Recommendation service method for site selection of investment attraction park based on three-dimensional GIS (Geographic Information System) map

Abstract

The invention discloses an investment attraction park site selection recommendation service method based on a three-dimensional GIS map, and the method comprises the steps: obtaining park information, park demands and investment demands, and carrying out the calculation according to the park information through employing a point aggregation method, and obtaining a park three-dimensional GIS map; constructing a park recommendation model, training the park recommendation model, obtaining a park demand feature vector after training is completed, inputting the demand feature vector to the park recommendation model, and outputting a park recommendation result by the park recommendation model; and constructing a site selection recommendation model, inputting an investment demand to the site selection recommendation model, and outputting a site selection recommendation result. According to the method, the three-dimensional GIS map is generated through the park address information, calculation recommendation is performed according to the park demand and the investment demand of the user, and the panoramic space is generated according to the recommendation result, so that multiple factors are comprehensively considered, recommendation is visually performed for the user, the recommendation result is visualized, and investors are helped to better understand the park condition and potential advantages.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a site selection recommendation service method for an investment park based on a three-dimensional GIS map. Background Art

[0002] As an important platform for attracting investment and promoting industrial development, the rationality of the site selection of the park investment attraction platform is directly related to the competitiveness and attractiveness of the park. However, traditional site selection methods often rely on experience and limited data support, and it is difficult to fully consider various factors and potential risks of the park. In addition, with the acceleration of urbanization and the increasing shortage of land resources, how to scientifically and efficiently select sites has become an urgent problem to be solved. The disadvantages of using traditional two-dimensional investment attraction park site selection recommendation services are as follows:

[0003] 1. Unable to quickly grasp the supporting environment such as the topography, transportation network, surrounding facilities, and environmental conditions around the park, and lack a more intuitive experience and feeling of the soft power and supporting resources of the park's surrounding environment;

[0004] 2. Traditional site selection methods lack scientific quantitative evaluation tools, making it difficult to accurately evaluate the key indicators of the park;

[0005] 3. The display effect of traditional recommended site selection is not novel enough, and it is impossible to display the recommended parks more intuitively and eye-catchingly on the website according to the needs and preferences of investors.

[0006] Therefore, there is a need for a method that can quantitatively combine multiple factors of the park to comprehensively and efficiently select a site and can intuitively display the site selection. Summary of the invention

[0007] To solve the above problems, the present invention provides a site selection recommendation service method for investment park based on three-dimensional GIS map. By generating three-dimensional GIS map and DeepFM model, site selection recommendation is made according to user needs, and various factors of the park are integrated to solve the problems that the existing site selection method is not intuitive enough and the evaluation method is unscientific.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for recommending a location for an investment park based on a three-dimensional GIS map, comprising:

[0010] Obtain park information, park demand and investment demand, and use point aggregation method to calculate the park 3D GIS map based on the park information;

[0011] Construct a park recommendation model, input a data set into the park recommendation model, train the park recommendation model, extract features of park demand after training, obtain a park demand feature vector, input the demand feature vector into the park recommendation model, the park recommendation model outputs the park recommendation result, and annotates the park recommendation result in the park 3D GIS map;

[0012] According to the park recommendation results, the investment resource information of the park recommendation results is obtained, a site selection recommendation model is constructed, and the site selection recommendation model is trained. After the training is completed, the investment demand is input into the trained site selection recommendation model, and the site selection recommendation results are output.

[0013] Furthermore, the three-dimensional GIS map of the park is obtained by calculating the park information using the point aggregation method, and the specific implementation method is as follows: the park information includes address data and building data, and the building data includes model data and material data of each building, and the address data is converted into a three-dimensional GIS coordinate system;

[0014] Draw a grid on the three-dimensional GIS coordinate system, match and classify the grid according to the address data, calculate the centroid point of each grid after classification, and obtain the park grid map;

[0015] According to the park grid map and building data, a three-dimensional model is generated on the three-dimensional GIS coordinate system to obtain the three-dimensional GIS map of the park.

[0016] Furthermore, the park recommendation model is a DeepFM model.

[0017] Furthermore, the data set is input into the park recommendation model to train the park recommendation model, and its specific implementation method includes: the data set includes a park supporting data set and a corresponding park evaluation data set, and the data set is preprocessed, and the preprocessing includes data cleaning and word segmentation; the park recommendation model is trained using the preprocessed data set, and the parameters of the park recommendation model are adjusted according to the output results of the park recommendation model.

