59results about How to "Save memory" patented technology

The invention discloses a mobile robot HDL global relocation method and system. According to the invention, the initial position of the robot is determined by using the UWB signal intensity and the time difference, and the position probability distribution function is calculated. In combination with an RANSAC algorithm, a position probability distribution function is multiplied by initial uniform distribution of the RANSAC to obtain a fusion position distribution probability, the RANSAC algorithm is guided to carry out key sampling near a UWB positioning result, and a global repositioning pose is output through multiple iterations and point cloud matching. According to the method, the positioning time of the robot in the global repositioning stage is shortened, the positioning precision is improved, and the calculated amount and memory occupation are reduced. According to the invention, the advantages of UWB positioning and RANSAC point cloud matching are combined, and the global repositioning precision and efficiency of the mobile robot in a complex environment are improved.

The invention provides a lightweight remote sensing image change detection method based on a binary neural network and large kernel stripe convolution. The implementation steps are as follows: obtaining a training sample set and a test sample set; constructing a lightweight remote sensing image change detection network model based on a binary neural network and large-kernel stripe convolution, and performing iterative training on the lightweight remote sensing image change detection network model; and obtaining a lightweight remote sensing image change detection result. In the process of carrying out iterative training on a lightweight remote sensing image change detection network model and obtaining a change detection result, deep feature extraction is carried out on a dual-tense feature map through ResNet18 networks in the two branch networks, the weight and activation can be quantified as + 1 and-1, 32 times of memory saving and 58 times of CPU acceleration are achieved, and the method has the advantages of being simple in structure, convenient to operate and high in practicability. The defect that in the prior art, the number of parameters is large is overcome, the detection efficiency is effectively improved, and meanwhile the defect of accuracy caused by binary quantization is overcome through the large-kernel stripe convolutional network.

The invention discloses a false news detection method and system based on a partition filtering network, and the method comprises the steps: carrying out the fusion of news text features and comment text features, and obtaining event features; inputting the news text and the comment text into a dual-emotion acquisition module to obtain dual-emotion features; inputting the word vector of the comment text into a partition filtering network to obtain an initial vertical feature and an initial rumor feature; fusing the event feature with the initial vertical feature to obtain a corrected vertical feature; classifying the corrected vertical features to obtain a first classification result; fusing the initial rumor feature, the dual emotion feature and the corrected standing feature to obtain a corrected rumor feature; classifying the corrected rumor features to obtain a second classification result; and fusing the first classification result and the second classification result to obtain a final classification result.

The invention discloses a lower limb motion function evaluation method, system and device based on deep learning image processing and a storage medium, and the method comprises the steps: shooting a lower limb motion process of a patient through a common mobile device, and collecting a motion image; performing image preprocessing to obtain a preprocessed action image data set; inputting the motion image data set into the trained progressive mode model, extracting feature points of a joint three-dimensional space, and calculating parameters related to a lower limb motion function according to the feature points; the parameters obtained through real-time calculation and set parameters are subjected to fitting calculation to calculate the overlap ratio, rehabilitation training is evaluated in real time according to the overlap ratio, and an optimized motion curve is given according to the real-time evaluation result to guide a patient to conduct rehabilitation; an image processing algorithm is optimized, and a line-by-line processing mode is adopted to replace a current full-image processing mode, so that evaluation and analysis of lower limb movement under shooting of a common camera in a mobile device are realized, and remote medical diagnosis is completed quickly and efficiently.

This application relates to a method, apparatus, device, and medium for adaptive parameter pruning of a deep learning model. The method includes: acquiring a pre-set training set of a deep learning model; pre-training the deep learning model based on the training set to determine the parameters of each layer in the deep learning model; capturing the low-rank structure of the parameters of each layer in the deep learning model based on a unified tensor network model, and decomposing the parameters into multiple ring factor tensors and a core factor tensor; constructing an adaptive parameter pruning model based on a sparse Bayesian learning algorithm according to the prior distribution results of the Gaussian distributions corresponding to the ring factor tensors and the core factor tensor, and the prior distribution results of the Gamma distribution corresponding to the noise level; and pruning irrelevant components of the parameters based on the adaptive parameter pruning model that has reached convergence to determine the corresponding compressed parameters. This application enables the deep learning model to maintain a low number of parameters while still maintaining high accuracy.

