251results about How to "Achieve efficiency" patented technology

The invention relates to the technical field of traffic safety management, in particular to a vehicle abnormal driving recognition and early warning method and system. The method comprises the following steps: firstly, acquiring vehicle position, speed, direction and driver operation data in real time, then correcting time sequence data by adopting a smooth denoising method of adaptive segmented clustering, generating an information confirmation code through standardization, and then carrying out weighted fusion on multi-source data to obtain a multi-source data fusion model; the strategy decision-making module is used for extracting vehicle speed change rate, lane deviation degree and driver behavior characteristic key indexes, identifying and grading abnormal driving behaviors by setting a threshold value, calculating a comprehensive abnormal score, and generating a grading intervention strategy for slight, medium and serious anomalies according to a comparison result of the abnormal score and the set threshold value. And finally, combining automatic intervention in the vehicle with global coordinated intervention of a road manager. According to the invention, the intervention strategy is generated based on multi-source data acquisition, processing, analysis and decision, so that the vehicle driving safety and the road management efficiency are improved.

The invention provides a multi-AGV path planning and obstacle avoidance method based on an improved artificial bee colony algorithm, and aims to solve the key technical challenges of multi-AGV system path planning and obstacle avoidance in modern industrial automation. According to the method, a dynamic updatable gridding environment model is constructed to accurately represent a starting point, an end point, static obstacles and dynamic obstacles in a working environment, an improved artificial bee colony algorithm is adopted to globally optimize path planning of multiple AGVs, constraint conditions of obstacle avoidance and multi-AGV path conflicts are met, and the optimal path planning of the multiple AGVs is realized. And the efficient operation of the system and the path collision-free property are ensured. In addition, the method further comprises path tracking control and conflict detection and obstacle avoidance control so as to ensure that the AGVs can accurately run along the planned path, conflict points in the paths of the multiple AGVs are detected in real time, and different coping strategies are adopted to achieve path coordination and scheduling. Finally, the cooperation efficiency and safety of the multi-AGV system can be effectively improved.

The invention discloses a visual detection system for an automobile door panel assembly, and relates to the technical field of automatic detection, and the visual detection system comprises the steps: installing a high-precision industrial camera, and carrying out the synchronous collection of the multi-view image data of an automobile door panel; preprocessing the acquired multi-view image data of the automobile door panel; carrying out detail enhancement on the preprocessed multi-view image data of the automobile door panel by adopting a super-resolution algorithm; integrating the enhanced multi-view image data of the automobile door panel into an automobile door panel panorama through binocular stereoscopic vision matching, and constructing an automobile door panel three-dimensional model; the automobile door panel three-dimensional model is compared with a standard automobile door panel three-dimensional model, and the buckle defect in the automobile door panel is recognized; and carrying out position marking and defect type analysis on the identified defect points, and generating an automobile door panel detection report according to the defect type analysis. According to the invention, details are enhanced by using a super-resolution algorithm, defects are accurately positioned by using a three-dimensional model, and the detection efficiency and accuracy are improved.

The invention discloses a wireless communication network data secure transmission method and system, and relates to the technical field of secure transmission. During system operation, agricultural monitoring data are collected in real time through various sensor devices arranged in an agricultural environment, encryption processing is performed by using a symmetric encryption algorithm, and a dynamic key length adjustment mechanism is supported; a random number generator is integrated, unified management is performed on key generation, distribution, updating and destruction in a system, a timestamp or a random serial number is embedded into each data packet, the uniqueness of the timestamp or the serial number is verified at a receiving end, so that replay attacks are resisted, and the data are comprehensively analyzed and evaluated according to a set algorithm, so that the replay attack is prevented. The method comprises the following steps: acquiring a security data transmission performance index SDTP through calculation, generating a monitoring report, detecting abnormal conditions in a data transmission process, including channel quality reduction, data loss or verification failure, and providing a visual monitoring result.

The invention discloses a method for performing efficient lossless text compression based on a large language model, which relates to the technical field of text compression and comprises the following steps of: preprocessing to-be-compressed text data to obtain an initial token sequence; selecting and loading a pre-trained big language model, analyzing the current context information by using the big language model, obtaining a probability value of each token, forming a probability distribution list, and recording a ranking sequence number of the current token; the context window is moved backwards by one token, and context information is updated until all the tokens are processed; and recording the ranking sequence numbers of all the tokens in sequence to form a sequence number list so as to obtain a final text compression result. According to the method, the compression rate is remarkably increased, the requirements for storage space and transmission bandwidth are reduced, the high processing speed is kept, and the method is suitable for real-time compression and decompression of large-scale text data.

