91 results about "Spatial density" patented technology

The invention provides a vehicle mutual-aid claim settlement method and system based on multi-modal data fusion. Wherein vision and depth sensing data and maintenance records in a vehicle accident are integrated, part types and collision mechanical parameters are extracted, and surface microscopic deformation and internal hidden cracks are captured through dual-band scanning; and carrying out time-space association on the two to construct a comprehensive damage map. And generating a standard damage sequence by using the generative adversarial network based on a normal damage rule of the component, comparing the standard damage sequence with accident damage data layer by layer, extracting an abnormal region deviating from an evolution rule, and calculating a fraud risk index according to the spatial density of the abnormal region. And when the index exceeds a dynamic threshold value, triggering multi-source data to cooperatively verify feature consistency, and outputting a claim settlement result with a fraud mark. According to the technical scheme provided by the invention, accurate identification of maintenance fraud and collaborative optimization of claim settlement risk control efficiency are realized.

The invention relates to the technical field of project management, in particular to a building engineering project construction quality management method and system, and the method comprises the following steps: obtaining a drawing node sequence index and comparing a process stage number, screening compliant nodes, judging the reasonability of a task configuration time sequence, analyzing the deviation of an included angle between a crack angle and a load direction, and recognizing structural abnormality. And aggregating the partition abnormal nodes to form a construction project quality risk distribution section mapping result. According to the method, the structure drawing node sequence is compared with the construction process, it is guaranteed that input data logic is unified, the task configuration reasonability is improved through input time and stage sequence cross check, a stress deviation judgment mechanism is established in the crack direction through vector included angle and load comparison, and structural logic verification of inspection data is achieved; abnormal nodes are identified according to double standards of time compliance and stress deviation, a spatial density model is constructed by combining partition aggregation with an area ratio, building quality risk distribution is accurately presented, and the problem focusing and processing priority judgment capability is enhanced.

The invention relates to the technical field of image analysis and detection, in particular to a glass bottle label defect detection method and system based on image processing, and the method comprises the steps: obtaining a high-resolution image through a fixed-angle light source and near-infrared irradiation, carrying out the noise suppression processing through gradient enhancement filtering, detecting the label edge, and recognizing the label region boundary. And obtaining a label area image. According to the invention, high-resolution image acquisition is combined with gradient enhanced filtering, so that the identification precision of the label area is improved. Local detection unit division and fiber arrangement direction analysis enable direction abnormal region positioning to be more accurate. Spatial density analysis of mutation point distribution effectively distinguishes the fiber fracture degree, and misjudgment is avoided. And multi-scale calculation is combined with consistency fluctuation rate and density change rate cross validation, so that the defect detection stability is enhanced. And through gradient analysis of the fluctuation accumulation amplitude and the direction concentration deviation value, the defect classification accuracy is improved, and the defects of scratches, tearing, falling, label dislocation and the like can be effectively distinguished.

The invention relates to the technical field of three-dimensional reconstruction, in particular to a lightweight three-dimensional reconstruction method and system based on a three-dimensional Gaussian model, and the method comprises the steps: constructing the three-dimensional Gaussian of a target scene based on a multi-view original image of the target scene through employing a 3DGS algorithm; identifying the minimum resolution of each Gaussian primitive in the three-dimensional Gaussian, setting a spherical region according to the minimum resolution, taking the intersection result of the spherical region and the Gaussian primitive as a region evaluation score, and trimming the redundancy region geometry in the three-dimensional Gaussian in combination with the opacity; and taking the maximum zoom scale of the Gaussian primitive as a spatial density proxy variable, and dynamically distributing spherical harmonic coefficient orders according to a local spatial density proxy variable so as to meet the rendering requirements of different three-dimensional Gaussian geometric regions. On the premise that the image synthesis quality is not lost, the model scale and the transmission overhead can be remarkably compressed, the resource constraint condition of the edge computing equipment can be effectively adapted, and the higher training speed and the better rendering effect are achieved.

The invention discloses a distributed photovoltaic power generation anomaly positioning optimization method, and particularly relates to the technical field of power generation anomaly positioning, and the method comprises the steps: carrying out the space-time alignment of the static information, environmental parameters and dynamic power generation data of multiple stations in a target region, carrying out the spatial density clustering based on the three-dimensional distance of the stations, and dividing spatial sub-clusters; recognizing a space-time hot spot region in combination with a decoupling model of a historical loss rate and irradiance; constructing a space loss gradient field for the hot spot region, and reversely tracing to a normal threshold boundary point to realize pollution propagation path reconstruction; according to the method, wavelet packet decomposition, variational mode decomposition and short-time Fourier transform are adopted for a non-hotspot area to extract multi-scale attenuation and noise features, artifacts are removed through neighborhood slope difference, macroscopic error shielding and microcosmic anomaly positioning of meteorological and component differences are achieved, and the checking accuracy and the operation and maintenance efficiency are improved.

