Mountain town space and traffic diversified characteristic data processing method

Abstract

The invention provides a mountain town space and traffic diversified feature data processing method, and relates to the technical field of electric digital data processing, remote sensing data is collected at different sampling heights, building features and traffic features of mountain town are extracted, hierarchical clustering is performed on all the building features and traffic features, and the data of the mountain town space and traffic diversified features is obtained. Obtaining a building feature cluster set and a traffic feature cluster set; determining reward heterogeneity of the mountain cities and towns in the spatial dimension under different sampling heights based on a feature reward mechanism of reinforcement learning; performing spectral difference analysis on the remote sensing data at different sampling heights to obtain spectral response differences; and determining a reward factor of multivariate feature enhancement through reward heterogeneity and spectral response difference, and fusing all remote sensing data according to the reward factor to obtain fused features of mountain urban space and traffic distribution. By adopting the scheme of the invention, enhanced fusion of multi-element features at different sampling heights can be realized, so that the accuracy of mountain town space and traffic feature analysis is improved.

Description

Technical Field

[0001] The present application relates to the technical field of electronic digital data processing, and more specifically, to a method for processing data on diversified characteristics of mountain town space and transportation. Background Art

[0002] Due to their complex topography and unique ecological environment, mountain towns face many challenges in urban planning, traffic design and spatial layout. The spatial and traffic characteristics of mountain towns are often diversified, including multiple clusters, multiple centers, multiple dimensions and multiple levels. These characteristics not only affect the overall function of the city, but also have a profound impact on transportation, economic production, social culture and ecological environment protection.

[0003] Traditional data processing methods have certain limitations in capturing the characteristics of mountainous towns. Although the early manual mapping method can reflect the adaptability of land through manual analysis, it is difficult to effectively cope with the dynamic changes of complex terrain and large-scale data. With the widespread application of GIS (Geographic Information System) technology, the ability of spatial data analysis has been significantly improved, but conventional GIS methods mostly focus on two-dimensional feature extraction and it is difficult to fully integrate high-dimensional and multi-source data, especially when facing multi-sampling height and multi-spectral data. The ability to express feature fusion and heterogeneity is insufficient. By analyzing the spectral response at different sampling heights and fusing multi-source features, the insufficient expression of complexity in traditional methods can be effectively solved, and the accuracy of spatial and traffic feature analysis can be improved. Therefore, how to achieve the enhanced fusion of multi-features at different sampling heights, so as to improve the accuracy of spatial and traffic feature analysis in mountainous towns. Summary of the invention

[0004] The present application provides a method for processing data on diversified characteristics of mountain town space and transportation, which can realize the enhanced fusion of multi-dimensional characteristics at different sampling heights, thereby improving the accuracy of analysis of mountain town space and transportation characteristics.

[0005] The present application provides a method for processing mountain town space and traffic diversified feature data, the data processing method comprising the following steps:

[0006] Collect remote sensing data of mountain towns at different sampling heights and map all remote sensing data into a multivariate feature space;

[0007] Extracting architectural features and traffic features of mountain towns at different sampling heights from the multivariate feature space, and then hierarchically clustering all architectural features and traffic features to obtain architectural feature clusters and traffic feature clusters;

[0008] Determine the reward information between the building feature cluster and the traffic feature cluster based on the feature reward mechanism of reinforcement learning, and determine the reward heterogeneity of mountain towns in the spatial dimension at different sampling heights through the reward information;

[0009] According to the distribution difference of remote sensing pixels in each remote sensing data, the spectral difference analysis is performed on the remote sensing data collected at different sampling heights to obtain the spectral response differences at different sampling heights;

[0010] The reward factor for multivariate feature enhancement is determined by the reward heterogeneity and the spectral response difference, and all collected remote sensing data are fused according to the reward factor to obtain the fusion characteristics of mountain town space and traffic distribution.

[0011] In this embodiment, multispectral remote sensing technology is used to collect remote sensing data of mountain towns at different altitudes.

[0012] In this embodiment, mapping all remote sensing data into a multivariate feature space specifically includes:

[0013] Align all remote sensing data to obtain all aligned remote sensing data;

[0014] The principal component analysis algorithm is used to map all the remote sensing data into a low-dimensional feature space to obtain a multivariate feature space.

