Evacuation path planning system based on spontaneous geographic information

Abstract

The invention discloses an evacuation path planning system based on spontaneous geographic information, and relates to the field of evacuation path planning. The invention aims to solve the problem of low evacuation efficiency caused by an evacuation path obtained by an existing evacuation path planning method. The method comprises the following steps: searching and acquiring basic data of a target area on a social media platform by utilizing a place name of the target area; the target area basic data comprises a description text and image data; the basic data preprocessing module is used for preprocessing basic data of a target area; obtaining a text element set and a picture element set by utilizing the preprocessed basic data of the target area; constructing an element library by utilizing the text element set and the picture element set; and assigning weights to the elements in the element library, thereby setting an evacuation path according to the element weights. The method is used for drawing the evacuation path.

Description

Technical Field

[0001] The present invention relates to the field of evacuation route planning, and in particular to an evacuation route planning system based on spontaneous geographic information. Background Art

[0002] With the rapid development of science and technology, spatial information technology has become an important part of urban planning and construction. The planning of urban development depends on multiple factors such as space, population, industry, and events. Space is the carrier of the corresponding association of various spatial element information. The mapping of real geographic space objects in the computer is called spatial elements. There are inherent associations between different spatial elements. The preference for spatial elements can affect people's choice of space and the path to the space. Studies have shown that the organization of spatial elements will affect people's movement patterns in space, thereby affecting the choice of paths in space. In emergency route planning, human flow is an important factor. The gathering and distribution of human flow depends on people's preference for spatial elements. Therefore, the study of spatial element preferences is very important for controlling spatial human flow and planning emergency routes.

[0003] Traditional evacuation path planning methods mainly include distance-based path planning methods and graph theory-based path planning methods. These two methods only focus on the length of the evacuation path. However, in areas with large traffic flow such as scenic spots and campuses, the flow of people is not evenly distributed, but is closely related to elements such as landscape buildings in the area. Therefore, it is difficult to evacuate according to the pre-planned shortest path in the face of emergency evacuation. The existing methods do not consider the impact of spatial elements on evacuation. Evacuation according to the evacuation path obtained by the existing technology will result in low evacuation efficiency. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that obtaining an evacuation path using an existing evacuation path planning method will result in low evacuation efficiency, and propose an evacuation path planning system based on spontaneous geographic information.

[0005] An evacuation route planning system based on spontaneous geographic information, comprising: a basic data acquisition module, a basic data preprocessing module, an element set acquisition module, an element library construction module, and a spatial element preference acquisition module;

[0006] The basic data acquisition module uses the location name of the target area to search and acquire the basic data of the target area on the social media platform;

[0007] The target area basic data includes: description text and image data;

[0008] The basic data preprocessing module is used to preprocess the basic data of the target area;

[0009] The element set acquisition module acquires a text element set and a picture element set using the preprocessed target area basic data;

[0010] The element library construction module constructs an element library using a text element set and a picture element set;

[0011] The spatial element preference acquisition module is used to assign weights to elements in the element library, thereby obtaining element preferences in the target area;

[0012] The evacuation path planning module plans a safe evacuation path in the target area by utilizing the element preferences in the target area.

[0013] Furthermore, the basic data acquisition module uses the location name of the target area to search and acquire the basic data of the target area on the social media platform, specifically:

[0014] Use the location name and coordinates of the target area to retrieve data on social media platforms and open source map platforms, use the automated web crawler program WebSpider to crawl the basic data of the target area, and convert the description text in the basic data of the target area into csv format;

[0015] The target area basic data includes: description text and image data.

[0016] Furthermore, the basic data preprocessing module is used to preprocess the basic data of the target area, specifically:

[0017] A1. Preprocessing the description text in the basic data of the target area to obtain the preprocessed description text;

[0018] A2. Use the image management tool VisiPics to filter the image data in the basic data of the target area, remove the image data that exceeds the preset comparison similarity threshold, and obtain the pre-processed image data.

