Search

CN-122022049-A - Urban landscaping optimal planning system based on multisource remote sensing data

CN122022049ACN 122022049 ACN122022049 ACN 122022049ACN-122022049-A

Abstract

The invention discloses an urban landscaping optimal planning system based on multi-source remote sensing data, which comprises a multi-source remote sensing data processing module, a space-time ground object three-dimensional coding module, a space recognition module, a false recognition risk assessment module, an ecological evolution trend assessment module, a greening plot optimization module, a greenbelt path network construction module and an optimal path planning module, wherein the multi-source remote sensing data processing module is used for generating a multi-source remote sensing time sequence data set, the space recognition module is used for generating vegetation expression data, the false recognition risk assessment module is used for determining a false recognition high risk area set, the ecological evolution trend assessment module is used for generating an ecological evolution trend distribution map, the greening plot optimization module is used for carrying out iterative search through multi-directional antenna detection and synaptic weight update based on an improved beetle optimization algorithm, and generating a final greening plot set, the greenbelt path network construction module is used for constructing a greenbelt path network structure, and the optimal path planning module is used for outputting an optimal urban landscaping greening planning path. The invention combines remote sensing intelligent recognition with improved beetle optimization algorithm to realize the optimal planning of urban greening paths.

Inventors

  • MENG FANLU
  • CHEN XIAN
  • LI XIN

Assignees

  • 菏泽市牡丹区自然资源局
  • 青州市园林绿化和环卫中心
  • 菏泽市牡丹区国有经济林场

Dates

Publication Date
20260512
Application Date
20260203

Claims (10)

