DBScan 之 PostGIS ST_ClusterDBSCAN 和 Java 的 DBSCANClusterer 两种实现

avatar 2019年09月20日17:50:56 6 4005 views
博主分享免费Java教学视频，B站账号：Java刘哥 ，长期提供技术问题解决、项目定制：本站商品点此
在做热力图的时候，我们需要对所有的坐标进行聚合计算，比如对数据库里的所有数据的坐标进行聚合，以100米为一组，即经纬度计算100米范围内(通常我们是以经纬度，如每 0.00009度)的所有数据为一组，并统计每组的坐标数量，该组的经纬度我们可以取所有点的平均值。

最后把所有组的经纬度和每组的坐标数量返回给前端，前端就能展示了，数量多的可以显示更颜色可以更亮和图形更大，数量小的话颜色显示浅些。

现在的问题是怎么进行聚类，如何对坐标进行聚类。

目前我知道的有两种方法，一种是通过 PostgreSQL 的插件 PostGis 来计算，换言之通过数据库SQL语句来执行；另一种是通过Java代码来实现聚类，目前还在探索中，好像apache有相关的类。

下面分别来介绍：

PostGIS 的 ST_ClusterDBSCAN


先说以下通过 PostGis 来实现。主要是用到 ST_ClusterDBSCAN这个函数

下面是一个例子
  1. SELECT
  2.     cid,
  3.     COUNT ( cid ),
  4.     AVG ( latitude ) AS latitude,
  5.     AVG ( longitude ) AS longitude
  6. FROM
  7.     (
  8. SELECT
  9.   -- 聚类计算
  10.     st_clusterdbscan ( geom, 0.00009, 1 ) OVER () AS cid,
  11.     -- 将 geom 字段经纬度 
  12.     split_part( st_aslatlontext ( geom, 'DD.DDDDDD' ), ' ', 1 ) :: NUMERIC AS latitude,
  13.     split_part( st_aslatlontext ( geom, 'DD.DDDDDD' ), ' ', 2 ) :: NUMERIC AS longitude
  14. FROM
  15.     your_table_name
  16. WHERE
  17.   -- 条件
  18.     ST_Intersects (
  19.     geom,
  20.     st_geometryfromtext ( #{ boundWkt }, 4326 )) = TRUE
  21.     ) sq
  22. GROUP BY
  23.     cid