The invention belongs to the technical field of waste resource recycling, and particularly relates to a waste resource recycling plastic product manufacturing and forming device which comprises a base, an injection mechanism used for plastic plasticizing and moving is arranged at the top of the base, and a first feeding hopper is arranged at the top of the injection mechanism; a mold closing mechanism connected with a discharging port of the injection mechanism is arranged at the top of the base and used for conducting pressing molding on molten plastic, a treatment box is arranged at the position, located on one side of the injection mechanism, of the top of the base, and a smashing mechanism used for smashing waste resources is arranged in the treatment box; the two driving crushing cutters are driven by the first motor to synchronously and relatively rotate, and then are matched with the fixed crushing cutter to crush waste plastics; according to the waste plastic crushing and cleaning device, multiple functions of crushing, screening, cleaning, drying and the like can be ingeniously integrated, the crushing effect and efficiency of waste plastic are improved, and the powder cleaning effect and the powder quality are also improved.

The invention discloses a high-frequency data remote transmission method and system, and the method comprises the steps: dynamically adjusting the collection frequency and mode according to environment conditions, and obtaining high-frequency data in real time; based on the data type and the transmission network state, a compression algorithm, a filtering mode and a coding strategy are automatically selected; the congestion condition of the transmission network is predicted through a built-in machine learning model, and a packaging strategy and transmission parameters are dynamically adjusted according to network conditions; in the data receiving process, the bit error rate in the transmission process is monitored in real time, abnormity in transmission is automatically detected, repair is carried out through an error correction algorithm, and if repair fails, the system automatically requests the data packet again; and the recovered high-frequency data is classified and stored according to the importance, real-time performance and storage cost of the data, and automatic storage strategy adjustment is realized in a remote database. According to the technical scheme, the high efficiency, reliability and safety of remote transmission of high-frequency data are improved.

The invention relates to the technical field of signal encryption, in particular to an aviation data bus signal encryption method and system based on a field programmable gate array (FPGA), which comprises the following steps of: based on input three-dimensional tensor signal data, extracting slices from a time index of an aviation data bus according to a fixed interval, progressively sequencing amplitudes of signal channels, and then segmenting; according to the method, through processing logic of matrix decomposition and feature curve generation, the feature extraction and dimension reduction process is optimized, the distinction degree and the expression capacity of feature information are improved, the randomness and the anti-cracking capacity of a key are enhanced through a key generation mode combining a dynamic disturbance factor and an encryption parameter, and the safety of signal transmission is guaranteed; through combination of amplitude superposition and frame-by-frame storage, dynamism and high efficiency of a signal encryption process are realized, signal tampering and eavesdropping risks are reduced, and through feature reconstruction and point-by-point matching processing in a decryption stage, accurate restoration of signal amplitude is ensured, and data integrity and anti-interference capability are guaranteed.

The invention provides a task scheduling method and device, electronic equipment and a computer readable storage medium, and relates to the technical field of computers. The method comprises the following steps: adjusting the current priority of each task to be scheduled in a main queue according to the current resource load; under the condition that the priority adjustment frequency corresponding to the to-be-scheduled task reaches a preset frequency and the duration of the to-be-scheduled task entering the main queue exceeds a first preset duration, the to-be-scheduled task is stored in a retention queue as a retention task; for the detention tasks with overlong waiting time in the detention queue, the detention tasks are added into the main queue again after the priorities are adjusted, so that the tasks are scheduled to be dequeued and executed according to the dynamically adjusted priority sequence, the tasks with high priorities can be executed as soon as possible, and the situation that the tasks with low priorities cannot be processed for a long time can be avoided; the efficiency and the fairness of task processing are considered under the condition of high concurrency.

The invention provides a coalescence-selective filtration-deep bed filtration efficient oil-water separation system which comprises a coalescence system, a selective filtration system and a deep bed filtration system, oily wastewater enters a coalescer, and tiny oil droplets are coalesced into larger oil droplets through wetting coalescence and collision coalescence; the coalesced wastewater enters a selective filter, oil droplets and water respectively flow out through an oil channel and a water channel, and primary separation is realized; according to the present invention, the water after the primary separation enters the deep bed filter to achieve the deep separation, the efficient oil-water separation and the water quality standard discharge or reuse are achieved through the combination process of coalescence, selective filtration and deep bed filtration, and the system has advantages of high treatment efficiency, low energy consumption, less carbon emission, oil and water recycling, or water standard discharge.