[0015] In this embodiment, extracting architectural features and traffic features of mountain towns at different sampling heights from the multivariate feature space specifically includes:

[0016] For each sampling height, a convolutional neural network model is used to identify edge features of buildings in the multivariate feature space to obtain architectural features of mountain towns at the sampling height;

[0017] Traffic features of mountain towns at the sampling height are extracted from the multivariate feature space using a road detection algorithm.

[0018] In this embodiment, all building features and traffic features are hierarchically clustered to obtain building feature clusters and traffic feature clusters, which specifically include:

[0019] All building features and traffic features are clustered based on the minimum distance mechanism to obtain building feature clusters and traffic feature clusters at different density levels.

[0020] Obtain building feature clusters through all building feature clusters;

[0021] The traffic feature clusters are obtained through all the traffic feature clusters.

[0022] In this embodiment, the feature reward mechanism based on reinforcement learning determines the reward information between the building feature cluster and the traffic feature cluster, specifically including:

[0023] Determining a correlation matrix between the building feature cluster and the traffic feature cluster;

[0024] Based on the feature reward mechanism of reinforcement learning, reward adjustment is performed on matrix elements in the association matrix that are greater than a preset threshold, and penalty adjustment is performed on matrix elements in the association matrix that are less than the preset threshold;

[0025] The adjusted correlation matrix is ​​used to represent the reward information between the building feature cluster and the traffic feature cluster.

[0026] In this embodiment, determining the reward heterogeneity of mountain towns in the spatial dimension at different sampling heights through the reward information specifically includes:

[0027] Obtain the difference in spatial distribution of mountain towns at different sampling heights;

[0028] The reward heterogeneity of mountain towns in the spatial dimension at different sampling heights is determined based on the difference in the reward information and the spatial distribution.

[0029] In this embodiment, the spectral difference analysis is performed on the remote sensing data collected at different sampling heights according to the distribution difference of the remote sensing pixels in each remote sensing data, and the spectral response differences at different sampling heights are obtained, which specifically include:

[0030] The spectral characteristics of remote sensing pixels are used to determine the spectral response value of remote sensing images at each sampling height;

[0031] The spectral response differences at different sampling heights are determined by the distribution differences of all spectral response values ​​and remote sensing pixels in each remote sensing data.

[0032] In this embodiment, all the collected remote sensing data are fused according to the reward factor to obtain the fusion characteristics of mountain town space and traffic distribution, which specifically include:

[0033] The remote sensing data at different sampling heights are weightedly fused according to the enhancement factor to generate a fused image with high spatial resolution;

[0034] Based on the pre-trained feature extraction model, fusion features of mountain town space and traffic distribution are extracted from the fused image.

[0035] In this embodiment, before mapping all remote sensing data into a multivariate feature space, it also includes performing noise reduction processing on all remote sensing data.

[0036] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0037] By collecting remote sensing data of mountain towns at different sampling heights, all remote sensing data are mapped into a multivariate feature space; the architectural features and traffic features of mountain towns at different sampling heights are extracted from the multivariate feature space, and then all architectural features and traffic features are hierarchically clustered to obtain architectural feature clusters and traffic feature clusters; the feature reward mechanism based on reinforcement learning determines the reward information between the architectural feature clusters and the traffic feature clusters, and the reward heterogeneity of mountain towns in the spatial dimension at different sampling heights is determined through the reward information; the remote sensing data collected at different sampling heights are subjected to spectral difference analysis according to the distribution difference of remote sensing pixels in each remote sensing data, and the spectral response difference at different sampling heights is obtained; the reward factor for multivariate feature reinforcement is determined through the reward heterogeneity and the spectral response difference, and all the collected remote sensing data are fused according to the reward factor to obtain the fusion characteristics of the space and traffic distribution of mountain towns.

