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

A system and method is provided for variable bit rate encoding using a complexity ratio. Quantization parameter is calculated using a complexity ratio, which is equal to a local complexity divided by a global complexity. Complex pictures are allocated a larger bit budget relative to simple pictures. With the larger bit budget the quality of complex pictures can be maintained while reducing the overall size of the encoded video stream.

A system and method is provided for vectorizing misaligned references in compiled code for SIMD architectures that support only aligned loads and stores. In this framework, a loop is first simdized as if the memory unit imposes no alignment constraints. The compiler then inserts data reorganization operations to satisfy the actual alignment requirements of the hardware. Finally, the code generation algorithm generates SIMD codes based on the data reorganization graph, addressing realistic issues such as runtime alignments, unknown loop bounds, residual iteration counts, and multiple statements with arbitrary alignment combinations. Loop peeling is used to reduce the computational overhead associated with misaligned data. A loop prologue and epilogue are peeled from individual iterations in the simdized loop, and vector-splicing instructions are applied to the peeled iterations, while the steady-state loop body incurs no additional computational overhead.

The invention relates to the technical field of computer graphics and image processing, in particular to a 3D human face modeling method based on front silhouette, which mainly models the 3D human face in a fully automatic manner. The method comprises the following steps: a general human face grid model is built by 113 feature points and 184 tri patch information arrays which are connected with the 113 feature points based on Candide parameterization; 2D position information and depth information of a specific human face are determined by virtue of human eyes location algorithm and side nose tip location algorithm by inputting two 2D images, the general face model is modified and transformed into a specific human face model; an interpolating is calculated for the human body patches by Delaunay Triangulation Technology and texture mapping is carried out by the Bilinear Interpolation Technology, thus causing the human face image to be appeared more delicate and accurate; and the method solves the problems such as low efficiency, high calculation expense, and the like, existing in the prior art.

The invention discloses a data encryption/decryption checking method and system. The method comprises the following steps that: an encryption device performs Xor operation on each data section in plaintext data at an encryption end to acquire a checking code, combines the plaintext data at the encryption end with the checking code to acquire combined data, encrypts the combined data to acquire a ciphertext and outputs the ciphertext to a decryption end; and a decryption device decrypts the received ciphertext to acquire decrypted data, separates out the plaintext data at the decryption end and the checking code from the decrypted data, and checks the checking code. According to the invention, the checking code is generated through the simple Xor operation, and the data is encrypted by adopting one-time encryption transformation, so that the encryption transformation frequency is reduced, the calculation overhead is prevented from being multiplied, and the encryption/decryption efficiency is improved effectively; and by checking the correctness of the checking code, whether the data is wrong or maliciously changed during encryption or decryption and transmission or storage can be detected; and the confidentiality and correctness of the data are guaranteed at the same time.

One embodiment of the present invention provides a system that allocates memory for immutable data on a computing device. During operation, the system receives an application to be executed on the computing device. Next, the system allocates a memory region on the computing device to store immutable data for the application. Note that this memory region is smaller than the immutable data for the application. When the system subsequently receives a request to access a block of immutable data for the application, the system allocates space in this memory region for the block, and proceeds to load the block into the memory region. However, if at a later time the space occupied by this first block is needed for another block, the system unloads and discards the first block. If a subsequent operation needs to use information in the first block, the system regenerates the block by transforming raw data associated with the block into a form that can be directly accessed by the application, and then reloads the block into the memory region.

The invention discloses a message authentication method based on a semi-trusted management center in the vehicle networking. The user communicates with RSU and other vehicles arranged on both sides ofthe road through OBU, and extracts, analyzes and effectively utilizes the message, so that the vehicle runs more safely, and the user can obtain more efficient service at the same time. The inventionis based on the environment of a semi-trusted management center, more realistic, at the same time the vehicle's pseudonym is generated by the vehicle, RSU and STA together, RSU and STA trace the identity of the vehicle jointly, A single party cannot obtain the true identity of the vehicle, thus the safety is improved.The invention adopts the self-healing key distribution method to exclude the revoked vehicles from the group, so that the vehicles do not need to store and query the revoked list, the storage space, the resources are calculated and the searching time are reduced, and the efficiency of the message authentication is improved.

A program storage device, readably by a machine, tangibly embodying instructions to perform method steps for constructing a call graph whereby for each method M, a set of types SM of objects that may occur in method M is determined for each field F, a set of types SF of objects is determined that may be stored in field F, the method comprising: determining the allocation sites inside the body of method M; determining the set of directly called methods M' inside the body of method M; and determining the set of virtually called methods M'' inside the body of method M.

The invention discloses a method for protecting the vehicular ad hoc network conditional privacy based on a registration list. The method comprises the following steps: (1) offline registration, including trusted association TA initialization, road-side unit RSU initialization and on-board unit OBU initialization; and (2) driving process, including mutual authentication, traffic information release, message authentication, notice message generation and message receiving. According to the method, a vehicle user carries out cooperative communication with RSU which are deployed on two sides of aroad through the OBU, so that the driving safety can be effectively improved, and users can conveniently and quickly acquire related service.

