Java地理工具: Snap to line identifiying line that was snapped to

Question

我正在尝试编写一个 Java 程序，它将大量 GPS 坐标捕捉到线形文件(道路网络)，并且不仅返回新坐标，还返回捕捉到的线段的唯一标识符。该标识符是否是 FID、其他语言中使用的“索引”(即，其中 1 是第一个特征等)或属性表中的任何列都无关紧要。

我在 R 中使用 maptools::snapPointsToLines 完成了此操作函数，但考虑到我需要处理的数据量，这是不可扩展的，因此我正在寻找 Java 来更快地处理数据，以便在 R 中进行分析。

我的代码(如下)目前与用于捕捉的 geotools 教程非常相似，其中的细微差别是我在 GPS 点的(1900 万行)CSV 中读取而不是生成它们，并编写了结果的 CSV。它捕捉得很好，而且比我得到的要快得多，但我不知道如何识别捕捉到的线。可用的文档似乎涵盖了功能集的查询和过滤，我无法特别适用于此代码创建的索引行对象，并且我的代码中的现有函数 toString() 返回一些对于我的目的而言难以理解的内容，例如 com.vividsolutions.jts.linearreff.LocationIndexedLine@74cec793。

基本上，我只是希望 lineID 字段生成任何其他 GIS 软件或语言都可以与特定路段匹配的内容。

package org.geotools.tutorial.quickstart;

import java.io.*;
import java.util.List;
import java.util.Arrays;

import com.vividsolutions.jts.geom.Coordinate;
import com.vividsolutions.jts.geom.Envelope;
import com.vividsolutions.jts.geom.Geometry;
import com.vividsolutions.jts.geom.LineString;
import com.vividsolutions.jts.geom.MultiLineString;
import com.vividsolutions.jts.index.SpatialIndex;
import com.vividsolutions.jts.index.strtree.STRtree;
import com.vividsolutions.jts.linearref.LinearLocation;
import com.vividsolutions.jts.linearref.LocationIndexedLine;

import org.geotools.data.FeatureSource;
import org.geotools.data.FileDataStore;
import org.geotools.data.FileDataStoreFinder;
import org.geotools.feature.FeatureCollection;
import org.geotools.geometry.jts.ReferencedEnvelope;
import org.geotools.swing.data.JFileDataStoreChooser;
import org.geotools.util.NullProgressListener;
import org.opengis.feature.Feature;
import org.opengis.feature.FeatureVisitor;
import org.opengis.feature.simple.SimpleFeature;
import com.opencsv.*;

public class SnapToLine {

    public static void main(String[] args) throws Exception {

        /*
         * Open a shapefile. You should choose one with line features
         * (LineString or MultiLineString geometry)
         * 
         */
        File file = JFileDataStoreChooser.showOpenFile("shp", null);
        if (file == null) {
            return;
        }

        FileDataStore store = FileDataStoreFinder.getDataStore(file);
        FeatureSource source = store.getFeatureSource();

        // Check that we have line features
        Class<?> geomBinding = source.getSchema().getGeometryDescriptor().getType().getBinding();
        boolean isLine = geomBinding != null 
                && (LineString.class.isAssignableFrom(geomBinding) ||
                    MultiLineString.class.isAssignableFrom(geomBinding));

        if (!isLine) {
            System.out.println("This example needs a shapefile with line features");
            return;
        }
         final SpatialIndex index = new STRtree();
        FeatureCollection features = source.getFeatures();
        //FeatureCollection featurecollection = source.getFeatures(Query.FIDS);
        System.out.println("Slurping in features ...");
        features.accepts(new FeatureVisitor() {

            @Override
            public void visit(Feature feature) {
                SimpleFeature simpleFeature = (SimpleFeature) feature;
                Geometry geom = (MultiLineString) simpleFeature.getDefaultGeometry();
                // Just in case: check for  null or empty geometry
                if (geom != null) {
                    Envelope env = geom.getEnvelopeInternal();
                    if (!env.isNull()) {
                        index.insert(env, new LocationIndexedLine(geom));
                    }
                }
            }
        }, new NullProgressListener());
 /*

 /*
         * We defined the maximum distance that a line can be from a point
         * to be a candidate for snapping 
         */

        ReferencedEnvelope bounds = features.getBounds();
        final double MAX_SEARCH_DISTANCE = bounds.getSpan(0) / 1000.0;



        int pointsProcessed = 0;
        int pointsSnapped = 0;
        long elapsedTime = 0;
        long startTime = System.currentTimeMillis();
        double longiOut;
        double latiOut;
        int moved;
        String lineID   = "NA";

        //Open up the CSVReader. Reading in line by line to avoid memory failure.

