AI-based GIS surveying and mapping data analysis processing system and method

Abstract

The AI-based GIS surveying and mapping data analysis processing system and method of the application relates to the technical field of artificial intelligence, obtains original surveying and mapping image data and pre-processes to obtain surveying and mapping image data; carries out shadow segmentation and binarization on the surveying and mapping image data to obtain a shadow mask image, removes the shadow from the surveying and mapping image data based on the shadow mask image to obtain shadow-free image data; extracts a time-weighted feature map based on the shadow-free image data and a time weight function, converts the time-weighted feature map into a change detection image by presetting a change probability threshold; inputs the surveying and mapping image data into a constructed ground object classification task, outputs a ground object category probability map, optimizes the change detection image based on the ground object category probability map to obtain an optimized change detection image, and improves the geographical change detection precision based on surveying and mapping data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an AI-based GIS surveying and mapping data analysis processing system and method. BACKGROUND

[0002] GIS surveying and mapping data is data about geographic spatial information collected and processed by using geographic information system (GIS) technology. The collection approaches of GIS surveying and mapping data include field measurement, remote sensing, map digitization, etc. These data provide an important basis for geographic spatial change analysis, decision making, etc.

[0003] Chinese patent application with publication number CN111339225A discloses a method and system for dynamic updating of urban geographic spatial data. The method includes discovering a change event, labeling, collecting evidence and reporting the change event; updating management, comparing the change event information, confirming whether to establish an update task, and publishing and tracking the update task; information recycling, receiving the update task published in the updating management step, and collecting, mapping and recycling the update task information that needs to be changed; incremental updating, extracting, updating and publishing the special incremental information of the recycled information.

[0004] Buildings in urban geographic environment are various and dense, and the buildings block each other, with obvious shadow effect. The shadow between ground objects changes with time, and the existing algorithm is difficult to accurately distinguish the shadow change and the actual ground object change, resulting in a decrease in change detection accuracy. SUMMARY

[0005] The present application aims to at least solve one of the technical problems in the related art. To this end, one object of the present application is to provide an AI-based GIS surveying and mapping data analysis processing system and method, which improves the geographic change detection accuracy based on surveying and mapping data.

[0006] One aspect of the present application provides an AI-based GIS surveying and mapping data analysis processing method, comprising:

[0007] Step S100: obtaining original surveying and mapping image data and preprocessing to obtain surveying and mapping image data;

[0008] Step S200: performing shadow segmentation and binarization on the surveying and mapping image data to obtain a shadow mask image, and performing shadow removal on the surveying and mapping image data based on the shadow mask image to obtain shadow-free image data;

[0009] Step S300: extracting a time-weighted feature map based on the shadow-free image data and a time weight function, predefining a change probability threshold, and converting the time-weighted feature map into a change detection image;

[0010] Step S400: input the survey image data into the constructed ground object classification task, output a ground object category probability map, and optimize the change detection image based on the ground object category probability map to obtain an optimized change detection image.

[0011] The specific method of obtaining the original survey image data and preprocessing to obtain the survey image data is:

[0012] Step S110: determine the time range and spatial range of the monitoring area;

[0013] Step S120: obtain the original survey image data within the time range and spatial range, and preprocess it to obtain survey image data I t .

[0014] The specific method of performing shadow segmentation and binarization on the survey image data to obtain a shadow mask map, and removing shadows from the survey image data based on the shadow mask map is:

[0015] Step S210: select a deep learning model as an initial shadow segmentation model, use survey image data as input, and use a shadow probability map as output;

[0016] Step S220: obtain a shadow segmentation training set, the shadow segmentation training set includes survey image data and its corresponding shadow annotation mask, train the shadow segmentation model using a cross-entropy loss function, minimize the cross-entropy loss function between the true value of the shadow annotation mask of a pixel and the shadow probability predicted by the pixel, optimize the model parameters, and when the value of the cross-entropy loss function converges, obtain the trained shadow segmentation model;

[0017] Step S230: use the trained shadow segmentation model to perform shadow segmentation on survey image data I t , and output a shadow probability map P t ;

[0018] Step S240: preset a shadow probability threshold τ, binarize the shadow probability map, determine the pixels with a shadow probability greater than or equal to the shadow probability threshold as shadow pixels, set their shadow mask value to 1, determine the pixels with a shadow probability less than the shadow probability threshold as non-shadow pixels, set their shadow mask value to 0, and obtain a shadow mask map of the shadow probability map The shadow pixels constitute a shadow region.

[0019] Step S250: remove shadows from the shadow region in survey image data I t based on the shadow mask map to obtain shadow-free image data

[0020] The shadow removing of the shadow region in the surveying image data I t is performed according to the shadow mask map, which comprises the following steps: when the shadow mask value of the kth pixel in the shadow probability map is equal to 0, it indicates that the shadow probability of the pixel is less than the shadow probability threshold, so the pixel is judged as not belonging to the shadow region, and the pixel value of the pixel is unchanged; otherwise, the pixel is judged as belonging to the shadow region, and the original pixel value of the shadow region is recovered through the shadow removing.

