Meta-Learning-Based Settlement Site Selection Environment Pattern Identification Method, System, Device and Medium

Abstract

The present invention discloses a method, system, device and medium for identifying the settlement site selection environment pattern based on meta-learning, including: obtaining the remote sensing image and DEM data of the settlement to be identified; inputting the remote sensing image and DEM data of the settlement to be identified into the trained settlement ground object segmentation model to obtain the environmental ground object distribution map of the settlement; inputting the environmental ground object distribution map of the settlement into the trained settlement site selection environment pattern identification model to obtain the type of the site selection environment pattern of the settlement. The method, system, device and storage medium can automatically identify the site selection environment pattern of traditional settlements under the condition of only providing a small number of settlement samples.

Description

Technical Field

[0001] The present invention belongs to the field of machine vision, and relates to a method, system, device and medium for identifying settlement site selection environment patterns based on meta-learning. Background Art

[0002] Traditional villages are the historical products of Chinese agricultural culture. They are production and living residences built by people according to certain organizational rules in specific natural conditions. Their morphological characteristics such as scale, layout, texture, and structure are the manifestations of specific natural laws and economic cultures in the physical space, and also the concentrated carriers of regional landscape and cultural context values. Traditional villages formed in different historical periods and specific geographical conditions reflect social factors such as economy, politics, and culture in their development process. However, in the current wave of urbanization and industrialization, profound changes and reconstructions have occurred in the settlement economic structure, social lifestyle, and environmental pattern. With the intervention of human manipulation such as new village construction and rural tourism, the slow, organic, bottom-up growth process and renewal process of traditional villages have been disrupted, bringing development opportunities while inevitably causing many problems such as settlement environment change, ecological imbalance, and cultural context fragmentation. There are a large number of traditional villages in China, which are widely and scattered, showing complex and diverse style types, facing different protection and development problems. However, the current village construction practice often leads to problems such as the same appearance of thousands of villages and the disappearance of characteristics due to simple and extensive management models. How to scientifically and efficiently address these problems so that thousands of traditional villages can inherit regional landscapes and characteristic cultural contexts and achieve sustainable development in the modernization process is the core topic of rural construction at home and abroad. At present, "continuing the texture and protecting the original environmental pattern" is the consensus principle and key strategy in the field of settlement planning and design. Planners and designers must more scientifically and quantitatively understand the environmental characteristics of settlement space based on different regional characteristics, analyze the development laws of different settlement environmental patterns, and provide more efficient and intelligent analysis tools for rural planning practice.

[0003] The research on the site selection environment pattern of traditional settlements has been carried out for a long time in disciplines such as architecture, urban and rural planning, and geography. The classic research method is qualitative summary and inductive description of phenomena, taking the environmental features around the settlement as the main research object, and combining internal and external factors such as the terrain conditions and cultural connotations of the settlement to qualitatively describe and summarize the settlement environmental pattern and give artificial classification results. However, there are a large number of settlements in China, which are widely and scattered. Settlements in different regional environments have complex and changeable environmental landscapes and site selection patterns. Using the above methods to carry out investigation and research requires a lot of time and expert knowledge, and the research results have certain regional limitations and are difficult to meet the construction planning requirements across the country. Therefore, exploring new technologies and methods to achieve intelligent identification of settlement site selection environment patterns in a wider range and more categories is an innovative research challenge in this field. Summary of the Invention

[0004] The object of the present invention is to overcome the above-mentioned shortcomings of the prior art, and provide a method, system, device and medium for identifying the settlement location environment pattern based on meta-learning. The method, system, device and medium can automatically identify the settlement location environment pattern of traditional settlements under the condition of only providing a small number of settlement samples.

[0005] To achieve the above object, the method for identifying the settlement location environment pattern based on meta-learning described in the present invention includes:

[0006] Obtain the remote sensing image and DEM data of the settlement to be identified;

[0007] Input the remote sensing image and DEM data of the settlement to be identified into the trained settlement ground object segmentation model to obtain the environmental ground object distribution map of the settlement;

[0008] Input the environmental ground object distribution map of the settlement into the trained settlement location environment pattern identification model to obtain the type of the settlement location environment pattern.

