Skip to main content

Nearest Neighbor Search Using KdTree





//Input should be in input.txt file
//10(2.1,4.3)(3.3,1.5)(4.7,11.1)(4.9,1.6)(5.0,12.3)(5.1,1.2)(6.7,3.3)(19.2,5.4)(20.5,7.9)(100.3,52)
//The first number is the number of points, n, and following it are the coordinates of the n points in ascending order of the x-coordinate.


import java.util.*;
import java.lang.*;
import java.io.*;
import java.util.StringTokenizer;

class KDNode {
    int axis;
    double[] x;
    int id;
    boolean checked;
    boolean orientation;

    KDNode Parent;
    KDNode Left;
    KDNode Right;

    public KDNode(double[] x0, int axis0) {
        x = new double[2];
        axis = axis0;
        for (int k = 0; k < 2; k++)
            x[k] = x0[k];

        Left = Right = Parent = null;
        checked = false;
        id = 0;
    }

    public KDNode FindParent(double[] x0) {
        KDNode parent = null;
        KDNode next = this;
        int split;
        while (next != null) {
            split = next.axis;
            parent = next;
            if (x0[split] > next.x[split])
                next = next.Right;
            else
                next = next.Left;
        }
        return parent;
    }

    public KDNode Insert(double[] p) {
        x = new double[2];
        KDNode parent = FindParent(p);
        if (equal(p, parent.x, 2) == true)
            return null;

        KDNode newNode = new KDNode(p, parent.axis + 1 < 2 ? parent.axis + 1 : 0);
        newNode.Parent = parent;

        if (p[parent.axis] > parent.x[parent.axis]) {
            parent.Right = newNode;
            newNode.orientation = true; //
        } else {
            parent.Left = newNode;
            newNode.orientation = false; //
        }

        return newNode;
    }

    boolean equal(double[] x1, double[] x2, int dim) {
        for (int k = 0; k < dim; k++) {
            if (x1[k] != x2[k])
                return false;
        }

        return true;
    }
    double distance2(double[] x1, double[] x2, int dim) {
        double S = 0;
        for (int k = 0; k < dim; k++)
            S += (x1[k] - x2[k]) * (x1[k] - x2[k]);
        return S;
    }
};

class KDTree {
    KDNode Root;

    int TimeStart, TimeFinish;
    int CounterFreq;

    double d_min;
    KDNode nearest_neighbour;

    int KD_id;

    int nList;

    KDNode CheckedNodes[];
    int checked_nodes;
    KDNode List[];

    double x_min[], x_max[];
    boolean max_boundary[], min_boundary[];
    int n_boundary;

    public KDTree(int i) {
        Root = null;
        KD_id = 1;
        nList = 0;
        List = new KDNode[i];
        CheckedNodes = new KDNode[i];
        max_boundary = new boolean[2];
        min_boundary = new boolean[2];
        x_min = new double[2];
        x_max = new double[2];
    }

    public boolean add(double[] x) {
        if (nList >= 2000000 - 1)
            return false; //can't add more points

        if (Root == null) {
            Root = new KDNode(x, 0);
            Root.id = KD_id++;
            List[nList++] = Root;
        } else {
            KDNode pNode;
            if ((pNode = Root.Insert(x)) != null) {
                pNode.id = KD_id++;
                List[nList++] = pNode;
            }
        }

        return true;
    }

    public KDNode find_nearest(double[] x) {
        if (Root == null)
            return null;

        checked_nodes = 0;
        KDNode parent = Root.FindParent(x);
        nearest_neighbour = parent;
        d_min = Root.distance2(x, parent.x, 2);;

        if (parent.equal(x, parent.x, 2) == true)
            return nearest_neighbour;

        search_parent(parent, x);
        uncheck();

        return nearest_neighbour;
    }

    public void check_subtree(KDNode node, double[] x) {
        if ((node == null) || node.checked)
            return;

        CheckedNodes[checked_nodes++] = node;
        node.checked = true;
        set_bounding_cube(node, x);

        int dim = node.axis;
        double d = node.x[dim] - x[dim];

        if (d * d > d_min) {
            if (node.x[dim] > x[dim])
                check_subtree(node.Left, x);
            else
                check_subtree(node.Right, x);
        } else {
            check_subtree(node.Left, x);
            check_subtree(node.Right, x);
        }
    }

    public void set_bounding_cube(KDNode node, double[] x) {
        if (node == null)
            return;
        int d = 0;
        double dx;
        for (int k = 0; k < 2; k++) {
            dx = node.x[k] - x[k];
            if (dx > 0) {
                dx *= dx;
                if (!max_boundary[k]) {
                    if (dx > x_max[k])
                        x_max[k] = dx;
                    if (x_max[k] > d_min) {
                        max_boundary[k] = true;
                        n_boundary++;
                    }
                }
            } else {
                dx *= dx;
                if (!min_boundary[k]) {
                    if (dx > x_min[k])
                        x_min[k] = dx;
                    if (x_min[k] > d_min) {
                        min_boundary[k] = true;
                        n_boundary++;
                    }
                }
            }
            d += dx;
            if (d > d_min)
                return;

