19results about How to "Reduce calls" patented technology

The invention discloses an intelligent question answering system and method based on question decomposition and community semantic search. According to the method, firstly, a complex question is decomposed into a plurality of single-hop sub-questions through an original loop question decomposition technology, then information needed for answering each sub-question is obtained based on communities in a vector retrieval knowledge graph, and the sub-questions are answered by using a large language model. And finally, generating a final question and answer for the original complex question by combining the LLM with the answers of all the sub-questions. Compared with other methods in the industry, the method has the advantages that the information integration capability is remarkably improved, information fragmentation is avoided through community-level search and abstract generation, and high-quality and logically consistent answers are provided. When a complex query is processed, the effect of the method is superior to that of a general method in the industry and is improved by 29% at most; meanwhile, on the basis of ensuring a better question and answer effect, the efficiency is remarkably improved, and on the premise that the effect exceeds that of an existing optimal method GraphRAG, the token consumption is only 3.2% of that of the existing optimal method GraphRAG.

The application discloses a general-purpose learning method and system for a physical robot, which comprises a special talent model library, a monitor set, a general talent model and a multi-robot distributed exploration cluster. The special talent model library contains L pre-trained special talent models. The monitor set contains L monitors, each of which is arranged with a special talent model and used for evaluating whether a current input state belongs to the use range of the special talent model. The multi-robot distributed exploration cluster contains multiple physical robots, each of which is arranged in a different task scene. In the application, multiple robots are introduced to perform different tasks in parallel, the monitor is used to dynamically determine when the special talent model needs to be involved, the appropriate special talent model is selected for guidance, the DAGGER type iterative data aggregation mechanism is adopted, the general talent model continuously learns on the state distribution generated by self-exploration in multiple scenes, and finally, the general talent model which can adapt to multiple scenes and surpass a single special talent is obtained.

The invention provides a camera data transmission method and device based on a kernel driver and a related product, and the method and device can create an independent memory space for a camera based on the kernel driver. According to the method, the camera data is acquired and cached to the pre-created memory space, and finally the camera data is acquired in the memory space, that is, the camera data can be acquired in the pre-created memory space, and the data packet is transferred to the business layer through the kernel driving layer, so that system call caused by traditional use of socket is avoided, and the system performance is improved. 80000 + times of system calling can be reduced for a single picture, 80000000 + CPU instructions can be reduced, system calling is reduced, switching between a user mode and a kernel mode is reduced, and CPU consumption is reduced.

The invention provides an information processing method and device, electronic equipment and a storage medium, which can be applied to the technical field of finance and medical treatment, and the method comprises the following steps: determining a first time period of first receiving time of receiving target token information relative to a preset time point; performing secret key generation processing according to the first time period to obtain a first secret key; performing decryption processing on the target token information through the first secret key, and if decryption fails, determining that verification of the target token information fails; wherein the target token information is generated in the following mode: determining a second time period of second receiving time of receiving a request for sharing the target information relative to a preset time point; performing secret key generation processing according to a second time period to obtain a second secret key; and performing token information generation based on the second secret key according to the target information to obtain target token information. According to the embodiment of the invention, the influence of calling the secret key on the interface performance of the storage medium in the decryption process of the encrypted token information can be reduced.

The invention discloses a method and device for recognizing and reconstructing vector characters of a CAD drawing, and relates to the technical field of data processing. The method comprises the steps that a PDF file is analyzed, and vector elements and attribute information of the vector elements are extracted; character stroke candidate lines are screened out according to attribute grouping; carrying out connectivity analysis and density-based spatial clustering on the lines, and identifying a text region; rendering the vector lines of the text region into an image with high resolution, carrying out single character segmentation, and recognizing characters by combining drawing-level font library template matching with a single character recognition model; combining the single characters into word groups according to the positions and the context information, and performing word order correction by adopting a language model; the recognition result is written back into an editable CAD text object; according to the method, vector geometric information is directly utilized, distortion of bitmap sampling is avoided, the small character recognition rate and the anti-jamming capability are effectively improved, and end-to-end semantic recovery from vector lines to searchable and editable texts is achieved.

The present application relates to a video data processing method, device, equipment and storage medium, which are applied in the field of image processing, wherein the method comprises: obtaining a video image, storing the video image in a preset resource cache area, reading preset image data in a flash memory, superimposing the image data into the video image, obtaining a processed video image, and sending the processed video image to a video output channel for display. The technical effect of the present application is: improving the performance of video superposition processing and reducing the time consumption of video superposition processing.