The invention relates to the technical field of data mining, in particular to an elevator emergency rescue station site selection decision optimization method, system and device and a medium, and the method comprises the steps: obtaining elevator related data of a target area, and carrying out the preprocessing of the elevator related data; based on the preprocessed elevator related data, space density distribution of the elevators in the target area is extracted; constructing a quantitative evaluation model, and extracting a rescue difficulty quantitative result of the target area based on spatial density distribution; and constructing a site selection decision model based on stochastic programming, and solving an emergency rescue station site selection decision result. The invention aims to improve the rationality and rescue efficiency of site selection of the elevator emergency rescue station.

The invention discloses a business district drawing method and device based on geographic space aggregation, and the method comprises the steps: carrying out the spatial density clustering of all network points in a to-be-drawn space, and obtaining a plurality of network point clusters; correspondingly generating a plurality of convex hulls based on the plurality of dot clusters; and performing boundary optimization on the convex hull according to the collected river network data, and determining the optimized convex hull as the business circle in the space to be drawn. According to the method, the problems of fuzzy boundary, data lag, poor strategy adaptability and the like in the traditional technology can be effectively solved, accurate definition of the contour of the business district and standardized output of grade division are realized, and then the efficiency of drawing the business district is improved.

The invention discloses a power distribution equipment point cloud multi-target classification extraction method, system, equipment and medium, and belongs to the technical field of power distribution line three-dimensional visualization, and the method comprises the steps: obtaining three-dimensional laser point cloud data covering a power distribution line region, and carrying out the noise point elimination and spatial density resampling, and obtaining first point cloud data; extracting ground points in the first point cloud data, and constructing a digital elevation model; removing the ground points to obtain second point cloud data containing the power distribution equipment and the environment target; executing clustering processing, and extracting structural feature parameters of each cluster; inputting the extracted structural feature parameters into a pre-trained classification model to complete classification; and correcting a classification result, and outputting a power distribution equipment point cloud classification data set with a category label. The method has the beneficial effects that in the ground point identification and elimination stage, a sliding window area consistent with the distribution line layout characteristics is constructed, an initial ground point set is identified by using a height gradient trend, and a misjudgment point set with obvious terrain disturbance or local ground characteristics is screened out through joint verification of a fitting residual error and a ground structure template, so that the accuracy of the ground point identification and elimination is improved. And the stability of subsequent structure extraction and the precision of ground object separation are enhanced.

The invention discloses an XR large space exhibition interaction method based on an LBE technology, and relates to the technical field of augmented reality, and the method comprises the steps: carrying out the space-time alignment of user position coordinates and environment three-dimensional point cloud data through an ICP algorithm, carrying out the noise filtering through Gaussian filtering, and forming a user position matrix; generating a spatial density distribution diagram from the user position matrix through a PPO algorithm, and performing dynamic grid segmentation on the spatial density distribution diagram to generate dynamic grid layout parameters; and inputting the dynamic grid layout parameters into a chaos prediction model, simulating a user movement mode through an improved Lorenz equation, extracting an obstacle position through an RANSAC algorithm, and performing multi-dimensional feature mapping on the user movement mode and the obstacle position to generate a collision risk thermodynamic diagram. Through the chaos prediction model and the RANSAC algorithm, the prediction precision of the user moving mode and the recognition capability of obstacles in the changing environment are improved, so that the interactive experience is smooth and natural, and the safety of the user is guaranteed.

The invention relates to the technical field of warehousing of bulk grains in a horizontal warehouse, discloses a warehousing method and system for bulk grains in a horizontal warehouse, and aims to solve the problems of uneven accumulation, potential safety hazards and low efficiency in an existing method. A three-dimensional warehouse body coordinate system is constructed through multi-dimensional analysis, and spatial density distribution is simulated; establishing a dynamic accumulation model based on density data to simulate a spatial-temporal trajectory, calculating acting force among particles through finite element analysis to obtain stress distribution, and constructing a stability evaluation system; the impurity rate is corrected through a coupling model by utilizing environment parameters of the internet of things, and a stability evaluation result is dynamically compensated by adopting a multi-factor matrix; a deep reinforcement learning algorithm is applied to optimize a control strategy, an intelligent model is deployed at an edge node, and a joint control automation system realizes closed-loop management. According to the bulk grain warehousing device, the uniformity, stability and efficiency of bulk grain warehousing operation are improved.

