budget. The time budget provides a time period for the completion of tasks by selected processing devices independent of clock frequencies employed by the processing devices for processing the tasks.

The invention discloses a message analysis-based protocol format automatic inferring method, which is a method used for analyzing the input and output messages of a protocol entity so as to infer the specific format of the protocol messages. The method comprises the following steps of: firstly, according to the displayable attribute of characters, segmenting the obtained network message in bytes, and carrying out primary clustering based on a format sequence presented by segmentation; secondly, carrying out multi-sequence comparison on the message samples with similar structure by taking the segment as a unit, realizing alignment and length unification of message segments, thereby mastering the basic structure of the message segments and obtaining the whole structure of the message; and finally, carrying out a semantic inferring phase, based on the structure of the message, according to the value and the change features of each field in the sample, following the identification strategies of various semantics, and using the semantic inferring flow of interval field, data field, serial number field, length field and format identifier field. The accuracy and the efficiency of semantic inferring are improved.

The invention discloses a passive type moving target track mapping method based on compressed sensing. The method comprises the steps that firstly, a wireless sensor network is deployed and configured; secondly, a target tracking model is established; thirdly, a moving object is tracked; fourthly, the target track vector is solved. The advantage of the compressed sensing in the aspect of sparse recovery is utilized, the track of the target is accurately mapped only through a small number of observation data, mapping is finished only through one-time calculation, and the problem that due to frequent positioning in an existing algorithm, calculating expenditure is high is avoided. The moving target track mapping method is very suitable for tracking the track of targets in large-scale scenes where wild animals live, and high-precision target track mapping can be achieved.

The invention discloses a single-pixel photon counting three-dimensional imaging system and method based on multi-resolution wavelet approximation. The imaging system comprises a structure light projection system, a photon receiving system and a synchronous control and signal processing system; a target scene is imaged through adopting a single-photon camera structure formed by combining a single-pixel photon detector and a digital micro-mirror device (DMD). According to a multi-resolution wavelet approximation principle, detail information from a low resolution to a high resolution ratio of the target scene is gradually obtained from an initial resolution ratio image, and is used for reconstructing a final resolution ratio three-dimensional image. According to the single-pixel photon counting three-dimensional imaging system and method, the number of times of sampling is effectively reduced and the imaging time is shortened; the single-pixel photon counting three-dimensional imaging system and method are applicable to high resolution ratio three-dimensional imaging application; meanwhile, the computation cost of a CS (Chirp Scaling) algorithm is avoided and the time needed by reconstruction is shortened; the single-photon camera structure formed by combining the single-pixel photon detector and the DMD is utilized so that the system size is reduced and the system structure is simplified; the single-pixel photon counting three-dimensional imaging system and method have the characteristics of simple structure, high reliability and low cost.

The embodiment of the invention provides an LRCN network-based behavior identification method and apparatus, a device and a medium, wherein the method comprises the steps of obtaining a to-be-identified video frame sequence and a corresponding optical flow graph; inputting the video frame sequence to be identified and the corresponding optical flow graph into a long-time recursion convolutional network model; obtaining a behavior category label of the to-be-identified video frame sequence; inputting each adjacent preset number of frames in the to-be-identified video frame sequence into a first convolutional neural network in a long-time recursive convolutional network model, and inputting the optical flow graph corresponding to the preset number of frames into a second convolutional neural network in the long-time recursive convolutional network model, wherein the convolutional neural network performs convolutional layer sharing on the preset number of frames and the optical flow graph in a data fusion manner. According to the scheme, sharing is introduced between the convolutional layers, so that the behavior recognition is performed after a large amount of redundancy of image information between adjacent frames is reduced, and the overall calculation amount of the network is reduced.

The invention relates to a video processing method and device, a computer readable storage medium and computer equipment. The method comprises the steps of obtaining video data obtained by shooting atarget object; reading a demonstration sub-file displayed when the target object is shot from a demonstration file; obtaining interaction information generated by the demonstration sub-file in a display process; when the video data is encoded, using the demonstration sub-file and the interaction information as supplementary enhancement information to be packaged with the video data, and obtaininga video data packet; and transmitting the video data packet to a corresponding receiving end. According to the scheme provided by the invention, occupation of a large number of transmission resourcesdue to adoption of a double-channel transmission mode in video transmission can be avoided.

The invention discloses a defining method based on electric vehicle travelling track characteristics and a data mining technology. The defining method comprises a common track modeling method of track starting point and finishing point, a track envelope curve and driving characteristics and an efficient common track mining algorithm in massive electric vehicle travelling track data, wherein the common track modeling method comprises track starting point and finishing point modeling, track envelope curve modeling and driving characteristic modeling; and the common track mining algorithm comprises a track starting point and finishing point space-time clustering algorithm, a common track finding and track envelope curve determining algorithm and a driving characteristic extracting algorithm. The invention provides a modeling method of a parameterized common track and the track characteristics are parameterized and abstracted; and on the basis of guaranteeing the accuracy of describing a common track rule by the common track, the simplicity of track description is further improved and common modeling is carried out on diversified tracks in the aspect of a macroscopic space-time scale so that the extraction of dynamic information characteristics in a travelling process is realized and identification and classification of track data are realized.