[0038] It can be seen that this application can improve the accuracy of spatial and traffic feature analysis in mountainous towns by strengthening the integration of multivariate features at different sampling heights; first, by mapping remote sensing data at different sampling heights to form a multivariate feature space, and then extracting building and traffic features and performing hierarchical clustering, it is possible to automatically identify functional units and traffic networks at different spatial scales from the data. The use of hierarchical clustering can ensure that the differences between different features can be clearly distinguished and effectively utilized, thereby improving the applicability of the model; then, the feature reward mechanism based on reinforcement learning provides intelligent adjustment for the relationship between building features and traffic features, ensuring that the matching correlation between different feature clusters is fully explored. This mechanism is not It can not only enhance the expression ability of spatial features, but also improve the processing ability of various data differences in complex environments by rewarding heterogeneity, further optimizing the effect of feature fusion. In addition, spectral response difference analysis provides a scientific basis for feature fusion by identifying the distribution differences of remote sensing pixels at different sampling heights, ensuring the accurate capture of spectral features during the fusion process; finally, by combining reward heterogeneity and spectral response differences to determine the reward factor of multivariate features, it can achieve weighted fusion of multi-source remote sensing data and generate fusion features with high spatial resolution. This fusion feature can not only accurately reflect the spatial layout and traffic distribution of mountain towns, but also adapt to dynamically changing environments, providing more accurate and detailed analysis results. In summary, the present application scheme can achieve the enhanced fusion of multivariate features at different sampling heights, thereby improving the accuracy of spatial and traffic feature analysis of mountain towns. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 is an exemplary flow chart of a method for processing spatial and transportation diversified characteristic data of mountainous towns provided in this application;

[0041] Figure 2 is an exemplary flow chart for determining a multivariate feature space according to the present application;

[0042] Figure 3 is an exemplary flow chart for determining spectral response differences according to the present application. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0044] The embodiment of the present application provides a method for processing multi-dimensional characteristic data of mountain town space and traffic, the core of which is to collect remote sensing data of mountain town at different sampling heights, map all remote sensing data into multi-dimensional characteristic space; extract architectural characteristics and traffic characteristics of mountain town at different sampling heights from the multi-dimensional characteristic space, and then hierarchically cluster all architectural characteristics and traffic characteristics to obtain architectural characteristic clusters and traffic characteristic clusters; determine the reward information between the architectural characteristic clusters and the traffic characteristic clusters based on the feature reward mechanism of reinforcement learning, and determine the reward heterogeneity of mountain town in spatial dimension at different sampling heights through the reward information; perform spectral difference analysis on remote sensing data collected at different sampling heights according to the distribution difference of remote sensing pixels in each remote sensing data, and obtain the spectral response difference at different sampling heights; determine the reward factor of multi-dimensional characteristic reinforcement through the reward heterogeneity and the spectral response difference, and fuse all collected remote sensing data according to the reward factor to obtain the fusion characteristics of mountain town space and traffic distribution. The above scheme can realize the enhanced fusion of multi-dimensional characteristics at different sampling heights, thereby improving the accuracy of mountain town space and traffic characteristic analysis.

[0045] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown in FIG. 1 , this figure is an exemplary flow chart of a method for processing mountain town space and traffic diversified feature data according to this embodiment of the present application, and the data processing method includes the following steps:

[0046] In step S1, remote sensing data of mountain towns are collected at different sampling heights, and all remote sensing data are mapped into a multivariate feature space.

[0047] In specific implementation, multispectral remote sensing technology can be used to collect remote sensing data of mountain towns at different altitudes. It should be noted that the remote sensing data at different altitudes in this application provide different spatial resolutions and viewing angles, allowing mountain towns to be observed from both macro and micro perspectives, wherein the remote sensing data include spectral data and spatial data of different bands; it should also be noted that the present application also includes denoising all remote sensing data before mapping them into a multivariate feature space, and mean filtering technology can be used to perform denoising on all remote sensing data, which will not be elaborated here.

[0048] Preferably, in this embodiment, reference Figure 2 As shown, this figure is an exemplary flow chart of determining a multivariate feature space in an embodiment of the present application. In this embodiment, all remote sensing data can be mapped into a multivariate feature space in the following manner, namely:

[0049] First, in step S11, all remote sensing data are aligned to obtain all aligned remote sensing data;

[0050] Then, in step S12, a principal component analysis algorithm is used to map all the aligned remote sensing data to a low-dimensional feature space to obtain a multivariate feature space.

[0051] In the specific implementation, first, ground control points can be used for image registration, and all remote sensing data can be projected into a unified geographic coordinate system to obtain spatially unified corresponding remote sensing data; then, principal component analysis technology is used to flatten all aligned remote sensing data (converting from a multidimensional image matrix to a two-dimensional matrix form, with each row representing a pixel and each column representing a feature, such as a spectral band, texture feature, etc.), and then the covariance matrix of all features is calculated to quantify the correlation between features, and then the covariance matrix is ​​decomposed by eigenvalues, and the main feature components are screened out through the correlation between features to obtain multivariate feature data. Finally, the multivariate feature data is constructed into a low-dimensional multivariate feature space according to the dimensions of different sampling heights.