The invention provides a structural surface identification method based on rock mass outcrop point cloud data, and relates to the technical field of rock mass engineering, and the method comprises the steps: obtaining the rock mass outcrop point cloud data, carrying out the partitioning of the rock mass outcrop point cloud data based on an octree principle, obtaining points in each partition, traversing the points in each partition, and carrying out the recognition of the structural surface of the rock mass outcrop point cloud data. The method comprises the steps of obtaining a normal vector of each point based on principal component analysis, determining a clustering center of the normal vectors based on kernel density estimation, classifying the normal vectors of the points to clustering centers based on the clustering centers of the normal vectors by adopting a fuzzy C mean value method, and obtaining point cloud data forming each group of structural surfaces by representing each group of structural surfaces. Secondary clustering is carried out on the point cloud data of the structural plane based on the spatial density algorithm, the boundary of the structural plane is recognized, the problem that existing structural plane recognition is low in efficiency is solved, and the method is suitable for structural plane recognition.

The invention discloses a main earthquake group intelligent identification method considering a micro-earthquake contour coefficient and spectrogram clustering. The method comprises the following steps: 1, acquiring micro-earthquake monitoring data; 2, acquiring a spatial similarity matrix, an energy similarity matrix and a time similarity matrix, and weighting to form a comprehensive similarity matrix; 3, forming a feature matrix based on the comprehensive similarity matrix, and mapping data points in the feature matrix into two-dimensional space coordinates; 4, clustering data points in the two-dimensional space coordinates by adopting a DBSCAN algorithm, and obtaining a neighborhood radius and a minimum sample number optimization combination based on a micro-seismic contour coefficient; and 5, under the optimization combination of the neighborhood radius and the minimum sample number, clustering data points in the two-dimensional space coordinates by adopting a DBSCAN algorithm to obtain a plurality of micro-seismic event optimization clusters. And 6, screening out a micro-seismic event cluster with the highest comprehensive feature score as a main seismic group. The method is reasonable in design, and the main earthquake group is selected by fusing evaluation indexes of space density, energy characteristics and time characteristics.

The invention belongs to the technical scheme of human-computer interaction interface optimization, and particularly relates to a self-adaptive region English text compression display method and system. The method comprises the following steps of: 1, acquiring dynamic display parameters, and preprocessing the acquired parameters; 2, performing semantic structure analysis on the input English text through a semantic segmentation module; 3, based on the display parameters preprocessed in the step 1 and the English text subjected to semantic structure analysis in the step 2, calculating a text compression level based on a spatial density evaluation matrix, and adopting a three-level progressive compression strategy; and 4, original text data and character rendering are stored through a reversible compression marking system. The method is used for solving the problems of semantic integrity and cross-scene adaptability when the English text is displayed on the intelligent terminal equipment in a limited area.

The invention discloses a three-dimensional geological modeling method, system and device and a storage medium, and the method comprises the steps: obtaining first point cloud data and second point cloud data of a to-be-modeled region, splicing the first point cloud data and the second point cloud data through a spatial interpolation method, and obtaining third point cloud data; performing sparse region detection on the first point cloud data and the second point cloud data through a spatial density analysis method to obtain fourth point cloud data; based on the third point cloud data and the fourth point cloud data, performing prediction through a preset graph convolutional neural network to obtain a predicted geological boundary label; and based on the third point cloud data, the fourth point cloud data and the predicted geological boundary label, modeling is carried out through a Delaunay triangulation method to obtain a three-dimensional geological model of the to-be-modeled area, and the accuracy of three-dimensional geological model construction is improved.

The invention discloses a dynamic region division method based on movable coordinate data, belongs to the field of big data processing, and aims to realize efficient processing and dynamic region identification of coordinate data. According to the method, firstly, activity coordinate data of a plurality of activity subjects in a target area within a certain time period are collected, abnormal values and noise interference are removed through data preprocessing, and data accuracy is ensured; thirdly, preliminarily dividing the activity area into a plurality of sub-areas according to distribution characteristics of coordinates, performing sub-area fusion in combination with time and space density factors on the basis, and dynamically updating sub-area boundaries so as to adapt to a behavior mode of an activity subject changing along with time; according to the method, the area can be accurately and flexibly divided in real time according to the actual activity condition, the space utilization efficiency is effectively improved, the method can be widely applied to various fields such as indoor space optimization layout, and powerful support is provided for reasonable resource allocation.

