CN108197134A - Grouped point object automatic Synthesis algorithm under big data support - Google Patents

Abstract

In the automatic synthesis algorithm of point group targets supported by big data, the weight of points is an important parameter of point group automatic synthesis, but the existing algorithms lack sufficient theoretical basis for its determination. Big data can provide rich and real-time data references for point weights. To this end, an automatic synthesis algorithm of point group targets supported by big data is proposed. The basic principles are as follows: firstly, two factors, the scope of influence and the population of influence, are introduced as the weight index of points, and the relevant data are obtained, processed and visualized; Based on the quantity, the selection operation of the points is realized by using the normalization and "concentric circle" methods; finally, the comprehensive results and the visual graphics are compared and analyzed and verified by experiments. Experiments show that this algorithm not only inherits the advantages of existing algorithms, but also scientifically calculates the weight information of point groups and applies it to the automatic synthesis of point groups. The results obtained are more reasonable and have strong real-time performance.

Description

Automatic synthesis algorithm of point group targets supported by big data

technical field

The invention belongs to the field of cartography and geographic information science and technology, and is an automatic synthesis algorithm of point group targets based on big data.

Background technique

Point group synthesis is an important part of map synthesis. Its purpose is to correctly express the overall information of the point group as much as possible while the number of points is reduced, that is, to use a certain model or algorithm to extract from the original point group subcollection. The weight of a point reflects the individual importance of a point in the overall spatial point group, and is an important parameter for point group synthesis. Among the existing point group synthesis algorithms, there is a class of algorithms, such as algorithms based on Voronoi diagrams, which do not take into account the weight information of points in the synthesis process; although the other class of algorithms take into account the weight information of points, they still have the following Two defects: (1) The determination of the weight value is not precise enough and lacks sufficient calculation basis. For example, the algorithm based on the weighted Voronoi diagram is a relatively complete point group synthesis algorithm so far, but the weight value of the midpoint of the algorithm is obtained through expert experience. Given, for example, set the weight of Grade A hospital to 2, the weight of Grade B hospital to 1, the weight of irrigation wells to 2, and the weight of other wells to 1; (2) The weight value is given in advance and There will be no change. However, in the actual geographical space, the weight value of the point will often change with the passage of time. For example, a Grade B hospital represented by a point in the map aims at the current situation of a large number of cardiovascular and cerebrovascular cases. The introduction of talents vigorously develops corresponding specialties. As a result, its scale and the number of medical patients have surpassed many first-class hospitals, and its weight value should also change accordingly.

Big data can provide rich real-time information for the target things, not only can provide scientific and effective reference for the determination of the weight value of the point, but also make it possible to change the weight value of the point in real time.

Contents of the invention

In view of the above situation, this paper introduces big data as the basis for calculating point group weights, and proposes an automatic synthesis algorithm for point group targets on this basis.

Point weight data selection

Point groups are an important part of spatial geographic objects. When the scale of the map is reduced, many spatial objects appear as point groups, such as residential areas, hospitals, schools, supermarkets, restaurants, trees, etc. Most of the above point groups have their own scope of influence and other characteristics. Point groups such as hospitals, schools, and supermarkets have their own scope of influence and affected groups. In view of this, this paper uses the influence range and influence population of the point as the index to measure the weight of the point.

These two factors are selected as the weight index of the point because the scope of influence of the point and the affected population and the relationship between the two reflect the two properties of the point group:

(1) The level of the point object. For a point object, there is usually a service range spreading around it, and this range can reflect the level of the point object. If the range is relatively large and the number of affected people is large, it means that the level of this point is high. For example, a hospital with a wide range of influence and a large number of affected people can indicate that the hospital has a large scale, advanced medical equipment, and a high level of health services. In daily life, the level corresponding to this kind of hospital is also higher; on the contrary, if the point of influence corresponding to the point is small and the number of affected people is small, it means that the level of this point is low;

(2) The local density of the point group. If a point object has a wide range of influence and a small number of people, it means that the number of similar point groups near the point is small, and the local density is small.