[0018] Furthermore, the park recommendation model outputs park recommendation results, and marks the park recommendation results in the three-dimensional GIS map of the park. The specific implementation method includes: the park recommendation model outputs the matching degree between each park and the park demand, and the three-dimensional models of several parks with the highest matching degree are highlighted and displayed in the three-dimensional GIS map of the park.

[0019] Furthermore, after marking the park recommendation results in the three-dimensional GIS map of the park, it also includes: generating a panoramic space of the park recommendation results, specifically including: obtaining a panoramic picture of the recommended park recommendation results, cutting the panoramic picture according to quadtree slicing to obtain a number of slices, creating a rendering object, the rendering object is a sphere or a hexahedron, and loading and rendering the slices inside the rendering object according to the slicing order to obtain a panoramic space.

[0020] Furthermore, slice loading and rendering are performed inside the rendering object in a multi-threaded asynchronous manner according to the slice order.

[0021] Furthermore, after generating the panoramic space of the park recommendation results, it also includes: determining the longitude and latitude of the center point of the panoramic space, establishing a coordinate system based on the center point of the panoramic space, obtaining building information of the park recommendation results, the building information including the longitude and latitude of each building in the park, the enterprises in the buildings and the address information of the enterprises, calculating the vertical angle and azimuth angle of each enterprise with the center point based on the center point and the address information of the enterprise, and marking the position of each enterprise in the panoramic space using the vertical angle and the azimuth angle.

[0022] Furthermore, the investment demand is input into the trained site recommendation model, and the site recommendation result is output. The specific implementation method is as follows: the investment demand is quantified, and the quantified investment demand is converted into an investment feature vector through feature extraction, and the investment feature vector is input into the site recommendation model. The site recommendation model calculates the matching degree between the investment resource information and the investment demand according to the investment feature vector, and recommends the investment resource according to the matching degree to obtain the site recommendation result.

[0023] Furthermore, in the panoramic space, the labels of corresponding enterprises are highlighted according to the site recommendation results.

[0024] Through the above technical scheme, the present invention has the following beneficial effects: a three-dimensional GIS map is generated through the park address information, and calculation and recommendation are performed according to the user's park needs and investment needs, and a panoramic space is generated according to the recommendation results, thereby comprehensively considering various factors, intuitively recommending to users, and visualizing the recommendation results to help investors better understand the situation of the park and its potential advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The present invention is a schematic diagram of the overall process of a method for recommending a location for investment park based on a three-dimensional GIS map. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] See also Figure 1 , a site selection recommendation service method for investment attraction park based on three-dimensional GIS map, comprising:

[0029] Obtain park information, park demand and investment demand, and use point aggregation method to calculate the park 3D GIS map based on the park information;

[0030] Construct a park recommendation model, input a data set into the park recommendation model, train the park recommendation model, extract features of park demand after training, obtain a park demand feature vector, input the demand feature vector into the park recommendation model, the park recommendation model outputs the park recommendation result, and annotates the park recommendation result in the park 3D GIS map;

[0031] According to the park recommendation results, the investment resource information of the park recommendation results is obtained, a site selection recommendation model is constructed, and the site selection recommendation model is trained. After the training is completed, the investment demand is input into the trained site selection recommendation model, and the site selection recommendation results are output.

[0032] In an optional embodiment, the three-dimensional GIS map of the park is calculated based on the park information using the point aggregation method, and its specific implementation method is as follows: the park information includes address data and building data, and the building data includes model data and material data of each building, and the address data is converted into a three-dimensional GIS coordinate system;

[0033] Draw a grid on the three-dimensional GIS coordinate system, match and classify the grid according to the address data, calculate the centroid point of each grid after classification, and obtain the park grid map;

[0034] According to the park grid map and building data, a three-dimensional model is generated on the three-dimensional GIS coordinate system to obtain the three-dimensional GIS map of the park.

[0035] The point set aggregation algorithm based on grid distance is used to obtain the information of parks with investment demand and supporting facilities around the parks, and points are aggregated on the three-dimensional GIS map according to the latitude and longitude information of the parks and supporting facilities around the parks.

