11results about How to "The processing result is accurate" patented technology

The embodiment of the invention discloses a task cooperative processing method and device based on multiple agents. The method comprises the steps that a to-be-processed task is disassembled into a plurality of sub-tasks based on a large model; determining a candidate agent corresponding to each sub-task of the plurality of sub-tasks from a preset agent library; evaluating the candidate agent corresponding to each sub-task of the plurality of sub-tasks, and determining a comprehensive score of the candidate agent corresponding to each sub-task of the plurality of sub-tasks; based on the comprehensive score of the candidate agent corresponding to each sub-task of the plurality of sub-tasks and the real-time state of the candidate agent corresponding to each sub-task of the plurality of sub-tasks, determining the agent corresponding to each sub-task of the plurality of sub-tasks by adopting a greedy algorithm; and processing the plurality of sub-tasks based on the intelligent agent corresponding to each of the plurality of sub-tasks to obtain a processing result of the to-be-processed task.

The invention relates to the technical field of image recognition processing, in particular to a logistics order invoice registration recognition method and device, equipment and a storage medium. The logistics order invoice registration identification method comprises the following steps: obtaining an unregistered invoice picture, and carrying out the preprocessing of the unregistered invoice picture, so as to obtain a preprocessed picture; constructing a character recognition model, and processing the preprocessed picture by adopting the character recognition model to extract character information in the preprocessed picture; and establishing a verification rule, verifying the character information according to the verification rule, marking the character information which does not pass the verification, and associating the character information which passes the verification with the logistics order. According to the logistics order invoice registration identification method, invoice registration can be automatically completed, invoice with problems can be accurately identified, invoice registration efficiency can be improved, errors caused by manual registration can be reduced, financial risks can be effectively reduced, and problem invoices can be avoided.

The present invention relates to the field of data processing technology, and in particular to a fire monitoring method based on a mine optical fiber temperature measurement system, including: obtaining a temperature monitoring data sequence at multiple positions in a mine tunnel and the distance between any two positions; obtaining the degree of change of the temperature monitoring data sequence at each sampling moment at each position; obtaining the confidence at each sampling moment; obtaining the connectivity threshold at each sampling moment, and combining the confidence at each sampling moment to obtain the optimal connectivity threshold; according to the optimal connectivity threshold, through cluster analysis based on dynamic splitting of a connectivity graph, obtaining the reference of each temperature monitoring data, and then obtaining the predicted temperature value of each position at the next moment, and comparing it with the early warning value to realize accurate monitoring of the fire in the mine tunnel. The present invention realizes accurate monitoring of the fire based on the mine optical fiber temperature measurement system by adjusting the connectivity graph node weight coefficient according to the regional variation characteristics of the temperature data.

The embodiment of the present invention provides an update method, device, server and storage medium of an AI service platform, wherein the server in the AI ​​service platform includes a first type of AI model and a second type of AI model, wherein the first type of AI model includes multiple first AI models, each of which does not have a function to be updated and all users of the AI ​​service platform can use its function, and the second type of AI model includes multiple second AI models, each of which has a function to be updated and only some users can use its function, and the method includes: obtaining business data generated by a user when using the function to be updated of at least one second AI model; and updating at least one second AI model based on the business data. The business data generated by the user is real data of a real application scenario, and the second AI model trained by the server based on the business data can output accurate processing results for the application scenario, which can improve the accuracy of the processing results.

The present invention discloses a data processing method, device, equipment and medium based on a graph convolutional neural network, the method comprising: obtaining first graph structure data to be processed and inputting it into a pre-trained graph convolutional neural network; calling the graph convolutional neural network to perform neighbor node sampling operations and neighborhood information aggregation operations on each first node in the first graph structure data to update each first node; calling the graph convolutional neural network to obtain the processing result of the first graph structure data based on each updated first node. The present invention enables the graph convolutional neural network to mine the differences in the relationship between nodes, so that a more accurate processing result can be made based on the more sufficient information mined.

The embodiment of the invention relates to a photoetching mask, a photoetching alignment mark and a same-layer exposure error compensation method, and the photoetching mask comprises an exposure unit which is used for carrying out stepping exposure on a wafer and comprises at least one group of photoetching alignment patterns consisting of a first pattern and a second pattern, the first pattern and the second pattern are respectively positioned on two sides, far away from each other, of the exposure unit along the stepping direction, and the first pattern and the second pattern are open patterns; a first mark and a second mark are correspondingly formed when the first pattern and the second pattern are transferred to a wafer, and in two adjacent exposure patterns, the first mark in one exposure pattern and the second mark in the other exposure pattern can be combined in an overlapping region to form a photoetching alignment mark containing a closed pattern; the closed pattern comprises at least three closed areas, the centers of the at least three closed areas are not collinear, and the photoetching alignment mark is used for realizing alignment when the next photoetching process is performed on the wafer. Therefore, accurate alignment can be realized.