In the point synthesis process, the points with higher grades have relatively higher importance and should be retained; the points with lower local density should also be retained in order to maintain the structural characteristics of the spatial point group as much as possible.

Therefore, in the algorithm, the data acquired and processed are mainly the influence range and the affected population of the spatial point group.

Acquisition and processing of big data

In this algorithm, two weight measurement indicators are selected, and their acquisition methods are as follows:

(1) Scope of Influence: The first step in obtaining data on the scope of influence is to obtain the source information of the affected people. This location information can be obtained through mobile phone trajectory tracking data on the cellular network, extracted using web crawlers, or obtained based on network data streams, etc. , of course, can also be obtained through the internal information of the corresponding point. For example, for the scope of influence of a hospital, the trajectory can be tracked starting from the hospital, and the end point of the trajectory can be recorded as the source of the affected population, which can also be obtained through the registration information of the hospital.

(2) Number of affected people: With the popularity of the Internet and social software, data on the number of affected people can be crawled through Tencent location data or Weibo check-in data, or can be obtained by using the internal information of the site. For example, the data on the number of people affected by a hospital can be approximated by using Tencent’s location big data, or can be approximated by using Weibo check-in data within a specified range around a certain point within a certain period of time.

In order to obtain more accurate and effective data, it is required to clean the original data. The algorithm in this paper mainly extracts valid data from the following two aspects: (1) Check the validity. Check the validity of the data to see if it is within a reasonable range, for example, the number of people affected cannot be negative; (2) Check the applicability. For the collected data points in the scope of influence, import them into the ArcGIS analysis platform according to their latitude and longitude, overlap with the scope of the study area, and delete the points outside the study area.

Representation of Influence Area: The Influence Area is represented by an Influence Area Polygon around a point. This algorithm stipulates that the polygon of the scope of influence is represented by the distribution boundary polygon formed by the points projected onto the map from the source of the affected people ^[ . The construction process of the influence area polygon is as follows:

Step1: There will be a lot of chance to affect the source of the crowd. In order to avoid the impact of this chance on the range, record the frequency of the collected data points and sort them in descending order, discard the 10% points with the lowest frequency, and put the remaining The collected points are projected to plane coordinates as the original point group.

Step2: Scan the original point group, use the constrained Delaunay triangulation to construct its constrained Delaunay triangulation, and then introduce the dynamic threshold "peeling" method to construct its distribution boundary polygon.

Store the original point group into the array pointArray, obtain the area of the distribution boundary polygon corresponding to each point in the array and store it in the array areaArray in turn.

(2) Representation of the number of affected people: the size of the affected population is determined by the number of people who visit the area where the point is located within a period of time. In order to facilitate the representation and reduce the error, this algorithm uses the hierarchical clustering algorithm to cluster the data of the affected people at each point, so that the difference in the number of affected people at points in the same cluster is small, and the number of affected people at points between different clusters is small. The number of affected populations varies greatly. Express it on the map, which is represented by the gradient filling color of the affected area polygon. The darker the color, the larger the number of people affected by the corresponding point, and vice versa, the smaller the number of people affected by the corresponding point.

Similarly, the number of affected people corresponding to each point in the point array pointArray is stored in the array numArray.

3 trade-offs

The weight of the points in the algorithm is determined by two factors: the polygonal area of the influence range and the number of people affected. On this basis, the points are divided into three types: high-level must be retained (Type I), low-level discarded directly (Type II), Participate in selection competition between the two (type Ⅲ). Define two selection constraints:

(1) Level constraint conditions. According to the scope of influence and the number of affected people, three types of points are distinguished. During the selection process, type I points are retained, and type II points are directly deleted. in:

TypeⅠ={ P _i | P _i corresponds to a large range of influence or a large number of people affected by it or (a large range of influence and a large number of people affected by it)}

TypeⅡ={ P _i | P _i corresponds to a small area of influence and a small number of affected people}

