9 results about "Scale selection" patented technology

The invention discloses a dense small target detection method for an unmanned aerial vehicle aerial photography scene, and belongs to the technical field of computer vision and target detection. In order to solve the problems of small target scale dynamic change and feature expression weakening caused by flight height change, imaging resolution difference and scene complexity in aerial photography of an unmanned aerial vehicle, the invention provides a detection framework fusing an attention scale selection (AGSS) module and a dynamic local self-attention (DPSA) module. The method specifically comprises the following improvements: (1) an AGSS module enhances the significance and discrimination ability of small targets in multi-scale features through global context modeling and a dynamic weight distribution mechanism; and (2) a DPSA module introduces a sparse selection mechanism in a channel dimension, and focuses computing resources on a channel sensitive to a small target, so that efficient and lightweight attention modeling is realized. The above modules cooperate with each other, so that high reasoning efficiency is maintained, and small target detection precision and robustness in a complex background, low illumination and dense target scene are significantly improved. Experimental results show that on typical unmanned aerial vehicle aerial photography data sets such as VisDrone-DET2019 and the like, the method is superior to an existing mainstream method in multiple indexes such as the average precision (mAP), the accuracy rate and the recall rate, especially has obvious advantages in the aspects of integrity and stability of small target detection, and has good practical application value and popularization prospects.

The invention discloses a YOLO11s trunk small target feature extraction method based on attention guidance scale selection and gating detail injection, and the method comprises the steps: introducing an attention guidance scale injection ASI module into a high-resolution feature layer of which the output step length of a YOLO11s trunk is 8, and extracting detail features through a multi-scale depth divisible convolution branch by the ASI module, original features and multi-scale abstracts are fused to construct a channel-space attention map, on one hand, multi-scale branch weights are generated in a self-adaptive mode through a unified attention map, on the other hand, pixel-by-pixel space gating is conducted on weighted detail streams, the gated detail streams are injected into trunk features in a residual mode through 1 * 1 convolution, and a multi-scale branch weight is generated; only multi-scale details associated with small targets are enhanced. According to the method, the small target feature expression capability of a high-resolution layer is remarkably improved on the premise of small increase of parameter quantity and calculation quantity, false activation of complex background textures is inhibited, and more discriminative small target features are provided for subsequent detection or segmentation tasks.

The present application relates to a kind of scale switching and volume adjustment method, device and equipment, method includes: obtaining air pressure difference value;According to the preset scale selection strategy, the scale pronunciation corresponding to the air pressure difference value is selected;According to the preset volume acquisition strategy, the volume value corresponding to the air pressure difference value is obtained;The volume of the note of the scale is adjusted based on the volume value.The present application can automatically switch scale according to air pressure difference value, and adjust volume, it is convenient and easy.

The invention discloses a rainfall numerical prediction correction method based on adaptive spatial-temporal scale selection, and the method comprises the steps: investigating a target region and a surrounding region around the target region at a spatial scale, and carrying out the design of first real-time data and second real-time data for the input of an NWP model at a time scale, a first future time period and a second future time period are designed for the output of the NWP model. In the execution process of the technical scheme, a computer can execute a computer program based on the steps of the method, namely, automatic rainfall numerical forecasting correction based on self-adaptive spatial-temporal scale selection can be achieved, data can be used in the aspect of data selection no matter in the time scale or the spatial scale, and the accuracy of rainfall numerical forecasting correction is improved. Errors and distortion caused by excessive intervention of human factors due to human personal preferences are avoided. And the application of an electric digital data processing technology in the related technical field of rainfall numerical value generation can be realized.