The invention discloses a video virtual fitting framework in a real scene based on a controllable image diffusion model. The video virtual fitting framework comprises a data preprocessing module used for constructing feature expressions of figures and clothes; the single-stage image fitting network is used for generating a virtual fitting image based on the preprocessed data and performing implicit deformation and fusion of clothes by using a cross attention mechanism; the diffusion guiding module is used for enhancing the time sequence consistency of the generated video, a self-supervision model Video MAE and a self-supervision image model DINO-V2 are introduced, the video is masked by the self-supervision model Video MAE, and a complete video is reconstructed through inter-frame information; and the long video generation module is used for generating a long video and keeping the consistency of the video, splitting the complete video into a plurality of short video clips, and reducing the jitter at the clip joints by using a time sequence common noise reduction strategy. The diffusion guiding module is designed, efficient time sequence consistent video virtual fitting is achieved through self-supervised learning, and the method has the advantages of being excellent in shielding processing, stable in long video generation and the like.

The invention discloses a data transmission and control method and system based on a Beidou short message, and relates to the technical field of remote monitoring and communication, and the method comprises the steps: a Beidou slave station receives remote signaling and telemetering data, carries out the data processing of the received data, packages the data into a Beidou short message format, carries out the data transmission through a multi-mode command transmission protocol, and carries out the data transmission through a Beidou short message transmission protocol; the method comprises the following steps: receiving command data converted by a Beidou master station, monitoring a system in real time through a command execution state, feeding back an execution progress and state to the master station in real time, automatically executing a recovery process through a fault self-healing mechanism when data transmission is abnormal, and automatically adjusting data transmission frequency and power by utilizing a self-adaptive energy management algorithm, so as to realize real-time data transmission. And through the data transmission prediction model, predicting a data transmission demand, and adjusting a data refreshing strategy in advance. Through the adaptive energy management algorithm and the prediction model, the data transmission frequency and power are dynamically adjusted, the energy consumption is optimized, and the data transmission efficiency and real-time performance are improved.

The invention relates to the technical field of maritime ship target beyond visual range monitoring, in particular to a compact type ground wave radar maritime ship track prediction method and system. AIS track data are obtained and converted into distance, speed and azimuth angles under a radar coordinate system to serve as characteristic parameters; the multi-dimensional target features which are segmented into n historical time steps through preprocessing and a sliding window serve as input of a prediction model, and the model outputs a prediction result of a future single time step; in the prediction model, each encoder layer extracts time sequence features of different depths from input features, and modeling is performed on the time sequence features by using a plurality of different attention mechanisms; and the output of the last layer of encoder passes through a linear mapping layer, and the high-dimensional time sequence representation is converted into the output prediction characteristics of the future time step. The multi-head self-attention mechanism is utilized to dynamically distribute the weights of the multi-dimensional target features, the complex association among different features is captured, the relationship among different time steps is analyzed from multiple angles, and the anti-interference capability is enhanced.

The invention discloses a power grid transient voltage coordination control method under a distributed photovoltaic storage high permeability working condition, and the method comprises the steps: detecting the voltage change condition of a grid-connected point where a distributed photovoltaic power station is located, and combining the short-circuit ratio of the grid-connected point where the distributed photovoltaic power station is located to a collection point; carrying out pre-estimation calculation on the reactive compensation amount required near the grid-connected point where the key point is located; and through optical storage cooperative control and consideration of different photovoltaic reactive power margin conditions, the reactive power compensation capability of the photovoltaic inverter is developed, smooth processing of the output power of the photovoltaic inverter is realized, and the transient voltage stability of a grid-connected point is improved.

The invention relates to a noise feature vector classification model construction method, a speech enhancement method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: coding penalty parameters and kernel function parameters of a support vector machine; taking the penalty parameter and the kernel function parameter as optimization variables of a genetic algorithm, setting an initial population, and constructing a fitness evaluation function based on classification accuracy; performing selection operation, crossover operation and mutation operation on the initial population, and updating an iterative population according to a fitness evaluation function to obtain an optimal penalty parameter and an optimal kernel function parameter; and training a support vector machine through the optimal penalty parameter and the optimal kernel function parameter, and constructing a noise feature vector classification model. In a noisy environment, the classification effect of the noise feature vector is improved under the condition that nonlinear and non-stationary noise exists, and the noise feature classification requirement in a complex sound signal processing scene is met.