[0021] The method for obtaining the shadow-free image data comprises the following steps:

[0022] Step S251: performing connected domain labeling on the shadow mask map, marking the mutually connected shadow pixel groups as the same shadow connected domain, and extracting the attribute features of each shadow connected domain S a , including the area A a , the perimeter P a , the shape complexity PAR a , and the image statistical features, including the mean μ a , the variance

[0023] Step S252: according to the attribute features of the shadow connected domain, presetting an area threshold, a complexity threshold and a variance threshold, designing a decision function, selecting a removing strategy for the shadow region according to the decision function, and executing the corresponding removing strategy for each shadow connected domain to obtain the shadow-free image data.

[0024] The decision rule of the decision function is: when the area of the shadow connected domain is less than the area threshold and the shape complexity is less than the complexity threshold, performing shadow correction on the shadow region; when the area of the shadow connected domain is greater than or equal to the area threshold and the variance is less than the variance threshold, performing shadow compensation on the shadow region; otherwise, performing shadow matching on the shadow region.

[0025] The shadow correction specifically refers to: establishing a shadow correction function by using the pixel values inside and outside the shadow region, and mapping the pixel values inside the shadow region to the shadow-free state.

[0026] The calculation formula of the shadow correction function is: wherein, I a k and I a k' are the pixel values of the kth shadow pixel in the a th shadow connected domain before and after the shadow correction, μ a and σ a are the mean and standard deviation of the pixel values of the shadow pixels in the shadow connected domain, μ o and σ o ​is the mean and standard deviation of pixel values of the pixels outside the shadow region, and (k) represents the kth pixel;

[0027] The shadow compensation adopts a Poisson image editing method, estimates the non-shadow pixel values in the shadow region by solving a Poisson equation with the pixels outside the shadow region as boundary conditions, and the calculation formula of the Poisson image editing method is:

[0028]

[0029]

[0030] wherein, is a Laplacian operator, is the boundary of the shadow region;

[0031] The shadow matching adopts a patch-based shadow matching method, and the calculation formula of the patch-based shadow matching method is: wherein, P(k) and P(k') are image patches centered at the kth pixel and the k'th pixel respectively, D(·) is the Euclidean distance between the image patches, and Ω is the image space domain, represents the pixel (k') that makes the objective function take the minimum value, Ω\S a represents the part of the image space domain excluding the a th shadow connected domain;

[0032] The acquisition method of the image patch is P(k) = {I(x+xi, y+yj) | -r≤xi, yj≤r}, wherein x and y represent the coordinate positions of the kth pixel in the surveying image data, I(x+xi, y+yj) represents the pixel value at the coordinate position (x+xi, y+yj) in the surveying image data, xi and yj represent the coordinate offset of the image patch relative to the kth pixel, and r represents the half side length of the image patch.

[0033] The specific method for extracting the time-weighted feature map based on the non-shadow image data and the time weight function is:

[0034] Step S310: constructing a spatio-temporal feature extraction network, taking the non-shadow image data as input and taking the spatio-temporal feature map as output F t (k);

[0035] Step S320: defining a time weight function, adding the time weight function to the spatio-temporal feature extraction network, applying the time weight to the spatio-temporal feature map, and obtaining a time-weighted feature map The time weight function selects an exponential decay function.

[0036] The specific method for converting the time-weighted feature map into a change detection image based on the preset change probability threshold is:

[0037] Step S330: set a change probability threshold δ, and convert the time-weighted feature map into a change region mask B t If the value of the kth pixel in the time-weighted feature map is greater than or equal to the change probability threshold, the corresponding change region mask B t (k) is set to 1, otherwise, the corresponding change region mask B t (k) is set to 0.

[0038] Step S340: for the kth pixel, if the value of the change region mask thereof is 1, it belongs to a change region, and the value of the kth pixel in the time-weighted feature map is assigned to the change probability C t (k) in the change detection image, otherwise, the change probability of the pixel in the change detection image is assigned to 0.

[0039] Step S350: output the change region mask B t and the change detection image C containing the change probability t .

[0040] The specific method of inputting the survey image data into the constructed land feature classification task, outputting a land feature category probability map, and optimizing the change detection image based on the land feature category probability map to obtain an optimized change detection image is as follows:

[0041] Step S410: a multi-task learning network is established, including a change detection task and a land feature classification task, the input of the change detection task is a change detection image, and the output is an optimized change detection image, the input of the land feature classification task is survey image data, and the output is a land feature category probability map;

[0042] The multi-task learning network is composed of a shared feature extraction layer, a task-specific layer, a total loss function, and an optimization algorithm;

[0043] Step S420: the total loss function of the multi-task learning network is obtained by weighted summation of a change detection loss function and a land feature classification loss function, the change detection loss function is defined as a binary cross-entropy loss, and the land feature classification loss function is defined as a cross-entropy loss;

[0044] Step S430: the land feature category probability map output by the multi-task learning network is used to optimize the change detection image, for each change region, the land feature categories therein are counted, the change characteristics of different land feature categories are obtained, and whether the change region is a real change is judged according to the change characteristics, the change probability C t (k) of a false change region is updated to 0 to obtain an optimized change detection image.

[0045] One aspect of the present application provides an AI-based GIS surveying and mapping data analysis and processing system, comprising:

[0046] The surveying data acquisition module is used to acquire raw surveying image data and preprocess it to obtain surveying image data.