[0019] Furthermore, the step A1 of preprocessing the description text in the target area basic data to obtain the preprocessed description text is specifically as follows:

[0020] A1-1. Represent the description text in the target area basic data in matrix form to obtain a description text matrix, and then use the toarray function to convert the description text matrix into an array form, and use the array form of the description text matrix as a description text vector;

[0021] A1-2. Remove duplicate description text vectors to obtain preprocessed description text:

[0022] The scikit-learn machine learning library of Python is used to calculate the cosine similarity between every two description text vectors. If the cosine similarity value of any two description text vectors is greater than the similarity threshold, any vector of the current two description text vectors is eliminated.

[0023] Furthermore, the CountVectorizer function is used to represent the description text in the target area basic data in a matrix form in A1-1.

[0024] Furthermore, the element set acquisition module acquires the text element set using the preprocessed target area basic data, specifically:

[0025] B1-1. Segment each description text after preprocessing, and connect each word into a new string through spaces;

[0026] B1-2. Calculate the term frequency-inverse document frequency TF-IDF of each word in each description text, so as to obtain the TF-IDF value vector of each description text;

[0027] B1-3. Use the get_feature_names_out() function and the preprocessed description text to obtain the feature vocabulary;

[0028] B1-4, obtain the intersection of the word set in the current description text and the word set in the feature vocabulary, sort each word in the intersection from large to small according to the TF-IDF value, and use the first n words as keywords for the description text i;

[0029] B1-5. Integrate all keywords describing the text and manually delete keywords irrelevant to spatial elements to obtain a text element set.

[0030] Furthermore, the picture element set is obtained by:

[0031] B2-1. Use Google Vision API to identify spatial elements in pre-processed image data and use the spatial elements as labels for the corresponding pre-processed image data;

[0032] The spatial elements include: objects, scenes, and activities;

[0033] B2-2. Count the frequency of label occurrence of all preprocessed image data, delete labels with a frequency threshold less than the threshold, and manually remove labels irrelevant to spatial elements to obtain a picture feature set.

[0034] Furthermore, the element library construction module constructs an element library using the text element set and the picture element set, specifically:

[0035] First, the labels in the image feature set are compared with the text feature set to obtain common features and establish a feature library;

[0036] Then, the elements in the element library are classified and stored according to the element type;

[0037] Element types include: buildings, roads, natural landscapes, and indoor spaces.

[0038] Furthermore, the spatial element preference acquisition module is used to assign weights to elements in the element library, specifically:

[0039]

[0040] The larger the weight of the current type of element is, the higher the preference for the current type of element is.

[0041] Furthermore, the evacuation path planning module plans a safe evacuation path in the target area using the element preference in the target area, specifically:

[0042] First, sort the elements of the target area from large to small according to their weights;

[0043] Then, multiple evacuation paths are set, the weights and sums of the elements included in each evacuation path are obtained, and the path consisting of the elements with the highest weight sum is obtained as the evacuation path.

[0044] The beneficial effects of the present invention are:

[0045] The present invention widely obtains spontaneous data from different users by utilizing spontaneous geographic information (VGI), thereby ensuring a large and more balanced sample size. The present invention uses the basic data of the target area to obtain a text element set and a picture element set, and then constructs an element library, and then assigns weights to the elements in the element library, thereby planning an evacuation path based on the element weights. In the process of planning the evacuation path, the present invention takes into account the impact of the elements in the study area on the flow of people, which can avoid the problem that areas with large flow of people cannot be evacuated according to the preset evacuation path when an actual emergency occurs, thereby reducing the time for people to find evacuation routes and improving the efficiency of evacuation. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of the present invention;

[0047] Figure 2 A framework diagram was constructed for the spatial element preference evaluation model of college students based on spontaneous geographic information. DETAILED DESCRIPTION

[0048] Specific implementation method 1: Figure 1As shown, in this embodiment, an evacuation path planning system based on spontaneous geographic information includes: a basic data acquisition module, a basic data preprocessing module, an element set acquisition module, an element library construction module, a spatial element preference acquisition module, and an evacuation path planning module;