The invention discloses a video super-resolution reconstruction method. The video super-resolution reconstruction method comprises the following steps: acquiring a plurality of low-resolution trainingimage blocks which are continuous in time and a high-resolution training image block corresponding to at least one low-resolution training image block; training a convolutional neural network model by taking the plurality of low-resolution training image blocks as input data and taking the high-resolution training image blocks as truth value tags; and performing super-resolution reconstruction onthe low-resolution to-be-reconstructed video by using the trained convolutional neural network model. According to the method, the time-related characteristics of the video are fully considered in the training process of the model, so that unnecessary calculation overhead caused by frame-by-frame processing can be effectively avoided through subsequent super-resolution reconstruction.

The invention discloses a disease diagnosis method on the basis of medical image data augmentation and belongs to the technical field of disease diagnosis methods. The method includes: 1) mapping original image data to a deep characteristic space and extracting the characteristic of highly-linearized semantic information; 2) according to the distribution of the data, corresponding to different types in the medical image, in the characteristic space, acquiring a characteristic covariance matrix for data augmentation; 3) calculating loss function of the data augmentation and continuously optimizing the loss to acquire a model with a higher characteristic extraction capability. The data augmentation method only works in model training, so that there is no more calculation load and calculationtime introduced during the prediction of the medical image data by using the model. By means of the method, fixed medical image data is effectively augmented, so that the demand on the quantity of labeled patient data during training of a deep neural network. The method effectively relieves the problems that the medical image data is difficult to acquire and is high in labeling cost and further increases disease diagnosis accuracy.

The embodiment of the invention provides a bone flexible-body realization method and device and electronic equipment. The method includes: creating a bone joint tree corresponding to a target object,wherein the bone joint tree is formed by multiple bones; for each bone with a father bone thereof, obtaining a current-frame position and a next-frame position of the bone, and obtaining a current-frame position and a next-frame position of the father bone of the bone; calculating and obtaining the amount of rotation of the bone relative to the father bone according to the current-frame position and the next-frame position of the bone and the current-frame position and the next-frame position of the father bone; and controlling a bone chain between the bone and the father bone to rotate aroundthe father bone according to the amount of rotation to enable the bone to move from the current-frame position to the next-frame position. According to the bone flexible-body realization scheme, movement of the bone is converted into rotation of the bone and the bone chain before the father bone, massive calculation overheads caused by one-by-one calculation on vertices in the prior art are avoided, and running efficiency of a flexible body is improved.

The invention discloses a data processing method and device, equipment and a medium. The method comprises steps of obtaining sample data which comprises user data and article data; determining a userembedding vector of the user data locally at the first participant, and receiving an article embedding vector sent by a server, the article embedding vector being generated locally at the server basedon the article data; and based on the user embedding vector and the article embedding vector, obtaining a preset prediction model of the first participant through federated learning training. The method is advantaged in that a technical problem of low federation efficiency of an end device model in the prior art is solved.

The invention provides a method for constructing an HBase cluster full-text index, a data reading method and a data writing method. The method for constructing the HBase cluster full-text index includes the steps that Zookeeper is configured in an HBase cluster, and a full-text index Table_Index table is built; the HBase cluster configures all nodes of the cluster into a queue based on Zookeeper, and reading WALs of the nodes are stored in the queue; individual threads are started at corresponding node bits in the Table_Index table, and the WALs of the nodes are made to correspond to the node bits of the Table_Index table according to the sequence of the queue, wherein the reading WAL and the started thread of one HBase cluster node are set on each node bit. The method for constructing the full-text index for HBase cluster data has high timeliness and avoids storage and calculation expenditure brought by extra data processing.

The invention discloses an adaptive reinforcement fusion real-time instance segmentation method based on a camera and a laser radar. The method comprises the steps: respectively extracting the image features of a camera image and a laser radar projection image of a target through employing a convolutional neural network, and respectively obtaining a first image feature and a second image feature;adaptively allocating weights of the first image feature and the second image feature, weighting the first image feature according to the allocated first weight to obtain a third image feature, weighting the second image feature according to the allocated second weight to obtain a fourth image feature, and performing enhanced fusion on the third image feature and the fourth image feature; According to the fused image features, outputting the category, confidence, bounding box and mask of the target by using a real-time instance segmentation network, and obtaining an instance segmentation result of the target. According to the method, target instance segmentation can be accurately and robustly realized in real time in a complex environment, and the method has a wide application prospect inthe field of intelligent network connection vehicle perception.