[0009] It also includes:

[0010] Construct a settlement environment pattern data set;

[0011] Construct a settlement ground object segmentation model;

[0012] Construct a settlement location environment pattern identification model;

[0013] Use the settlement environment pattern data set to train the settlement ground object segmentation model to obtain the trained settlement ground object segmentation model;

[0014] Use the settlement environment pattern data set to train the settlement location environment pattern identification model to obtain the trained settlement location environment pattern identification model.

[0015] The specific process of constructing the settlement environment pattern data set is as follows:

[0016] Use satellite maps to download a certain number of settlement remote sensing images R = {R1, R2,..., R i ,...,} and digital elevation models D = {D1, D2,..., D i ,...} across the country, manually mark the settlement location environment pattern categories and the surrounding mountains, water systems, forests, farmlands, vegetation and wastelands to obtain the environmental ground object classification mark map G and the settlement environment pattern category label L, and then construct the settlement environment pattern data set through the remote sensing image R, elevation model D, environmental ground object classification mark map G and settlement environment pattern category label L.

[0017] The settlement ground object segmentation model is constructed based on the semantic segmentation method;

[0018] The settlement site selection environmental pattern identification model is constructed based on the meta - learning method.

[0019] The settlement ground object segmentation model is based on Deeplab - v3 +. Among them, a bypass E is additionally designed in the feature extraction part of the backbone network θ , to support the semantic learning of terrain features. Finally, the ground objects and terrain features are fused and sent into the classifier for the segmentation of the settlement and its surrounding environmental ground objects. Among them, Deeplab - v3 + is a semantic segmentation network model.

[0020] The settlement site selection environmental pattern identification model consists of an encoder f composed of a ResNet12 θ and a meta - learning classifier . And the training process of the settlement site selection environmental pattern identification model is divided into a full - classification training stage and a meta - training stage. Among them, the encoder f θ is an encoder composed of ResNet12.

[0021] The settlement site selection environmental pattern identification system described in the present invention includes:

[0022] An acquisition module, used to acquire the remote - sensing image and DEM data of the settlement to be identified;

[0023] A first calculation module, used to input the remote - sensing image and DEM data of the settlement to be identified into the trained settlement ground object segmentation model to obtain the environmental ground object distribution map of the settlement;

[0024] A second calculation module, used to input the environmental ground object distribution map of the settlement into the trained settlement site selection environmental pattern identification model to obtain the type of the settlement site selection environmental pattern.

[0025] It further includes:

[0026] A first construction module, used to construct a settlement environmental pattern data set;

[0027] A second construction module, used to construct a settlement ground object segmentation model;

[0028] A third construction module, used to construct a settlement site selection environmental pattern identification model;

[0029] A first training module, used to train the settlement ground object segmentation model by using the settlement environmental pattern data set to obtain the trained settlement ground object segmentation model;

[0030] A second training module, used to train the settlement site selection environmental pattern identification model by using the settlement environmental pattern data set to obtain the trained settlement site selection environmental pattern identification model.

[0031] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the settlement site selection environment pattern identification method are implemented.

[0032] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the settlement site selection environment pattern identification method are implemented.

[0033] The present invention has the following beneficial effects:

[0034] When the settlement site selection environment pattern identification method, system, device, and medium based on meta-learning according to the present invention are specifically operated, the remote sensing image and DEM data of the settlement to be identified are input into the trained settlement ground object segmentation model to obtain the environmental ground object distribution map of the settlement. Then, the environmental ground object distribution map of the settlement is input into the trained settlement site selection environment pattern identification model to obtain the type of the settlement site selection environment pattern, so as to automatically identify the settlement site selection environment pattern of traditional settlements. In addition, the present invention adopts a settlement ground object segmentation model and a settlement site selection environment pattern identification model, and only a small number of settlement samples are required during the identification process.