The invention provides a super-lens reverse design method based on a selective dual-channel perception fusion network, and the method comprises the steps: obtaining a picture-phase spectrum data set through obtaining K pairs of original images and corresponding original phase spectrums; inputting the data to an adaptive difference frequency feature extraction network ADFFNet; inputting a selective non-common feature interaction network SFIFNet to respectively obtain image frequency features and spectral frequency features; the fusion reconstruction features of the image and the fusion reconstruction features of the spectrum are obtained; obtaining a picture generation result and a spectrum generation result; according to the method, the non-common frequency characteristics between the image and the corresponding spectrum can be deeply mined and deeply fused, the mapping accuracy between the image and the corresponding spectrum response can be improved, and the generation accuracy and efficiency of the image and the spectrum can be improved.

The invention provides a workpiece defect detection method and device based on visual inspection, and the method comprises the steps: obtaining the position information of a to-be-detected workpiece, carrying out the image collection of the to-be-detected workpiece based on the position information, and obtaining a plurality of groups of detection images of the to-be-detected workpiece under different visual inspection conditions; and carrying out workpiece defect detection on the plurality of groups of detection images by adopting a comparison judgment and fusion judgment method. The invention provides a workpiece defect visual detection method based on a CCD camera image, and aims to improve the efficiency and accuracy of workpiece defect detection and prevent visual errors caused by manual detection so as to effectively avoid missing detection and false detection of defective workpieces.

The invention relates to the technical field of smart communities, in particular to a future community construction method and system based on government big data, and the method comprises the following steps: collecting power use data, water resource consumption data and public transport utilization rate data in a community, synchronizing the data, and carrying out the standardization processing, and generating a community resource use data set. According to the method, through a time sequence analysis method, the resource demand of a community in a future week can be effectively predicted, a resource flow simulation model is established, prediction data is input in real time, model parameters are adjusted, influences of policies are continuously simulated and evaluated, optimal allocation of community resources is achieved, and the community resource allocation efficiency is improved by constructing a resource allocation diagram and determining an optimal path. The method achieves the high efficiency and fairness of resource distribution, continuously tracks the distribution effect and resident feedback, further guarantees the adaptability and continuity of resource distribution, not only optimizes the community resource management, but also improves the welfare of residents, and enhances the sustainable development of communities.

The invention discloses a screen cabinet adding method based on BIM and a computer vision technology. The method comprises the following steps: 1) establishing a BIM model template; 2) carrying out 2D visual detection and ROI extraction; 3) acquiring and downsampling 3D point cloud data; 4) processing the downsampled 3D point cloud data by using point cloud correction filtering, including removing point cloud outliers by using an improved DBSCAN algorithm and performing point cloud completion based on a grid method, thereby improving the quality of the point cloud; and 5) registering the point cloud and the BIM model template, and verifying whether the installation space meets the installation requirements of the screen cabinet or not. According to the method, 2D and 3D visual technologies are combined, efficient and accurate space information acquisition and screen cabinet adding feasibility evaluation are achieved, the labor cost is reduced, the design accuracy is improved, and technical support is provided for efficient transformation of the urban power distribution room.

The invention relates to the field of earthquake early warning, and discloses an earthquake early warning system based on a marine government platform, and the system comprises a data collection module which is used for obtaining the earthquake information of a historical earthquake, and the earthquake information comprises epicentral coordinates, epicentral depths, earthquake grades, and corresponding intensity attenuation models and seismic zones; the intensity prediction module is used for constructing a generalization intensity prediction model based on the epicentral coordinates, the epicentral depth and the earthquake grade, and performing differentiation correction on the generalization intensity prediction model according to the seismic zone to which the generalization intensity prediction model belongs, so as to obtain a fitting intensity prediction model; the output of the fitting intensity prediction model represents the prediction intensity of any coordinate about epicenter coordinates; the region division module is used for dividing the earthquake region into a first early warning region to an Nth early warning region based on the epicenter depth and the earthquake grade by taking epicenter coordinates as a center and combining a fitting intensity prediction model; and the early warning sending module is used for distributing early warning sending priorities to the first early warning area to the Nth early warning area according to the earthquake grade.