        CSVReader csvReader = new CSVReader(new FileReader(new File("fakedata.csv")));
        String[] rowIn;



        //open up the CSVwriter
        String outcsv = "fakedataOUT.csv";
        CSVWriter writer = new CSVWriter(new FileWriter(outcsv));



        while ((rowIn = csvReader.readNext()) != null) {

            // Get point and create search envelope
            pointsProcessed++;
            double longi = Double.parseDouble(rowIn[0]);
            double lati  = Double.parseDouble(rowIn[1]);
            Coordinate pt = new Coordinate(longi, lati);
            Envelope search = new Envelope(pt);
            search.expandBy(MAX_SEARCH_DISTANCE);

            /*
             * Query the spatial index for objects within the search envelope.
             * Note that this just compares the point envelope to the line envelopes
             * so it is possible that the point is actually more distant than
             * MAX_SEARCH_DISTANCE from a line.
             */
            List<LocationIndexedLine> lines = index.query(search);

            // Initialize the minimum distance found to our maximum acceptable
            // distance plus a little bit
            double minDist = MAX_SEARCH_DISTANCE + 1.0e-6;
            Coordinate minDistPoint = null;

            for (LocationIndexedLine line : lines) {
                LinearLocation here = line.project(pt);
                Coordinate point = line.extractPoint(here);
                double dist = point.distance(pt);
                if (dist < minDist) {
                    minDist = dist;
                    minDistPoint = point;
                    lineID = line.toString();
                }
            }


            if (minDistPoint == null) {
                // No line close enough to snap the point to
                System.out.println(pt + "- X");
                longiOut = longi;
                latiOut  = lati;
                moved    = 0;
                lineID   = "NA";
            } else {
                System.out.printf("%s - snapped by moving %.4f\n", 
                        pt.toString(), minDist);
                longiOut = minDistPoint.x;
                latiOut  = minDistPoint.y;
                moved    = 1;        
                pointsSnapped++;
            }
    //write a new row

    String [] rowOut = {Double.toString(longiOut), Double.toString(latiOut), Integer.toString(moved), lineID}; 
    writer.writeNext(rowOut);
        }

        System.out.printf("Processed %d points (%.2f points per second). \n"
                + "Snapped %d points.\n\n",
                pointsProcessed,
                1000.0 * pointsProcessed / elapsedTime,
                pointsSnapped);
        writer.close();
    }
}

我不仅是 Java 新手，而且只是在领域特定语言(如 R)方面进行了自学；我与其说是一个编码员，不如说是一个使用代码的人，所以如果解决方案看起来很明显，我可能缺乏基本理论!

p.s 我知道有更好的 map 匹配解决方案(graphhopper 等)，我只是想开始 eas!

谢谢!

Answer 1

我会尽量避免在 JTS 兔子洞中走得太远，并坚持使用 GeoTools(当然我是一个 GeoTools 开发人员，所以我会这么说)。

首先我会使用SpatialIndexFeatureCollection保存我的行(假设它们适合内存，否则 PostGIS 表是可行的方法)。这使我不必建立自己的索引。

然后我会使用CSVDataStore避免从 GPS 流中解析我自己的点(因为我很懒，而且那里也有很多出错的地方)。

这意味着大部分工作都归结为这个循环，DWITHIN 查找指定距离内的所有要素:

try (SimpleFeatureIterator itr = pointFeatures.getFeatures().features()) { 
  while (itr.hasNext()) {
    SimpleFeature f = itr.next();
    Geometry snapee = (Geometry) f.getDefaultGeometry();
    Filter filter = ECQL.toFilter("DWITH(\"the_geom\",'" + writer.write(snapee) + "'," + MAX_SEARCH_DISTANCE + ")");
    SimpleFeatureCollection possibles = indexed.subCollection(filter);
    double minDist = Double.POSITIVE_INFINITY;
    SimpleFeature bestFit = null;
    Coordinate bestPoint = null;
    try (SimpleFeatureIterator pItr = possibles.features()) {
      while (pItr.hasNext()) {
        SimpleFeature p = pItr.next();
        Geometry line = (Geometry) p.getDefaultGeometry();

        double dist = snapee.distance(line);
        if (dist < minDist) {
          minDist = dist;
          bestPoint = DistanceOp.nearestPoints(snapee, line)[1];
          bestFit  = p;
        }
      }
    }

在循环结束时，您应该知道距离线最近的特征(bestFit)(包括其 ID 和名称等)、最近的点(bestPoint)和移动的距离(minDist)。

我可能也会使用 CSVDatastore 将功能写回。

如果您有数百万个点，我可能会考虑使用 FilterFactory 直接创建过滤器，而不是使用 ECQL 解析器。