The invention provides an image processing system and method for petroleum pipeline leakage detection, and belongs to the technical field of image processing.The method comprises the steps that a global edge feature map of a petroleum pipeline is extracted from a pipeline monitoring image, and a pipeline structure feature point set of the petroleum pipeline is extracted from a contour curve in the global edge feature map; extracting an oil stain identification mask of the petroleum pipeline according to a gradient modulation coefficient matrix determined by the spatial density distribution of the pipeline structure feature point set; initializing a closed evolution curve surrounding the suspected petroleum leakage area, performing evolution on the closed evolution curve by using an edge constraint function constructed by the pipeline structure feature point set and the oil stain identification mask, and detecting a spatial relationship between the current position of the closed evolution curve and the pipeline structure feature point set; and segmenting a petroleum leakage area of the petroleum pipeline according to the spatial relationship. According to the technical scheme, segmentation of the petroleum leakage boundary can be achieved in the state that the contrast ratio of the weak edge of the oil stain diffusion area and the pipeline metal surface is insufficient.

The application relates to the technical field of machine vision and image processing, and discloses a plastic valve visual defect real-time detection method and system based on deep learning, which comprises the following steps: acquiring a gray image containing a plastic valve, positioning the centroid of a plastic valve main body area by using geometric moments; converting the gray image to a polar coordinate based on the centroid, applying one-dimensional high-pass filtering, and determining a dynamic segmentation threshold to extract abnormal gray pixel points; performing spatial density clustering after reverse mapping to determine a suspected defect area; calculating the Euclidean distance between the centroid of the suspected defect area and the centroid of the plastic valve main body area, synchronously inputting the distance and image features into a double-input convolutional neural network to output a plastic valve defect detection result. The application filters out cloud and fog-shaped diffuse reflection interference, and improves the accuracy of defect detection.

The invention provides an optical cable fault intelligent analysis and positioning method based on big data and AI, and the method comprises the steps: extracting a spatial density feature value from a latitude and longitude coordinate set, judging the coordinate as a high-density region coordinate if the spatial density feature value exceeds a preset threshold value, and obtaining a preliminary spatial distribution point set; performing grouping processing on the preliminary spatial distribution point set by adopting a clustering algorithm, fusing Euclidean distance measurement in the longitude and latitude coordinate set, determining a cluster of fault points, and obtaining a distribution mode cluster group; and performing prediction processing on a time sequence of the high-incidence area identification set through a time sequence analysis algorithm, fusing trend data of the time evolution sequence segment, determining a prediction value of a future evolution rule, and obtaining a prediction rule value set.

The invention provides a sensor data processing method and device, electronic equipment and a storage medium, and relates to the technical field of data processing. According to the specific implementation scheme, n to-be-processed data sets of a target vehicle are acquired; determining m candidate data sets in the n to-be-processed data sets according to a type screening condition; and determining at least one target data set in the m candidate data sets based on the spatial density coefficient. According to the technical scheme, secondary cleaning of multiple types of sensor data can be realized based on the data type composition of the acquired sensor data and the spatial density distribution of the data acquisition position, so that the problem of large redundant data volume caused by sensor data cleaning only based on a single-type dimension is avoided; and the usability of the processed sensor data in an automatic driving simulation test scene is improved.

This invention discloses a target detection method based on variable threshold and matrix calculation, relating to the fields of computer vision and artificial intelligence application technology; including: Step 1: For detection images containing small targets, preprocess all detection boxes output by the detection model; Step 2: Preset the IoU value as the density threshold. If and only if the IoU between two detection boxes is greater than or equal to the density threshold, it is determined that the two detection boxes constitute a spatial proximity relationship. Based on this proximity relationship, for each detection box, count the total number of neighboring detection boxes that meet the density threshold condition, and define the statistical value as the local spatial density value of the detection box; Step 3: Map the local spatial density value one-to-one to the dynamic confidence threshold corresponding to the detection box, and expand the N-dimensional vector of the dynamic confidence threshold of N detection boxes into a threshold matrix T with dimension N×N; Step 4: Compare the IoU matrix I with the threshold matrix T position by position; Step 5: Compare the summation vector S with the confidence vector C element by element: For each detection box i, only the detection boxes that satisfy the condition S[i]≤C[i] are retained as the output result for subsequent target detection processes.

