13results about How to "Meet deployment needs" patented technology

The invention, which relates to the field of FPGA and machine learning, discloses a high-flow-degree MambaV2 hardware acceleration system with an optimized calculation sequence based on FPGA, comprising a preceding stage module, a tensor expansion module, a hidden layer state calculation module, an output state calculation module, a residual multiplication normalization module and an output linear layer module. The calculation process of the MambaV2 algorithm is divided into a plurality of pipeline levels, and a data stream is output in a pipeline form. According to the method, an original algorithm structure is not changed, the operation sequence of a hidden layer state equation in an original algorithm is changed, an algorithm for optimizing the calculation sequence is provided, the number of times of operation is reduced, cache consumption is reduced, the MambaV2 model is deployed and reasoned on an FPGA, the edge scene deployment requirement is met, customized acceleration of the FPGA is utilized, and data streams are adjusted, so that the edge scene deployment efficiency is improved. The pipeline design of the whole algorithm is realized, the utilization rate of computing resources is improved, the problem of data blocking is solved, and computing delay and cache consumption are reduced.

The invention relates to a high-confidence intelligent detection method and device for a cluster-dense small target, and a medium. The method comprises the steps: constructing an HLS-YOLO-P network model which comprises an HWD down-sampling module based on wavelet transformation, a global feature extraction module C2f-LSKA and a task alignment detection head STADH; and performing multi-layer feature fusion and context information enhancement processing on the image data containing the target scene to obtain a detection result containing a target category, a position and confidence. And carrying out post-processing on the detection result, wherein the post-processing comprises intersection-to-union ratio and confidence threshold adjustment. And finally, generating a corresponding inspection report. Compared with the prior art, the method has the advantages of high detection precision, light weight, wide application range and the like.

The invention provides a low-resource-demand mobile phone number big data general database construction method which comprises the following steps: acquiring a data type strongly associated with a mobile phone number, configuring a multi-dimensional general interface based on the data type, and automatically loading and generating a general database based on a configuration result; splitting the mobile phone number based on a preset number splitting rule, and taking a number segment obtained by splitting as an addressing address; performing address marking on the general database based on the addressing address, and performing multi-dimensional management configuration on the general database after address marking based on the multi-dimensional general interface; and converting the universal database into a deployment file based on the multi-dimensional management configuration result, and performing multi-category terminal difference deployment on the deployment file based on the deployment requirement. The deployment requirements of different environments are met, meanwhile, the resource demand quantity is reduced, and the storage efficiency and the management effect of the mobile phone number related data are improved.

The embodiment of the present application provides a method and device for predicting the charging demand of an electric vehicle, which is applied to an electric vehicle charging demand prediction server. The method includes: constructing a test data set; training a load prediction model through the test data set; inputting the test data set into the trained load prediction model to obtain a prediction result. The deep neural network electric vehicle charging demand prediction method based on improved probabilistic sparse self-attention and contrastive learning adopted in the embodiment of the present application can overcome the problem of large long-term demand prediction errors and the need for additional information correction, and can meet the deployment requirements of scenarios such as mobile terminal use, edge device operation, and local area networks in remote areas and network security restrictions. In particular, it can be deployed in the energy storage system of the charging site or even in the charging pile for independent operation, providing more accurate prediction results and more flexible deployment methods, so that existing charging sites have room for intelligent upgrades.

The present invention proposes a human resource data processing method and system based on a recursive blockchain. It belongs to the technical field of human resource information management; the human resource data processing method comprises: after receiving human resource information, the blockchain node obtains dynamic evaluation parameters through the first dynamic evaluation coefficient, the second dynamic evaluation coefficient and the third dynamic evaluation coefficient of the blockchain node itself; the target blockchain node corresponding to the blockchain node is determined by combining the dynamic evaluation parameters with the dynamic evaluation parameters of the candidate sub-nodes of the consensus chain of the blockchain node, and the human resource information is shared. By adopting the recursive blockchain method, data is shared and stored on multiple nodes, which enhances the security of the system; even if a node fails or is attacked, other nodes can still maintain the integrity and availability of the data to ensure that the data will not be lost or tampered with.

The invention discloses a karst cave stalactite damage degree intelligent assessment method based on image recognition, and relates to the cross technical field of computer vision and geological resource investigation. The method comprises the steps of collecting stalactite images in a karst cave, dividing the images into a labeled data set and a non-labeled data set, preprocessing the images in the labeled data set and the non-labeled data set, constructing a depth separable convolutional neural network model fusing region proposal and alignment technologies, and constructing a deep separable convolutional neural network model; a mixed training strategy combining supervision and semi-supervision is adopted, and training and parameter optimization are carried out on the network model by using the preprocessed image; and inputting a to-be-evaluated stalactite image into the trained network model for feature extraction, candidate region generation and feature alignment, and synchronously outputting a stalactite existence judgment result, a damaged region position and a damage degree classification result through a built-in three-detection-head structure of the model. Therefore, automatic and non-contact detection of stalactite damage assessment is realized.