The application discloses a laser vision real-time weld seam tracking method based on a lightweight target detection network, and the method comprises the following steps: before welding, an industrial camera collects an initial weld seam image and transmits the initial weld seam image to an embedded industrial computer, pixel coordinate values of initial weld seam feature points are acquired, and a starting position of the weld seam is obtained; a lightweight target neural network is constructed; during welding, the industrial camera continuously collects weld seam images and sends the weld seam images to the embedded industrial computer, pixel coordinate values of weld seam feature points in the weld seam images are extracted by using the lightweight target neural network; a welding position to be reached by a robot is obtained through the pixel coordinate values of the weld seam feature points; a difference between the welding position of the robot and a current position of the robot is calculated, and a deviation value is sent to a control cabinet, so that the welding gun of the robot is controlled to automatically track and move along the weld seam. The weld seam image is detected by using the lightweight target neural network, and the problem that the real-time rate of the existing weld seam tracking method is not high on a resource-limited device is solved.

The present application relates to a virtual scene rendering method, device, electronic device and storage medium. The method counts the meshes of the models in all bounding volumes of a model cluster, deduplicates the primitives referenced by the meshes based on the material data corresponding to the meshes, and finally generates a unified material data and texture data for all bounding volumes in the model cluster. Since a large number of duplicate primitives are removed from the texture data, the number of primitives contained in the final generated texture data is reduced, thereby avoiding excessive occupation of memory resources and saving memory.

The present invention discloses a time-dependent continuous and discontinuous number method for non-global contact identification and a system, the method comprising the following steps: creating a geometric body numerical model, applying boundary conditions, calculating unit forces, identifying units that may generate contact in the entity units at the boundary and the discrete fracture network unit list, storing and updating the potential contact entity unit list, and then performing contact detection in the potential contact entity unit list, updating the state of each unit and repeating the steps S300-S600 in the attached figure until the simulation is completed. The time-dependent continuous and discontinuous number method for non-global contact identification provided by the present invention can reflect the gradual increase of potential contact pairs with crack extension in real time, can effectively inherit the unit forces of failed joint units, maintain the numerical stability of the numerical model, and has sufficient parallel capability, which is of great significance to improving the operational efficiency of continuous and discontinuous numerical models and improving simulation accuracy.

This invention provides a method for UAV obstacle avoidance trajectory planning based on digital elevation maps, including digital elevation map conversion, initial trajectory loading, trajectory pre-planning, and trajectory re-planning. Using this invention, UAV obstacle avoidance trajectory planning can fully consider terrain variations during low-altitude flight missions to achieve ground-hugging flight. Simultaneously, this method first performs offline trajectory pre-planning and uploads the planned trajectory to the flight controller to provide flight guidance, effectively solving the problem of low real-time performance. When obstacles or threat areas appear on the pre-planned trajectory, the method provided by this invention can find the vertices of obstacles with complex shapes, handle obstacle overlap issues, and thus quickly plan a flight path.

The invention discloses a key value cache processing method and device, equipment and a storage medium, the method is applied to edge equipment in communication connection with remote storage, when a large language model service is used for reasoning, the first capacity of a local key value cache is detected, and when a preset capacity threshold value is reached, the first capacity of the local key value cache is determined. And creating and maintaining a sliding window covering a plurality of key value cache entries in a local memory of the edge device. New tokens are continuously generated along with large language model service reasoning, and local cache content is updated through the sliding window: on one hand, the oldest key value cache entry in the sliding window is removed to release memory space; on the other hand, a first cache entry corresponding to the newly generated token is synchronously obtained from the remote storage and stored in the local memory. According to the invention, communication traffic and delay with remote storage are effectively reduced, and realization of low-memory, low-delay and high-performance LLM reasoning service is facilitated. The technical scheme of the invention can be widely applied to the technical field of artificial intelligence.

A data transmission method and device based on rate control, and a communication device, the method comprising: receiving one or more data to be sent, determining that there is data to be sent that is restricted from being sent among the one or more data to be sent; if the number of data supported by the terminal is greater than or equal to the number of data allowed to be sent, determining the first time difference between the sending time of the first sent data and the receiving time of the data to be sent, and judging whether the first time difference is less than the rate control time unit; if the first time difference is less than the rate control time unit, determining that the data to be sent that is restricted from being sent needs to be discarded, and the data to be sent that is not restricted from being sent can be sent. The present invention can effectively reduce system overhead and effectively improve control accuracy.

Provided are a communication method, a terminal device and a network device. The method comprises: a terminal device sending a first message to a network device, the first message being used for requesting to download at least one piece of model data. The terminal device notifies the network device of one or more pieces of model data to be downloaded by sending the first message to the network device, so as to achieve the transmission of the model data between the terminal device and the network device. In one aspect, the effective management of model data is achieved, and by sending the first message to the network device, the terminal device may download the required model data as needed; and in another aspect, the memory of the terminal device is saved, and the terminal device does not need to provide a large amount of storage space to store large amounts of model data.