[0047] The shadow segmentation and removal module is used to segment and binarize the surveying and mapping image data to obtain a shadow mask map. Based on the shadow mask map, the surveying and mapping image data is then used to remove shadows to obtain shadow-free image data.

[0048] The change detection and conversion module is used to extract time-weighted feature maps based on shadowless image data and a time weighting function, preset change probability thresholds, and convert the time-weighted feature maps into change detection images.

[0049] The change detection optimization module is used to input surveying and mapping image data into the constructed land cover classification task, output a land cover category probability map, and optimize the change detection image based on the land cover category probability map to obtain an optimized change detection image.

[0050] The AI-based GIS surveying and mapping data analysis and processing system and method proposed in this application have the following advantages over existing technologies:

[0051] This application employs a deep learning model for shadow segmentation, automatically learning and extracting features of shadow regions to improve the accuracy of shadow segmentation. It adaptively selects different shadow removal strategies, adopts the optimal removal method for different types of shadow regions, improves the effect of shadow removal, effectively restores the original pixel values ​​of shadow regions, obtains high-quality shadow-free image data, suppresses the interference of shadow regions in change detection, and improves detection accuracy.

[0052] This application constructs a spatiotemporal feature extraction network, takes shadowless image data as input, extracts spatiotemporal feature maps, captures the change information of surveying and mapping data in the time dimension, and assigns different weights to feature maps at different time points to highlight the importance of recent changes, suppress the interference of long-term changes, improve the time sensitivity of change detection, and effectively integrates the spatiotemporal features of multi-temporal image data through a time weighting mechanism, thereby enhancing the ability to capture changes in actual ground features.

[0053] This application establishes a multi-task learning network to simultaneously perform change detection and land cover classification tasks. It fully utilizes the complementary information between the two tasks to improve the accuracy of change detection, optimizes the change detection results using land cover classification results, and judges the authenticity of change areas based on the change characteristics of different land cover categories, effectively removing false changes and improving the reliability of change detection. Attached Figure Description

[0054] Figure 1 A flowchart illustrating the AI-based GIS surveying and mapping data analysis and processing method provided in this application;

[0055] Figure 2 Flowchart of the method for acquiring shadowless image data provided in this application;

[0056] Figure 3 Flowchart of the method for acquiring change detection images provided in this application;

[0057] Figure 4 Functional module diagram of the AI-based GIS surveying and mapping data analysis and processing system provided in this application. Detailed Implementation

[0058] To better understand this application, various aspects of this application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely illustrative of exemplary embodiments of this application and are not intended to limit the scope of this application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0059] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The drawings are for illustrative purposes only and are not strictly to scale. As used herein, the terms “approximately,” “about,” and similar terms are used to indicate approximation, not degree, and are intended to illustrate inherent deviations in measured or calculated values ​​that will be recognized by one of ordinary skill in the art. Furthermore, the order in which the steps are described in this application does not necessarily indicate the order in which these steps occur in actual operation, unless otherwise expressly defined or deduced from the context.

[0060] It should also be understood that expressions such as "comprising," "including," "having," "containing," and / or "comprising" are open-ended rather than closed-ended expressions in this specification, indicating the presence of the stated features, elements, and / or components, but not excluding the presence of one or more other features, elements, components, and / or combinations thereof. Furthermore, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features, not just individual elements in the list. Additionally, when describing embodiments of this application, the word "may" is used to mean "one or more embodiments of this application." And the term "exemplary" is intended to refer to examples or illustrations.

[0061] Unless otherwise specified, all terms used herein (including engineering and technical terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that, unless expressly stated herein, terms defined in common dictionaries shall be interpreted as having the meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or overly formalized meaning.

[0062] It should be noted that, where there is no conflict, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0063] Example 1

[0064] like Figure 1 As shown, the AI-based GIS surveying and mapping data analysis and processing method provided in this application includes:

[0065] Step S100: Acquire raw surveying image data and preprocess it to obtain surveying image data;

[0066] The specific method for obtaining raw surveying and mapping image data and preprocessing it to obtain surveying and mapping image data is as follows:

[0067] Step S110: Determine the time and spatial range of the monitoring area;

[0068] Step S120: Acquire the original mapping image data within the time and spatial range, and preprocess it to obtain mapping image data.

[0069] The original mapping image data comes from different data sources, including remote sensing imagery, aerial imagery, and ground survey data; the mapping image data is I t Where t is the timestamp;

[0070] The preprocessing includes: performing geometric correction and spatial registration on the original surveying and mapping image data to ensure consistent spatial reference and pixel alignment; converting the data format of the original surveying and mapping image data into a standard format; and unifying the coordinate systems of different data sources into the target coordinate system.

[0071] The above step S100 acquires the original surveying and mapping image data, providing a data foundation for subsequent tasks such as shadow removal and change detection.

[0072] Step S200: Perform shadow segmentation and binarization on the surveying image data to obtain a shadow mask map. Based on the shadow mask map, remove shadows from the surveying image data to obtain shadow-free image data.

[0073] like Figure 2The diagram shown is a flowchart of the method for acquiring shadowless image data provided in this application.