The invention belongs to the technical field of information safety, and discloses an inter-satellite networking authentication system and method suitable for a double-layer satellite network. The system comprises a ground authentication server, a high-orbit satellite authentication client and a low-orbit satellite authentication client, wherein the ground authentication server is used for completing initialization of a satellite authentication system, namely generating and distributing identity information, a secret key and track parameters which are needed for authentication between the satellites; the high-orbit satellite authentication client and the low-orbit satellite authentication client are main bodies of inter-satellite networking authentication, and the inter-satellite identity authentication and key negotiation are realized through the interaction of authentication parameters. An authentication precomputation mechanism is designed by utilizing the characteristics that a satellite network clock is highly synchronous and a node operation trajectory can be predicted, so that the authentication efficiency between the satellites is effectively improved. According to the system and the method, safe and efficient identity authentication and secret key negotiation of a high-orbit satellite and a low-orbit satellite in a networking stage can be realized, and the system and the method can be applied to networking authentication between the high-orbit satellite and the low-orbit satellite.

The invention provides a distributed duplication elimination system-oriented data routing method, which comprises the following steps that a server categorizes fingerprints of all data blocks forming data, and transmits the fingerprints of different categories to different abstract storage nodes for storing data abstracts of the fingerprints of the corresponding categories respectively; the received fingerprints in the abstract storage nodes are queried to obtain hit scores of the fingerprints in each duplication elimination node; the hit scores are returned to the server; the server obtains a summery score of each duplication elimination node according to the hit scores of the fingerprints in each duplication elimination node, and determines a target duplication elimination node by combining the summery scores and the storage conditions of the duplication elimination nodes. According to the method, duplication elimination effect and the storage utilization rate are balanced, the communication and calculation overhead in a fingerprint query process is effectively inhibited, and the data routing expandability of a distributed duplication elimination system is improved.

The invention provides a convolutional neural network structure simplification and image classification method based on an evolutionary strategy. The structure between a convolution model feature extraction part and a full connection layer is simplified. The method comprises the following steps: firstly, pre-training a convolution model to fix a model weight, and performing feature extraction on an image data set by using a convolution part; then, randomly generating a plurality of binary sequences to serve as an initial population, wherein each binary sequence corresponds to a structure between a feature extraction part and a full connection layer; then, under each structure of the current population, using the pre-trained weight to classify the feature vectors of the test set, using theclassification accuracy as the fitness of the corresponding structure, obtaining a new-generation structure population through crossover, variation and selection operations, and continuously iteratingto be equal to the number of iterations; and finally, obtaining a structure between the simplified feature extraction part and the full connection layer, and performing fine adjustment on the pre-training weight by applying the structure.

The invention discloses an image segmentation method, device and equipment and a storage medium. The method comprises the steps of obtaining an original picture; and inputting the original picture into an image segmentation model to obtain an image semantic segmentation image of the original picture, and generating the image segmentation model based on joint training of a superpixel segmentation algorithm and a full convolutional neural network. According to the embodiment of the invention, when image semantic segmentation is carried out on an original image by using an image segmentation model generated by training a super-pixel segmentation algorithm and a full convolutional neural network, the low-level features of the original picture are identified; the generated image semantic segmentation image is accurate in edge positioning and good in space consistency; the super-pixel segmentation algorithm only participates in the training generation process of the image segmentation modelbut not participates in the process of generating the image semantic segmentation graph by adopting the image segmentation model, so that compared with the image segmentation model generated only based on full convolutional neural network training, the prediction precision of the model is improved on the basis of not increasing the calculation overhead.

A method for designing scene picture facing to picture drawing platform includes combining multiple Leaf nodes being brothers to each other with multiple Group nodes being brothers to each other if they have the same state control information and so on, combining LOD nodes being brothers to each other and being under Group node if they are near center, operating static Transform with its children and grandchildren nodes in advance and discard Transform node in scene picture, eliminating empty node and redundant node.

A camera device includes an optronics system and an image capture control unit, for acquiring a sequence of images of a surrounding region of a vehicle. The optronics system includes a wide-angle optical system and a high-resolution image acquisition sensor. The optronics system and the image capture control unit are configured to generate a reduced-resolution binned image of the entire capture region of the optronics system, or to capture an unbinned high-resolution image of a subregion of the capture region, respectively for each individual image of the sequence of images, depending on a current traffic and/or surrounding situation. A height and a width of the subregion are set depending on the current situation. A size of the subregion is set such that the pixel count of the high-resolution image of the subregion is no greater than the pixel count of the reduced-resolution image of the entire capture region.

The embodiment of the invention discloses a position information hiding method and a position server. The method comprises the following steps: after receiving a current query request aiming at a target user, acquiring real position information of the target user, and false position information fed back aiming at last query request of the target user; and when the false position information satisfies a preset anonymity condition, hiding the real position information through the false position information.