  1. 1. Urban landscaping optimal planning system based on multi-source remote sensing data is characterized by comprising the following modules: the multi-source remote sensing data processing module is used for acquiring multi-source remote sensing data covering a target city area, and performing space registration processing under a unified coordinate system to generate a multi-source remote sensing time sequence data set; The space-time ground object three-dimensional coding module is used for receiving the multi-source remote sensing time sequence data set, respectively constructing a three-dimensional expression structure for each grid unit in the target city area, and generating vegetation expression data; the space recognition module is used for inputting the vegetation expression data into a space recognition network, carrying out greening type semantic recognition on each grid unit, and generating a semantic result vector and a reconstruction residual vector; The false recognition risk assessment module is used for generating a false recognition risk distribution map based on the semantic result vector and the reconstruction residual vector and combining vegetation index time sequence data corresponding to each grid unit, and determining a false recognition high risk area set; The ecological evolution trend evaluation module is used for acquiring carbon sink indexes, surface temperature indexes and soil moisture indexes of each grid unit, modeling time sequences and generating an ecological evolution trend distribution map; The greening plot optimization module is used for taking the ecological evolution trend distribution map as optimization input data, taking the false recognition high risk area set as a space constraint condition, carrying out iterative search through multi-directional antenna detection and synaptic weight updating based on an improved beetle optimization algorithm introducing a nerve mimicry synaptic drift mechanism, and generating a final greening plot set; The greenbelt path network construction module is used for constructing a greening node set based on the final greening land block set, and establishing a communication relation between greening nodes according to the ecological evolution trend distribution diagram and the space distance to construct a greenbelt path network structure; and the optimal path planning module is used for screening and sequencing the candidate paths of the green land path network structure and outputting an optimal urban landscaping planning path.
  2. 2. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the multi-source remote sensing data comprises optical remote sensing images, synthetic aperture radar images, laser radar point cloud data and thermal infrared remote sensing data.
  3. 3. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the space-time ground object three-dimensional coding module is specifically: Performing regular grid division on the multi-source remote sensing time sequence data set according to a set spatial resolution to obtain a plurality of grid units covering a target city area, wherein each grid unit corresponds to a unique spatial coordinate range; For each grid unit, respectively reading an optical remote sensing image pixel value, a synthetic aperture radar back scattering value, a laser radar point cloud height statistic value and a thermal infrared remote sensing bright temperature value in a corresponding space coordinate range at each time node to form a multi-source original feature vector of the grid unit under the time node; Sequentially arranging the multisource original feature vectors obtained by the same grid unit under different time nodes according to time sequence to form a time sequence feature sequence of the grid unit; respectively calculating the mean value and standard deviation of each dimension feature in the time dimension, subtracting the mean value from the feature value corresponding to each time node according to a Z score standardization method, and dividing the feature value by the standard deviation to obtain a standardized time sequence feature sequence; Based on the standardized time sequence characteristic sequence, respectively calculating a corresponding vegetation index value, a corresponding ground surface temperature value and a corresponding ground surface structure height value at each time node, and marking the calculation result as ground feature intensity information of the grid unit under the time node; Splicing the space coordinate index, the time node index and the corresponding ground object intensity information of each grid unit under the same time node in sequence to generate a three-dimensional data slice of each grid unit under the time node; stacking three-dimensional data slices generated by the same grid unit under all time nodes along the time dimension to form a three-dimensional expression structure of the grid unit; And collecting the three-dimensional expression structures corresponding to all grid units in the target city area to obtain vegetation expression data.
  4. 4. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the spatial recognition module is specifically: Stacking the ground object intensity information corresponding to each grid unit at all time nodes according to time sequence based on the vegetation expression data, taking the ground object intensity information as an input sample, and inputting the input sample into a space recognition network; In the coding path of the space recognition network, sequentially executing two-dimensional convolution operation on each input sample, and performing nonlinear mapping processing on a convolution result through a ReLU activation function after each two-dimensional convolution operation to obtain coding feature mapping; Performing downsampling operation on the encoded feature map to obtain a downsampled feature map, wherein the spatial resolution of the downsampled feature map is lower than the spatial resolution of the input samples; Inputting the downsampling feature map to a first decoding path and a second decoding path, respectively; In the first decoding path, up-sampling operation and two-dimensional convolution operation are sequentially carried out on the down-sampling feature mapping, nonlinear mapping processing is carried out through a ReLU activation function after each two-dimensional convolution operation, wherein the number of convolution kernels of the two-dimensional convolution operation of the last layer is set as the number of greening types, and semantic result vectors corresponding to the greening types are output; In the second decoding path, up-sampling operation and two-dimensional convolution operation which are in the same sequence as the first decoding path are sequentially carried out on the down-sampling feature mapping, nonlinear mapping processing is carried out through a ReLU activation function after each two-dimensional convolution operation, wherein the number of convolution kernels of the last layer of two-dimensional convolution operation is set as the number of feature channels of the input sample, and reconstruction feature mapping consistent with the spatial resolution of the input sample is obtained; Performing subtraction operation on the reconstructed feature map and the input sample element by element at the same spatial position to obtain a reconstructed residual feature map, and performing global average pooling operation on the reconstructed residual feature map along the spatial dimension to obtain a corresponding reconstructed residual vector; and correspondingly storing the semantic result vector and the reconstructed residual vector according to grid units.
  5. 5. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the false recognition risk assessment module is specifically: Based on the semantic result vector, respectively reading corresponding greening type component values for each grid unit, and recording the component value with the largest numerical value in the semantic result vector as a semantic confidence value of the grid unit; subtracting the semantic confidence value from 1 to obtain the corresponding semantic uncertainty of the grid unit; based on the reconstructed residual vector, respectively reading corresponding residual difference values of each grid unit, taking absolute values of the residual component values, carrying out summation operation, and dividing the sum by the number of residual components to obtain average residual magnitudes of the grid units; acquiring vegetation index values corresponding to each grid unit at a plurality of time nodes, and arranging the vegetation index values according to time sequence to form a vegetation index time sequence of the grid units; Performing item-by-item differential operation on vegetation index values of adjacent time nodes in the vegetation index time sequence to obtain a vegetation index change sequence; taking absolute values of all differential values in the vegetation index change sequence, and then carrying out summation operation, and dividing the absolute values by the number of differential terms to obtain the average change amplitude of the vegetation index of the grid unit; the semantic uncertainty, the average residual amplitude and the vegetation index average variation amplitude are respectively subjected to Z-score standardization processing, and the semantic uncertainty, the average residual amplitude and the vegetation index average variation amplitude after the standardization processing are summed to obtain a false recognition intensity value corresponding to each grid unit; mapping the false recognition intensity values according to the space positions of the grid units to generate a false recognition risk distribution map; and in the false recognition risk distribution diagram, the false recognition intensity values corresponding to all grid units are sorted in descending order according to the numerical value, and the grid units positioned in the preset quantity proportion in the sorting result are selected as a false recognition high risk area set.