The invention discloses a material irradiation defect storage method suitable for multi-element SRSCD simulation. The material irradiation defect storage method comprises the following steps: creating a Defect-Reaction List; randomly selecting a reaction, searching defects of which the types are the same as those of reactants in the defect-reaction list, and executing corresponding operation according to the number of the searched defects; aiming at the selected reaction, searching whether a defect of which the type is the same as that of a product exists in a defect-reaction list or not, and executing the corresponding operation, and the invention relates to the field of nuclear material irradiation damage simulation and the field of high-performance calculation.

The application discloses a modbus address optimization method based on a small stack chip, and specifically comprises the following steps: receiving a modbus reading command, determining a starting address and a reading length; traversing a structure body, searching for a module parameter according to a reading data address; saving the reading data to a reply frame buffer, adding one to the reading data address and adding one to the reading data length, returning to the previous step until the reading data is completed, and entering the next step; generating a reply data frame and sending the reply data frame. The application changes the original several global array methods into a plurality of global arrays defined according to parameter modules, and further, when reading a reserved address space, only 0 can be read, and when writing the reserved address space, any data can be written but no actual physical address is written, that is, about 40% of the original stack space is saved, and the saved stack space can be used in the development of logical codes while the modbus protocol can be normally used.

The invention discloses a device configuration method, an identification method and electronic equipment. The method comprises the steps that after an attribute file opening signal is received, configuration information of a to-be-adjusted device is acquired according to device tree description; searching a configuration value in an IO port line correspondingly connected with the to-be-adjusted device according to the configuration information, and associatively storing the configuration value and the IO port line; obtaining rewriting values of the to-be-configured device, wherein the rewriting values are in one-to-one correspondence with the states of the IO port lines; replacing a configuration value with the rewriting value; and after receiving an attribute file closing signal, restoring the rewriting value to the configuration value. For products of the same type, if different device model configurations occur, the whole device tree does not need to be rewritten due to the limitation of the number of IO port lines, and only the configuration value of the IO port line corresponding to the unused to-be-adjusted device needs to be replaced by the rewriting value corresponding to the to-be-configured device used by the product, so that the device configuration efficiency is improved. Therefore, the identification of the to-be-adjusted device is changed into the identification of the to-be-configured device.

The application discloses a simulation method and device of a multi-physical field coupling resonator, a storage medium and a terminal, wherein the method comprises the following steps: acquiring a resonator model to be solved; identifying a plurality of physical fields in the resonator model to be solved; performing cubic grid division on the resonator model to be solved, and dividing the resonator model to be solved after the grid division into a plurality of grid areas; and acquiring system matrices of all the grid areas by using a finite element method; wherein a degree of freedom vector in the system matrices comprises physical field components of the plurality of physical fields in the resonator model to be solved; the system matrices of all the grid areas are sequentially spliced based on a coupling mode of the same physical field components in the matrices, so as to acquire the system matrices of the resonator model to be solved; and the system matrices of the resonator model to be solved are solved, so as to acquire a target solving physical quantity. The application realizes the simulation combination of the plurality of physical fields in the resonator, reduces the solving time, and also reduces the memory required for calculating the matrices.

The application provides a log processing method of a database, a storage medium and equipment. The log processing method comprises: obtaining log configuration information, and determining whether log configuration has been completed according to the log configuration information; if not, assigning a preset level to a log entry to be written; generating an entry number according to the log level; and writing the log into a pre-set shared memory according to the entry number. Therefore, even if the log configuration has not been completely parsed, the log can be written into the pre-set shared memory by assigning a preset level to the log entry, thereby avoiding the problem of log loss before the log configuration is completely parsed. Moreover, because the log exists in the shared memory, even if the program is interrupted, the log in the shared memory will not be lost, and the log that has been written in advance can still be obtained.

This invention proposes a lightweight remote sensing image change detection method based on a binary neural network and large-kernel strip convolution. The implementation steps are as follows: obtaining training and testing sample sets; constructing a lightweight remote sensing image change detection network model based on a binary neural network and large-kernel strip convolution and iteratively training it; and obtaining the lightweight remote sensing image change detection results. In the process of iteratively training the lightweight remote sensing image change detection network model and obtaining the change detection results, this invention uses ResNet18 networks in two branches to perform deep feature extraction on the feature maps of both time periods. This allows the weights and activations to be quantized to +1 and -1, achieving a 32-fold memory saving and a 58-fold CPU acceleration, avoiding the drawback of existing technologies with a large number of parameters, and effectively improving detection efficiency. Simultaneously, the large-kernel strip convolution network compensates for the accuracy deficiency caused by binary quantization.