The application discloses a bridge dynamic deformation measurement method fusing ground-based interferometric radar and inertial vision, and the method comprises the following steps: establishing a common coordinate system, sampling the measurement data in the ground-based interferometric radar coordinate system and the inertial vision technology coordinate system to the common coordinate system respectively to obtain registration data of the ground-based interferometric radar and the inertial vision technology in the common coordinate system, interpolating the registration data in the inertial vision technology to obtain interpolation registration data matched with the spatial density of the ground-based interferometric radar data, respectively calculating the spatial model matrix of the ground-based interferometric radar line-of-sight deformation, the inertial vision technology vertical deformation and the inertial vision technology bridge lateral deformation according to the interpolation registration data and the registration data of the ground-based interferometric radar, calculating the fusion weight matrix of both by using an iterative almost unbiased estimation method, and calculating the multi-dimensional dynamic deformation time sequence of the full span of the bridge by using a least square method according to the fusion weight matrix, so that the bridge deformation monitoring is more accurate.

The invention provides an aggregate grain shape and grading identification method and device based on image processing and a storage medium, and the method comprises the steps: obtaining a dynamic falling image flow of an aggregate grain group, separating a closed contour image of a single aggregate grain from the dynamic falling image flow, carrying out the boundary constraint harmonic field construction to generate harmonic field distribution, and carrying out the stable critical structure extraction based on the harmonic field distribution. The method comprises the following steps: identifying a critical turning point of field intensity value change, encoding the critical turning point as a topology invariant sequence, calling a shape semantic mapper calibrated by concrete interface performance, performing semantic mapping processing on the topology invariant sequence, and outputting quantitative representation of single aggregate particle on the aggregate-slurry interface occlusion capability. Spatial density weighted fusion is carried out according to the quantitative characterization of the whole batch of aggregate particles, and after normalization processing is carried out, a structural efficiency index reflecting the aggregate particle shape functional characteristics is generated through weighted fusion. The method provided by the invention can comprehensively reflect the grain shape functional characteristics of the whole batch of aggregate, and provides a direct effectiveness evaluation basis for optimization of aggregate grading in concrete mix proportion design.

The invention discloses a self-adaptive ICESat-2 glacier photon denoising method, which comprises the following steps of: extracting a potential photon center point coordinate from ATL03 original photon data, fitting a curve by using the center point coordinate and constructing an envelope line; performing coarse denoising on the original photon data through the envelope line to obtain coarse screening photon point data; introducing an LOF strategy to construct a self-adaptive ellipse search window; calculating the slope of the curve to determine the search direction of the self-adaptive elliptical search window; a density clustering method is adopted, a grid adjustment mechanism is introduced to optimize density estimation, and photon space density is calculated; and based on the adaptive elliptical search window, the search direction and the photon space density, performing fine denoising on the coarse screening data to obtain signal photons. The method can accurately identify signal photons and reduce elevation calculation errors, so that the glacier surface height inversion precision is improved, meanwhile, the operation time can be greatly shortened, the algorithm operation efficiency is improved, and the actual requirement for large-scale data rapid processing is met.

The invention discloses a ship navigation trajectory prediction method and system, and the method comprises the steps: obtaining the inertial navigation data of a ship, carrying out the inertial navigation calculation based on the inertial navigation data, and obtaining a first trajectory prediction result; obtaining historical navigation data of the ship, and obtaining a second trajectory prediction result through a trajectory prediction model based on the historical navigation data; carrying out credibility evaluation on the historical navigation data in spatial density and mode similarity dimensions to obtain a credibility index of the historical data; dynamically determining data fusion weights of the first trajectory prediction result and the second trajectory prediction result based on a credibility index evaluation result; and fusing the first trajectory prediction result and the second trajectory prediction result according to the data fusion weight to generate a final ship navigation trajectory prediction result.

The present invention relates to a robust multi-sensor multi-target tracking method in a non-uniform clutter background, comprising: first, when a sensor with a limited detection range has no targets within the observation area, estimating the clutter rate of the non-uniform clutter, then estimating the spatial density of the non-uniform clutter to obtain an estimated clutter rate and an estimated spatial density; estimating the detection probability by integrating the amplitude likelihood function within a signal-to-noise ratio confidence interval based on the target signal echo amplitude characteristics within the observation area to obtain an estimated detection probability; passing the estimated clutter rate, estimated spatial density, and estimated detection probability into a standard TPMB filter to obtain a robust TPMB filter; and estimating the state of the multi-target tracks using the robust TPMB filter. In scenarios with unknown non-uniform clutter and detection probability and limited sensor detection range, this method utilizes multi-field-of-view fusion to achieve tracking of multiple targets that appear and disappear randomly, addressing the problem of mismatch in fixed detection models and providing continuous target track information.