[0049] The basic data acquisition module acquires the basic data of the target area on the social media platform using the location name of the target area;

[0050] The target area basic data includes: description text and image data;

[0051] The target area basic data may also include: basic information of the information publisher and coordinate data of the information publisher;

[0052] If the current space is a university campus, the basic information of the information publisher includes: the education and age of the information publisher; the basic information of the information publisher and the coordinate data of the information publisher can assist in screening the information of college students, thereby determining the current campus space;

[0053] The basic data preprocessing module is used to preprocess the basic data of the target area;

[0054] The element set acquisition module acquires a text element set and a picture element set using the preprocessed target area basic data;

[0055] The element library acquisition module constructs an element library using a text element set and a picture element set;

[0056] The spatial element preference acquisition module is used to assign weights to elements in the element library, thereby obtaining element preferences in the target area;

[0057] The evacuation route planning module plans a safe evacuation route using element preferences in the target area.

[0058] Specific implementation method 2: The basic data acquisition module uses the location name of the target area to acquire the basic data of the target area on the social media platform, specifically:

[0059] Use the location names of the target area to retrieve data on social media platforms and open source map platforms, use the automated web crawler program WebSpider to crawl the basic data of the target area, and convert the data in the basic data of the target area except the image data into csv format;

[0060] The target area basic data includes: description text, image data

[0061] The present invention retrieves texts related to space preferences on social media platforms and open source map platforms based on relevant keywords such as space names, and uses an automated web crawler program WebSpider to crawl the corresponding texts to obtain its related spontaneous basic data. The expected content includes: ① description text; ② picture data (if any); ③ basic information of the information publisher; ④ coordinate data of the information publisher. The data except the picture data is integrated into csv format to facilitate the next step of screening.

[0062] Specific implementation method three: The basic data preprocessing module is used to preprocess the basic data of the target area, specifically:

[0063] A1. Preprocess the description text in the target area basic data to obtain the preprocessed description text:

[0064] A1-1. Use the CountVectorizer function to represent the description text in the target area basic data as a matrix, then use the toarray function to convert the description text matrix into an array form, and use the array form of the description text matrix as the description text vector;

[0065] A1-2. Remove duplicate description text vectors to obtain preprocessed description text:

[0066] Use Python's scikit-learn machine learning library to calculate the cosine similarity between every two description text vectors in the target area basic data. If the cosine similarity value of any two description text vectors is greater than the similarity threshold, then remove any of the current two description text vectors.

[0067] The similarity threshold is 0.8. When the cosine similarity between two description text vectors is greater than 0.8, it means that the two current description text vectors are duplicate texts.

[0068] A simple example of step A1 code is as follows:

[0069] #Step 1: Import the library

[0070] #Use CountVectorizer in scikit-learn to convert text into vectors

[0071] #Use cosine_similarity to calculate the cosine similarity between texts

[0072] from sklearn.feature_extraction.text import CountVectorizer

[0073] from sklearn.metrics.pairwise import cosine_similarity

[0074] #Step 2: Define the text

[0075] #Define two sample texts "Description Text 1" and "Description Text 2";

[0076] text1="Description text 1"

[0077] text2="Description text 2"

[0078] #Step 3: Convert text into vectors using CountVectorizer

[0079] #CountVectorizer maps the words in the text into feature vectors for cosine similarity calculation

[0080] vectorizer=CountVectorizer().fit_transform([text1,text2])

[0081] #Step 4: Vectorize the text

[0082] #.toarray() converts the sparse matrix obtained by CountVectorizer into an array form to facilitate subsequent calculations

[0083] vectors=vectorizer.toarray()

[0084] #Step 5: Calculate cosine similarity

[0085] #Use cosine_similarity to calculate the cosine similarity between two vectors (text)

[0086] #The cosine similarity range is between -1 and 1. The closer it is to 1, the more similar the texts are.