[0035] Furthermore, the present invention directly uses the settlement environment pattern data set to train the settlement ground object segmentation model VOSM and the site selection environment pattern identification model VPRM. On this basis, the remote sensing image and elevation data of the settlement to be identified are sent into the VOSM model to obtain the environmental ground object distribution map of the settlement. Then, the environmental ground object distribution map of the settlement is input into the site selection environment pattern identification model VPRM for pattern recognition, which can identify the environmental pattern of settlements of untrained types under the condition of only providing a support set, has good identification robustness, and avoids the problem that the identification model needs to reconstruct the data set and train the model again when applied to other unknown areas, providing a scientific and efficient intelligent tool for the analysis of the settlement site selection pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a structural framework diagram of the settlement environmental ground object segmentation model VOSM;

[0037] Figure 2 It is a structural framework diagram of the settlement environment pattern identification model VPRM;

[0038] Figure 3 It is a training flow diagram of the settlement environment pattern identification model VPRM. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments, and are not intended to limit the scope of the present invention disclosure. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessarily confusing the concepts disclosed in the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0040] The schematic structural diagrams according to the disclosed embodiments of the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and for the purpose of clear expression, some details are enlarged and some details may be omitted. The shapes of various regions and layers shown in the figures and their relative sizes and positional relationships are only exemplary. In practice, there may be deviations due to manufacturing tolerances or technical limitations, and those skilled in the art can design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0041] Reference Figure 1 、 Figure 2 and Figure 3 The method for identifying the settlement location environment pattern based on meta-learning according to the present invention includes the following steps:

[0042] 1) Use the 91 satellite map to download the remote sensing images R = {R1, R2,..., R i ,...} and digital elevation models D = {D1, D2,..., D i ,...} (hereinafter referred to as DEM data) of traditional villages (hereinafter referred to as settlements) nationwide, and manually mark the settlement location environment pattern categories and the surrounding mountains, water systems, forests, farmlands, vegetation, and wastelands to obtain the environmental feature classification mark map G and the settlement environment pattern category label L. Construct a settlement environment pattern dataset through the remote sensing image R, elevation model D, environmental feature classification mark map G, and settlement environment pattern category label L;

[0043] 2) Construct a settlement object segmentation model based on semantic segmentation (Village Object Segmentation Model, hereinafter referred to as VOSM), and use the settlement environment pattern dataset constructed in step 1) to train the settlement object segmentation model;

[0044] After experiments, the settlement environment pattern dataset is divided into a training set, a validation set, and a test set according to the quantitative relationship of 7:2:1, and then the settlement object segmentation model is trained with this, and the recognition results of the environmental features are shown in Table 1.

[0045] Table 1

[0046]

[0047] 3) Build a Village Patterns Recognition Model based on Meta - learning (hereinafter referred to as VPRM);

[0048] 4) Use the settlement environment pattern dataset constructed in step 1) to train the settlement site selection environment pattern recognition model;

[0049] After experiments, the settlement environment pattern dataset is divided into a training set, a validation set and a test set according to the quantitative relationship of 7:2:1, and then the settlement site selection environment pattern recognition model is trained with this. The experimental results of the settlement site selection environment pattern recognition model VPRM after training are shown in Table 2.

[0050] Table 2

[0051] <L

[0052] 5) Input the remote sensing image and DEM data of the settlement to be identified obtained into the trained settlement ground object segmentation model to obtain the environmental ground object distribution map of the settlement, and then input the environmental ground object distribution map of the settlement into the trained settlement site selection environment pattern recognition model to obtain the site selection environment pattern type of the settlement.

[0053] Specifically, the specific operation of step 1) is as follows:

[0054] Use the "List of Chinese Traditional Villages" issued by the Ministry of Housing and Urban - Rural Development to select representative typical cases and download the sky - earth (remote sensing) images and DEM data of the settlement and its surrounding areas in the 91 satellite map. Among them, the accuracy of the remote sensing image is selected as 0.53 meters, the scale is selected as 2560*2560 pixels, the accuracy of the DEM data is selected as 8.5 meters, the scale is selected as 160*160 grids, and the remote sensing image and the DEM data are coordinate - aligned. Subsequently, manually mark the ground object distribution map G of the settlement, mountains, water systems, forests, farmlands and wastelands i [[ID=3)] and the settlement environment pattern category label L i , and then construct a settlement environment pattern dataset.

[0055] Among them, based on the remote sensing image and DEM data of the settlement, use the labelme semantic segmentation annotation program to manually mark the areas where the six types of environmental ground objects of the settlement, mountains, water systems, forests, farmlands and wastelands are located to form the environmental ground object distribution annotation map G of the settlement i .