The invention belongs to the technical field of 5G communication transmission, and provides a method for improving the performance of an integrated station and user equipment thereof, which are used for optimizing GTPU message transmission among an NGC, a CU and a DU. The method comprises the following steps: firstly, defining a shared memory area, dividing the area into a plurality of buffer memory pools with 2048 bytes, and distributing an index number for each buffer; in the communication process, data are not directly transmitted, but index values of the buffers are transmitted. In order to improve the efficiency, a batch processing mechanism is adopted, that is, only when 16 buffers are accumulated, system calling sending is carried out once, or when 16 buffers are not accumulated in a certain time, sending is carried out, so that data timeout is prevented. In the downlink process, after the NGC fills the message, the index is transmitted to the CU through the FIFO, the index is transmitted to the DU after being processed by the CU, and finally the memory is released by the DU. The uplink process starts from the DU, passes through the CU and is finally processed by the NGC, and the memory is released. According to the invention, the system calling frequency is reduced, the data replication overhead is reduced, and the communication efficiency is improved.

The application discloses a lossless compression method and system for image sensor raw data, and relates to the technical field of image sensors.The application fully utilizes the statistical characteristics that the pixel value difference between adjacent frames under the same exposure is extremely small, combines the inherent color arrangement rule of the Bayer format image, separates the color channels first and then performs time domain difference calculation, greatly reduces the data entropy value, and obtains a compression ratio much higher than that of a general lossless compression algorithm on the premise of completely lossless original pixel values; meanwhile, independent compression and packaging are performed with the minimum service unit as the granularity, target data can be directly located and accessed without decompression and large-scale package, the data retrieval and calculation overheads are reduced, the traceability of the original data is retained, the statistical confidence of the test result is not affected, and the batch data storage requirement in the image sensor production test scene is adapted.

The invention discloses a multi-terminal control right management method and system, and the method comprises the following steps: matching control elements of other terminals according to a control element trigger signal when any terminal outputs the control element trigger signal, removing the control right of the matched control element, and keeping the control right of the unmatched control element. The method has the advantages that it is ensured that the same function of the same device cannot be controlled by multiple terminals, signal conflicts are avoided, the safety of the controlled device is ensured, control elements with different functions in other terminals are reserved, a user can still control other functions of the device through other terminals, and the user experience is improved. The flexibility of multi-terminal control is kept under the condition of avoiding multi-terminal control conflicts, and the user experience is improved.

The application provides a kind of multi virtual power plant transaction decision method and device for day-ahead electricity market, it is related to virtual power plant optimization operation technical field.The method comprises: according to distribution network operator and multiple virtual power plant systems, distribution network operator-virtual power plant master-slave game double-layer structure is established;Objective function and constraint condition of distribution network operator pricing game model are constructed;Objective function and constraint condition of deterministic virtual power plant robust transaction model are constructed;The uncertainty of wind, photovoltaic and controllable load is considered, and the deterministic virtual power plant robust transaction model is converted into uncertainty virtual power plant robust transaction model, and then a distributed robust scheduling model is obtained;Dynamic kriging element model is used to solve, and local search mechanism is introduced to obtain transaction decision;Wherein, column and constraint generation algorithm is used to solve distributed robust scheduling model.The application can help virtual power plant aggregator to significantly improve the income in day-ahead transaction.

The invention discloses a Yarn application process tree construction method and device and a storage medium, and relates to the technical field of Hadoop YARN resource management and monitoring. A method for constructing a Yarn application process tree comprises the steps that a monitoring program is started, and a monitor marks monitored application processes as different codes according to configuration; executing corresponding monitoring logic according to the code to obtain an application process tree construction result; the monitoring logic comprises a traditional global scanning construction mode, a single-process mode, a sub-process interface construction mode and an incremental updating mode. The method is simple in implementation mode and suitable for process monitoring and resource metering optimization of the node side of the YARN node-level resource manager, self-adaptive selection for different types of containers is achieved by introducing a multi-mode switchable monitoring framework, and resource waste and performance bottleneck caused by traditional fixed full scanning are avoided.

The embodiment of the invention provides a calling method and device for standard part resources in industrial design software, computer equipment and a medium, and the method comprises the steps: receiving a matching condition input by a user on a design software platform, screening out a standard part three-dimensional model meeting the matching condition according to the matching condition, and storing the standard part three-dimensional model in the design software platform; taking the model as a candidate standard component three-dimensional model; selecting an adaptive standard component three-dimensional model from the candidate standard component three-dimensional models according to actual installation requirements as a target standard component three-dimensional model; recommending a matched standard component three-dimensional model according to the target standard component three-dimensional model; and calling the target standard component three-dimensional model and the matched standard component three-dimensional model for installation, and displaying attribute information on the target standard component three-dimensional model and the matched standard component three-dimensional model. According to the scheme, the design process is optimized through standard part full-life-cycle tracing and global consistency management, and the standard part calling and assembling time is shortened.