[0087] cosine_sim=cosine_similarity(vectors)

[0088] #Step 6: Output the results

[0089] #Print the cosine similarity between text 1 and text 2, cosine_sim[0][1] represents their similarity value

[0090] print(f"Cosine Similarity between text1 and text2:{cosine_sim[0][1]}")

[0091] #Step 7: Determine duplicate text

[0092] #If the similarity is greater than or equal to 0.8, the text is considered duplicate and will be removed

[0093] if cosine_sim[0][1]>=0.8:

[0094] print("The text has high similarity and is a duplicate text. It is recommended to remove it.")

[0095] else:

[0096] print("The text similarity is low and is not a duplicate text.")

[0097] A2. Use the image management tool VisiPics to filter the image data in the basic data of the target area, remove the image data that exceeds the preset comparison similarity threshold, and obtain the pre-processed image.

[0098] Specific implementation method 4: The element set acquisition module uses the pre-processed target area basic data to acquire the text element set and the picture element set, specifically:

[0099] B1. Get the text element set:

[0100] B1-1. Use jieba.cut() in Python to segment each description text after preprocessing, and connect each word into a new string through spaces;

[0101] B1-2. Use the scikit-learn library to calculate the term frequency-inverse document frequency (TF-IDF) of each word in each description text, so as to obtain the TF-IDF value vector of each description text;

[0102] B1-3. Use the get_feature_names_out() function and the preprocessed description text to obtain the feature vocabulary;

[0103] B1-4, obtain the intersection of the word set in the current description text and the word set in the feature vocabulary, sort each word in the intersection from large to small according to the TF-IDF value, and use the first n words as keywords for the description text i;

[0104] Wherein, n=5;

[0105] B1-5. Integrate all the keywords describing the text and manually delete the keywords irrelevant to the spatial elements based on experience to obtain the text element set;

[0106] The keywords include: square, stairs, trees, etc.;

[0107] The code for step B1 can be simply illustrated as follows:

[0108] #Step 1: Import pandas to process the dataset, import TfidfVectorizer for TF-IDF calculation, and import jieba library for Chinese word segmentation

[0109] import pandas as pd

[0110] from sklearn.feature_extraction.text import TfidfVectorizer

[0111] import jieba

[0112] #Step 2: Import text data and define text lists

[0113] texts = ["Description text 1","Description text 2",]

[0114] #Step 3: Segment Chinese text

[0115] #Use jieba.cut() to perform word segmentation and concatenate the word segmentation results into a new string with spaces

[0116] texts_segmented=[”.join(jieba.cut(text))for text in texts]

[0117] #Step 4: Calculate TF-IDF value

[0118] #Use scikit-learn's TfidfVectorizer to calculate the word frequency-inverse document frequency of the segmented text

[0119] vectorizer=TfidfVectorizer()

[0120] X=vectorizer.fit_transform(texts_segmented)

[0121] #Step 5: Extract feature words and TF-IDF matrix

[0122] #Use get_feature_names_out() to get the feature word list and convert the sparse matrix into a dense matrix

[0123] feature_names=vectorizer.get_feature_names_out()

[0124] dense = X.todense()

[0125] denselist = dense.tolist()

[0126] #Step 6: Output the TF-IDF score for each text

[0127] #Sort by score and extract the top 5 keywords

[0128] for i,text in enumerate(denselist):

[0129] tfidf_scores=dict(zip(feature_names,text))

[0130] sorted_keywords = sorted(tfidf_scores.items(), key = lambda x:x[1], reverse = True) print(f"Keywords of text {i+1}:",[kw[0]for kw in sorted_keywords[:5]])

[0131] B2. Get the image feature set:

[0132] B2-1. Use Google Vision API to identify spatial elements (objects, scenes, activities, etc.) in pre-processed image data, and use the spatial elements as labels for the corresponding pre-processed image data;

[0133] The code of step B2-1 can be simply illustrated as follows:

[0134] #Step 1: Import the Google Cloud Vision API library

[0135] #Import vision API related modules from google.cloud.vision_v1 to allow access to Google CloudVision's image analysis functions