[0052] In step S2, the architectural features and traffic features of mountain towns at different sampling heights are extracted from the multivariate feature space, and then all architectural features and traffic features are hierarchically clustered to obtain architectural feature clusters and traffic feature clusters.

[0053] In this embodiment, the architectural features and traffic features of mountain towns at different sampling heights can be extracted from the multivariate feature space in the following manner, namely:

[0054] For each sampling height, a convolutional neural network model is used to identify edge features of buildings in the multivariate feature space to obtain architectural features of mountain towns at the sampling height;

[0055] Traffic features of mountain towns at the sampling height are extracted from the multivariate feature space using a road detection algorithm.

[0056] It should be noted that the convolutional neural network model in the present application is used to extract the edge features of buildings. In the specific implementation, for each sampling height, first, the convolutional layer of the convolutional neural network model can be used to automatically extract spatial features, such as the geometric shape and edge features of the building, and then the edge features of the building features are output by the convolutional neural network model, and the extracted edge features are used as the building features of mountain towns, and the building features include building geometry information, building density information and building function information; then, the Canny edge detection algorithm in the road detection algorithm can be used to extract traffic features from the multivariate feature space, and the traffic features include road geometry information, road network topology information and traffic flow information.

[0057] In this embodiment, all building features and traffic features are hierarchically clustered to obtain building feature clusters and traffic feature clusters in the following manner:

[0058] All building features and traffic features are clustered based on the minimum distance mechanism to obtain building feature clusters and traffic feature clusters at different density levels.

[0059] Obtain building feature clusters through all building feature clusters;

[0060] The traffic feature clusters are obtained through all the traffic feature clusters.

[0061] In specific implementation, each building feature can be regarded as a separate cluster, and the building density in the building feature can be used as a clustering indicator, that is, the distance between the building densities of the building features is calculated, and based on the minimum distance principle, the two nearest clusters are gradually merged until the preset conditions are met, and multiple building feature clusters are obtained, and then all the building feature clusters are combined into a set as a building feature cluster set. Similarly, the traffic flow in the traffic feature can be used as a clustering indicator, and based on the minimum distance principle, the two nearest clusters are gradually merged until the preset conditions are met, and multiple traffic feature clusters are obtained, and then all the traffic feature clusters are combined into a set as a traffic feature cluster set. It should be noted that the preset conditions in the present application can be set to divide all features into 4 clusters. In other embodiments, the preset conditions can also be set to others, which are not limited here.

[0062] In step S3, the feature reward mechanism based on reinforcement learning determines the reward information between the building feature cluster and the traffic feature cluster, and the reward heterogeneity of the mountain town in the spatial dimension at different sampling heights is determined through the reward information.

[0063] In this embodiment, the feature reward mechanism based on reinforcement learning can be used to determine the reward information between the building feature cluster and the traffic feature cluster in the following manner, namely:

[0064] Determining a correlation matrix between the building feature cluster and the traffic feature cluster;

[0065] Based on the feature reward mechanism of reinforcement learning, reward adjustment is performed on matrix elements in the association matrix that are greater than a preset threshold, and penalty adjustment is performed on matrix elements in the association matrix that are less than the preset threshold;

[0066] The adjusted correlation matrix is ​​used to represent the reward information between the building feature cluster and the traffic feature cluster.

[0067] In specific implementation, first, the building feature cluster and the building feature cluster can be input into a pre-trained association model, and the association model is used to output the association matrix between the building feature cluster and the building feature cluster, and the association value in the association matrix describes the matching relationship between the building feature cluster in the building feature cluster and the traffic feature cluster in the traffic feature cluster, such as the matching degree between traffic flow and building density; then, a reinforcement learning model is defined, and the current association matrix is ​​used as the initial state parameter of the reinforcement learning model. The reinforcement learning model rewards the matching relationship between the building and the traffic through a customized reward mechanism, and sets an association threshold to distinguish between matrix elements with high association and low association, and further rewards the association matrix through the reinforcement learning model, that is, reward adjustment is performed on the matrix elements in the association matrix that are greater than the association threshold, and penalty adjustment is performed on the matrix elements in the association matrix that are less than the association threshold. It should be noted that the adjustment mechanism in the present application can be dynamically adjusted according to the matching degree between the building and the traffic, and the reward adjustment refers to increasing the value based on the current data, and the penalty adjustment refers to decreasing the value based on the current data; finally, the adjusted association matrix is ​​used as the expression form of the reward information between the building feature cluster and the traffic feature cluster.