(2) Proximity constraint conditions. For type III points, sort them in ascending order according to the number of affected people per unit area of influence area obtained by formula (1) and delete them one by one. If a point is deleted, it will be adjacent to its area of influence polygon The point corresponding to the polygon is "solidified", and the next non-"solidified" point is found and processed until the deletion end condition is met; if all points within the range are "solidified" and the deletion end condition is still not met, then unfreeze ", and perform the second and third delete operations in turn.

point trade-off algorithm

According to the above constraints, the type II point is a point with a relatively small influence range and the number of affected people in the point group, and its selection is simpler than that of type I and type III points. Strategy:

(1) According to the square root law, select a specified number of type I points and delete them, and the remaining points constitute the comprehensive result;

(2) In order to realize the comparison between different dimensional data, the polygonal area of the scope of influence and the number of people affected are converted into dimensionless scalars by normalization;

(3) Express the normalized results in a plane coordinate system with the number of affected people as the abscissa and the polygon area of the affected area as the ordinate. At this time, the point with a smaller weight value (type II point) is closer to the origin of the coordinate. Therefore, the method of making 1/4 concentric circles with the origin as the center is used, and the points with smaller weight values are selected in turn, and the judgment and deletion operations are performed on them. This is called the "concentric circle" method later.

The specific steps of point deletion are described as follows:

(1) Obtain the number n of pre-deleted points in the synthesis process according to the square root law;

(2) Establish a plane Cartesian coordinate system with the number of people affected as the horizontal axis and the polygonal area of the affected area as the vertical axis, normalize the elements in the array numArray of the number of people affected and the areaArray of the polygon area of the affected range respectively, and calculate the results One-to-one correspondence means in the coordinate system, at this time each point in the coordinate system corresponds to the weight attribute value of a point in the original point group;

(3) Use the coordinate origin as the origin, draw a 1/4 circle in the coordinate system with the minimum value of the weight attribute value point and the coordinate origin as the initial radius, and add a delete flag "flag='D' to the point on the arc , store it in the "Type II" point array secondType, and add a reserved flag "flag='R'" to the points corresponding to the neighbor polygons of the affected range polygon;

(4) Take the minimum plane distance between the weight attribute value points in the plane coordinate system as an increment, update the radius value, make a 1/4 concentric circle with the origin as the center, and connect the arc and the previous concentric circle Add deletion marks to the points without any marks in the formed circle, and still store them in the array secondType, and add reservation marks to the points corresponding to the neighbor polygons of the affected range polygon;

(5) Compare n and the size of the number of elements in the secondType array, if the value of n is large, return to (4); if the two values are the same, go to (6); otherwise, remove the points added to the secondType array in the previous round from the array and remove the deletion marks of the corresponding points, calculate the number of people affected per unit area n _p ^' of these points according to the above formula 2, sort them in ascending order, and give the one with the smallest n _p ^' value in the sequence without any mark from front to back Add a deletion mark to the point, store it in the array secondType, and add a reservation mark to its polygon neighbors in the scope of influence, until the number of elements in the secondType array is the same as the n value, then go to (6);

(6) Delete the points corresponding to all elements in the secondType array in the original point group, and the remaining point group is the comprehensive result, and the algorithm ends.

Description of drawings

Figure 1 is the polygon construction of the scope of influence

Figure 2 is the clustering of affected populations

Figure 3 is the algorithm flow chart

Figure 4 is the comprehensive process of the supermarket point group

Figure 5 is the point deletion process

Figure 6 is the comprehensive process of some hospitals in a certain place

Figure 7 is the comprehensive process of a gas station in a certain place

Figure 8 is a comprehensive process of traffic lights in a certain place.

