35 results about "Local scale" patented technology

The invention discloses a three-dimensional fracture network seepage and solute transport numerical simulation method based on topological pruning, which comprises the following steps of: analyzing a fracture network topological structure from global and local scales by applying a graph theory, establishing a quantitative relation between fracture connectivity and an underground flow field, and accurately describing an influence mechanism of topological characteristics on a seepage field; on the basis of network topology analysis, the key fracture sub-network is extracted through a pruning method, the key fracture sub-network is used for replacing a complete network for high-precision numerical simulation, and the calculation efficiency is remarkably improved while the calculation precision is guaranteed. The invention provides an efficient and reliable solution for the problems of seepage and solute transport in a hard rock fractured medium.

The invention provides an optical and acoustic image registration method for underwater structure crack detection, and the method comprises the steps: collecting a multi-mode image through the vision field and time synchronization, building a one-to-one correspondence relation according to the pose and time metadata, and improving the image quality through the denoising, distortion correction and other means; optical and acoustic multi-scale features are respectively extracted through a dual-channel network, and cross-modal semantic alignment is realized in combination with a shared weight and a channel attention mechanism; generating a high-resolution scale offset field by using local cross-correlation and micro-upsampling, and introducing manifold regularization constraint to ensure that a vector field is smooth and continuous; according to the method, the local scale field and the affine parameters are combined, non-rigid space mapping is achieved through thin-plate spline interpolation, feedback iterative optimization based on edge structure consistency is assisted, the registration precision of a key structure is enhanced, the automatic registration effect of the underwater multi-modal image is improved, and the method has high robustness and practical application value.

The invention provides a method for identifying forest tree species by using laser point cloud data, and the method comprises the following steps: collecting three-dimensional laser point cloud data of a forest region, setting an elevation threshold, filtering ground points, and extracting a point cloud sample object; respectively extracting VFH, CVFH and ESF feature descriptors from the sample point cloud, and constructing three types of geometric feature vectors; performing supervised classification on the features by adopting a random forest and a support vector machine learning classifier; the output of each classifier is fused through strategies such as weighted voting, an average method or a stacking method, and a final tree species identification result is obtained; according to the method, three types of global or semi-global feature descriptors of VFH, CVFH and ESF are extracted for a point cloud sample object of a single tree, feature modeling is carried out on tree species from three dimensions of spatial attitude, local scale structure and global shape distribution, the advantage of real restoration of a target structure by using point cloud data is utilized, and the feature modeling efficiency is improved. And the problem of projection distortion of image features under multiple view angles is avoided.

The invention discloses an ecological security pattern key area identification method and device, a medium and a product, and relates to the field of ecological security pattern construction.The method comprises the steps that according to geospatial data, a morphological spatial pattern identification technology, an area threshold value and connectivity analysis are adopted, and an ecological source land is determined; generating a comprehensive ecological resistance surface of the ecological source land; according to the ecological source land and the comprehensive ecological resistance surface, determining an ecological corridor by adopting a minimum cost path model; constructing an ecological security pattern network diagram based on the ecological source and the ecological corridor; according to the ecological security pattern network diagram, carrying out ecological security pattern network topology feature analysis; determining a comprehensive ecological pressure index according to the ecological pressure index factor; and determining an ecological security pattern partition according to the network topology structure and the comprehensive ecological pressure index. According to the method, the interaction and relation of the ecological security pattern in the local scale can be fully considered.

The invention discloses a complex flow field prediction-oriented dual-scale attention neural network and method, and relates to the technical field of dual-scale attention neural networks. In order to solve the defect that high-precision prediction of a local area is difficult to consider while global prediction is ensured in the prior art, the technical scheme provided by the invention comprises the following steps: carrying out flow field analogue simulation; collecting and processing the time series data; respectively inputting the sequence and the subsequences into an upper branch and a lower branch of a double-scale attention neural network, calculating attention information in the sequence by the upper branch, calculating attention information among the subsequences by the lower branch, performing attention fusion at the tail end of the network, and outputting flow field features considering global and local scales; and inputting the fused flow field features into a full connection layer, predicting the speed and pressure in the target area, and performing optimization training by adopting a weighted loss function based on a sequence to complete neural network training and verification. The method is suitable for rapid and high-precision prediction work of speed and pressure in a complex flow field.