[0068] It should be noted that this application sets up an incentive mechanism to evaluate the coordination ability of urban spatial layout and functions, which can provide a good data basis for the optimization of urban spatial layout.

[0069] In this embodiment, the reward heterogeneity of mountain towns in the spatial dimension at different sampling heights can be determined by the reward information in the following manner, namely:

[0070] Obtain the difference in spatial distribution of mountain towns at different sampling heights;

[0071] The reward heterogeneity of mountain towns in the spatial dimension at different sampling heights is determined based on the difference in the reward information and the spatial distribution.

[0072] In the specific implementation, first, the remote sensing data of different sampling heights can be subjected to multivariate feature mapping to obtain the feature space corresponding to the sampling height, and then the average value of the similarity between the feature space at each sampling height and the feature space at other sampling heights is taken as the difference in the spatial distribution of mountain towns at the sampling height, where the similarity can be quantified using the Euclidean distance between the feature spaces, which is not specifically limited here; then, for each sampling height, the inverse of the difference in the spatial distribution of mountain towns at the sampling height is taken as the reward value of the sampling height, and then the reward values ​​of all sampling heights are obtained, and then all the reward values ​​and the association matrix in the reward information are taken as the reward The input parameters of the model can be used to learn the differences in the spatial dimensions of different features (i.e., architectural features and traffic features) in mountainous towns through the reward model. The reward model can output the difference values ​​of the rewards of mountainous towns in the spatial dimensions at different sampling heights, and the difference values ​​can be used to describe the reward heterogeneity of mountainous towns in the spatial dimensions at different sampling heights. The reward heterogeneity can intuitively reflect the feature change characteristics of mountainous towns at different sampling heights, providing a key basis for optimizing feature fusion. It should also be noted that the reward model in this application is obtained through convolutional neural network training, and other algorithm structures can also be used in other embodiments, which are not limited here.

[0073] It should be noted that the reward heterogeneity in this application is a characteristic indicator used to quantify the degree of change in the distribution of architectural features and traffic features at different sampling heights in the spatial dimension. By calculating the reward heterogeneity and quantifying the differences in the spatial dimension of mountain towns, it can provide guidance for the weight allocation in the subsequent multi-feature fusion, and further improve the accuracy and rationality of remote sensing data processing.

[0074] In step S4, spectral difference analysis is performed on the remote sensing data collected at different sampling heights according to the distribution difference of the remote sensing pixels in each remote sensing data, so as to obtain the spectral response difference at different sampling heights.

[0075] Preferably, in this embodiment, reference Figure 3 As shown in FIG. 1 , this figure is an exemplary flow chart for determining the spectral response difference in an embodiment of the present application. In this embodiment, the spectral difference analysis is performed on the remote sensing data collected at different sampling heights according to the distribution difference of the remote sensing pixels in each remote sensing data. The spectral response difference at different sampling heights can be obtained in the following manner, namely:

[0076] First, in step S41, the spectral response value of the remote sensing image at each sampling height is determined using the spectral characteristics of the remote sensing pixel;

[0077] Then, in step S42, the spectral response differences at different sampling heights are determined through all spectral response values ​​and the distribution differences of remote sensing pixels in each remote sensing data.

[0078] In specific implementation, first, for each sampling height, the spectral characteristics of each pixel can be extracted from the remote sensing data, and the spectral characteristics include multi-band reflectance values ​​(such as visible light, near infrared, thermal infrared, etc.), and then the spectral characteristics of all pixels are obtained, the spectral characteristics of all pixels are statistically analyzed, and the mean spectrum and standard deviation spectrum of the spectral characteristics of all pixels are calculated, and then the ratio between the mean spectrum and the standard deviation spectrum can be used as the spectral response value of the remote sensing image at the sampling height; then, the distribution characteristic values ​​of the remote sensing pixels in the remote sensing data at each sampling height can be used as the distribution difference of the remote sensing pixels in the corresponding remote sensing data, wherein the distribution characteristic values ​​can be quantified using the distribution gradient, which will not be elaborated here, and finally the product between the spectral response value of the remote sensing image at each sampling height and the distribution difference of the remote sensing pixels in the corresponding remote sensing data is used as the spectral response difference at each sampling height, and then the array composed of the spectral response differences at all sampling heights is used as the spectral response difference at different sampling heights.