[0074] The specific method for performing shadow segmentation and binarization on the surveying image data to obtain a shadow mask map, and then removing shadows from the surveying image data based on the shadow mask map to obtain shadow-free image data is as follows:

[0075] Step S210: Select a deep learning model as the initial shadow segmentation model, with the mapping image data as input and the shadow probability map as output;

[0076] The shadow probability map refers to an image output by the shadow segmentation model that is the same size as the input mapping image data, where the value of each pixel represents the probability that the pixel belongs to the shadow.

[0077] Step S220: Obtain a shadow segmentation training set, which includes survey image data and its corresponding shadow annotation mask. Use the cross-entropy loss function to train the shadow segmentation model, minimize the cross-entropy loss function between the true value of the shadow annotation mask of a pixel and the predicted shadow probability of the pixel, optimize the model parameters, and obtain the trained shadow segmentation model when the value of the cross-entropy loss function converges.

[0078] The formula for calculating the cross-entropy loss function is as follows: Among them, L s Let K be the cross-entropy loss function, K be the total number of pixels, k be the pixel index, and M be the number of pixels. t (k) represents the true value of the shadow annotation mask for the k-th pixel, P t (k) represents the predicted shadow probability of the k-th pixel, and log(·) is the logarithmic function;

[0079] Step S230: Use the trained shadow segmentation model to process the survey image data I t Perform shadow segmentation and output the shadow probability map P. t ;

[0080] Step S240: Set a preset shadow probability threshold τ, binarize the shadow probability map, determine pixels with shadow probabilities greater than or equal to the shadow probability threshold as shadow pixels, and set their shadow mask values. A value equal to 1 indicates that pixels with a shadow probability less than the shadow probability threshold are classified as non-shadow pixels, and their shadow mask values ​​are set accordingly. When the value equals 0, the shadow mask map of the shadow probability map is obtained. The shadow pixels constitute the shadow area;

[0081] The formula for calculating the shadow mask value is:

[0082] The value of the shadow probability threshold is set by those skilled in the art based on experience.

[0083] Step S250: Apply the shadow mask to the survey image data I t Shadow removal is performed on the shadowed areas to obtain shadow-free image data.

[0084] The pixel values ​​of the shadowless image data are represented as follows: in, Let φ(·) be the pixel value of the k-th pixel in the shadowless image data after shadow removal, and let I be the shadow removal function. t (k) represents the pixel value of the k-th pixel in the mapping image data;

[0085] The mapping image data I based on the shadow mask. t Shadow removal in the shadow region refers to: when the shadow mask value of the k-th pixel in the shadow probability map... When the value is 0, it means that the shadow probability of the pixel is less than the shadow probability threshold. In this case, the pixel is judged not to belong to the shadow area, so its pixel value remains unchanged. Otherwise, the pixel is judged to belong to the shadow area, and the original pixel value of the shadow area is restored by shadow removal.

[0086] The method for acquiring the shadowless image data is as follows:

[0087] Step S251: Perform connected component labeling on the shadow mask image, marking interconnected groups of shadow pixels as the same shadow connected component, and for each shadow connected component S... a Extract attribute features, including the area A of the shaded connected components. a Perimeter P a Shape complexity PAR a and image statistical features, the image statistical features including the mean μ of the pixel values ​​of shadow pixels within the shadow connected region. a ,variance

[0088] The area of ​​the shaded connected region is equal to the number of pixels in the shaded connected region, A a Let represent the area of ​​the a-th shaded connected region;

[0089] The perimeter of the shaded connected region is the number of boundary pixels of the shaded connected region, P a Let represent the perimeter of the a-th shaded connected region;

[0090] The formula for calculating the shape complexity of the shaded connected component is: Among them, PAR a Indicate the shape complexity of the a-th shaded connected component;

[0091] The method for marking shaded connected components employs a two-pass scanning algorithm, including a first scan and a second scan. The specific method for the first scan is as follows: initialize the shaded mask, mark the label values ​​of all pixels as 0, traverse the shaded mask, and for each shaded pixel, check the label values ​​of its upper and left neighboring pixels. If the neighboring pixels are already marked shaded pixels, then mark the current shaded pixel with an equivalent label value; if there are multiple different label values ​​among the neighboring pixels, then record these label values ​​as the same label value; if there are no already marked shaded pixels among the neighboring pixels, then mark the current shaded pixel with a new label value.

[0092] The second scan method is as follows: traverse the shadow mask map, and for each non-zero label value, replace it with the smallest equivalent label value according to the record of its equivalent label value.

[0093] Different label values ​​in the shaded mask diagram after the connected components are labeled correspond to different shaded connected components;

[0094] Step S252: Based on the attribute characteristics of the shadow connected components, preset the area threshold, complexity threshold and variance threshold, design the decision function, select the shadow removal strategy according to the decision function, and execute the corresponding removal strategy for each shadow connected component to obtain shadow-free image data;

[0095] The decision rule of the decision function is as follows: when the area of ​​the shaded connected region is less than the area threshold and the shape complexity is less than the complexity threshold, the shaded region is shaded and corrected; when the area of ​​the shaded connected region is greater than or equal to the area threshold and the variance is less than the variance threshold, the shaded region is shaded and compensated; otherwise, the shaded region is shaded and matched.