The invention discloses an arrhythmia detection model modeling method based on multi-scale spatial-temporal feature fusion, and the method comprises the steps: designing a multi-scale selection fusion embedding module, and achieving the efficient fusion of multi-scale time features in an ECG signal; an arrhythmia detection model is designed, a space-time agency attention module is introduced into one branch, space-time features of ECG signals in input leads are extracted, fused and enhanced, and in the other branch, the cosine similarity of approximation coefficients between all lead pairs is calculated and a threshold value is applied, so that the time-space characteristics of the ECG signals in the leads are extracted, fused and enhanced. Constructing a graph structure of the correlation between the leads to quantify a feature dependency relationship between the leads, and performing operation depth extraction on spatial topological features between the leads through a multi-layer graph convolutional network module; in-lead multi-scale spatial-temporal features and inter-lead spatial features obtained by two branches of the arrhythmia detection model are fused cooperatively, so that the network can understand ECG signals more comprehensively, and obtained mixed features are input into a full-connection layer for prediction and classification.

The application provides a preferred scale selection method for ecosystem service research, and relates to the technical field of ecosystem service evaluation. The method comprises the following steps: calculating the values of each ES index of each analysis unit under each scale according to the multi-scale division results and geographic data of a target region; constructing a spatial weight matrix according to the adjacent relationship of the analysis units under the same scale, and determining the Moran I index value of each ES according to the spatial weight matrix and the values of all ES indexes under the same scale, and then constructing a target decision matrix; determining the information entropy of each ES index according to the target decision matrix, and then determining the entropy weight weight of each ES index; constructing a weighted normalized decision matrix according to the target decision matrix and the determined entropy weight weight, and then determining the comprehensive distance between each scale and the ideal solution; determining the relative closeness of each scale to the ideal solution according to the comprehensive distance of each scale, and then determining the global optimal scale. The application aims to improve the selection accuracy of the universal scale.

This application discloses a precipitation numerical forecast correction method based on adaptive spatiotemporal scale selection. Spatially, it examines the target area and its surrounding areas. Temporally, it designs first and second real-time data for the input of the NWP model, and designs first and second future time periods for the output of the NWP model. During the execution of this application's technical solution, a computer can execute a computer program based on the method steps described in this application, thus achieving automated precipitation numerical forecast correction based on adaptive spatiotemporal scale selection. Data selection is based on sound criteria at both the temporal and spatial scales, avoiding errors and distortions caused by excessive human intervention due to personal preferences. It also enables the application of digital data processing technology in the field of precipitation numerical correlation technology.

The invention provides a laryngoscope vocal cord focus detection method based on multi-scale selection and attention fusion. An improved laryngoscope vocal cord focus detection network model based on YOLO11 is researched and designed. The method comprises the steps that in order to effectively improve the detection capability of a model on laryngoscope vocal cord lesions, an MSSFA (Multi-Scale Selective Fusion Attention) attention mechanism is researched and designed, and the core idea of the method is that different scale features are subjected to adaptive pooling and then fusion guidance to generate attention weights, so that the capture capability of the model on multi-scale information is enhanced. After the MSSFA is used for a C3K2 module of a backbone network, the model can capture detail information of different scales while establishing a long-distance dependency relationship. The laryngoscope vocal cord focus detection method based on multi-scale selection and attention fusion is high in accuracy, the detection effect and robustness of the model are effectively improved, and powerful support is provided for the medical field of laryngoscope examination.

The invention provides an optimal scale selection method for ecosystem service research, and relates to the technical field of ecosystem service assessment, and the method comprises the steps: calculating each ES index value of each analysis unit under each scale according to a multi-scale division result and geographic data of a target region; constructing a spatial weight matrix according to an adjacent relation of the analysis units under the same scale, determining a corresponding Moran I index value of each ES according to the spatial weight matrix and all ES index values under the scale, and constructing a target decision matrix according to the Moran I index value; determining the information entropy of each ES index according to the target decision matrix, and determining the respective entropy weight according to the information entropy; constructing a weighted normalized decision matrix according to the target decision matrix and the determined entropy weight, and determining a comprehensive distance between each scale and an ideal solution according to the weighted normalized decision matrix; and according to the comprehensive distance of each scale, determining the relative closeness between each scale and the ideal solution, and determining the global optimal scale according to the relative closeness. The objective of the invention is to improve the selection accuracy of a universal scale.