[0036] In an optional embodiment, it also includes: obtaining the viewport range of the current viewport, obtaining the viewport range according to the current viewport, performing a matrix conversion of the three-dimensional GIS geographic location in the scene from the screen to three-dimensional coordinates through the four points at the corners of the terminal where the viewport is located, filtering out grid range data according to the obtained three-dimensional geographic coordinates of the viewport, loading the grid data, and thus displaying a three-dimensional GIS map to the user.

[0037] The vector data that can generate the model obtained from the spatial database includes data such as park height, material attributes, and building type. This type of data is assembled through dynamic loading to construct three-dimensional spatial data. The system presets various types of three-dimensional model related data such as residential buildings, apartments, shopping malls, factories, office buildings, etc. The three-dimensional model related data is preset to level L2. The dynamic data is matched with the preset value and then returned to the client for rendering by the web page three-dimensional engine as follows:

[0038] Get the height and material properties of the park and surrounding facilities, prepare the 3D model in advance, and load the 3dTiles basic model to be rendered on the client in advance;

[0039] Through type matching, the vector data to be generated and the preset values ​​of the three-dimensional model data such as the building type and material mapping are integrated, constructed and returned;

[0040] After receiving the height and material data information of the model, the loader of the three-dimensional model engine is used to assemble the vertex data corresponding to the model data and load the material map rendering of the model surface.

[0041] In an optional embodiment, the campus recommendation model is a DeepFM model.

[0042] The DeepFM model consists of DNN (deep neural network) and FM (factorization machine) parts. The factorization machine learns a representation vector for each dimension of the feature, and then sets the weight of the product of the representation vectors of two features to the dot product of their respective representation vectors. By decomposing the cross-feature coefficients, different cross-terms are no longer independent, so the data of one cross-term can assist in predicting another cross-term. The FM algorithm can handle large-scale sparse data and maintain sufficiently good generalization performance. At the same time, it can automatically learn the information brought by feature cross-talk.

[0043] The neural network part and the factorization machine part in the DeepFM model share the same input vector, which enables DeepFM to learn low-order and high-order feature interactions from the original input features. It has linear features, low-order cross features and high-order features, and does not require manual low-order cross features, which can be trained faster and more accurately.

[0044] In an optional embodiment, the data set is input into the park recommendation model to train the park recommendation model, and its specific implementation method includes: the data set includes a park supporting data set and a corresponding park evaluation data set, and the data set is preprocessed, and the preprocessing includes data cleaning and word segmentation; the park recommendation model is trained using the preprocessed data set, and the parameters of the park recommendation model are adjusted according to the output results of the park recommendation model.

[0045] The park supporting data set includes the park's land area, plot ratio, completion time, number of buildings, floor height, floor weight, number of objects, object load, number of freight elevators, freight elevator load, whether there is central air conditioning, etc. After data preprocessing, the park supporting data set is converted into standard data and stored in the ES database through ETL. During training, the park recommendation model is trained by loading data through the ES database.

[0046] In an optional embodiment, the park recommendation model outputs park recommendation results, and marks the park recommendation results in the three-dimensional GIS map of the park. The specific implementation method includes: the park recommendation model outputs the matching degree between each park and the park demand, and the three-dimensional models of several parks with the highest matching degree are highlighted and displayed in the three-dimensional GIS map of the park.

[0047] The park requirements include region, rent, industry, area, floor height, load-bearing capacity, elevator, etc. After data preprocessing, the above information is converted into feature vectors through feature extraction. The park requirement feature vector is sent to the park recommendation model, and the recommendation score of each park is obtained through similarity and score calculation as the matching degree between the end user's needs and the park, and recommendations are made according to the matching degree.

[0048] In the 3D GIS map of the park, the 3D models of several parks with the highest matching degree are highlighted and selected for display, including: the position of the model in the scene can be obtained by loading the model in the 3D scene through the model's position information, the model object is added and marked with a bounding box, the highlight data of the corresponding model is read, and the overall material of the model is modified using the material modification data.

[0049] In an optional embodiment, after marking the park recommendation results in the three-dimensional GIS map of the park, it also includes: generating a panoramic space of the park recommendation results, specifically including: obtaining a panoramic image of the recommended park recommendation results, cutting the panoramic image according to quadtree slicing to obtain a number of slices, creating a rendering object, the rendering object is a sphere or a hexahedron, and loading and rendering the slices inside the rendering object according to the slicing order to obtain a panoramic space.