The invention relates to the technical field of machine vision, and particularly discloses a length measurement method and system based on machine vision, and the method comprises the steps: collecting the geometric data of a sequence image recognition clamp to form a geometric constraint set, and extracting the candidate edge of the end part of a pole piece; executing constraint deformation regression based on the geometric constraint set to obtain time-varying distortion mapping and generate a geometric correction image; triggering a focal plane perturbation to collect an auxiliary image, calculating a parallax field, solving a local scale factor, and generating a mapping relation from pixels to actual lengths; executing sub-pixel edge fitting in the geometric correction image to obtain pole piece endpoint coordinates; calculating the extension amount of the measured pole piece; according to the method, time-varying distortion is eliminated through online correction of geometric drive of the clamp, depth-related scale change is compensated by using active focal plane perturbation, and the precision problem of length measurement in a cross-medium environment is effectively solved by combining sub-pixel positioning and time sequence processing, and extra calibration equipment is not needed.

The embodiment of the invention provides a vectorized map coding method and system, electronic equipment and a medium, in the method, a map is coded into a structured form containing a topological relation between a lane section and a target node, and geometric information and semantic information are endowed, so that a basis is established for multi-scale feature fusion. Then, lane line global features are generated through the first graph structure data. Construction of node features focuses on detail mining of local scales, and it is ensured that microscopic geometric and semantic features of lane lines are not missed; and the global topological relation of the vectorized map can be accurately captured through the construction of the global features of the lane lines. And finally, carrying out fusion processing on the node features of the local scale and the global features of the lane line of the global scale, and through cross-scale feature association, enabling the features of each target node to not only reserve geometric details of the lane section where the target node is located, but also integrate global topology semantics of the whole vectorized map. And the accuracy of track prediction in the automatic driving system is obviously improved.

The present invention discloses a target recognition and tracking method based on the fusion of a ship-borne navigation radar and a dome camera, comprising the following steps: constructing and training a ship type recognition model and a ship license plate detection model; calculating the latitude and longitude coordinates of the target; calculating the angle required for the dome camera to turn and capture the target direction, and calculating the focal length required for the target to reach a set image percentage in the camera image; predicting the target longitude and latitude at the current time; converting the predicted target longitude and latitude coordinates to the position in the current camera PTZ angle image; fusing the target in the image with the target identified by the radar ID into the same target; calculating the local scale size, the ship's direction of travel, and the bow position, and calculating the corresponding camera PTZ value; performing ship type recognition and ship license plate detection, and identifying the text on the ship license plate. The present invention models the target's navigation trajectory using historical radar data, predicts the target's current position, and improves maritime target recognition and tracking methods.

The present invention relates to a method for classifying land features based on the fusion of satellite remote sensing multispectral and panchromatic images. The method specifically comprises the following steps: S1. Acquire multispectral and panchromatic images of a target area, extract EMAP spatial features from the multispectral image, and concatenate them with the multispectral image to obtain shallow spatial spectral features; S2. Use the shallow spatial spectral features as input to a deep network and the panchromatic image as the reconstructed output, and obtain multi-resolution hidden layer features through network training; S3. Extract low-resolution hidden layer features, medium-resolution hidden layer features, and high-resolution hidden layer features, respectively, and perform feature fusion and classification training through a convolutional neural network to obtain land feature classification results within the target area. Compared with the prior art, the present invention combines shallow features at the global scale of the image with deep features at the local scale, fully exploiting the multi-resolution hidden layer feature expression between MS and PAN images, and has the advantages of high classification accuracy and high robustness.

The application discloses an ecological security pattern key area identification method, equipment, medium and product, relates to the field of ecological security pattern construction, and comprises the following steps: determining an ecological source according to geographical space data, adopting a morphological space pattern identification technology, an area threshold and connectivity analysis; generating a comprehensive ecological resistance surface of the ecological source; determining an ecological corridor according to the ecological source and the comprehensive ecological resistance surface, adopting a minimum cost path model; constructing an ecological security pattern network graph based on the ecological source and the ecological corridor; performing ecological security pattern network topology feature analysis according to the ecological security pattern network graph; determining a comprehensive ecological pressure index according to an ecological pressure index factor; and determining ecological security pattern partitioning according to the network topology structure and the comprehensive ecological pressure index. The application can fully consider the interaction and connection of the ecological security pattern at a local scale.