[0096] The values ​​of the area threshold, complexity threshold, and variance threshold are set by those skilled in the art based on experience.

[0097] The shadow correction specifically refers to: using the radiation information inside and outside the shadow area to establish a shadow correction function, and mapping the pixel values ​​in the shadow area to a shadow-free state;

[0098] The formula for calculating the shadow correction function is as follows: in, and Let μ be the pixel value of the k-th shadow pixel in the a-th shadow connected component before and after shadow correction. a and σ a Let μ be the mean and standard deviation of the pixel values ​​of the shadow pixels within the shadow connected region. o and σ oLet (k) represent the mean and standard deviation of the pixel values ​​of the neighboring pixels outside the shaded area, and (k) represent the k-th pixel.

[0099] The shadow compensation specifically refers to: using pixel information outside the shadow area to estimate the value of the unshadowed pixels within the shadow area; specifically, the estimation of the unshadowed pixel value within the shadow area can be achieved through interpolation or padding methods, for example, using the Poisson image editing method, by solving the Poisson equation, using pixels outside the shadow area as boundary conditions to estimate the value of the unshadowed pixels within the shadow area;

[0100] The calculation formula for the Poisson image editing method is as follows:

[0101]

[0102] in, For the Laplace operator, The boundary of the shaded area;

[0103] The shadow matching specifically refers to: finding a matching region outside the shadow region that is similar to the image information inside the shadow region, and filling the shadow region with the pixel values ​​of the matching region by copying or compositing, for example, using a patch-based shadow matching method;

[0104] The calculation formula for the patch-based shadow matching method is as follows: Where P(k) and P(k') are image patches centered at the k-th and k'-th pixels, respectively, D(·) is the Euclidean distance between image patches, and Ω is the image spatial domain. Ω represents the pixel (k') that minimizes the objective function. a This represents the portion of the image spatial domain after removing the a-th shaded connected component;

[0105] The image patch refers to a local influence area centered on a pixel; the method for obtaining the image patch is: P(k)={I(x+xi,y+yj)|-r≤xi,yj≤r}, where x and y represent the coordinate positions of the k-th pixel in the mapping image data, I(x+xi,y+yj) represents the pixel value at coordinate position (x+xi,y+yj) in the mapping image data, xi and yj represent the coordinate offset of the image patch relative to the k-th pixel, and r represents the half-side length of the image patch;

[0106] The half-length of the image patch is set by those skilled in the art based on experience.

[0107] The image spatial domain refers to the entire image of the surveying and mapping image data. In digital image processing, the image is regarded as a two-dimensional function, and the spatial coordinates of the pixels are the domain of the two-dimensional function, which is the image spatial domain.

[0108] In mapping image data, there may be multiple disconnected shadow regions, which may differ in size, shape, texture and other characteristics. Dividing the shadow regions into shadow connected regions can distinguish these different shadow regions, which is convenient for subsequent analysis and processing.

[0109] For each shaded connected component, its attribute features are extracted and input into a decision function. Based on these features, the decision function selects the optimal shadow removal method for each shaded connected component. This connected component-based adaptive shadow removal strategy selects the optimal removal method for different shaded regions, improving the effectiveness and accuracy of shadow removal. Compared to processing the entire shaded region as a whole, the connected component-based approach better adapts to the diversity and complexity of shaded regions in mapping image data, achieving more intelligent and efficient shadow removal.

[0110] Step S200 above achieves accurate shadow segmentation through a deep learning model and employs an adaptive shadow removal strategy to effectively remove shadow effects and obtain high-quality shadow-free image data. This step resolves the interference caused by shadow effects and lays the foundation for accurately distinguishing shadow changes from actual ground feature changes.

[0111] Step S300: Based on the shadowless image data and the time weighting function, extract the time-weighted feature map, preset the change probability threshold, and convert the time-weighted feature map into a change detection image;

[0112] The specific method for extracting time-weighted feature maps based on shadowless image data and a time-weighted function is as follows:

[0113] Step S310: Construct a spatiotemporal feature extraction network, taking the shadowless image data as input and the spatiotemporal feature map as output F. t (k);

[0114] The spatiotemporal feature extraction network is ConvLSTM, which combines a CNN model and an LSTM model. The CNN model is responsible for extracting the spatial features of the shadowless image data, while the LSTM model is responsible for capturing the time series variation features. The input is multi-temporal shadowless image data, and the output is a spatiotemporal feature map through the forward propagation process of the spatiotemporal feature extraction network.

[0115] The construction process of the spatiotemporal feature extraction network is as follows:

[0116] The spatiotemporal feature extraction network selects the ConvLSTM model, which includes an input layer, a ConvLSTM layer, a pooling layer, and a fully connected layer.

[0117] The input layer is used to receive multi-temporal shadowless image data;

[0118] The ConvLSTM layer consists of n ConvLSTM units. The input of the ConvLSTM layer is the shadowless image data at the current timetamp t and the hidden state at the previous timetamp. The output is the hidden state at the current timetamp. The hidden state contains the change information of the shadowless image data in the time dimension.