The present invention relates to the technical field of fault warning, and in particular to a fault warning remote management method applied to an intelligent cleaning robot, the method comprising: a remote management terminal controls a data acquisition device to take a photo of a product, obtains a dirty image of the product before cleaning, inputs the dirty image into a pre-trained machine learning model, and outputs a decontamination image; imports the dirty image and the decontamination image into an image comparison strategy to calculate the contamination index of the product before cleaning, and marks the degree of contamination of different areas of the product before cleaning; obtains a later image of the product after cleaning, imports the later image and the decontamination image into an image comparison strategy to calculate the contamination index of the product after cleaning, and calculates the change in the degree of contamination of different areas of the product after cleaning; according to the change in the degree of contamination of different areas of the product, it is judged whether the intelligent cleaning robot has a fault and the feedback is sent to the remote management terminal. The data processing volume is small, and the processing result is accurate, more intuitive and effective.

The invention discloses a distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, and the device comprises a chaotic fiber laser, a linear Sagnac interferometer, and a computer. The chaotic fiber laser comprises a semiconductor pumping source, a wavelength division multiplexer, an erbium-doped fiber, a polarization controller, a fourth fiber coupler and an optical isolator which are connected in sequence, the optical isolator is connected with the wavelength division multiplexer, and the fiber coupler is connected with a linear Sagnac interferometer. According to the distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, a chaotic fiber sensing system and a Sagnac interference light path are combined, and a convolutional neural network is introduced to extract weak vibration signals, so that an effective signal frequency range can be extracted under a relatively low signal-to-noise ratio; the sensitivity is higher and the detection range is wider.

The application relates to the technical field of heat dissipation devices, in particular to an integrated asymmetric radiation envelope structure heat dissipation system for a computing power computer room, which comprises a heat dissipation module, a running module, a heat import optimization module, an image sensing module, a data acquisition module, a data processing module, a data analysis module and an execution module. The system preprocesses the running data of the received heat dissipation module, judges the heat dissipation efficiency of the system through the effective net radiation heat dissipation power density and a preset threshold value, adjusts the average mirror reflectivity of the photon mirror based on the judgment result, and further determines the standard effective net radiation heat dissipation power density according to the adjustment result. The application realizes the unified processing and analysis performance optimization of the heat dissipation system, effectively reduces the effective net radiation heat dissipation power density of the envelope structure heat dissipation system of the application, and improves the resource utilization rate.

The invention discloses a time correlation data processing method and device based on double time sequences. The method comprises the following steps: acquiring a plurality of pieces of time-related data from a plurality of data sources in real time, wherein each piece of time-related data comprises a first timestamp representing data occurrence time; for each piece of time-related data, recording the acquisition time of the piece of time-related data to obtain a second timestamp representing the data recording time; for each piece of time-related data, constructing a time identifier of the piece of time-related data according to the first timestamp and the second timestamp; based on the time identifiers of the multiple pieces of time-related data, multiple pieces of target time-related data with mutually matched time sequence characteristics are selected from the multiple pieces of time-related data; and processing the time-related business according to the multiple pieces of target time-related data. According to the data processing method disclosed by the embodiment of the invention, accurate and efficient processing of the time-related data is realized by utilizing the double-time-sequence index and the time-sequence feature matching.

The present application relates to a screen detection method, device, computer equipment and readable storage medium. The method comprises: obtaining a detection screen image of a target device currently being monitored, dividing the detection screen image through a plurality of different window specifications to obtain a plurality of detection image blocks, performing correlation processing based on the plurality of detection image blocks and a plurality of template image blocks in a template screen image, and determining an abnormal detection result of the detection screen image. The above method takes into account the image pixel level, divides the detection screen image through different window specifications to obtain a smaller detection image block in the detection screen image, and implements screen detection through a smaller detection image block and a template image block in the template screen image, thereby improving the accuracy of whether there are abnormal defects in the screen image displayed on the display interface of the BMC, and further improving the accuracy of whether there are flower screen defects in the screen image displayed on the display interface of the BMC.