The application discloses a ship type optimization method based on transfer learning, and relates to the technical field of ship type optimization.The application improves the precision and generalization performance of the final ship performance model by using a Two Stage TrAdaBoost.R2 regression transfer algorithm.The algorithm uses the knowledge of some known related data sets to assist the target ship type data modeling, so as to improve the precision of the model.The application is divided into two stages for ship type sampling and modeling: a coarse proxy model is constructed based on global scale sampling data, and a fine proxy model is constructed based on the data obtained through the global and local scale sampling stages, and the two models jointly assist the subsequent optimization process.A double proxy assisted optimization method is proposed for the optimization process.The iteration process of the adopted optimization algorithm is improved, and the single proxy model is no longer updated independently, but the fitness values of the two proxy models are considered simultaneously in the optimization process, and the optimal one is selected for iteration.

The application discloses a guiding auxiliary device and a control system thereof, and belongs to the technical field of feature registration. A target object surface model is constructed through multi-model access and three-dimensional point cloud reconstruction. The registration problem caused by the difference between global and local scales is solved by using curvature feature matching and potential area screening. A closed operation area is formed through edge point mapping, and precise spatial alignment of the design model and the actual scene is realized by combining centroid deviation adjustment. Key points of equipment positioning are planned based on a rigid transformation matrix, ensuring geometric alignment of the operation end and the functional feature points, and a collision-free path is generated through safety verification. With the aid of light source visualization guidance and real-time deviation monitoring, dynamic feedback of the operation process is realized. After operation, effect evaluation is performed through comparison of the geometric features of the functional points, and global reset is avoided through local correction. The device operation precision and safety in the precise medical implantation scene are effectively improved.

The invention discloses a typesetting method and typesetting equipment capable of printing while shooting, a medium and a program product, and relates to the technical field of image processing and intelligent printing. On the basis of converting an original image into structured data and performing global mapping, the typesetting device divides consumables into a main visual area, an auxiliary visual area and an avoidance area, constructs a semantic adhesion cluster in combination with spatial adjacency between objects, and can detect whether an image-text module with close logic association can be split by a fold line or shielded by a hole. If conflicts occur, the limitation of global scaling is broken, and local scaling and offset coordinates are independently calculated for the adhesion clusters for adjustment. According to the derivation process of multi-dimensional feature interaction, the problems of visual blind areas and semantic segmentation of traditional two-dimensional typesetting in a complex three-dimensional pasting scene are effectively avoided, and the integrity and transmission accuracy of graphic and text information of special labels (such as flag-shaped labels and winding labels) in actual physical application are improved.

Embodiments of the present application provide a vector map encoding method and system, electronic equipment and medium. In the method, the map is encoded into a structured form containing the topological relationship of lane segments and target nodes, and is given geometric information and semantic information, thereby establishing a basis for multi-scale feature fusion. Subsequently, the first graph structure data is used to generate lane line global features. The construction of node features focuses on local scale detail mining, ensuring that the microscopic geometry and semantic features of the lane line are not missed. The construction of lane line global features can accurately capture the global topological relationship of the vector map. Finally, the node features of the local scale and the lane line global features of the global scale are fused and processed, and through cross-scale feature association, the features of each target node not only retain the geometric details of the lane segment, but also integrate the global topological semantics of the entire vector map, significantly improving the accuracy of trajectory prediction in the autonomous driving system.

The application discloses a three-dimensional multi-plane scale estimation method based on three vanishing points and local scales, which comprises the following steps: selecting three groups of vanishing lines and marking known local scale line segments from a single monocular image, and calculating the coordinates of three vanishing points; performing monocular camera calibration of multi-planes according to the obtained coordinates of the three vanishing points and the known local scale line segment information; selecting a to-be-measured line segment from the multi-planes spanned by the three groups of vanishing lines on the monocular image. Depth estimation is performed by using the camera coordinates of the end points of the to-be-measured line segment and the coordinates of the shared vanishing point. Finally, multi-plane scale estimation of a three-dimensional space object is performed according to the camera internal and external parameters obtained by camera calibration and the depth estimation value. The application can perform multi-plane scale estimation of a three-dimensional space object by using a single monocular image, expands the scale estimation range, and improves the scale estimation accuracy.