The application provides a kind of method and system for intelligent management and control of national space planning based on multi-source data fusion, relating to spatial data processing technical field, the method comprises: obtaining remote sensing image data, geographic information vector data and ecological environment monitoring data of target area;The remote sensing image data and geographic information vector data are parsed, and surface cover elements and planning control boundaries are extracted;The intersection area of the surface cover element and the planning control boundary is calculated, and the spatial overlap area data is generated;The spatial overlap area data is executed kernel density estimation operation, and the element distribution density in continuous spatial domain is solved, to generate spatial density mapping result;The change trajectory of the spatial density mapping result on continuous time sequence is extracted.The application improves the accuracy of national space planning control.

The present invention discloses an exhibition interaction method for large-scale XR spaces based on LBE technology, which relates to the technical field of extended reality. The method comprises the following steps: aligning user location coordinates and environmental three-dimensional point cloud data in time and space using an ICP algorithm, filtering noise using a Gaussian filter, and forming a user location matrix; generating a spatial density distribution map using a PPO algorithm from the user location matrix, and dynamically meshing the spatial density distribution map to generate dynamic mesh layout parameters; inputting the dynamic mesh layout parameters into a chaotic prediction model, simulating user movement patterns using an improved Lorenz equation, extracting obstacle positions using a RANSAC algorithm, and simultaneously performing multi-dimensional feature mapping on user movement patterns and obstacle positions to generate a collision risk heat map. The present invention utilizes a chaotic prediction model and a RANSAC algorithm to improve the accuracy of user movement pattern prediction and the ability to identify obstacles in a changing environment, making the interactive experience smooth and natural while ensuring user safety.

The invention discloses an overhead distribution line completion acceptance method based on multi-view solution and fast space reconstruction. The method comprises the steps of collecting a visible light image and image information of each base tower of an overhead distribution line to be accepted; a target detection algorithm is adopted to quickly identify and mark the pole tower, each power device and the key structure point in the visible light image, and an identification image is determined and comprises identification frames of the pole tower, each power device and the key structure point; performing spatial density clustering analysis on the spatial position of the recognition image to obtain a clustering result of each tower; based on the image information, carrying out multi-view solution and rapid spatial reconstruction on the clustering result of each tower to obtain spatial coordinate information of each tower; and according to the space coordinate information of each tower, determining the size information of the power equipment and the key structure points on the tower, and determining whether the power equipment has defects according to the size information.

The present invention relates to the technical field of image recognition, and discloses a classification method, device, equipment and storage medium for hand-drawn sketches. The method includes: extracting features of the hand-drawn sketch to be recognized through a target dilated convolution to obtain multi-scale stroke structures and blank image regions; densely extracting the hand-drawn sketch to be recognized through a target neural network to obtain the spatial density of the hand-drawn sketch; recognizing the multi-scale stroke structures and blank image regions in the channel dimension according to the spatial density of the hand-drawn sketch to obtain the design type of the hand-drawn sketch to be recognized; classifying the hand-drawn sketch to be recognized according to the design type; through the above method, extracting the hand-drawn sketch to be recognized respectively by the target dilated convolution and the target neural network, then splicing in the channel dimension according to the spatial density of the hand-drawn sketch to obtain the contour of the hand-drawn sketch to be recognized, and then classifying according to the design type corresponding to the contour, which can effectively improve the accuracy of classifying hand-drawn sketches.

The invention discloses an image semantic self-adaptive compression transmission method and system for a mobile device, and relates to the technical field of image transmission communication.The method comprises the steps that original inertial motion data and interactive operation data of the device are obtained, and an interference image index and an interactive operation fitting broken line are obtained through analysis; when the interference is significant, mapping the interference into a global basic compression ratio, and extracting a target region point from a broken line; calculating a preset radius based on the target area point and the interference degree, delimiting an interactive focus area, and calculating an adjustment coefficient according to an intersection relation of each area; a high-quality first compression parameter for an interactive focus area and a second compression parameter for saving bandwidth for other areas are obtained, and differential coding and transmission are carried out according to the first compression parameter and the second compression parameter; based on spatial density analysis of historical interaction data, a focus area range is dynamically corrected for a subsequent period. According to the method, the self-adaptive clear transmission of the user concerned area and the efficient compression of the background area are realized under the condition of equipment shaking.