[0136] from google.cloud import vision_v1

[0137] #Step 2: Define function to get image label

[0138] #Define the function `get_image_labels`, which receives the path of the image as input and returns the labels in the image

[0139] def get_image_labels(image_path):

[0140] client=vision_v1.ImageAnnotatorClient()

[0141] image = vision_v1.Image()

[0142] image.source.image_uri=image_path

[0143] response=client.label_detection(image=image)

[0144] labels=[label.description for label in response.label_annotations]

[0145] return labels

[0146] #Step 3: Specify the image path

[0147] #Define the image path to be analyzed

[0148] image_path="path / to / image.jpg"

[0149] #Step 4: Call the function and get the image label

[0150] #Call the `get_image_labels` function, pass in the image path, and get the labels in the image

[0151] labels=get_image_labels(image_path)

[0152] Step 5: Output image labels

[0153] #Print image labels to show the identified spatial elements

[0154] print("Image Labels:",labels)

[0155] B2-2. Count the frequency of label occurrence of all preprocessed image data, delete labels with a frequency threshold less than the threshold, and manually remove labels irrelevant to spatial elements to obtain a picture feature set.

[0156] Specific implementation method 5: The element library acquisition module uses the text element set and the picture element set to construct the element library, specifically:

[0157] Compare the labels in the image element set and the text element set horizontally to obtain common elements and establish an element library;

[0158] The elements in the element library are stored according to the element type classification;

[0159] Element types include: buildings, roads, natural landscapes, and indoor spaces.

[0160] Specific implementation method six: The spatial element preference acquisition module is used to assign weights to elements in the element library, so as to obtain element preferences in the target area, specifically:

[0161]

[0162] The larger the weight of the current type of element is, the higher the preference for the current type of element is.

[0163] Specific implementation method seven: The evacuation path planning module plans a safe evacuation path using the element preference in the target area, specifically:

[0164] First, graphical tools are used to transform the comprehensive evaluation results of spatial element preferences into easy-to-understand graphical displays;

[0165] Then, sort the elements of the target area from large to small according to the element weight;

[0166] Then, multiple evacuation paths are set, the weights and sums of the elements included in each evacuation path are obtained, and the path consisting of the elements with the highest weight sum is obtained as the evacuation path.

[0167] Finally, an evacuation path model was established based on anylogic software, and the evacuation effect was visualized based on the density map to verify the evacuation effect.

[0168] In this step, spatial preference elements are identified based on the spatial element comprehensive evaluation model. According to different evacuation scenarios such as fire evacuation, earthquake evacuation, and flood evacuation, the correlation between elements and evacuation is demonstrated, and spatial elements suitable for evacuation are manually screened. For example, fire evacuation should avoid densely populated areas, choose open public spaces and areas with dense transportation facilities, and flood evacuation should avoid low-lying areas. Identify the core elements in different emergency scenarios and assign weights to spatial element categories based on the spatial element comprehensive evaluation model. Then, use the high-weight spatial elements in the preference evaluation model as the target area for route selection, select the shortest path, safest path, fastest path, etc. as the path selection criteria according to the evacuation scenario, calculate the evacuation route weight, and comprehensively consider the spatial element weights and preference values ​​of the route to select the route with the highest weight. Finally, an evacuation path model is established based on anylogic software, and the evacuation effect is visualized based on the density map to verify the evacuation effect; a high weight means that this element has a greater potential to be selected in the selection scenario. The reason for choosing a high weight is that the route formed by connecting high-weight elements is more in line with the user's perception of spatial choice (i.e., the intuition formed by element preference) or more recognizable, which can eliminate the problem that the designed evacuation route is inconsistent with the route selected by the user in the actual emergency situation, affecting the evacuation.