The application relates to the technical field of e-commerce marketing compliance auditing, and discloses a marketing content compliance auditing method based on a domain large model; the method first acquires marketing materials to be audited, a delivery context, supervision rules and platform auditing rules, splits the rules into auditing requirements, and establishes a mapping relationship between the auditing requirements and violation categories; carrier recognition, content extraction, format normalization and fact fragment association are performed on texts, pictures, videos and landing page links; requirement level matching, exception situation verification and disposition level aggregation are performed according to the deterministic auditing rules, the confidence interval and the domain large model, and an auditing conclusion and a disposition instruction are generated; the application can make the multi-modal marketing content auditing conclusion locatable, traceable and reviewable, and reduce the amount of manual review.

The invention relates to the technical field of positioning instrument image processing, and discloses an image processing method of a positioning instrument, the method adopts an optical filter with an inner and outer ring structure to collect an infrared image, the inner ring is specially designed for a lens, and the outer ring filters an infrared lamp. According to the method, graying processing is carried out on an acquired infrared image, a Sobel operator is used for calculating horizontal and vertical gradients, and then the gradient amplitude and direction are determined. Through comparison with a set threshold value, edge pixel points of the object image are accurately identified, and finally a clear edge image is output. Compared with a traditional method, noise and miscellaneous points are effectively reduced through the inner-ring optical filter and the outer-ring optical filter, the image preprocessing step is simplified, and the calculation complexity is reduced. Meanwhile, due to the clear input image, the Sobel operator does not need to be additionally adjusted during gradient calculation, and the calculation process is further simplified.

This application relates to a video rendering processing method, apparatus, electronic device, and storage medium. The method includes: requesting a preset number of window handles and storing the preset number of window handles as a handle pool; when rendering video, obtaining an available window handle from the handle pool, creating the necessary rendering resources for the available window handle, and starting the video rendering process; when stopping video rendering, searching the handle pool for the window handle corresponding to the stopped video rendering, and clearing the rendering resources of the found window handle. This application solves the problem in the prior art where, during frequent and continuous multi-channel video rendering without restarting, the terminal rendering software causes a continuous accumulation of window handles, thus affecting system stability.

The invention discloses a web of thins autonomous management and control system and method for environment control, and belongs to the technical field of Internet of Things and computer application. The management and control system mainly comprises an equipment registration module (1), a user strategy module (2), an application strategy module (3), a sensor management and control module (4), an actuator management and control module (5) and a reasoning decision module (6). According to the method, the object description specification proposed by the W3C is expanded, so that equipment in different application scenes and fields can be more comprehensively described, and meanwhile, the situation that calling of the equipment is influenced due to different descriptions of the same function caused by common sense cognition in the specification is avoided.

The application belongs to the field of rapid design and manufacture of production equipment, and particularly relates to a robot gripper structure parameter rapid simulation optimization method, which comprises the following steps: constructing a robot gripper structure parameterization model; simulating the stress and deformation of the robot gripper structure parameterization model under different sizes and different load conditions by using statics finite element simulation software to obtain stress and deformation simulation data; constructing a stress and deformation prediction model; optimizing the size parameters that meet the stress and deformation constraint conditions under the actual load conditions by establishing a particle swarm optimization algorithm, obtaining the optimal value of the size parameters, and substituting the optimal value into the stress and deformation prediction model to obtain the size parameters under the actual conditions and the stress and deformation under the load, so as to realize the optimization of the robot gripper structure parameters. The application utilizes the Kriging surrogate model to predict the stress and deformation, reduces the calling of the finite element model, and improves the optimization efficiency.

The invention provides a resource search method and device, equipment, a storage medium and a program product, and the method comprises the steps: carrying out the intention recognition of a resource search request inputted by a user based on a pre-trained large language model, and obtaining a request resource entity and a request resource type; determining a plurality of target search engines from a plurality of search engines based on the request resource type, and expanding search terms based on the request resource type; calling each target search engine for searching based on the request resource entity and the expanded search term; and verifying a search result returned by the target search engine, and generating a response of the resource search request based on the verified search result. Multiple search engines are used for parallel retrieval, so that the coverage degree and the search success rate are improved; by automatically verifying the search result, the search result filtering is realized, the user operation is simplified, and the convenience of resource downloading is improved.