[0119] The pooling layer spatially downsamples the output of the ConvLSTM layer;

[0120] The fully connected layer transforms the output of the pooling layer to generate the final spatiotemporal feature map.

[0121] Multi-temporal shadowless imagery data was collected as training data. Variation regions were labeled within the training data, generating variation region masks as training labels. A binary cross-entropy loss function was used to measure the difference between the predicted variation probability and the actual variation region mask training labels. The multi-temporal shadowless imagery data was input into a ConvLSTM model. Forward propagation was used to calculate the output spatiotemporal feature map, and the binary cross-entropy loss function was also calculated. The model parameters were updated using a backpropagation algorithm. Training was complete when the value of the binary cross-entropy loss function converged. The trained ConvLSTM model was then used to extract spatiotemporal feature maps from the shadowless imagery data.

[0122] Step S320: Define a time weighting function, add the time weighting function to the spatiotemporal feature extraction network, apply the time weights to the spatiotemporal feature map, and obtain the time-weighted feature map. The time weighting function is selected as an exponential decay function;

[0123] The formula for calculating the time weight is: w(t)=exp(-λ×(Tt)); where λ is the decay factor, T is the latest timestamp, and w(t) is the time weight;

[0124] The time weighting function is used to measure the importance of timestamp t for change detection. The further away a time point is from the latest timestamp, the smaller its weight. Integrating time weight information into the spatiotemporal feature map makes the feature contributions of different time points different.

[0125] The decay factor is used to control the decay rate of the time weight and is set by those skilled in the art according to actual needs.

[0126] The formula for calculating the time-weighted feature map is: Among them, F t (k) represents the value of the spatiotemporal feature map at the k-th pixel under timestamp t. This represents the value of the time-weighted feature map at the k-th pixel under timestamp t;

[0127] like Figure 3 The diagram shown is a flowchart of the method for acquiring change detection images provided in this application.

[0128] The specific method for converting the time-weighted feature map into a change detection image using the preset change probability threshold is as follows:

[0129] Step S330: Set the change probability threshold δ and convert the time-weighted feature map into a change region mask B. t If the value of the k-th pixel in the time-weighted feature map is greater than or equal to the change probability threshold, then the corresponding change region mask is set to 1; otherwise, the corresponding change region mask is set to 0.

[0130] The formula for calculating the change region mask is: Among them, B t (k) represents the value of the mask for the k-th pixel in the changed region at timestamp t;

[0131] The value of the change probability threshold is set by those skilled in the art based on experience.

[0132] When the value of a pixel in the time-weighted feature map is greater than or equal to the change probability threshold, it indicates that the pixel belongs to the change region; otherwise, it indicates that it does not belong to the change region.

[0133] Step S340: For the k-th pixel, if the value of its change region mask is 1, it belongs to the change region. Assign its value in the time-weighted feature map to the change probability in the change detection image. Otherwise, assign the change probability of the pixel in the change detection image to 0.

[0134] The calculation formula for the change-detected image is: Among them, C t (k) represents the probability of change of the k-th pixel in the image at time stamp t;

[0135] In change detection images, only pixels belonging to the changed area retain their change probability value, while the change probability value of other pixels is set to 0 to highlight the changed area, while retaining the change probability information for subsequent analysis.

[0136] Furthermore, the contour and area of ​​the changed region are extracted based on the changed region mask. The contour information can be used to calculate the geometric features of the changed region, and the area information can be used to evaluate the significance of the change, eliminate small noise and false changes, and improve the reliability of change detection. The extraction of the contour and area of ​​the changed region based on the changed region mask first involves morphological processing of the changed region mask, such as dilation and erosion. The dilation operation can expand the changed region and fill small holes; the erosion operation can shrink the changed region and eliminate small noise. Then, the contour and area of ​​the changed region are extracted. The morphological processing can be implemented using image processing libraries such as OpenCV.

[0137] Step S350: Output the change region mask B t and change detection images containing change probability C t .

[0138] Step S300 above, through a spatiotemporal feature extraction network and a time-weighted mechanism, fuses the spatiotemporal features of multi-temporal image data, enhancing the change detection model's ability to capture changes in actual ground features. This step effectively addresses the problem of shadows between ground features changing over time, improving the temporal sensitivity and accuracy of change detection.

[0139] Step S400: Input the surveying and mapping image data into the constructed land cover classification task, output the land cover category probability map, optimize the change detection image based on the land cover category probability map, and obtain the optimized change detection image;

[0140] The specific method for inputting surveying and mapping image data into a constructed land cover classification task, outputting a land cover category probability map, and optimizing the change detection image based on the land cover category probability map to obtain an optimized change detection image is as follows:

[0141] Step S410: Establish a multi-task learning network, including a change detection task and a land cover classification task. The input of the change detection task is a change detection image, and the output is an optimized change detection image. The input of the land cover classification task is mapping image data, and the output is a land cover category probability map. The multi-task learning network consists of a shared feature extraction layer, a task-specific layer, a total loss function, and an optimization algorithm.