[0169] Embodiment: Taking the design of the evacuation path of the university campus space of the second campus of Harbin Institute of Technology as an example, the feasibility of the present invention is verified. Figure 2 As shown:

[0170] Step 1: Based on social media platforms such as Weibo, Xiaohongshu, WeChat public accounts, and open source map retrieval platforms such as Baidu Maps and Amap, search with words such as "Harbin Institute of Technology Second Campus" and "Harbin Institute of Technology Second Campus" as keywords to obtain texts related to the spatial element preferences of college students at Harbin Institute of Technology Second Campus. Use the automated web crawler program WebSpider to crawl the corresponding texts and obtain relevant spontaneous geographic information, including: ① description text; ② image data (if any); ③ basic information of the information publisher; ④ coordinate data of the information publisher. Convert all data except the image data into csv format and store them.

[0171] Step 2: The obtained text is represented as a vector, and the cosine similarity of the text is calculated using the scikit-learn machine learning library of Python. The cosine angle between the texts is calculated to measure their similarity, and duplicate texts with too high similarity are eliminated. Image sorting software is used to eliminate duplicate image data. The image management tool VisiPics is used to filter duplicate images and eliminate image data with too high similarity. After the screening is completed, the basic data of the Second Campus of Harbin Institute of Technology is obtained.

[0172] Step 3: Process the sample text in the basic data of the second campus of Harbin Institute of Technology into segmented text data, load the scikit-learn and pandas machine learning toolkits in Python and import the segmented text data, and use the jieba library in Python to segment the text. Use the scikit-learn library to calculate the term frequency-inverse document frequency (TF-IDF), filter and sort the texts according to the TF-IDF value from large to small, manually remove keywords that are not related to spatial elements, extract the top five keywords that appear frequently in the text, and obtain the text feature set.

[0173] Step 4: Sample spontaneous geographic information data. Access the Google Vision API and use Google Vision to identify campus spatial elements in the spontaneous geographic information data image sampling samples to obtain the generated relevant labels. Preprocess the obtained label data and filter to obtain the label set.

[0174] Step 5: Count the frequency of the obtained label results, delete the label categories with too few frequencies, and classify the remaining labels according to the differences in spatial element types, so as to obtain the spatial element library of Harbin Institute of Technology Second Campus University Students, which can be counted and analyzed according to the categories of spatial elements to obtain the distribution of each category.

[0175] Step 6: Calculate the element weights based on the spatial element database of the second campus of Harbin Institute of Technology. For each spatial element category, use the relative frequency to calculate its weight according to its quantity in the element database, and assign values ​​to the preference dimensions. Integrate the calculated comprehensive evaluation values ​​to establish the spatial element preference evaluation model of the second campus of Harbin Institute of Technology.

[0176] Step 7: Use graphical tools to convert the evaluation results of the spatial element preference evaluation model of the second campus of Harbin Institute of Technology into easy-to-understand graphical displays. Based on the graphical results, the preferences of college students for different spatial element categories of the second campus of Harbin Institute of Technology are obtained.

[0177] Step 8: Plan the emergency evacuation route of the campus according to the fire evacuation situation. In the fire evacuation, open public squares and areas with convenient transportation should be given priority, and densely populated areas should be avoided. The route planning aims at the lowest risk evacuation, and the elements include corridors, squares in front of the school, school gates, etc. According to the actual impact of each element on the evacuation efficiency, the weight is assigned to obtain the emergency evacuation route plan. An AnyLogic agent model is established to simulate the evacuation scenario, evaluate the evacuation time, safety, and congestion, and obtain the final emergency evacuation route plan.

Claims

1. An evacuation route planning system based on spontaneous geographic information, characterized in that The system includes: a basic data acquisition module, a basic data preprocessing module, an element set acquisition module, an element library construction module, and a spatial element preference acquisition module; The basic data acquisition module uses the location name of the target area to search and acquire the basic data of the target area on the social media platform; The target area basic data includes: description text and image data; The basic data preprocessing module is used to preprocess the basic data of the target area; The element set acquisition module acquires a text element set and a picture element set using the preprocessed target area basic data; The element library construction module constructs an element library using a text element set and a picture element set; The spatial element preference acquisition module is used to assign weights to elements in the element library, thereby obtaining element preferences in the target area; The evacuation path planning module plans a safe evacuation path in the target area by utilizing the element preferences in the target area.