The invention discloses a digital signature quick generation method of SM2 public key cryptography. The method comprises a step of performing optimization based on a SM2 signature generation algorithmin GM/T 0003.2-2012 SM2 elliptic curve public key cryptography standard on a SM2 signature server platform, and the method is a quick computation method of the SM2 signature generation algorithm, thesignature formation in the SM2 digital signature algorithm and the quick computation on the 256-bit prime number field in the SM2 recommendation parameter can be quickly realized. The optimization isperformed for the signature scheme and the finite field computation in the existing method, compared with the general optimization method on the elliptic curve, the method disclosed by the inventionbelongs to the specific optimization for the SM2 algorithm; on the SM2 signature server platform, especially a server platform using the recommended SM2 parameter and with long private key life cycle,the computation overhead is small, and the speed increasing can be effectively realized.

The invention discloses a writing and block granularity compressing and combining method and system of a key value storage system based on an OCSSD, and belongs to the field of computer storage. Through selective compression and combination of data block granularity, read-in and write-back of non-overlapping data blocks are avoided, and the problem of read-write amplification in the LSM-tree compression and combination process can be reduced; the data blocks of the same file are uniformly distributed in all channels of the OCSSD, so that the multi-channel parallel characteristic of the OCSSD can be explicitly utilized, and the throughput of a key value storage system is improved; by directly managing a bottom storage medium at a user layer and bypassing intermediate software layers such asa file system, consumption of host resources is reduced, and semantic features of application software are conveniently transmitted. Generally speaking, the problem of read-write amplification of thekey value storage system based on the LSM-tree can be effectively reduced, semantic isolation between upper-layer application software and bottom-layer storage equipment is eliminated, and the throughput of the whole key value of the system to operation is improved.

Provided is a reverse wave beam control method applied to a retrodirective discrete array. The method comprises the following steps: modulating guide information determined by a system to carrier frequency capable of real-time frequency adjusting to generate a reverse guide signal; sending the signal to the retrodirective array; carrying out demodulation on the received signal by the retrodirective array to obtain the guide information and calculating conjugate phase of the guide information; synthesizing a conjugate phase signal according to the conjugate phase obtained from calculation; and carrying out power amplification on the conjugate phase signal and emitting the signal through the retrodirective array. The method enables incoherent local oscillators which are independent mutually and are in a discrete state to generate the microwave signal which is coherent in phase and of which the phase is conjugate with that of the guide information, and realizes reverse wave beam control of the microwave beam backtracking along the incidence direction of the guide signal, thereby protecting system core frequency secret safety, avoiding malicious interference through frequency hopping and improving anti-interference capability of the system.

The invention discloses a multi-backup data holding certification method under a cloud storage environment. The method comprises the following steps: 1) a client generates a public and private key pair of a user and uploads the public key pair to a cloud service provider CSP; 2) the client generates n distinguishable copies of the file F; then dividing the file F into m file blocks and generatinglabels of each file block, dividing each file copy into m file blocks, and uploading the labels of the file F and its m file blocks and the file copies to the CSP; 3) the client sends a challenge to the CSP; 4) according to the received challenge, CSP calculates the set Q to be challenged by the user through the public key; and then generating evidence P according to the set Q and the public key and returning the evidence P to the client; 5) the client end adopts the same way to generate the set Q, and then utilizes the evidence P to verify and challenge the integrity of the replica data. Theverification mode of the invention is flexible, and has positive significance to the replica positioning.

The invention relates to a heterogeneous parallel computing method for updating sparse matrix LU decomposition rows. Based on the superlu algorithm, combined with the master-slave heterogeneity of Taihu Lake Light, the superlu solver transfers the large-scale computing task to the slave kernel in the process of matrix decomposition and row updating, and utilizes the efficient computing and data communication ability of the slave kernel to improve the computational ability of the super-large-scale sparse matrix and the overall performance of the solution. The invention divides tasks according to the data of the matrix, Firstly, a matrix space is created for each slave kernel, and each slave kernel is responsible for solving several matrix rows. In the process of solving the linear equationsof the lower triangular identity matrix type, each row of data is independent of each other, avoiding the conflict and dependence of the solution space, and smoothly realizing the solution of the equations.

Disclosed herein are systems and methods that can improve the quality and efficiency of document facsimile transmission. Quality can be improved by allowing a facsimile document sender to view the document as it will be transmitted to a receiving facsimile device. Efficiency of facsimile transmission is improved by providing systems and methods that allow the transmitting facsimile device to determine the capabilities and/or preferences associated with the receiving facsimile device and convert the facsimile document to match those capabilities and/or preferences before a call is placed.

The invention discloses a small target detection-oriented network and difficult sample mining method. The method comprises the following steps: step 1, extracting trunk features of a current image byusing a trunk network; 2, constructing a neck network, and generating a feature pyramid; 3, constructing a region generation network model, and generating a region of interest; 4, generating trainingdata based on difficult sample mining; 5, cutting the region features by using a region-of-interest alignment module; and step 6, classifying and positioning a target by using a head network. According to the method, the small target detection precision can be remarkably improved, and only little calculation overhead is increased.