[0142] The shared feature extraction layer is responsible for extracting shared feature representations from the input change detection imagery and mapping imagery data. This shared feature extraction layer can employ a convolutional neural network. The task-specific layer is designed independently for both the change detection and land cover classification tasks. The task-specific layer for each task performs corresponding transformations and predictions on the shared feature representations output by the shared feature extraction layer. For the change detection task, the task-specific layer includes convolutional layers and fully connected layers to convert the shared feature representations into optimized change detection images. For the land cover classification task, the task-specific layer includes convolutional layers, fully connected layers, and a softmax activation function to convert the shared feature representations into land cover category probability maps. The loss functions for the change detection and land cover classification tasks are binary cross-entropy loss and cross-entropy loss, respectively. The total loss function of the multi-task learning network is obtained by weighted summation of the loss functions for the change detection and land cover classification tasks. The multi-task learning network minimizes the loss function through an optimization function to update its model parameters. The optimization algorithm used is the Adam optimization algorithm.

[0143] Step S420: The total loss function of the multi-task learning network is obtained by weighted summation of the change detection loss function and the land cover classification loss function. The change detection loss function is defined as the binary cross-entropy loss, and the land cover classification loss function is defined as the cross-entropy loss.

[0144] The total loss function is: L = L b +α×L d Where α is a balance factor used to control the importance of change detection and land cover classification tasks, and L b For change detection loss function, L d The loss function for land cover classification;

[0145] The formula for calculating the change detection loss function is as follows: Where K is the total number of pixels, G t (k) represents the actual change detection label. The probability of change detection is predicted for the change detection task.

[0146] The formula for calculating the land cover classification loss function is as follows: Where J is the number of land cover categories, Y t (k,j) represents the true label of the k-th pixel in the survey image data at timestamp t, belonging to land cover category j, and takes a value of 0 or 1. t (k,j) represents the predicted probability value of the k-th pixel of the survey image data belonging to the land cover category j at timestamp t, and the value range is [0,1].

[0147] Step S430: The change detection image is optimized using the land cover category probability map output by the multi-task learning network. For each change area, the land cover categories are statistically analyzed to obtain the change characteristics of different land cover categories. Based on the change characteristics, it is determined whether the change area is a true change, and the change probability C of false change areas is set. t (k) is updated to 0, resulting in the optimized change detection image.

[0148] For example, regarding changes in vegetation areas, whether a change represents a true change is determined by the magnitude of the change in the vegetation index. The formula for calculating the magnitude of the change in the vegetation index is as follows: Where t1 and t2 are the timestamps of the change detection images before and after the change, respectively. These are the vegetation indices before and after the change; if the change in the vegetation index is less than the preset change threshold, then the corresponding change area is a false change, and the change probability of the corresponding change area on the change detection image needs to be updated to 0.

[0149] The above steps optimize and improve change detection results using land cover classification information. Land cover classification information can help identify and filter out genuine change areas, eliminate false change areas, and improve the accuracy and reliability of change detection.

[0150] Example 2

[0151] like Figure 4 As shown, the AI-based GIS surveying and mapping data analysis and processing system provided in this application includes:

[0152] The surveying data acquisition module is used to acquire raw surveying image data and preprocess it to obtain surveying image data.

[0153] The shadow segmentation and removal module is used to segment and binarize the surveying and mapping image data to obtain a shadow mask map. Based on the shadow mask map, the surveying and mapping image data is then used to remove shadows to obtain shadow-free image data.

[0154] The change detection and conversion module is used to extract time-weighted feature maps based on shadowless image data and a time weighting function, preset change probability thresholds, and convert the time-weighted feature maps into change detection images.

[0155] The change detection optimization module is used to input surveying and mapping image data into the constructed land cover classification task, output a land cover category probability map, and optimize the change detection image based on the land cover category probability map to obtain an optimized change detection image.