[0056] In addition, the analysis of the settlement site selection environment pattern focuses on the planar position and spatial relationship between natural resources such as mountains, waters, forests, and fields and the artificial building residential areas. Urban and rural planning professionals have conducted field investigations and research analyses on representative typical settlements in the "List of Chinese Traditional Villages", and given the classification principles for five major types of site selection environment patterns, namely the mountain-ringed water-surrounded type, mountain type, plain type, river-side type, and mountain-relying type. Therefore, six types of environmental feature distribution maps of typical case settlements are marked out using remote sensing images and DEM data, and then the category labels of the site selection environment patterns are analyzed and determined by urban and rural planning professionals. Then, the settlement environment pattern samples composed of remote sensing image R i , elevation model D i , environmental feature distribution annotation map G i , and environmental pattern category label L i are formed. By operating in this way, a settlement environment pattern dataset for the five major types of settlements in the "List of Chinese Traditional Villages" is generated.

[0057] The specific operation of step 2) is as follows:

[0058] Adopt a multi-modal semantic segmentation method to establish a settlement feature segmentation model VOSM. Its main structure is based on deeplab-v3+, but a bypass E θ is added to the feature extraction part of the backbone network to support the semantic learning of terrain features such as mountains. Finally, the features of the ground objects and terrain are fused and sent into the classifier for the segmentation of the settlement and its surrounding environmental features. Its specific structure is as Figure 1 shown.

[0059] It should be noted that Deeplab-v3+ is a semantic segmentation network model. It combines an encoder-decoder structure with an Atrous Spatial Pyramid Pooling (ASPP) to achieve refined ground object segmentation. The mean Intersection over Union (MIoU) in the PASCAL Visual Object Classes dataset reaches 89%, and it has the advantages of high efficiency, accuracy, and simplicity. The backbone feature extraction network of the Deeplab-v3+ network model used in this invention is ResNet-50, which has a total of 4 layer layers. The output of the first layer layer is used as low-level features and sent into a 1×1 convolution to adjust the channels, and then sent into subsequent modules for feature fusion. The output of the last layer layer is used as high-level features and sent into the ASPP module.

[0060] In addition, E θ uses 3 convolutional layers to extract the terrain features in the input data. The output of convolutional layer -3 is used as the terrain features and sent into a 1×1 convolution to adjust the channels, and then sent into subsequent modules for feature fusion.

[0061] Specifically, the specific operation of step 3) is as follows:

[0062] Build a settlement location environment pattern discrimination model VPRM using meta - learning. The settlement location environment pattern discrimination model consists of an encoder f composed of a ResNet12 θ and a meta - learning classifier The input data is the support set S i and the query set Q i , and the output is the settlement environment pattern corresponding to the query set data. S i and Q i are composed of the settlement environment map G. Before entering the network, G is scaled to a size of 256×256. Through f θ a 512 - dimensional feature vector V is obtained. V is sent into the classifier to perform a similarity measurement with the center of S i . The one with the largest similarity is the discrimination result of this environmental pattern. The classifier calculates the probability that the input feature belongs to each type of settlement location environment pattern in a metric learning manner, and selects the cosine similarity function to judge the distance between the query set feature and the support set center. The specific structure of VPRM is as Figure 2 shown

[0063] Specifically, the specific operation of step 4) is as follows:

[0064] Divide the training process of the settlement location environment pattern discrimination model VPRM into two stages: full - classification training and meta - training, as Figure 3 shown. In the full - classification training stage, use the base - class data D of all settlement location environment patterns base to perform full - classification training, and remove the last classifier C to obtain the encoder f θ . The encoder f θ is used to map the input data to the feature space. In the meta - training stage, divide the base - class data D base into a series of episodes (the division scheme of the support set and the query set). Among them, each episode is regarded as a classification task T of N - class - K - sample i ; for each episode, perform a cosine similarity measurement between the query set feature and the mean of the support set features, and optimize the parameters of the VPRM discrimination model by minimizing the N - way prediction loss in the query set to obtain the trained VPRM model. Finally, use multiple un - trained episodes in the test set to evaluate the discrimination performance of the model

[0065] It should be noted that the encoder f θ is an encoder composed of ResNet12. The input is the environmental feature distribution map G of the settlement, and the output is a 512 - dimensional feature vector V

[0066] Additionally, for the classification task T of N classes and K samples i , in few-shot learning, a model is usually trained on a classification task of a set of N classes - K samples, denoted as which is also regarded as a series of episodes. Specifically, each episode contains a support set and a query set . Among them, the support set contains N different classes, and there are K labeled samples in each class, consisting of N×K samples for training; the query set has the same N classes, and each class has unlabeled samples to be identified for their types. In the training and testing phases, episodes are usually constructed in the same way.