The invention provides an environment variable speculation method based on multi-scale spatial-temporal feature fusion, a medium, equipment and a product, and relates to the technical field of environment variable speculation, and the method comprises the steps: carrying out the meshing of environment variable data, carrying out the feature extraction of static and dynamic data of environment auxiliary variable data of a single grid, a local grid and an overall grid, and carrying out the fusion, obtaining time sequence feature representations of an individual scale, a local scale and a global scale; obtaining a multi-scale fusion feature from the three-scale time sequence feature representation through a multi-head attention mechanism; calculating the Gaussian distance weight between each target grid and the grid where each monitoring station is located, multiplying the Gaussian distance weight by the environment target variable data of the corresponding monitoring station, and normalizing to obtain air quality features; and based on the air quality features and the multi-scale fusion features, a spatial distribution diagram of the environment target variables is obtained through a gating unit. According to the method, the problem that air quality feature extraction in environment target variable speculation is not comprehensive is solved.

The application provides a packaging box anomaly detection method and device based on semantic constraints and geometric priors, and relates to the technical field of defect detection. The method extracts local geometric description information from three-dimensional point cloud data, inputs the information into a geometric perception measurement module to generate a geometric reliability coefficient, and uses the coefficient to adaptively modulate a local scale parameter predicted based on a direction vector to construct a multi-modal anomaly measurement constrained by a geometric prior. The method inputs a sample to be detected into a trained anomaly detection model, calculates local anomaly measurement values of each region through the multi-modal anomaly measurement, generates an anomaly score map reflecting the degree of anomaly, and performs threshold segmentation and connected component analysis to realize detection and positioning of packaging box anomalies. The method can realize high-precision packaging box anomaly detection without relying on a large amount of manually labeled data, and significantly improves the reliability and efficiency of packaging box quality detection.

The invention provides a packaging box anomaly detection method and device based on semantic constraint and geometric prior, and relates to the technical field of defect detection.The method comprises the steps that local geometric description information is extracted from three-dimensional point cloud data and input into a geometric perception measurement module to generate a geometric reliability coefficient; performing adaptive modulation on a local scale parameter based on direction vector prediction by using the coefficient, and constructing a multi-modal anomaly metric constrained by geometric prior; and inputting a to-be-detected sample into the anomaly detection model obtained by training, calculating a local anomaly metric value of each region through the multi-modal anomaly metric, generating an anomaly score graph reflecting an anomaly degree, and performing threshold segmentation and connected domain analysis to realize detection and positioning of the anomaly of the packaging box. According to the method, high-precision packaging box anomaly detection can be realized under the condition of not depending on a large amount of manual annotation data, and the reliability and efficiency of packaging box quality detection are remarkably improved.

The invention provides a building group safety risk multi-scale analysis method fusing an InSAR technology and city big data. The method comprises the following steps: performing city-level wide-area scale deformation monitoring on a city-level wide-area scale level by using a PSI technology, verifying a deformation result based on GNSS monitoring data, identifying a risk area according to an abnormal deformation trend, then obtaining a continuous deformation field in the area, removing PS points with obvious temperature deformation, and finally obtaining a deformation result; according to the method, the interference of building temperature deformation on a deformation field in an area is reduced, vertical deformation and east-west deformation of a PS point are decomposed by utilizing a geometrical relationship, and finally, on a building-level local scale level, the building risk is comprehensively evaluated from the perspective of structural damage risk and rigid body deformation analysis according to four types of longitudinal settlement modes of a building. The method provides an accurate and comprehensive urban building group safety risk multi-scale analysis means, and has important application value for safety monitoring and risk early warning of urban building groups.

The invention discloses a guiding auxiliary device and a control system thereof, and belongs to the technical field of feature registration, and the method comprises the steps: building a target object surface model through multi-model access and three-dimensional point cloud reconstruction; curvature feature matching and potential region screening are utilized to solve the registration problem caused by global and local scale differences; a closed operation area is formed through edge point mapping, and precise space alignment of a design model and an actual scene is achieved by combining centroid deviation adjustment; planning equipment positioning key points based on a rigid transformation matrix, ensuring that an operation end is geometrically aligned with the functional feature points, and generating a collision-free path through safety verification; dynamic feedback of the operation process is achieved by means of light source visual guidance and real-time deviation monitoring, effect evaluation is conducted through functional point geometric feature comparison after operation, and global reset is avoided through local correction; and the equipment operation precision and safety in a precise medical implantation scene are effectively improved.