[0156] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0157] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An AI-based GIS surveying and mapping data analysis and processing method, characterized in that, The method comprises the following steps: obtain original survey image data and preprocess the survey image data to obtain survey image data; perform shadow segmentation and binarization on the survey image data to obtain a shadow mask image, remove shadows from the survey image data based on the shadow mask image, and obtain shadow-free image data; extract a time-weighted feature map based on the shadow-free image data and a time weight function, preset a change probability threshold, and convert the time-weighted feature map into a change detection image; input the survey image data into a constructed land feature classification task, output a land feature category probability map, optimize the change detection image based on the land feature category probability map, and obtain an optimized change detection image; the specific method for performing shadow segmentation and binarization on the survey image data to obtain a shadow mask image, removing shadows from the survey image data based on the shadow mask image, and obtaining shadow-free image data is as follows: select a deep learning model as an initial shadow segmentation model, use the survey image data as input, and use a shadow probability map as output; obtain a shadow segmentation training set, the shadow segmentation training set comprises survey image data and a corresponding shadow annotation mask, train the shadow segmentation model using a cross-entropy loss function, minimize the cross-entropy loss function between the true value of the shadow annotation mask of a pixel and the shadow probability predicted by the pixel, optimize the model parameters, and obtain a trained shadow segmentation model when the value of the cross-entropy loss function converges. Utilizing the trained shadow segmentation model on the mapping image data performing shadow segmentation, outputting a shadow probability map ; The preset shadow probability threshold τ is used to binarize the shadow probability map, pixels with a shadow probability greater than or equal to the shadow probability threshold are determined as shadow pixels, and the shadow mask value of the shadow pixels is set to 1; pixels with a shadow probability less than the shadow probability threshold are determined as non-shadow pixels, and the shadow mask value of the non-shadow pixels is set to 0, to obtain a shadow mask map of the shadow probability map , wherein the shadow pixels constitute a shadow region. , wherein the shadow pixels constitute a shadow region.​ Based on the shadow mask map, the survey image data Shadow removal is performed on the shadowed areas to obtain shadow-free image data. . 2.The AI-based GIS surveying and mapping data analysis processing method of claim 1, wherein, The mapping image data based on the shadow mask Shadow removal in the shadow region refers to: when the shadow mask value of the k-th pixel in the shadow probability map... When the value is 0, it means that the shadow probability of the pixel is less than the shadow probability threshold. Therefore, the pixel is judged not to belong to the shadow area, and its pixel value remains unchanged. Otherwise, the pixel is judged to belong to a shadow area, and the original pixel value of the shadow area is restored through shadow removal. 3.The AI-based GIS surveying and mapping data analysis processing method of claim 2, wherein, The method for obtaining shadow-free image data is as follows: The shadow mask image is labeled by connected domain, and the mutually connected shadow pixel groups are labeled as the same shadow connected domain. Attribute features are extracted, including area of the shadow connected domain , perimeter , shape complexity , and image statistical features including mean value of pixel values of the shadow pixels in the shadow connected domain , variance ; According to the attribute characteristics of the shadow connected domain, preset an area threshold, a complexity threshold and a variance threshold, design a decision function, select a removal strategy for the shadow area according to the decision function, and execute the corresponding removal strategy on each shadow connected domain to obtain shadow-free image data. 4.The AI-based GIS surveying and mapping data analysis processing method of claim 3, wherein, The decision rule of the decision function is as follows: when the area of the shadow connected domain is less than the area threshold and the shape complexity is less than the complexity threshold, the shadow area is corrected; when the area of the shadow connected domain is greater than or equal to the area threshold and the variance is less than the variance threshold, the shadow area is compensated; otherwise, the shadow area is matched. 5.The AI-based GIS surveying and mapping data analysis processing method of claim 4, wherein, The specific method for extracting a time-weighted feature map based on the shadow-free image data and a time weight function is as follows: Constructing a spatiotemporal feature extraction network, taking non-shaded image data as input and spatiotemporal feature map as output ; The time weight function is defined, a time weight function is added in the space-time feature extraction network, and the time weight is applied to the space-time feature map to obtain a time-weighted feature map The time weight function is an exponential decay function. 6.The AI-based GIS surveying and mapping data analysis processing method of claim 5, wherein, The specific method for presetting a change probability threshold and converting the time-weighted feature map into a change detection image is as follows: Set a change probability threshold δ, convert the time-weighted feature map into a change region mask When the value of the kth pixel in the time-weighted feature map is greater than or equal to the change probability threshold, set the corresponding change region mask to 1, otherwise, set the corresponding change region mask to 0; For the kthpixel, if the value of the change region mask thereof is 1, the pixel belongs to the change region, and a value of the pixel in the time-weighted feature map is assigned to a change probability of the pixel in the change detection image , otherwise, the change probability of the pixel in the change detection image is assigned as 0. Output change region mask And change detection imagery with change probability . 7.The AI-based GIS surveying and mapping data analysis processing method of claim 6, wherein, The specific method for inputting the survey image data into a constructed land feature classification task, outputting a land feature category probability map, optimizing the change detection image based on the land feature category probability map, and obtaining an optimized change detection image is as follows: establish a multi-task learning network comprising a change detection task and a land feature classification task, the input of the change detection task is a change detection image, and the output is an optimized change detection image, the input of the land feature classification task is survey image data, and the output is a land feature category probability map; the total loss function of the multi-task learning network is obtained by weighted summation of a change detection loss function and a land feature classification loss function, the change detection loss function is defined as a binary cross-entropy loss, and the land feature classification loss function is defined as a cross-entropy loss. The feature category probability map output by the multi-task learning network optimizes the change detection image, for each change region, the feature categories therein are counted, the change characteristics of different feature categories are obtained, and whether the change region is a real change is judged according to the change characteristics, and the change probability of the false change region is updated to 0, and an optimized change detection image is obtained.

8. The AI-based GIS surveying and mapping data analysis processing system for implementing the AI-based GIS surveying and mapping data analysis processing method according to any one of claims 1-7, characterized in that, ​ The survey data acquisition module is configured to acquire original survey image data and preprocess the original survey image data to obtain survey image data. The shadow segmentation removal module is configured to perform shadow segmentation and binarization on the survey image data to obtain a shadow mask image, perform shadow removal on the survey image data based on the shadow mask image, and obtain shadow-free image data. The change detection conversion module is configured to extract a time-weighted feature map based on the shadow-free image data and a time weight function, set a change probability threshold, and convert the time-weighted feature map into a change detection image. The change detection optimization module is configured to input the survey image data into a constructed land feature classification task, output a land feature category probability map, optimize the change detection image based on the land feature category probability map, and obtain an optimized change detection image.

Images