2. The evacuation route planning system based on spontaneous geographic information according to claim 1, characterized in that: The basic data acquisition module uses the location name of the target area to search and acquire the basic data of the target area on the social media platform, specifically: Use the location name and coordinates of the target area to retrieve data on social media platforms and open source map platforms, use the automated web crawler program WebSpider to crawl the basic data of the target area, and convert the description text in the basic data of the target area into csv format; The target area basic data includes: description text and image data.

3. The evacuation route planning system based on spontaneous geographic information according to claim 2, characterized in that: The basic data preprocessing module is used to preprocess the basic data of the target area, specifically: A1. Preprocessing the description text in the basic data of the target area to obtain the preprocessed description text; A2. Use the image management tool VisiPics to filter the image data in the basic data of the target area, remove the image data that exceeds the preset comparison similarity threshold, and obtain the pre-processed image data.

4. The evacuation route planning system based on spontaneous geographic information according to claim 3, characterized in that: The preprocessing of the description text in the target area basic data in A1 to obtain the preprocessed description text is specifically as follows: A1-1. Represent the description text in the target area basic data in matrix form to obtain a description text matrix, and then use the toarray function to convert the description text matrix into an array form, and use the array form of the description text matrix as a description text vector; A1-2. Remove duplicate description text vectors to obtain preprocessed description text: The scikit-learn machine learning library of Python is used to calculate the cosine similarity between every two description text vectors. If the cosine similarity value of any two description text vectors is greater than the similarity threshold, any vector of the current two description text vectors is eliminated.

5. The evacuation route planning system based on spontaneous geographic information according to claim 4, characterized in that: In A1-1, the description text in the target area basic data is represented as a matrix using the CountVectorizer function.

6. The evacuation route planning system based on spontaneous geographic information according to claim 5, characterized in that: The element set acquisition module acquires the text element set using the preprocessed target area basic data, specifically: B1-1. Segment each description text after preprocessing, and connect each word into a new string through spaces; B1-2. Calculate the term frequency-inverse document frequency TF-IDF of each word in each description text, so as to obtain the TF-IDF value vector of each description text; B1-3. Use the get_feature_names_out() function and the preprocessed description text to obtain the feature vocabulary; B1-4, obtain the intersection of the word set in the current description text and the word set in the feature vocabulary, sort each word in the intersection from large to small according to the TF-IDF value, and use the first n words as the keywords of the description text i; B1-5. Integrate all keywords describing the text and manually delete keywords irrelevant to spatial elements to obtain a text element set.

7. The evacuation route planning system based on spontaneous geographic information according to claim 6, characterized in that: The picture element set is obtained in the following way: B2-1. Use Google Vision API to identify spatial elements in preprocessed image data and use the spatial elements as labels for the corresponding preprocessed image data; The spatial elements include: objects, scenes, and activities; B2-2. Count the frequency of label occurrence of all preprocessed image data, delete labels with a frequency threshold less than the threshold, and manually remove labels irrelevant to spatial elements to obtain a picture feature set.

8. The evacuation route planning system based on spontaneous geographic information according to claim 7, characterized in that: The element library construction module constructs an element library using a text element set and a picture element set, specifically: First, the labels in the image feature set are compared with the text feature set to obtain common features and establish a feature library; Then, the elements in the element library are classified and stored according to the element type; Element types include: buildings, roads, natural landscapes, and indoor spaces.

9. The evacuation route planning system based on spontaneous geographic information according to claim 8, characterized in that: The spatial element preference acquisition module is used to assign weights to elements in the element library, specifically: The larger the weight of the current type of element is, the higher the preference for the current type of element is.

10. The evacuation route planning system based on spontaneous geographic information according to claim 9, characterized in that: The evacuation path planning module plans a safe evacuation path in the target area by using the element preference in the target area, specifically: First, sort the elements of the target area from large to small according to their weights; Then, multiple evacuation paths are set, the weights and sums of the elements included in each evacuation path are obtained, and the path consisting of the elements with the highest weight sum is obtained as the evacuation path.

Images