[0079] It should be noted that the spectral response difference in the present application measures the differences in the spectral characteristics of remote sensing images at different sampling heights; it should also be noted that the present application accurately characterizes the spectral difference characteristics of remote sensing data at different sampling heights through a comprehensive analysis of the spectral response value and the distribution difference, which can provide data support for the generation of fusion features and help improve the processing accuracy of spatial and traffic distribution characteristics of mountain towns.

[0080] In step S5, a reward factor for multivariate feature enhancement is determined by the reward heterogeneity and the spectral response difference, and all collected remote sensing data are fused according to the reward factor to obtain fused features of mountain town space and traffic distribution.

[0081] In this embodiment, the reward factor for multivariate feature reinforcement is determined by the reward heterogeneity and the spectral response difference in the following manner, namely:

[0082] The spectral response difference is used as an initialization parameter for processing spectral responses at different sampling heights in a reward model;

[0083] The reward heterogeneity is used as an initialization parameter in the reward model to handle the difference in multivariate feature rewards of mountain towns in the spatial dimension at different sampling heights;

[0084] The reward model outputs reward factors for multivariate feature enhancement at different sampling heights.

[0085] In the specific implementation, first, an array composed of the spectral response differences at all sampling heights and the reward difference values ​​of the multivariate features (such as building features and traffic features) of mountain towns in the spatial dimension at different sampling heights can be input into the reward model, and the reward factor is allocated to the multivariate features at different sampling heights through the reward model; it should be noted that a reward model based on a reinforcement learning reward mechanism (such as deep reinforcement learning) can be used in the present application. In other embodiments, other types of reward models can also be used, which are not limited here; it should also be noted that the reward factor allocation mechanism in the present application specifically includes: a feature combination with high spectral response differences, a higher reward factor, and a spatial feature with high reward heterogeneity, a higher reward factor.

[0086] In this embodiment, all the collected remote sensing data are fused according to the reward factor to obtain the fusion characteristics of mountain town space and traffic distribution, which can be specifically adopted in the following manner, namely:

[0087] The remote sensing data at different sampling heights are weightedly fused according to the enhancement factor to generate a fused image with high spatial resolution;

[0088] Based on the pre-trained feature extraction model, fusion features of mountain town space and traffic distribution are extracted from the fused image.

[0089] In the specific implementation, first, the remote sensing data of different sampling heights are weighted according to the enhancement factor, and the weights are assigned to each layer of remote sensing data, which are fused to generate high-spatial-resolution images. This step fully considers the spectral response differences and spatial feature reward heterogeneity of remote sensing data at different heights, ensuring that the fused image has both high detail expression capabilities and global consistency. Then, the fused high-resolution image is processed using a pre-trained deep learning feature extraction model (such as a model based on a convolutional neural network) to extract the spatial features of mountain towns such as buildings and roads, as well as the distribution information of the transportation network. Finally, the deep learning feature extraction model is used to output the fusion features, which cover the spatial layout and transportation network distribution of mountain towns, and can be used as the data basis for further analysis (such as spatial optimization planning and traffic load assessment), providing a scientific basis for urban development and planning, while improving the comprehensive utilization efficiency of multi-source data.

[0090] It can be seen that this application improves the accuracy of spatial and traffic feature analysis in mountainous towns by strengthening the fusion of multivariate features at different sampling heights; first, by mapping remote sensing data at different sampling heights to form a multivariate feature space, and then extracting building and traffic features, and performing hierarchical clustering on the features, it is possible to automatically identify functional units and traffic networks at different spatial scales from the data. The use of hierarchical clustering ensures that the differences between different features can be clearly distinguished and effectively utilized, thereby improving the refinement and applicability of the model; then, the feature reward mechanism based on reinforcement learning provides intelligent adjustment for the relationship between building features and traffic features, ensuring that the correlation between different feature clusters is fully explored. This mechanism The system can not only enhance the expression ability of spatial features, but also improve the processing ability of various data differences in complex environments by rewarding heterogeneity, further optimizing the effect of feature fusion. In addition, the spectral response difference analysis provides a scientific basis for feature fusion by identifying the distribution differences of remote sensing pixels at different sampling heights, ensuring the accurate capture of spectral features during the fusion process; finally, by combining reward heterogeneity and spectral response differences to determine the reward factor of multivariate features, it is possible to achieve weighted fusion of multi-source remote sensing data and generate fusion features with high spatial resolution. This fusion feature can not only accurately reflect the spatial layout and traffic distribution of mountain towns, but also adapt to dynamically changing environments, providing more accurate and detailed analysis results. In summary, the present application scheme can achieve the enhanced fusion of multivariate features at different sampling heights, thereby improving the accuracy of spatial and traffic feature analysis of mountain towns.