[0050] Krpano's slicing first requires a panoramic image file as input. Krpano will analyze the panoramic image to determine its basic information such as resolution and format. Users can specify parameters such as the size, format, and output directory of the slice. The slice will be set to a specific size, such as 256x256 pixels, and the size can be adjusted according to the view zoom level. The slicing principle is based on the quadtree and krpano classifies the orientation files of the panoramic angles. After calculating the segmentation size of the slice, the panoramic image input by the user is cut and stored in the pre-classified file. During the slicing process, the user's specific configuration can be personalized for the slice. Krpano uses the threejs engine at the bottom layer to render the sliced ​​panorama, create a rendering object, which is a sphere or a hexahedron, and loads and renders a texture inside the rendering object created by the engine according to the internally defined slice loading order.

[0051] In an optional embodiment, slice loading and rendering are performed in a multi-threaded asynchronous manner in the slice order within the rendering object.

[0052] Using asynchronous multithreading helps to make full use of computing resources, improve loading and rendering efficiency, and save rendering time.

[0053] In an optional embodiment, after generating a panoramic space of the park recommendation results, it also includes: determining the longitude and latitude of the center point of the panoramic space, establishing a coordinate system based on the center point of the panoramic space, obtaining building information of the park recommendation results, the building information including the longitude and latitude of each building in the park, the enterprises in the buildings and the enterprise address information, calculating the vertical angle and azimuth angle of each enterprise with the center point based on the center point and the enterprise address information, and marking the position of each enterprise in the panoramic space using the vertical angle and azimuth angle.

[0054] The enterprise address information includes the enterprise's address, building, floor, industry, enterprise label, etc.

[0055] The annotation implementation method is as follows: the panoramic scene loads the panoramic material, establishes the panoramic scene coordinate system, sets the direction of the current viewing angle center point to due north, and determines the coordinate scale of the current panoramic space; obtains the longitude, latitude and altitude information of the panoramic observation point, and obtains the geographic marker point data that needs to be marked on the three-dimensional map, wherein the geographic marker point data includes longitude, latitude and altitude; uses spherical distance and trigonometric functions to calculate the vertical angle between the panoramic space and the marker point based on the longitude, latitude and altitude data and the longitude, latitude and altitude data of the marker point; uses the direction angle algorithm to calculate the azimuth angle of the longitude position of the scene where the current marker point is located based on the longitude, latitude and altitude and the longitude and latitude of the marker point.

[0056] The coordinates are obtained by the above method to obtain the horizontal and vertical axis positions. Based on this method, the dynamic point recording interface or static writing of the mark point data can be used to realize batch automatic addition of mark points in the panoramic space.

[0057] In an optional embodiment, the investment demand is input into a trained site recommendation model, and the site recommendation result is output. The specific implementation method is: quantify the investment demand, convert the quantified investment demand into an investment feature vector through feature extraction, input the investment feature vector into the site recommendation model, and the site recommendation model calculates the matching degree between the investment resource information and the investment demand according to the investment feature vector, recommends the investment resource according to the matching degree, and obtains the site recommendation result.

[0058] The site recommendation model is a DeepFM model. The investment requirements include investment industry type, intended amount, preferential policies, floor height, orientation and other information. The rental information (rented enterprises and unrented) and the extracted park information are preprocessed through data cleaning, text segmentation, data conversion and other means, converted into standard data, and stored in the ES database through ETL. The data in the ES database is loaded, and the site recommendation model is trained based on the DeepFM model. The model structure and model training parameters are adjusted in combination with the model training results and test results to obtain a trained site recommendation model.

[0059] Obtain the user's investment intention and extract relevant information in the intention, such as industry type, intended amount, preferential policies, floor height, orientation, etc. After data preprocessing, convert the above information into feature vectors through feature extraction. Send the user's demand feature vector to the park building floor recommendation model, and obtain the recommendation score of each floor through similarity and score calculation as the matching degree between the final user's investment intention and the floor, and make recommendations according to the matching degree.

[0060] In an optional embodiment, in the panoramic space, the labels of corresponding enterprises are highlighted according to the site recommendation results.