相关文档

http://postgis.net/docs/manual-dev/ST_ClusterDBSCAN.html


DBSCAN Java实现


目前找到了 apache 的 DBSCANClusterer类
  1. import org.apache.commons.math3.ml.clustering.Cluster;
  2. import org.apache.commons.math3.ml.clustering.DBSCANClusterer;
  3. import org.apache.commons.math3.ml.clustering.DoublePoint;
  6. import java.util.Arrays;
  7. import java.util.List;
  9. /**
  10.  * @author saysky on 2019-9-20
  11.  */
  12. public class Demo {
  14.     public static void main(String[] args) {
  15.         final DoublePoint[] points = new DoublePoint[] {
  16.                 new DoublePoint(new double[] { 83.08303244924173, 58.83387754182331 }),
  17.                 new DoublePoint(new double[] { 45.05445510940626, 23.469642649637535 }),
  18.                 new DoublePoint(new double[] { 14.96417921432294, 69.0264096390456 }),
  19.                 new DoublePoint(new double[] { 73.53189604333602, 34.896145021310076 }),
  20.                 new DoublePoint(new double[] { 73.28498173551634, 33.96860806993209 }),
  21.                 new DoublePoint(new double[] { 73.45828098873608, 33.92584423092194 }),
  22.                 new DoublePoint(new double[] { 73.9657889183145, 35.73191006924026 }),
  23.                 new DoublePoint(new double[] { 74.0074097183533, 36.81735596177168 }),
  24.                 new DoublePoint(new double[] { 73.41247541410848, 34.27314856695011 }),
  25.                 new DoublePoint(new double[] { 73.9156256353017, 36.83206791547127 }),
  26.                 new DoublePoint(new double[] { 74.81499205809087, 37.15682749846019 }),
  27.                 new DoublePoint(new double[] { 74.03144880081527, 37.57399178552441 }),
  28.                 new DoublePoint(new double[] { 74.51870941207744, 38.674258946906775 }),
  29.                 new DoublePoint(new double[] { 74.50754595105536, 35.58903978415765 }),
  30.                 new DoublePoint(new double[] { 74.51322752749547, 36.030572259100154 }),
  31.                 new DoublePoint(new double[] { 59.27900996617973, 46.41091720294207 }),
  32.                 new DoublePoint(new double[] { 59.73744793841615, 46.20015558367595 }),
  33.                 new DoublePoint(new double[] { 58.81134076672606, 45.71150126331486 }),
  34.                 new DoublePoint(new double[] { 58.52225539437495, 47.416083617601544 }),
  35.                 new DoublePoint(new double[] { 58.218626647023484, 47.36228902172297 }),
  36.                 new DoublePoint(new double[] { 60.27139669447206, 46.606106348801404 }),
  37.                 new DoublePoint(new double[] { 60.894962462363765, 46.976924697402865 }),
  38.                 new DoublePoint(new double[] { 62.29048673878424, 47.66970563563518 }),
  39.                 new DoublePoint(new double[] { 61.03857608977705, 46.212924720020965 }),
  40.                 new DoublePoint(new double[] { 60.16916214139201, 45.18193661351688 }),
  41.                 new DoublePoint(new double[] { 59.90036905976012, 47.555364347063005 }),
  42.                 new DoublePoint(new double[] { 62.33003634144552, 47.83941489877179 }),
  43.                 new DoublePoint(new double[] { 57.86035536718555, 47.31117930193432 }),
  44.                 new DoublePoint(new double[] { 58.13715479685925, 48.985960494028404 }),
  45.                 new DoublePoint(new double[] { 56.131923963548616, 46.8508904252667 }),
  46.                 new DoublePoint(new double[] { 55.976329887053, 47.46384037658572 }),
  47.                 new DoublePoint(new double[] { 56.23245975235477, 47.940035191131756 }),
  48.                 new DoublePoint(new double[] { 58.51687048212625, 46.622885352699086 }),
  49.                 new DoublePoint(new double[] { 57.85411081905477, 45.95394361577928 }),
  50.                 new DoublePoint(new double[] { 56.445776311447844, 45.162093662656844 }),
  51.                 new DoublePoint(new double[] { 57.36691949656233, 47.50097194337286 }),
  52.                 new DoublePoint(new double[] { 58.243626387557015, 46.114052729681134 }),
  53.                 new DoublePoint(new double[] { 56.27224595635198, 44.799080066150054 }),
  54.                 new DoublePoint(new double[] { 57.606924816500396, 46.94291057763621 }),
  55.                 new DoublePoint(new double[] { 30.18714230041951, 13.877149710431695 }),
  56.                 new DoublePoint(new double[] { 30.449448810657486, 13.490778346545994 }),
  57.                 new DoublePoint(new double[] { 30.295018390286714, 13.264889000216499 }),
  58.                 new DoublePoint(new double[] { 30.160201832884923, 11.89278262341395 }),
  59.                 new DoublePoint(new double[] { 31.341509791789576, 15.282655921997502 }),
  60.                 new DoublePoint(new double[] { 31.68601630325429, 14.756873246748 }),
  61.                 new DoublePoint(new double[] { 29.325963742565364, 12.097849250072613 }),
  62.                 new DoublePoint(new double[] { 29.54820742388256, 13.613295356975868 }),
  63.                 new DoublePoint(new double[] { 28.79359608888626, 10.36352064087987 }),
  64.                 new DoublePoint(new double[] { 31.01284597092308, 12.788479208014905 }),
  65.                 new DoublePoint(new double[] { 27.58509216737002, 11.47570110601373 }),
  66.                 new DoublePoint(new double[] { 28.593799561727792, 10.780998203903437 }),
  67.                 new DoublePoint(new double[] { 31.356105766724795, 15.080316198524088 }),
  68.                 new DoublePoint(new double[] { 31.25948503636755, 13.674329151166603 }),
  69.                 new DoublePoint(new double[] { 32.31590076372959, 14.95261758659035 }),
  70.                 new DoublePoint(new double[] { 30.460413702763617, 15.88402809202671 }),
  71.                 new DoublePoint(new double[] { 32.56178203062154, 14.586076852632686 }),
  72.                 new DoublePoint(new double[] { 32.76138648530468, 16.239837325178087 }),
  73.                 new DoublePoint(new double[] { 30.1829453331884, 14.709592407103628 }),
  74.                 new DoublePoint(new double[] { 29.55088173528202, 15.0651247180067 }),
  75.                 new DoublePoint(new double[] { 29.004155302187428, 14.089665298582986 }),
  76.                 new DoublePoint(new double[] { 29.339624439831823, 13.29096065578051 }),
  77.                 new DoublePoint(new double[] { 30.997460327576846, 14.551914158277214 }),
  78.                 new DoublePoint(new double[] { 30.66784126125276, 16.269703107886016 })
  79.         };
  81.         final DBSCANClusterer<DoublePoint> transformer =
  82.                 new DBSCANClusterer<DoublePoint>(1, 0);
  83.         final List<Cluster<DoublePoint>> clusters = transformer.cluster(Arrays.asList(points));
  84.         System.out.println(clusters);
  85.     }
  86. }

第一个参数是表示最小半径，单位是度数，经纬度的度数，最小0，最大360。如当为1的时候，注意如果相邻的相邻的相邻都满足相差1个经纬度半径，也算在一个组的；

第二个参数时该组内至少有几个邻居(1表示该组至少要有2个点，0表示至少1个点)



文档如下

http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math4/ml/clustering/DBSCANClusterer.html
历史上的今天
九月
20日
  • 微信
  • 交流学习，资料分享
  • weinxin
  • 个人淘宝
  • 店铺名：言曌博客咨询部

  • (部分商品未及时上架淘宝)
avatar

发表评论

avatar 登录者：匿名
匿名评论，评论回复后会有邮件通知

  

已通过评论：0   待审核评论数：0