[0091] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0092] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0093] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A method for processing mountain town space and traffic diversified feature data, characterized in that: The data processing method comprises the following steps: Collect remote sensing data of mountain towns at different sampling heights and map all remote sensing data into a multivariate feature space; Extracting architectural features and traffic features of mountain towns at different sampling heights from the multivariate feature space, and then hierarchically clustering all architectural features and traffic features to obtain architectural feature clusters and traffic feature clusters; Determine the reward information between the building feature cluster and the traffic feature cluster based on the feature reward mechanism of reinforcement learning, and determine the reward heterogeneity of mountain towns in the spatial dimension at different sampling heights through the reward information; According to the distribution difference of remote sensing pixels in each remote sensing data, the spectral difference analysis is performed on the remote sensing data collected at different sampling heights to obtain the spectral response differences at different sampling heights; The reward factor for multivariate feature enhancement is determined by the reward heterogeneity and the spectral response difference, and all collected remote sensing data are fused according to the reward factor to obtain the fusion characteristics of mountain town space and traffic distribution.

2. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: Remote sensing data of mountain towns are collected at different altitudes using multispectral remote sensing technology.

3. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: Mapping all remote sensing data into a multivariate feature space specifically includes: Align all remote sensing data to obtain all aligned remote sensing data; The principal component analysis algorithm is used to map all the remote sensing data into a low-dimensional feature space to obtain a multivariate feature space.

4. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: The building features and traffic features of mountain towns at different sampling heights extracted from the multivariate feature space specifically include: For each sampling height, a convolutional neural network model is used to identify edge features of buildings in the multivariate feature space to obtain architectural features of mountain towns at the sampling height; Traffic features of mountain towns at the sampling height are extracted from the multivariate feature space using a road detection algorithm.

5. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: All building features and traffic features are hierarchically clustered, and the building feature clusters and traffic feature clusters obtained specifically include: All building features and traffic features are clustered based on the minimum distance mechanism to obtain building feature clusters and traffic feature clusters at different density levels. Obtain building feature clusters through all building feature clusters; The traffic feature clusters are obtained through all the traffic feature clusters.

6. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: Determining the reward information between the building feature cluster and the traffic feature cluster based on the feature reward mechanism of reinforcement learning specifically includes: Determining a correlation matrix between the building feature cluster and the traffic feature cluster; Based on the feature reward mechanism of reinforcement learning, reward adjustment is performed on matrix elements in the association matrix that are greater than a preset threshold, and penalty adjustment is performed on matrix elements in the association matrix that are less than the preset threshold; The adjusted correlation matrix is ​​used to represent the reward information between the building feature cluster and the traffic feature cluster.

7. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: Determining the reward heterogeneity of mountain towns in the spatial dimension at different sampling heights through the reward information specifically includes: Obtain the difference in spatial distribution of mountain towns at different sampling heights; The reward heterogeneity of mountain towns in the spatial dimension at different sampling heights is determined based on the difference in the reward information and the spatial distribution.

8. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: According to the distribution difference of remote sensing pixels in each remote sensing data, the spectral difference analysis of remote sensing data collected at different sampling heights is carried out, and the spectral response differences at different sampling heights are obtained, which specifically include: The spectral characteristics of remote sensing pixels are used to determine the spectral response value of remote sensing images at each sampling height; The spectral response differences at different sampling heights are determined by the distribution differences of all spectral response values ​​and remote sensing pixels in each remote sensing data.

9. A method for processing mountain town space and traffic diversified characteristic data as claimed in claim 1, characterized in that: According to the reward factor, all the collected remote sensing data are fused to obtain the fusion characteristics of mountain town space and traffic distribution, which specifically include: The remote sensing data at different sampling heights are weightedly fused according to the enhancement factor to generate a fused image with high spatial resolution; Based on the pre-trained feature extraction model, fusion features of mountain town space and traffic distribution are extracted from the fused image.

10. A method for processing mountain town space and traffic diversified feature data as claimed in claim 1, characterized in that: Before mapping all remote sensing data into multivariate feature space, all remote sensing data are also subjected to noise reduction processing.

Images