The invention relates to an acoustic shadow image detection system and method based on multi-dimensional feature fusion and structure guidance. The system comprises a first acquisition module for acquiring ultrasonic radio frequency data and a B-mode image obtained by performing logarithmic compression on the ultrasonic radio frequency data; the second acquisition module is used for acquiring a local scale parameter diagram and a local shape parameter diagram based on the ultrasonic radio frequency data, the local scale parameter diagram comprises local scale parameters corresponding to the positions respectively, and the local shape parameter diagram comprises local shape parameters corresponding to the positions respectively; the sound shadow probability graph acquisition module is used for acquiring a sound shadow probability graph based on the local scale parameter graph and the local shape parameter graph; and the correction module takes the B-mode image as a guide image, and performs sound shadow edge correction on the sound shadow probability graph to obtain a sound shadow detection image. According to the method, a low-echo pathological structure and a normal liquid dark region can be effectively distinguished, and high-precision boundary positioning of an acoustic shadow region can be realized based on refined analysis.

A multi-stage label propagation method combined with a discriminator model comprises the following steps: step 1, constructing an adaptive weight map based on a dynamic local scale so as to accurately model a similarity relationship between samples; 2, designing a staged label propagation strategy, propagating high-confidence labels in the initial stage, and optimizing propagation of low-confidence labels in the later stage; step 3, introducing an ELCTRA discriminator model, and optimizing a pseudo tag generation process by comparing loss; 4, a convergence check mechanism is introduced in the propagation process, and the stability of the algorithm is ensured by monitoring label vector changes; and step 5, generating final high-quality label prediction by integrating multi-stage optimization results. And the tag prediction accuracy and efficiency in semi-supervised learning are improved. An adaptive weight map is constructed by dynamically calculating local scales of sample points so as to accurately capture a relationship between samples, and labels are propagated in stages. High-confidence labels are rapidly propagated in the initial stage, low-confidence prediction is refined in the later stage, and the error propagation risk is remarkably reduced. A convergence check mechanism is introduced, the iteration stability is judged by monitoring label vector changes, and overfitting is avoided. The pre-trained ELCTRA discriminator model and the comparison loss are combined to optimize the pseudo tag generation process, so that the pseudo tag is more reliable, and the feature distinguishing capability of the model is improved. And the tag prediction accuracy and efficiency in semi-supervised learning are improved. The method is suitable for label generation and propagation of a large-scale data set, and has the characteristics of high efficiency and strong robustness.

The application provides a satellite remote sensing image block classification water body identification method based on a supervised learning algorithm, which takes a Sentinel-1 SAR image as basic data, and realizes water body identification based on a satellite remote sensing image through the steps of image preprocessing, feature data preparation, sample point preparation, classification model construction and identification, and image post-processing. The application takes the VV, VH, angle, VV / VH and sum bands of the SAR image as the training feature input of the classification model, which is conducive to improving the accuracy of the classification result; the hexagon is taken as the minimum block unit to block and calculate the large-scale image, which can eliminate the edge effect in the block calculation, more accurately capture the features of the ground objects on the local scale, and reduce the calculation amount of a single local classification model; through the image post-processing step, the non-water body noise pixels in the classified image can be eliminated, and the influence of noise such as vegetation and buildings can be weakened. The method has the characteristics of high classification and identification precision, and is especially suitable for the identification of large-scale water bodies.

The invention provides a modeling method and system for analyzing a statistical distribution rule of RSRP in a local space, and the method comprises the steps: respectively calculating approximate expressions of a direct path distance and a reflection path distance based on a two-dimensional plane geometrical relationship formed by a base station, a user and a single dominant reflection surface under the constraint that the user makes small displacement near an initial position; calculating a superposition result of the received signals, and deducing an approximate analytical expression of user side reference signal received power (RSRP) about a user displacement variable; deriving a probability density function PDF and / or a cumulative distribution function CDF of the RSRP in the local space on the basis of the approximate analytical expression under the condition that the user displacement is assumed to be uniformly distributed in the local area; and evaluating the coverage quality according to the spatial statistical distribution rule of the RSRP. According to the method, a local scale model which keeps physical interpretability and can export an approximate analytic solution is constructed, so that a quantitative basis is provided for engineering evaluation, measurement sampling interval selection and rapid simulation.

The invention discloses a large eddy simulation method based on diurnal variation characteristics of a local scale three-dimensional space wind-heat-humidity environment. The large eddy simulation method is specifically implemented according to the following steps: step 1, constructing a multi-field coupling model of the local scale wind-heat-humidity environment; step 2, acquiring parameters of the parameterized scheme; 3, setting a research area; 4, setting initial and boundary conditions; and 5, carrying out numerical simulation analysis on the daily change of the local scale wind-humidity-heat environment. According to the method, the dynamic-thermal effect of the urban vegetation layer is comprehensively considered through the integration method, a new large-eddy simulation model is constructed, and accurate numerical simulation under the local scale is achieved.