This invention proposes a target detection method for low-light environments, comprising the following steps: acquiring road traffic image data under nighttime or low-light conditions; annotating the road traffic image data with targets; constructing an enhanced detection network LLE-YOLO based on an improved version of YOLOv11n, where the C3k2 module in the backbone network is replaced with a CSP-PMSFA module, and the Conv module is replaced with a ContextGuidedBlock module; the C3k2 module in the neck network is replaced with a CSP-PMSFA module, and an SPDConv module is introduced; a DetectAux module is introduced in the head network; the low-light traffic perception dataset is divided into a training set, and the enhanced detection network LLE-YOLO is trained; after the enhanced detection network LLE-YOLO is trained, the auxiliary detection head and its related branches are removed, and end-to-end inference is performed on targets in low-light environments, outputting target category and location information. This method achieves lossless downsampling, adaptive enhancement of multi-scale features, effective suppression of background noise, and guidance of hard sample learning, thus solving the problem of target detection in low-light environments.

The embodiment of the invention provides a network planning method, a display method, equipment, a storage medium and a program product, and the method comprises the steps: enabling a network planning server to respond to a network planning request, and determining at least one network type; determining a target network configuration file matched with any network type from the plurality of network configuration files; the target network configuration file comprises network equipment information and a connection rule; generating first network planning information corresponding to the network type according to the network equipment information and the connection rule; under the condition that the at least one network type comprises a plurality of network types, determining a grid-connected part corresponding to the plurality of network types, and determining second network planning information corresponding to the grid-connected part according to a connection rule; and generating target network planning information based on the first network planning information and the second network planning information corresponding to the plurality of network types. According to the technical scheme provided by the embodiment of the invention, the network planning efficiency and accuracy are improved.

The utility model discloses a modularization expandable data collection ware belongs to environmental data collection technical field, solved the traditional data collection ware field maintenance efficiency low, the problem of poor expansion flexibility and poor environmental adaptability, including collection case, pluggable main control board subassembly, pluggable main control board subassembly includes double backboard zoning architecture, electromagnetic shield frame, battery, and main control board and battery are detachably installed in double backboard zoning architecture, and electromagnetic shield frame sets up between double backboard zoning architecture and the shell of collection case, double backboard zoning architecture includes upper backplate and lower backplate, in the utility model, pluggable main control board subassembly includes double backboard zoning architecture, and upper backplate can provide high density's I / O interface, and lower backplate constructs high -speed computing bus, and can easily expand to 128 way and above acquisition channel, and realizes the collaborative work of each component through the standardization bayonet, satisfies the deployment demand of large -scale ocean monitoring array.

The invention discloses an automatic driving target detection method based on gradient re-parameterization and center point estimation, and the method comprises the steps: carrying out the preprocessing of a collected image in automatic driving; performing feature processing on the preprocessed image through a gradient re-parametric network to obtain a multi-scale feature map; performing feature fusion on the multi-scale feature map to obtain a thermodynamic diagram; and performing target detection on the thermodynamic diagram through a central point estimation detection algorithm to obtain an output image containing a bounding box of the obstacle, the category of the obstacle and the confidence coefficient. The detection method based on the gradient re-parametric network and the central point is simple and efficient, does not need post-processing, and can adaptively detect objects with different scales, aspect ratios and positions; the algorithm based on gradient re-parameterization is free of network structure conversion, has the advantage of being friendly in quantization, and meets the deployment requirements of automatic driving edge equipment; the method has double efficiencies of training and reasoning, and is beneficial to the fields of limited computing resources, urgent need of delivery and rapid model iteration.

The invention provides a unified application store implementation system, which relates to the technical field of application management and comprises a management end, an application store, a support component and a runtime environment, the management end is used for providing a one-stop management front end, and a developer uploads and manages different versions of applications at a time; the application store is used for providing an application installation medium which is not bound with the running environment; full life cycle management of application installation media is provided; the application is published, and one-key installation of the application is completed based on the application installation medium; reversely generating a standardized application installation medium based on the running application; the supporting assembly is used for storing an application installation medium; storing application release record data; and the runtime environment is used for providing various runtime environments for the application. The standardized application installation medium is used, application installation, deployment and management and control of different environments and different operating system architectures are met, a user can easily manage one application installation medium, and the maintenance cost is reduced.

The present application belongs to the technical field of quantum computing and artificial intelligence, and particularly relates to a quantum neural network architecture search optimization method and system and application thereof in agricultural image classification. The present application adopts a quantum coarse screening-classical fine tuning method, efficiently excludes architectures with poor performance or not meeting resource constraints in the search space through Grover quantum pre-screening, quickly narrows the search funnel to the potential high performance area, avoids blind search of the classical evolutionary algorithm in a huge space, and greatly shortens the overall search period. The present application can finely optimize multiple targets such as precision, delay and resource consumption, and the obtained model architecture can effectively balance the problem of "global exploration" and "local exploitation" in the search process, so that the model architecture has the characteristics of high precision and low delay; meanwhile, the preset value lightweight constraint ensures the lightweight and high performance of the architecture, has good compatibility, and can completely compatible with the existing agricultural IT infrastructure and edge devices.

The invention discloses a substation equipment infrared target detection method, device and system, and a storage medium. The method comprises the following steps: S1, acquiring an infrared image of power equipment; s2, training an improved YOLOv9 network according to the infrared image of the power equipment; and S3, inputting a to-be-processed infrared image of the power equipment into the trained YOLOv9 network for target detection. According to the technical scheme, the problems that a traditional detection network is low in speed and large in size, the speed precision cannot be balanced, feature fusion is insufficient and the like are solved, the real-time detection requirement is met, meanwhile, the excellent detection precision is achieved, and the detection performance is improved.