[0061] After displaying the label of the result according to the system's intelligent recommendation, obtain the floor data in the panorama to identify the label that needs to be highlighted, create a dynamic image object supported by krpano, and load and display it in the panorama by adding dynamic images combined with the mark location information that needs to be highlighted.

[0062] The embodiment disclosed in this specification is only an example of a unilateral feature of the present invention, and the protection scope of the present invention is not limited to this embodiment, and any other functionally equivalent embodiments fall within the protection scope of the present invention. For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A method for recommending the location of an investment park based on a three-dimensional GIS map, characterized in that: include: Obtain park information, park demand and investment demand, and use point aggregation method to calculate the park 3D GIS map based on the park information; Construct a park recommendation model, input a data set into the park recommendation model, train the park recommendation model, extract features of park demand after training, obtain a park demand feature vector, input the demand feature vector into the park recommendation model, the park recommendation model outputs the park recommendation result, and annotates the park recommendation result in the park 3D GIS map; According to the park recommendation results, the investment resource information of the park recommendation results is obtained, a site selection recommendation model is constructed, and the site selection recommendation model is trained. After the training is completed, the investment demand is input into the trained site selection recommendation model, and the site selection recommendation results are output.

2. The method for recommending the location of an investment park based on a three-dimensional GIS map according to claim 1, characterized in that: The point aggregation method is used to calculate the three-dimensional GIS map of the park according to the park information, and its specific implementation method is: the park information includes address data and building data, and the building data includes model data and material data of each building, and the address data is converted into a three-dimensional GIS coordinate system; Draw a grid on the three-dimensional GIS coordinate system, match and classify the grid according to the address data, calculate the centroid point of each grid after classification, and obtain the park grid map; According to the park grid map and building data, a three-dimensional model is generated on the three-dimensional GIS coordinate system to obtain the three-dimensional GIS map of the park.

3. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 1, characterized in that: The park recommendation model is a DeepFM model.

4. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 1, characterized in that: The data set is input into the park recommendation model to train the park recommendation model. The specific implementation method includes: the data set includes a park supporting data set and a corresponding park evaluation data set, and the data set is preprocessed, and the preprocessing includes data cleaning and word segmentation; the park recommendation model is trained using the preprocessed data set, and the parameters of the park recommendation model are adjusted according to the output results of the park recommendation model.

5. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 1, characterized in that: The park recommendation model outputs park recommendation results, and marks the park recommendation results in the park's three-dimensional GIS map. The specific implementation method includes: the park recommendation model outputs the matching degree between each park and the park demand, and the three-dimensional models of several parks with the highest matching degree are highlighted and displayed in the park's three-dimensional GIS map.

6. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 1, characterized in that: After marking the park recommendation results in the three-dimensional GIS map of the park, it also includes: generating a panoramic space of the park recommendation results, specifically including: obtaining a panoramic picture of the recommended park recommendation results, cutting the panoramic picture according to quadtree slicing to obtain a number of slices, creating a rendering object, the rendering object is a sphere or a hexahedron, and loading and rendering the slices inside the rendering object according to the slicing order to obtain a panoramic space.

7. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 6, characterized in that: Slices are loaded and rendered in the rendering object in slice order in a multi-threaded asynchronous manner.

8. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 6, characterized in that: After generating the panoramic space of the park recommendation results, it also includes: determining the longitude and latitude of the center point of the panoramic space, establishing a coordinate system based on the center point of the panoramic space, obtaining the building information of the park recommendation results, the building information includes the longitude and latitude of each building in the park, the enterprises in the buildings and the address information of the enterprises, and calculating the vertical angle and azimuth angle of each enterprise with the center point according to the center point and the address information of the enterprise, and marking the position of each enterprise in the panoramic space using the vertical angle and the azimuth angle.

9. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 8, characterized in that: The method of inputting investment demand into the trained site recommendation model and outputting site recommendation result is specifically implemented as follows: quantifying the investment demand, converting the quantified investment demand into an investment feature vector through feature extraction, inputting the investment feature vector into the site recommendation model, and the site recommendation model calculates the matching degree between the investment resource information and the investment demand according to the investment feature vector, recommends the investment resource according to the matching degree, and obtains the site recommendation result.

10. The method for recommending the location of an investment attraction park based on a three-dimensional GIS map according to claim 9, characterized in that: In the panoramic space, the corresponding enterprise labels are highlighted according to the site recommendation results.

Images