The application discloses a biological data visualization and batch correction system based on deep manifold learning, which embeds a given data set into a two-dimensional or three-dimensional visualization space based on Euclidean manifold or hyperbolic manifold under the premise of considering specific task types, corresponding to biological data of the type "time fixed" or "time evolution" respectively. Specifically, the system learns the relationship between data points based on the local scale shrinkage technique through end-to-end training, converts the data to the visualization space while preserving the data geometric structure, and corrects the batch effect. The application has better effects in discovering complex biological data relationships, revealing time trajectories and processing complex batch factors.

The application discloses a caterpillar beam weld seam track fitting method based on an improved RANSAC algorithm and particularly relates to the technical field of weld seam track identification and automatic welding path planning. Standardized point cloud data is acquired. Welding spatter points are removed by combining the normal change rate of points with the extraction of weld protruding feature points through local height gradient analysis, a multi-scale point set structure of an overall scale point set and a local scale point set is constructed, a random consistency sampling is performed on the overall scale point set to fit a weld overall trend model, and a local track model set is generated on the local scale point set with the inner point set as a constraint. A track connection graph structure is constructed according to the local curvature change value and the direction continuity score, the screening and splicing of the local track model are completed, and a weld compensation track model is obtained. The compensation track model is subjected to continuous curve fitting to generate a weld center track curve. The application can realize high-precision weld seam track fitting in the presence of welding spatter and local deformation interference.

The application discloses a waveform-protected non-source wavelet local scale traveltime inversion method, relates to a non-source wavelet full waveform inversion method established by using a convolution and deconvolution method, and is combined with local scale traveltime inversion, and simultaneously solves the period jump of full waveform inversion and the source wavelet dependence problem. First, the influence of the source wavelet in the observation data and the simulation data is eliminated by using the convolution and deconvolution method, and the original seismic data waveform characteristics are protected. Secondly, local scale traveltime information of the observation data and the simulation data is extracted, and a waveform-protected non-source wavelet local scale traveltime inversion objective function is constructed. Then, the partial derivative of the waveform-protected non-source wavelet local scale traveltime inversion objective function to the velocity parameter is deduced in detail, and the corresponding underground velocity parameter gradient operator of the adjoint state method is deduced. Finally, the stable update of the velocity model parameter is realized by using an L-BFGS optimization algorithm. Finally, the effectiveness of the method is verified by the seismic data test.

This invention claims protection for an improved scale fusion-based brain tumor segmentation method, aiming to improve scale feature fusion by aggregating scale information within and between encoder layers, thereby achieving accurate brain tumor image segmentation. This method belongs to the field of computer vision. The method includes the following steps: Step 1. This invention proposes a global-local feature mixer, which achieves global-local scale information mixing by capturing local detail features and global context information in parallel, to solve the problem of uncertain brain tumor location. Step 2. This invention designs a multi-scale expansion initialization module, which can effectively capture multi-scale shape features, aiming to improve the network's ability to adapt to different brain tumor shapes. Step 3. This invention designs a multi-scale feature aggregation module, which can effectively aggregate multi-scale features from different encoder layers, helping to alleviate the problem of uncertain brain tumor size.

The invention provides a local-scale atmospheric pollution tracing method based on big data, and belongs to the technical field of atmospheric pollution tracing. The method comprises the following steps: S1, acquiring environmental basic information; s2, grid division is carried out on an area to be subjected to atmospheric pollution traceability, the distance x from each grid point to a pollution source in the wind direction axis direction is calculated, and the distance y from each grid point to the pollution source in the direction perpendicular to the wind direction axis is calculated; s3, calculating source intensity, a transverse diffusion coefficient and a longitudinal diffusion coefficient based on the environment basic information, the distance x and the distance y; s4, calculating the pollution contribution concentration of each pollution source to the grid points based on the environmental basic information, the distance y, the source intensity, the transverse diffusion coefficient and the longitudinal diffusion coefficient through a Gaussian diffusion model; and S5, carrying out local-scale atmospheric pollution tracing based on the pollution contribution concentration. The method has the advantages that the accuracy and comprehensiveness of local scale atmospheric pollution tracing are greatly improved.