This invention relates to a method for estimating the radiant energy emitted from the top of the atmosphere based on high-orbit satellite sensors. The method calculates the radiance of the radiant energy emitted from the top of the atmosphere using the following formula: where F is the radiance of the radiant energy emitted from the top of the atmosphere, and E... m Ω represents the energy emitted from the top of the atmosphere received by high-orbit satellite sensors. e To obtain the solid angle of observation of the top of the atmosphere by a high-orbit satellite sensor, the solid angle can be solved using a complete elliptic integral of the third kind. A rectangular coordinate system based on the solid angle is then constructed to calculate the boundary curve of the observation field of view. The corresponding geocentric and geofixed coordinates are obtained through a coordinate transformation matrix. This invention can accurately estimate the emitted energy from the top of the atmosphere, providing support for related research.

The invention provides a surface acoustic wave filter self-adaptive frequency sweeping rapid simulation method, which comprises the following steps of performing three-dimensional modeling on a surface acoustic wave filter, setting parameter information of each solving region, and extracting a corresponding calculation model; performing discretization processing on the surface acoustic wave filter calculation model by using a hexahedral mesh; a piezoelectric coupling discrete process is deduced, a piezoelectric coupling spectrum element frequency domain solver is constructed, the discrete domain and solving parameters of the surface acoustic wave filter are imported into the solver, and the solver is initialized; and constructing a self-adaptive frequency sweeping algorithm, updating a frequency point needing to be calculated in a self-adaptive updating manner, calling a piezoelectric coupling frequency domain solver to calculate admittance of the frequency point, and finally obtaining a frequency-admittance characteristic curve of the surface acoustic wave filter model under the broadband. Based on the spectral element method and the self-adaptive frequency sweeping algorithm, the method has the advantages of small calculation memory and high solving efficiency under the condition that the solving precision is met.

The application discloses a guided wave tomography method based on full waveform inversion of a recurrent neural network, comprising the following steps: step 1, time series forward modeling based on the acoustic formula of the finite difference method is realized by using sequence calculation of the recurrent neural network, and an initial model is obtained; step 2, the wave field value of each time step at each excitation position is calculated in turn, and a simulation value is generated; step 3, after actual observation values are processed, a least square target equation of the simulation value and the observation value is generated; step 4, the least square target equation is iteratively minimized by an inversion part, including an automatic differentiation method and an Adam algorithm, which are respectively responsible for gradient calculation of the target equation and iterative updating of a velocity model and a UNet weight; step 5, the updated velocity model is subjected to depth image prior UNet regularization; and step 6, steps 1 to 5 are repeated, and the next iteration calculation is performed until the target equation is lower than a specified threshold value, so that a final velocity model is obtained, and the final velocity model is converted into a thickness model by using a dispersion relation.

The present application relates to the field of data processing technology, and provides a data storage method, device, computer equipment and storage medium, wherein the method comprises: receiving data storage requests for at least two values ​​to be stored, the data storage request comprising the data identification of the values ​​to be stored; based on the data identification of the values ​​to be stored, searching for the numerical storage rules of the values ​​to be stored; if the numerical storage rules of the values ​​to be stored are not found, based on the maximum number of numerical digits of the target memory, and the maximum number of numerical digits and the number of the values ​​to be stored, processing the values ​​to be stored as target values, the number of the target values ​​being less than the number of the values ​​to be stored; obtaining the identification of the target values; based on the identification of the target values, storing the target values ​​in the target memory. The use of the present application can reduce the number of stored values, save the memory occupied by storage, and help improve the stability of the device.

Embodiments of the present disclosure disclose an adaptive memory conversion method and device for a distributed file system. A specific implementation of the method comprises: dividing a metadata set in a target directory to obtain a metadata subset group; setting a corresponding memory data structure form for each metadata subset; determining a target metadata subset; executing a corresponding access processing flow on the target metadata subset to obtain and output a final access result; obtaining access feature sets and system resource states of each metadata subset in the metadata subset group in real time to obtain a directory index set; determining memory data structure form conversion instructions corresponding to metadata subsets to be converted in the metadata subset group to obtain a conversion instruction set and a to-be-converted metadata subset group; and performing data reconstruction on the to-be-converted metadata subset group to obtain a converted metadata subset group. The implementation can improve memory utilization, reduce redundant memory occupation, and reduce unnecessary data access overhead.