The present invention proposes a structural safety monitoring method, device, medium and equipment based on artificial intelligence, which are applied to a client and a server. The server obtains data to be analyzed based on structural safety monitoring of a building sent by at least one client in real time; parses the data to be analyzed to obtain first data to be analyzed based on location features and second data to be analyzed based on early warning data; constructs a spatial coordinate system based on the first data to be analyzed, generates a BIM sub-model to be matched based on the data to be analyzed, and matches it with a pre-constructed BIM model. If a BIM sub-model exists, the BIM sub-model data is retrieved; the second data to be analyzed is matched with the BIM sub-model data to determine whether there is an abnormality. If so, an abnormal early warning message is sent to at least one client. The combination of improved BIM technology and artificial intelligence technology achieves high efficiency and accuracy in safety monitoring of building structures.

The invention discloses a fruit and vegetable quality nondestructive testing imaging control system which comprises the following components: S1, a multispectral light source module which captures spectral information inside and on the surface of the multispectral light source module; s2, a precision imaging module generates multispectral image data; s3, extracting key image features related to fruit and vegetable quality by an image data preprocessing module; s4, the quality analysis module adjusts algorithm parameters in real time and generates an analysis data set corresponding to the fruit and vegetable quality indexes; s5, a spectrum enhancement and adaptation module, which optimizes the sensitivity and resolution of the multispectral signal; s6, a distributed detection control module is used for controlling system parameters to be adjusted in real time, and the detection efficiency and detection are optimized; and S7, a data management and output module which generates a fruit and vegetable quality evaluation report through a computer system. The method has the advantages of being high in detection precision, high in adaptability, good in nondestructive property and high in detection efficiency.

The invention belongs to the technical field of image recognition and classification, and relates to an algae image classification system and method based on data enhancement. Dividing algae image data to be monitored into a training set and a test set, and performing data enhancement processing on images in the training set; and inputting the processed training set into a preset convolutional neural network to carry out algae cell identification and classification training, calculating the cross entropy loss of an algae real category in the training set and an algae prediction category output by the preset convolutional neural network, and carrying out network parameter updating on the cross entropy loss in combination with a back propagation and gradient descent algorithm. Obtaining an algae identification and classification model after parameter updating; and inputting the test set into the algae identification and classification model, and evaluating the accuracy of the real algae category in the test set and the algae output category output by the algae identification and classification model to obtain an optimized algae identification and classification model. According to the method, the algae images are classified, and high efficiency and accuracy of classification can be realized.

The invention provides a remote desktop configuration method and device and electronic equipment, and belongs to the technical field of computers. The method comprises the steps of obtaining employee portrait data, wherein the employee portrait data comprises at least two of basic information, organization information, work content information, work mode information, project participation information and confidential degree information; determining a first target configuration template matched with the employee portrait data based on a preset matching rule; the preset matching rule comprises remote desktop configuration templates corresponding to different employee portrait data; and sending the first target configuration template to a remote desktop server to perform remote desktop configuration based on the first target configuration template. According to the staff portrait data, a plurality of dimension features such as staff departments, work contents, levels, participation items and confidential degrees are comprehensively considered, the staff portrait data with the multi-dimension features are automatically matched with the remote desktop configuration template through the preset matching rule, and refinement, automation and efficiency of remote desktop configuration are achieved.

The invention relates to an electric vehicle charging scheduling system and method based on deep reinforcement learning and renewable energy sources, and belongs to the field of intelligent traffic systems. The system comprises an electric vehicle, a charging station, a roadside unit, an aggregator and a central controller, and the method comprises the following steps: firstly, predicting the production condition of renewable energy through a GAN, and constructing a short-term charging demand and energy supply prediction model in combination with the real-time state information of the charging station; and then, the forming process of the charging scheduling strategy is expressed as a Markov decision process (MDP), and path planning and charging mode optimization of multiple agents are carried out through an MADDPG algorithm. According to the invention, matching, path selection and charging mode selection of the charging station and the electric vehicle are optimized in real time. According to the method, multiple factors such as vehicle charging requirements, traffic conditions, renewable energy source supply and user green preference are comprehensively considered, the charging time and the waiting time of the electric vehicle are effectively shortened, and the overall utility of the charging scheduling system is remarkably improved.