A computer architecture and programming model for high speed processing over broadband networks are provided. The architecture employs a consistent modular structure, a common computing module and uniform software cells. The common computing module includes a control processor, a plurality of processing units, a plurality of local memories from which the processing units process programs, a direct memory access controller and a shared main memory. A synchronized system and method for the coordinated reading and writing of data to and from the shared main memory by the processing units also are provided. A hardware sandbox structure is provided for security against the corruption of data among the programs being processed by the processing units.

The invention proposes a weighted space-time graph-based aerospace information network routing method. According to the weighted space-time graph-based aerospace information network routing method, onthe basis that the orbital operation of a node has periodicity and regularity, the motion cycle of a spacecraft is divided into a plurality of static topologies according to the ephemeris; and a weighted space-time graph model is constructed with link bandwidth, delay, storage and other resources used as weights; a route is intelligently selected from a policy library on the basis of topology models, and therefore, routing can be more efficient, and efficient and reliable data transmission in an aerospace information network can be realized.

The invention provides an acquisition method of time variable performance data and a device thereof. The method comprises: a data source subsystem provided with a time variable performance counter registers data to an OAM subsystem for reading port parameters; the OAM subsystem acquires the data of the time variable performance counter according to data read port parameters and a preset time variable acquisition cycle, and converts the acquired data into time variable group corresponding to the time variable; the OAM subsystem converts the time variable group into a counter group according to the time preset variable group conversion relation and the time variable group and the measurement cycle of the preset measurement task, and reports. The device comprises the data source subsystem provided with a data read port parameter setting unit and a registering unit, and the OAM subsystem provided with a time variable data acquisition unit and a time variable conversion unit and a time variable group conversion unit. The platform of the time variable performance data acquisition can be realized by using the invention.

The embodiment of the invention discloses a security data aggregation method based on privacy protection and a wireless distributed sensing network. The method comprises the steps that data acquisition is performed by nodes, and the acquired data are determined to be the data to be uploaded; clusters determine the nodes of the data to be uploaded according to the data to be uploaded acquired by the nodes included in the clusters; the determined nodes of the data to be uploaded add camouflage data in the data to be uploaded of the nodes so that first data are generated and transmitted to convergence nodes; extremum aggregation processing is performed on all the received first data by the convergence nodes so that final aggregation data are generated and transmitted to a base station; and the camouflage data in the received final aggregation data are removed by the base station so that a data aggregation result is obtained. With application of the security data aggregation method based on privacy protection, the camouflage data are added in the data to be uploaded without encrypting the data to be uploaded so that consumption of energy resource is saved and additional computation overhead is avoided.

The invention belongs to the field of computer software application and provides an HBase full-text index asynchronous construction method. According to the method, HBase data in big data scenarios are subjected to asynchronous real-time construction of full-text indexes; by reading and analysis of HBase WAL logs, the full-text indexes of the data are constructed according to configurations of users. Therefore, storage and computation cost caused by extra data processing is avoided.

The invention provides an automatic driving automobile path planning method and system based on off-line incremental learning. The method comprises the following steps: determining a current state variable and target information of an automobile; utilizing a probability path prediction model based on vehicle kinematics to predict a state variable of the vehicle at the next moment in real time so as to plan a short-term path of the vehicle; constructing an artificial potential field based on the two-dimensional map information, and distributing different potential functions for different types of obstacles and road structures on the two-dimensional map; acquiring offline data of an automobile driving path through a linear quadratic adjustment strategy, and training the offline data by using a radial basis function neural network so as to predict the driving distance of the automobile driving path; planning a long-term driving path of the automobile based on a fast expansion random path planning algorithm of off-line learning, and determining an optimal obstacle avoidance path of automatic driving of the automobile in combination with a prediction result of the driving distance; according to the invention, vehicle trajectory prediction is efficiently and accurately realized.

The invention discloses a secure multi-party computing protocol algorithm fusing a Pedeersen commitment protocol and a Schnorr protocol, relating to the technical field of cryptology. Based on an existing Schnorr signature, improvement is carried out by fusing a Pedeersen commitment, when the submitted signature messages are different and not public, effective privacy protection can also be realized, a sealed auction scene is built, privacy techniques are adopted, an anonymous transaction is carried out by using a zk-SNARKs algorithm with zero knowledge proof on a chain, the identity of each transaction arbitration participant is determined by using an identity authentication protocol under the chain, and then secure multi-party calculation is carried out by using a newly proposed scheme of combining a Pedeersen commitment with a Schnorr signature. According to the secure multi-party computing protocol algorithm fusing the Pedeersen commitment and the Schnorr protocol, the security ofthe multi-party computing protocol algorithm is improved; by utilizing the characteristics of Pedersen addition homomorphism, under the condition that message commitments of all parties are not decrypted, combined operation is directly carried out on new message commitments in a ciphertext form, so that participating three parties are all added into actual transaction operation. And meanwhile, thesignature checking calculation time of the chain contract is also reduced.