  6. 6. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the ecological evolution trend evaluation module specifically comprises: Acquiring carbon sink indexes, surface temperature indexes and soil moisture indexes corresponding to each grid unit in a plurality of time nodes in a target city area; Arranging carbon sink indexes corresponding to the same grid unit at different time nodes according to time sequence to generate a carbon sink index time sequence, arranging earth surface temperature indexes according to time sequence to generate an earth surface temperature index time sequence, and arranging soil moisture indexes according to time sequence to generate a soil moisture index time sequence; Respectively carrying out time sequence modeling on the carbon sink index time sequence, the earth surface temperature index time sequence and the soil moisture index time sequence, wherein the time sequence modeling comprises the steps of carrying out first-order difference operation on each time sequence, and carrying out moving average operation on a difference result to obtain a carbon sink change trend sequence, an earth surface temperature change trend sequence and a soil moisture change trend sequence; respectively carrying out linear fitting on the carbon sink change trend sequence, the surface temperature change trend sequence and the soil moisture change trend sequence by taking a time node as an independent variable and a corresponding change trend value as a dependent variable, and respectively taking a slope value obtained by the linear fitting as a carbon joint performance change trend value, a temperature evolution trend value and a moisture evolution trend value of each grid unit; respectively calculating the average value and standard deviation of the carbon joint performance chemical trend value, the temperature evolution trend value and the moisture evolution trend value in all grid units, and carrying out standardization treatment by adopting a Z-fraction standardization method; Mapping the normalized carbon joint performance trend value, the normalized temperature evolution trend value and the normalized moisture evolution trend value according to the spatial position of the grid unit to generate an ecological evolution trend distribution map.
  7. 7. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the greening land block optimization module specifically comprises: Taking the ecological evolution trend distribution map as optimized input data, and taking each standardized evolution trend value corresponding to each grid unit as a basic evaluation value of the candidate greening land parcels; Marking grid units in the false recognition high risk area set as limited grid units by taking the false recognition high risk area set as a space constraint condition; initializing a beetle individual set based on a grid division result of a target city area, wherein each beetle individual corresponds to a candidate greening block combination, and the candidate greening block combination is formed by a preset number of grid unit indexes; In each iteration, for each beetle individual, taking each grid unit in the current candidate greening plot combination as a search primitive, performing multi-directional antenna detection in the adjacent eight neighborhood directions of the corresponding grid unit, and performing neighborhood replacement operation on a single grid unit index in the current candidate greening plot combination in each direction, wherein the neighborhood replacement operation is limited to select only adjacent grid unit indexes which do not belong to the current candidate greening plot combination for replacement so as to generate a new candidate greening plot combination; Reading a standardized carbon joint performance standardized trend value, a standardized temperature evolution trend value and a standardized moisture evolution trend value corresponding to each grid unit respectively for each candidate greening plot combination, subtracting the standardized temperature evolution trend value from 1 to obtain a temperature reversal value, summing the standardized carbon joint performance standardized trend value, the temperature reversal value and the standardized moisture evolution trend value, and multiplying the summation result by a set punishment coefficient when the candidate greening plot combination comprises a limited grid unit to obtain a corresponding combination evaluation value; Introducing a nerve mimicry synaptic drift mechanism into the improved beetle optimization algorithm, and setting a synaptic weight set corresponding to the multi-directional antenna detection direction for each beetle individual, wherein the synaptic weight set is used for representing the historical selection intensity of the beetle individual in different search directions, and each synaptic weight is set to be the same value during initialization; Based on the nerve mimicry synaptic drift mechanism, carrying out weighted calculation on the combined evaluation value corresponding to each direction and the synaptic weight in the corresponding direction in the synaptic weight set to obtain a direction weighted evaluation value, selecting a candidate greening plot combination with the largest direction weighted evaluation value in the current iteration as an updating position of a beetle individual, and updating the synaptic weight; After each iteration is completed, updating the historical optimal candidate greening plot combination corresponding to the beetle individual when the combined evaluation value of the current beetle individual is greater than the maximum combined evaluation value of the historical record; Repeatedly executing multidirectional antenna detection, synaptic weight updating and historical optimal candidate greening plot combination record updating operation on all beetle individuals until the preset maximum iteration times are reached; and selecting a grid unit index set with the largest combination evaluation value from the historical optimal candidate greening plots corresponding to all beetle individuals as a final greening plot set.
  8. 8. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 7, wherein the updating of synaptic weights is specifically: After the beetle individual completes the position update based on the direction weighted evaluation value, reading a combined evaluation value corresponding to the current position of the beetle individual, and reading a combined evaluation value corresponding to the previous iteration position of the beetle individual; When the combined evaluation value corresponding to the current position is larger than the combined evaluation value corresponding to the previous iteration position, multiplying the synaptic weight corresponding to the current moving direction in the nerve mimicry synaptic drift mechanism by a first updating coefficient to complete the enhancement updating of the synaptic weight; And when the combined evaluation value corresponding to the current position is smaller than or equal to the combined evaluation value corresponding to the previous iteration position, multiplying the synaptic weight corresponding to the current moving direction in the nerve mimicry synaptic drift mechanism by a second updating coefficient to finish the attenuation updating of the corresponding synaptic weight.
  9. 9. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the greenbelt path network construction module specifically comprises: Mapping each grid unit in the final greening land block set into a greening node to form a greening node set; For each greening node, respectively reading a standardized carbon joint performance trend value, a standardized temperature evolution trend value and a standardized moisture evolution trend value of a grid unit corresponding to the greening node, subtracting the standardized temperature evolution trend value from 1 to obtain a temperature reversal value, and summing the standardized carbon joint performance trend value, the temperature reversal value and the standardized moisture evolution trend value to obtain a node potential value of the greening node; calculating the space distance between two greening nodes for any two greening nodes, and dividing the space distance by the sum of node potential values of the two greening nodes to obtain the communication cost value between the two greening nodes; And establishing a communication relation between the greening nodes with the communication cost value smaller than the set cost threshold value, and integrating all the greening nodes and the corresponding communication relation to construct a greenbelt path network structure.
  10. 10. The urban landscaping optimal planning system based on multi-source remote sensing data according to claim 1, wherein the optimal path planning module is specifically: in the green land path network structure, one greening node is selected as a path starting point, a plurality of candidate paths are generated, and each candidate path is represented as a greening node sequence which is arranged according to a communication relation sequence; Counting the number of greening nodes of each candidate path, and only reserving candidate paths with the number of greening nodes being larger than a threshold value of the set minimum number of nodes and smaller than a threshold value of the set maximum number of nodes; Sequentially reading and summing the communication cost values between adjacent greening nodes in each reserved candidate path to obtain a path cost sum; Dividing the path cost sum by the total number of greening nodes of the candidate path to obtain an average path cost value corresponding to the candidate path; and sequencing all the candidate paths according to the corresponding average path cost values, and selecting the candidate path with the minimum average path cost value as the optimal urban landscaping planning path.