[0067] For the support set and the query set In meta-learning, the VPRM classification model needs to be trained on multiple tasks. The training set in each task is called the support set, and the test set is called the query set. The support set is composed of a small number of samples extracted from the base classes. The support set in each task contains N classes, and each class has K samples. The average feature W c of N classes is used as its centroid, and the calculation method of the centroid is shown in Equation (1).

[0068] For the metric module, in the meta-learning method based on similarity measurement, the feature encoder is optimized by the similarity measurement between the support samples and the query samples. Given a few-shot classification task with a support set S, where S c is the support set samples, the average representation w c is calculated according to Equation (1) as the center point of class c.

[0069]

[0070] For the query sample x in few-shot classification, according to the cosine similarity between the feature vector V of the sample x after passing through f θ and the center point of class c, the probability that the sample x belongs to class c is predicted, and its calculation method is shown in Equation (2).

[0071]

[0072] Specifically, the specific operation of step 5) is as follows:

[0073] For the settlement whose location environment pattern needs to be identified, first download its remote sensing image R and DEM data D, and input them into the trained settlement ground object segmentation model VOSM to obtain the settlement environment ground object distribution map G identified from multi-source data. Then input the settlement environment ground object distribution map G into the trained location environment pattern identification model VPRM to obtain the result of the location environment pattern category of this settlement.

[0074] Embodiment 2

[0075] The settlement location environment pattern identification system described in this invention includes:

[0076] An acquisition module, used to acquire the remote sensing image and DEM data of the settlement to be identified;

[0077] A first calculation module, used to input the remote sensing image and DEM data of the settlement to be identified into the trained settlement ground object segmentation model to obtain the settlement environment ground object distribution map;

[0078] A second calculation module, used to input the settlement environment ground object distribution map into the trained settlement location environment pattern identification model to obtain the location environment pattern type of this settlement.

[0079] It further includes:

[0080] A first construction module, used to construct a settlement environment pattern dataset;

[0081] A second construction module, used to construct a settlement ground object segmentation model;

[0082] A third construction module, used to construct a settlement location environment pattern identification model;

[0083] A first training module, used to train the settlement ground object segmentation model by using the settlement environment pattern dataset to obtain the trained settlement ground object segmentation model;

[0084] A second training module, used to train the settlement location environment pattern identification model by using the settlement environment pattern dataset to obtain the trained settlement location environment pattern identification model.

[0085] In the embodiments of this application, the division of modules is illustrative, only a logical function division. In actual implementation, there may be other division methods. In addition, in each embodiment of this application, each functional module can be integrated in one processor, or can exist separately physically, or two or more modules can be integrated in one module. The above integrated modules can be implemented in the form of hardware, or in the form of software functional modules.

[0086] Embodiment 3

[0087] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the settlement site selection environment pattern identification method are implemented. Among them, the memory may include internal memory, such as high-speed random access memory, and may also include non-volatile memory, such as at least one disk memory, etc.; the processor, network interface, and memory are interconnected through an internal bus, and this internal bus can be an Industry Standard Architecture bus, a Peripheral Component Interconnect standard bus, an Extended Industry Standard Architecture bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory is used to store programs. Specifically, the program can include program code, and the program code includes computer operation instructions. The memory can include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0088] Embodiment 4

[0089] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the settlement site selection environment pattern identification method are implemented. Specifically, the computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory can include random access memory (RAM) and / or cache memory, etc. The non-volatile memory can include read-only memory (ROM), hard disk, flash memory, optical disc, magnetic disk, etc.

[0090] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

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

[0092] These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the functions specified in one Figure 1 process or multiple processes and / or boxes Figure 1 a box or multiple boxes.