The invention relates to a resource allocation method for intensive tasks, and belongs to the technical field of wireless communication. Intensive tasks of ground terminal equipment are collected to a high-altitude platform to be processed or transmitted to a low-orbit satellite, and a system total energy consumption weighting and mathematical model is constructed; the method comprises the following steps: determining optimal path selection by constructing a bidirectional preference relationship between ground terminal equipment and a high-altitude platform and association between the high-altitude platform and a low-orbit satellite, and decoupling a system total energy consumption weighting and mathematical model into three sub-problems of a high-altitude platform selection association problem, task unloading and resource allocation for solving; joint iterative optimization is carried out on results of high-altitude platform selection association, task unloading decision and resource allocation, and a system total energy consumption weighted sum obtained by each iteration is used as a fitness evaluation standard until convergence is carried out to an optimal system total energy consumption weighted sum; the invention provides a hybrid unloading strategy with high energy efficiency, which not only can efficiently and flexibly allocate resources, but also can effectively improve the energy efficiency of the system.

The invention discloses a financial comprehensive auditing method and system, and the method comprises the steps: deploying a financial smart contract on a block chain, carrying out the modular design of the smart contract, and enabling each sub-module to correspond to a department; configuring an auditing rule template for each sub-module, and customizing auditing rules by corresponding departments based on the auditing rule templates; analyzing auditing rules of all departments based on a decision tree algorithm, and generating auditing check points; based on an intelligent contract event monitoring mechanism, auditing check points of all departments are monitored in real time, when a detection event does not accord with a corresponding auditing rule, early warning is carried out, and an auditing report is generated; the audit report is analyzed regularly, audit rules of all departments are updated based on analysis results, and the audit rules are encrypted and deployed on the block chain; and comprehensively auditing the finance of each department based on the updated auditing rule. According to the invention, high efficiency, flexibility and safety of financial auditing are realized through technical means of modular design, version control, real-time monitoring, data analysis and the like.

The invention discloses a cross-regional collaborative power spot market supply and demand balance adjusting system, and relates to the technical field of supply and demand balance management, and the system comprises a multi-stage cross-regional adjustment framework construction module; the future power supply and demand prediction module is used for performing future power supply and demand prediction and generating a supply and demand trend chart; the cross-regional supply and demand balance model construction module is used for constructing a cross-regional supply and demand balance model by integrating a multivariate optimization target and combining a multi-stage cross-regional market participation framework; the multi-stage collaborative optimization module is used for performing multi-stage collaborative optimization by taking the supply and demand trend chart as model input; and a cross-regional power distribution module. The technical problems that in the prior art, traditional power dispatching is mostly concentrated in a single region, cross-region cooperation capability is lacked, and flexible and efficient power balance adjustment is difficult to achieve are solved, and the technical effect of high efficiency and flexibility of cross-region power resource allocation is achieved through multi-stage cooperation optimization and a new energy priority mechanism.

The utility model provides a standardized cupboard system which comprises a standard kitchen cabinet, a standard cupboard module is arranged on one side of the standard kitchen cabinet, and a standard cupboard is arranged on the other side of the standard kitchen cabinet; a standard stove cabinet table is arranged on the top wall of the standard stove cabinet, a standard cabinet table is arranged on the top wall of each standard cabinet module, and a standard cabinet table is arranged on the top wall of each standard cabinet; vertical standard supporting tables are arranged above the two side edges of the standard kitchen range table, and a U-shaped structure is formed among the standard cupboard table, the standard kitchen range table, the vertical standard supporting tables and the standard cupboard table. The top wall of the standard cupboard module is flush with the top wall of the standard cupboard and is higher than the top wall of the standard kitchen cabinet, so that the height of operations such as vegetable cutting can be met, and the height of cooking operation can also be met; all parts can be standardized, the adaptive standard cabinet module is selected according to the actual size of a kitchen, the cabinet body installation efficiency is improved, the waiting period of a user is shortened, and the problem that an existing cabinet is long in customization period is solved.

The invention provides a multi-granularity GPU-CPU cooperative scheduling method for a heterogeneous computing system, and belongs to the technical field of heterogeneous computing resource management. The method comprises the following steps: S1, constructing a distributed cluster; s2, setting sampling parameters, and obtaining a working load time sequence of the distributed cluster; s3, resampling and aggregating the workload time sequence based on the sampling parameters to obtain a plurality of training subsets; s4, generating a workload prediction model by using each training subset and a time attention selection algorithm; s5, expanding the set evaluation function through the workload prediction model; s6, generating a scheduling rule based on the dynamic threshold parameter and the evaluation function; and S7, generating a scheduling decision based on the scheduling task and the scheduling rule, and performing scheduling based on the scheduling decision. The heterogeneous computing resource scheduling method has the advantages that the adaptive capacity, the resource utilization rate, the timeliness and the reliability of heterogeneous computing resource scheduling are greatly improved, and the energy consumption cost is greatly reduced.