The present invention discloses a license plate deblurring method, device, electronic device and storage medium. The method comprises acquiring a license plate image in real time; judging whether the license plate image is blurred, if so, entering a primary deblurring mode, if not, outputting the license plate image; in the primary deblurring mode, inputting the license plate image into a preset cyclic generative adversarial network model to obtain an optimized license plate image; judging whether the optimized license plate image is clear, if so, outputting the optimized license plate image, if not, entering a secondary deblurring mode; in the secondary deblurring mode, inputting the optimized license plate image into a preset super-resolution reconstruction network model to obtain a clear license plate image; and outputting a clear license plate image. The license plate deblurring method, device, electronic device and storage medium provided by the embodiments of the present invention can perform corresponding effective processing for license plates with different blur levels, thereby reducing the resources and time required for license plate deblurring and promoting the intelligent process of license plate deblurring processing.

The present invention discloses a single-channel unsupervised maternal-fetal mixed ECG signal R-peak detection method and device, storage medium, including: step S1, obtaining the maternal-fetal mixed ECG signal; step S2, grading multiple signal segments of the maternal-fetal mixed ECG signal, and determining the R-peak detection strategy of each signal segment; step S3, according to the R-peak detection strategy of each signal segment, determining the acquisition method of the reference R-peak sequence, and detecting the remaining R-peaks in each signal segment in a dynamic cycle extension manner based on the quasi-periodicity and morphological characteristics of the ECG signal; step S4, sequentially splicing the R-peak detection results of each signal segment as the final output. The technical solution of the present invention can simultaneously detect the mother's R peak and the fetal R peak in a single-channel unlabeled maternal-fetal mixed ECG signal, which is suitable for single-channel signal detection and does not need to rely on information interaction between multi-channel signals.

The invention relates to the technical field of Internet protocols, and discloses a method for storing overlapped IP segment types based on LPM and a storage medium, and the method comprises the steps: loading IP segments and VALUE information to an IPVALUE array; constructing an IP scale according to the IPVALUE array; the IPValue array is traversed, measurement is carried out according to the IP scale array, and the IP segment fragments are obtained; querying in an LPM unit according to the IP segment fragment, and when a corresponding fragment IP is found, taking a return value of the LPM unit as a VALUE set of the IP segment fragment; when the IP of the corresponding fragment is not found, creating a Value set, storing the Value information of the fragment, and storing the address of the IP of the fragment and the address of the Value set into an LPM unit; according to the method, the specified IP is queried based on the LPM unit, no return null is queried, the return set array address is queried, and storage and query processing of multiple pieces of overlapped IP segment information are effectively supported under the conditions that the memory is kept less and the query efficiency is high.

The present invention provides a method for encoding a Bitcoin wallet address string and a method for generating an address number table. The method for encoding a Bitcoin wallet address string includes: selecting substrings of different lengths from a number of Bitcoin wallet address strings, counting the number of unique substrings under each length, and calculating the repetition probability of repeated substrings under each length; selecting a lower probability value from each repetition probability, and determining the substring selection length corresponding to the lower probability value as the interception length of each Bitcoin wallet address string; unifying the starting interception position of each Bitcoin wallet address string, and intercepting the substrings of each Bitcoin wallet address string according to the interception length; according to the preset corresponding rules, each character of each substring is respectively corresponded to its own number, and each number of each substring is arranged in character order. The address number table generated by the encoding method of the present invention will greatly reduce the address space occupied by the prior art.

The present invention discloses an analysis method for composite amplified STR data, comprising the following steps: 1. Site reference information: input data; 2. Quality control (preprocessing): filter data; 3. STR typing method: a) find the best match; b) refer to internal reference data. Our method can not only ensure high typing accuracy and timeliness, but also realize automatic conversion of the CODIS system of STR sites. In terms of accuracy, our consistency with capillary electrophoresis data is as high as 100%, and in terms of timeliness, our analysis time is significantly less than the mainstream analysis software on the market. Therefore, as an important supplementary method for whole genome sequence analysis, this process has great practical significance for forensic genetics.

The invention discloses a driver passenger limb conflict behavior identification method and electronic equipment. The method comprises the following steps: S1, obtaining a monitoring picture of a cab; s2, detecting to obtain an arm detection frame in the monitoring picture; s3, detecting in all the arm detection frames to obtain an arm detection frame of the driver; and S4, when it is detected that the arm detection frame of a non-driver makes contact with the arm detection frame of the driver, it is judged that the passenger and the driver have limb conflicts. The method does not need multiple frames of pictures, can give an alarm once contact occurs, and has the performance of low delay. Compared with a video classification model, the detection model occupies less memory, is convenient to deploy in a vehicle-mounted machine, distinguishes the arms of a driver and a passenger when the driver and passenger limb conflict behaviors occur, and triggers an alarm only when the passenger makes contact with the arms of the driver, thereby being more accurate.