[0093] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 process or multiple processes and / or boxes Figure 1 a box or multiple boxes.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific implementation manners of the present invention, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. A method for identifying settlement site selection environment patterns based on meta-learning, characterized in that: include: Obtain remote sensing images and digital elevation models of the settlements to be identified; Input the remote sensing image and digital elevation model of the settlement to be identified into the trained settlement feature segmentation model to obtain the environmental feature distribution map of the settlement; Inputting the environmental feature distribution map of the settlement into the trained settlement site selection environmental pattern identification model to obtain the site selection environmental pattern type of the settlement; The settlement object segmentation model is based on deeplab-v3+, but an additional bypass E is required in the feature extraction part of the backbone network. θ , to support the semantic learning of terrain features, and finally integrate the terrain features with the multimodal features of the terrain and send them into the classifier to segment the settlement and its surrounding environmental features. The constructed multimodal settlement feature semantic segmentation network model is named the settlement feature segmentation model; The settlement site selection environment pattern identification model consists of an encoder f composed of a ResNet12 θ and a meta-learning classifier The training process of the settlement site selection environment pattern identification model is divided into the full classification training stage and the meta-training stage, wherein the encoder f θ is an encoder composed of ResNet12.

2. The method for identifying settlement site selection environment patterns based on meta-learning according to claim 1, characterized in that: Also includes: Construct a dataset of settlement environment patterns; Construct a settlement feature segmentation model; Construct a model to identify environmental patterns in settlement site selection; The settlement feature segmentation model is trained using the settlement environment pattern dataset to obtain a trained settlement feature segmentation model; The settlement environment pattern dataset is used to train the settlement site selection environment pattern identification model to obtain the trained settlement site selection environment pattern identification model.

3. The method for identifying settlement site selection environment patterns based on meta-learning according to claim 1, characterized in that: The specific process of constructing the settlement environment pattern dataset is as follows: Using satellite maps, we downloaded settlement remote sensing images R and digital elevation models D across the country. We manually labeled the settlement site environmental pattern categories and the surrounding mountains, water systems, woodlands, farmlands, vegetation, and wastelands to obtain the environmental feature classification label map G and the settlement environmental pattern category label L. We then constructed a settlement environmental pattern dataset using the remote sensing images R, elevation model D, environmental feature classification label map G, and settlement environmental pattern category label L.

4. The method for identifying settlement site selection environment patterns based on meta-learning according to claim 1, characterized in that: The settlement feature segmentation model is constructed based on semantic segmentation; The settlement site selection environmental pattern identification model is constructed based on meta-learning.

5. A settlement site selection environment pattern identification system based on meta-learning, characterized by: include: The acquisition module is used to obtain remote sensing images and DEM data of the settlement to be identified; The first calculation module is used to input the remote sensing image and DEM data of the settlement to be identified into the trained settlement feature segmentation model to obtain the environmental feature distribution map of the settlement; A second calculation module is used to input the environmental feature distribution map of the settlement into the trained settlement site selection environment pattern identification model to obtain the site selection environment pattern type of the settlement; The settlement object segmentation model is based on deeplab-v3+, but an additional bypass E is required in the feature extraction part of the backbone network. θ , to support the semantic learning of terrain features, and finally integrate the terrain features with the multimodal features of the terrain and send them into the classifier to segment the settlement and its surrounding environmental features. The constructed multimodal settlement feature semantic segmentation network model is named the settlement feature segmentation model; The settlement site selection environment pattern identification model consists of an encoder f composed of a ResNet12 θ and a meta-learning classifier The training process of the settlement site selection environment pattern identification model is divided into the full classification training stage and the meta-training stage, wherein the encoder f θ is an encoder composed of ResNet12.

6. The settlement site selection environment pattern identification system based on meta-learning according to claim 5 is characterized in that: Also includes: The first building module is used to construct a settlement environment pattern dataset; The second building module is used to build a settlement feature segmentation model; The third building block is used to construct a settlement site selection environmental pattern identification model; The first training module is used to train the settlement feature segmentation model using the settlement environment pattern dataset to obtain a trained settlement feature segmentation model; The second training module is used to train the settlement site selection environment pattern identification model using the settlement environment pattern data set to obtain the trained settlement site selection environment pattern identification model.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the settlement site selection environment pattern identification method based on meta-learning as described in any one of claims 1 to 4 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the settlement site selection environment pattern identification method based on meta-learning as described in any one of claims 1 to 4 are implemented.

Images