        }

        if (d < d_min) {
            d_min = d;
            nearest_neighbour = node;
        }
    }

    public KDNode search_parent(KDNode parent, double[] x) {
        for (int k = 0; k < 2; k++) {
            x_min[k] = x_max[k] = 0;
            max_boundary[k] = min_boundary[k] = false; //
        }
        n_boundary = 0;

        double dx;
        KDNode search_root = parent;
        while (parent != null && (n_boundary != 2 * 2)) {
            check_subtree(parent, x);
            search_root = parent;
            parent = parent.Parent;
        }

        return search_root;
    }

    public void uncheck() {
        for (int n = 0; n < checked_nodes; n++)
            CheckedNodes[n].checked = false;
    }

};

public class KDTNearest {

    public static void main(String args[]) throws IOException {
        BufferedReader in = new BufferedReader(new FileReader("input.txt"));
        String strLin;
        strLin = in .readLine();

        StringTokenizer strLi = new StringTokenizer(strLin, "(,) ");
        int numpoints = Integer.parseInt(strLi.nextToken());

        KDTree kdt = new KDTree(numpoints);

        while (strLi.hasMoreTokens()) {

            double x[] = new double[2];
            for (int i = 0; i < numpoints; i++) {

                x[0] = Double.parseDouble(strLi.nextToken());
                x[1] = Double.parseDouble(strLi.nextToken());
                kdt.add(x);
            }
        }
        System.out.println("Enter the co-ordinates of the point: (one after the other)");
        InputStreamReader reader = new InputStreamReader(System. in );
        BufferedReader br = new BufferedReader(reader);
        double sx = Double.parseDouble(br.readLine());
        double sy = Double.parseDouble(br.readLine());

        double s[] = {
            sx, sy
        };
        KDNode kdn = kdt.find_nearest(s);
        System.out.println("(" + kdn.x[0] + " , " + kdn.x[1] + ")");
    }
}

Comments

  1. What is the license for this code?
    Would like to use it in a project of mine.

    Greetings.

    ReplyDelete
  2. Thanks for the post.

    ReplyDelete

Post a Comment

Popular posts from this blog

Find an old archived link on web

In recent past, I'm need of a tool, to which the provider no longer provides the link to download. Previously I've blogged about Openlogic , which serves this purpose. Openlogic contains unofficial fork of the original repository, and to be precise, I haven't found the tool I'm in search of. I found Internet Archive : Wayback Machine , while searching for it. It has got 412 billion stored pages and you may look into older versions of any website or you can get link to any software/tool which used to be a website. All you have to do is, provide a link or part of a link to search. For example, to know how my website used to be in 2013, just type http://www.thechaithanya.com and you can see a calender with some of the dates highlighted, on which they had the snapshot of my website. You can even provide a part of link. For example, to get the link of older version of Android ADT, search using http://dl.google.com/android/adt/* You can find a set of links and...

Reverse Engineering : Extract contents from .img file

Unyaffs is a program to extract files from a YAFFS2 file system image. Currently it can only extract images created by mkyaffs2image. Download the source from here . Compiling : Extract the contents into a suitable place and run the following command make Usage : unyaffs [options] <image_file_name> [<extract_directory>] Options: -d detection of flash layout, no extraction -b spare contains bad block information -c <chunk size> set chunk size in KByte (default: autodetect, max: 16) -s <spare size> set spare size in Byte (default: autodetect, max: 512) -t list image contents -v verbose output -V print version Source: Official github repository

JavaScript - Singleton Pattern

The singleton design pattern is probably the simplest and most common pattern in JavaScript. So why should you use the singleton pattern? Encapsulation of members & functions Creates its own Namespace A singleton is a single instance object Encourages code reuse Improves readability because you can logically organise your code The point of a singleton is to only have one instance. A shopping cart is a good example of something that you may want only a single instance of at one time. The simplest form of Singleton is an object literal. This loose form of Singleton cannot be instantiated. All of the members are now accessible through the Singleton variable, accessible through dot notation. var myCart = {     self: this ,     totalCost: 0,     totalQty: 0,     cart: {},     getCart: function (){ },     updateCart: function (){ } }; alert( "Total cost: ...