Claims (1)

1. A point group target automatic synthesis algorithm under the support of a kind of big data, its feature comprises the following steps: (1) Acquisition and processing of weight data Scope of influence and data acquisition of affected population The scope of influence of the supermarket is based on the supermarket as the starting point, and is obtained by crawling the end point identification information of the mobile phone trajectory data and the mobile phone signal positioning information within one month; The sum of Bo sign-in data; data processing Clean the obtained information, delete the data points outside the study area and the wrong ones; perform hierarchical clustering analysis on the value of the number of people affected by all points in the original point group after processing, and group the points with similar numbers of people affected by them into one class, and increase the data difference between classes as much as possible; Mark the points where the supermarket is located on the ArcGIS platform, and construct the polygons of the influence range of these points; sort the number of affected people in descending order according to the clustering results and select the gradient color from deep to light to fill the polygons of the range of influence of the corresponding points; (2) trade-offs Definition and Expression of Constraints The weight of the points in the algorithm is determined by two factors: the polygonal area of the influence range and the number of people affected. On this basis, the points are divided into three types: high-level must be retained (Type I), low-level discarded directly (Type II), Participate in selection competition between the two (type Ⅲ); Define two selection constraints: Level constraint conditions, distinguish three types of points according to the scope of influence and the number of people affected, keep type I points during the selection process, and directly delete type II points; where: TypeⅠ={ P _{i |} A large number or (the scope of its influence is large and the number of people it affects is large)} TypeⅡ={ P _i | P _i corresponds to a small area of influence and a small number of affected people} Adjacent relationship constraint conditions, for type III points, they are sorted in ascending order according to the number of people affected by the unit area of influence area and then deleted one by one. And process the next non-"solidified" point until the deletion end condition is met; if all points within the range are "solidified" and the deletion end condition is still not met, unfreeze the "solidified" point, and perform the second pass and the third pass in sequence Delete operations over and over; point trade-off algorithm According to the above constraints, the type II point is a point with a relatively small influence range and the number of affected people in the point group, and its selection is simpler than that of type I and type III points. Strategy: a. According to the square root law, select a specified number of Type I points and delete them, and the remaining points constitute the comprehensive result fruit; b. In order to realize the comparison between data of different dimensions, the polygonal area of the scope of influence and the number of people affected are converted into dimensionless scalars by normalization method; c. Express the normalized results in a plane coordinate system with the number of affected people as the abscissa and the polygon area of the influence area as the ordinate. At this time, the point with a smaller weight value (Type II point) is closer to the origin of the coordinate, so Use the method of making 1/4 concentric circles with the origin as the center of the circle, select points with smaller weight values in turn, and perform judgment and deletion operations on them, which will be called the "concentric circle" method later; The specific steps of point deletion are described as follows: A. obtain the number n of pre-deleted points in the comprehensive process according to the square root law; b. Establish a plane Cartesian coordinate system with the number of affected people as the horizontal axis and the polygonal area of the affected area as the vertical axis, Normalize the elements in the array numArray of the number of people affected and the areaArray of the polygon area of the influence range respectively, and express the results one by one in the coordinate system. At this time, each point in the coordinate system corresponds to one of the original point groups. The weight attribute value of the point; c. Take the coordinate origin as the origin, draw a 1/4 circle in the coordinate system with the minimum value of the weight attribute value point and the coordinate origin as the initial radius, and add the deletion mark "flag='D' to the point on the arc, Store it in the "Type II" point array secondType, and add the reserved flag "flag='R'" to the points corresponding to the neighbor polygons of the affected polygons; d. Taking the minimum plane distance between weight attribute value points in the plane coordinate system as an increment, update the radius value, make a 1/4 concentric circle with the origin as the center, and form its arc and the previous concentric circle Points without any mark in the circle of the circle add a delete mark, still store it in the array secondType, and add a reserve mark to the point corresponding to the neighbor polygon of its influence range polygon, and go to (5); e. Compare n and the size of the number of elements in the secondType array, if the value of n is large, return (d); if the two values are the same, turn to (f); otherwise, remove the point added to the array secondType in the previous round from the array Delete, and remove the deletion mark of the corresponding point, calculate the number of people affected per unit area n _p ^' of these points, and sort them in ascending order, and add the deletion mark to the point without any mark with the smallest value of n _p ^' in the sequence from front to back , store it in the array secondType, and add reserved marks to the polygon neighbors of its influence range, until the number of elements in the secondType array is the same as the n value, then go to (6); f. Delete the points corresponding to all elements in the secondType array in the original point group, and the remaining point group is the comprehensive result, and the algorithm ends.