Description

Urban landscaping optimal planning system based on multisource remote sensing data Technical Field The invention relates to the technical field of urban planning and artificial intelligence, in particular to an urban landscaping optimal planning system based on multisource remote sensing data. Background Along with the continuous acceleration of the urban process and the continuous promotion of ecological civilization construction requirements, urban landscaping planning plays an increasingly important role in improving urban ecological environment, relieving heat island effect and improving resident life quality. In recent years, the application of remote sensing technology, geographic information system and intelligent optimization algorithm in urban green space planning is paid attention to gradually, and related research attempts provide decision support for urban landscaping layout through remote sensing image interpretation, ecological index evaluation and spatial analysis means. However, the following problems remain common in the practical application process in the prior art: The current urban landscaping planning method mainly relies on single-type or minority-type remote sensing data, is mainly based on optical remote sensing images, is easily influenced by illumination conditions, seasonal changes and urban complex ground object backgrounds, is difficult to stably distinguish real vegetation from artificial ground objects with similar spectral features in a high-density building area, and is unstable in greening recognition results and high in misjudgment rate, the multisource remote sensing data has obvious differences in spatial resolution, time resolution and observation mechanisms, the conventional method generally adopts a simple splicing or independent processing mode, lacks a unified space-time coding mechanism, is difficult to fully mine collaborative information of the multisource remote sensing data in time evolution and space structure layers, limits fine depicting capability of urban greening evolution processes, and is difficult to simultaneously consider ecological potential and recognition risk constraint in a greening planning optimization stage by a traditional site selection method based on rules or static weights, a population intelligent algorithm is generally provided with a problem of depending on instantaneous fitness in a searching process, is easy to sink into local optimum, and a stable and reliable greening path planning method is difficult to obtain in complex urban space. Therefore, how to provide an urban landscaping optimal planning system based on multi-source remote sensing data is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention aims to provide an urban landscaping optimal planning system based on multi-source remote sensing data, and the urban landscaping optimal planning system combines remote sensing intelligent recognition and improved beetle optimization algorithm to realize urban landscaping path optimal planning, utilizes the multi-source remote sensing data to carry out space-time ground object three-dimensional modeling and semantic recognition, constructs a false recognition risk distribution map and an ecological evolution trend distribution map, introduces a nerve mimicry synaptic drift mechanism and a path cost screening strategy, combines ecological potential and space compactness, and has the advantages of strong misjudgment avoidance capability, high optimization efficiency and strong path selection rationality. The urban landscaping optimal planning system based on multi-source remote sensing data, provided by the embodiment of the invention, comprises the following modules: the multi-source remote sensing data processing module is used for acquiring multi-source remote sensing data covering a target city area, and performing space registration processing under a unified coordinate system to generate a multi-source remote sensing time sequence data set; The space-time ground object three-dimensional coding module is used for receiving the multi-source remote sensing time sequence data set, respectively constructing a three-dimensional expression structure for each grid unit in the target city area, and generating vegetation expression data; the space recognition module is used for inputting the vegetation expression data into a space recognition network, carrying out greening type semantic recognition on each grid unit, and generating a semantic result vector and a reconstruction residual vector; The false recognition risk assessment module is used for generating a false recognition risk distribution map based on the semantic result vector and the reconstruction residual vector and combining vegetation index time sequence data corresponding to each grid unit, and determining a false recognition high risk area set; The ecological evolution trend evaluation module is used for acquiring carbon sin