/*
* Copyright 2013 Alibaba.com All right reserved. This software is the
* confidential and proprietary information of Alibaba.com ("Confidential
* Information"). You shall not disclose such Confidential Information and shall
* use it only in accordance with the terms of the license agreement you entered
* into with Alibaba.com.
*/
package com.alibaba.simpleimage.analyze.harris.match;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.alibaba.simpleimage.analyze.ModifiableConst;
import com.alibaba.simpleimage.analyze.harissurf.SURFInterestPointN;
import com.alibaba.simpleimage.analyze.kdtree.KDTree;
/**
* 类SurfMatchPoints.java的实现描述：TODO 类实现描述
*
* @author axman 2013-5-23 下午4:38:08
*/
public class SurfMatchPoints {
private static class _mylist extends ArrayList<SurfMatch> {
private static final long serialVersionUID = -1672787720681683109L;
public void removeRange(int formIndex, int toIndex) {
super.removeRange(formIndex, toIndex);
}
}
public static List<SurfMatch> findMatchesBBF(List<SURFInterestPointN> keys1, List<SURFInterestPointN> keys2) {
return findMatchesBBF(keys1, KDTree.createKDTree(keys2));
}
public static List<SurfMatch> findMatchesBBF(List<SURFInterestPointN> keys1, KDTree kd) {
List<SurfMatch> matches = new _mylist();
for (SURFInterestPointN sp : keys1) {
ArrayList<KDTree.BestEntry> kpNNList = kd.nearestNeighbourListBBF(sp, 2, 40);
if (kpNNList.size() < 2) throw (new IllegalArgumentException("BUG: less than two neighbours!"));
KDTree.BestEntry be1 = (KDTree.BestEntry) kpNNList.get(0);
KDTree.BestEntry be2 = (KDTree.BestEntry) kpNNList.get(1);
if ((be1.getDist() / be2.getDist()) > ModifiableConst.getTowPntDistRatio()) continue;
// XXX:最近邻点和次近邻点的距离比值
SURFInterestPointN kpN = (SURFInterestPointN) be1.getNeighbour();
if (Math.abs(kpN.getOrientation() - sp.getOrientation()) > ModifiableConst.getTowPntOrientationMinus())
// continue;
// if (Math.abs(kpN.getScale() - kp.getScale()) > ModifiableConst.getTowPntScaleMinus()) continue;
matches.add(new SurfMatch(sp, kpN, be1.getDist(), be2.getDist()));
}
return (matches);
}
/*
* public ArrayList FindMatches (ArrayList keys1, ArrayList keys2) { ArrayList matches = new ArrayList (); //KDTree
* kd = KDTree.CreateKDTree (keys2); foreach (Keypoint kp in keys1) { double distNearest = Double.PositiveInfinity;
* int nearest = -1; double dist2Nearest = Double.PositiveInfinity; int nearest2 = -1; for (int kn = 0 ; kn <
* keys2.Count ; ++kn) { Keypoint kp2 = (Keypoint) keys2[kn]; double dist = Math.Sqrt (KDTree.DistanceSq (kp, kp2));
* if (dist < distNearest) { nearest2 = nearest; dist2Nearest = distNearest; nearest = kn; distNearest = dist; } }
* if (nearest == -1 || nearest2 == -1) continue; if ((distNearest / dist2Nearest) > 0.6) continue; matches.Add (new
* Match (kp, (Keypoint) keys2[nearest])); Console.WriteLine ("({0},{1}) ({2},{3}) {4}", (int)(kp.X + 0.5),
* (int)(kp.Y + 0.5), (int)(((Keypoint) keys2[nearest]).X + 0.5), (int)(((Keypoint) keys2[nearest]).Y + 0.5),
* distNearest); } return (matches); }
*/
public static ArrayList<SurfMatch> filterJoins(List<SurfMatch> matches) {
Map<SURFInterestPointN, Integer> map = new HashMap<SURFInterestPointN, Integer>();
// Count the references to each keypoint
for (SurfMatch m : matches) {
Integer kp1V = map.get(m.getSp1());
int lI = (kp1V == null) ? 0 : (int) kp1V;
map.put(m.getSp1(), lI + 1);
Integer kp2V = map.get(m.getSp2());
int rI = (kp2V == null) ? 0 : (int) kp2V;
map.put(m.getSp2(), rI + 1);
}
ArrayList<SurfMatch> survivors = new ArrayList<SurfMatch>();
for (SurfMatch m : matches) {
Integer kp1V = map.get(m.getSp1());
Integer kp2V = map.get(m.getSp2());
if (kp1V <= 1 && kp2V <= 1) survivors.add(m);
}
return (survivors);
}
public static void filterNBest(ArrayList<SurfMatch> matches, int bestQ) {
Collections.sort(matches, new SurfMatch.MatchWeighter());
if (matches.size() > bestQ) {
((_mylist) matches).removeRange(bestQ, matches.size() - bestQ);
}
}
public static List<SurfMatch> filterFarMatchL(List<SurfMatch> matches, double minX, double minY) {
int arcStep = ModifiableConst.getSolpeArcStep();
if (matches.size() <= 1) return matches;
int max_vote_count = 0;
long max_vote = 0;
int[] ms = new int[90 / arcStep + 1]; // 用数据的索引拂过每个度数的key，不使用map来保存，性能优化
for (SurfMatch m : matches) {
if (Math.abs(m.sp2.getOrientation() - m.sp1.getOrientation()) > 0.1) {
continue;
}
double r = Math.atan((m.sp2.getY() + minY - m.sp1.getY()) / (m.sp2.getX() + minX - m.sp1.getX())) * 360
/ (2 * Math.PI);
if (r < 0) r += 90;
int idx = (int) r / arcStep; // 取整
ms[idx] = ms[idx] + 1;
if (ms[idx] > max_vote_count) {
max_vote_count = ms[idx];
max_vote = idx;
}
m.slopeArc = idx;
}
ArrayList<SurfMatch> survivors = new ArrayList<SurfMatch>();
for (SurfMatch m : matches) {
if (m.slopeArc == max_vote) survivors.add(m);
}
return survivors;
}
public static List<SurfMatch> filterFarMatchR(List<SurfMatch> matches, double minX, double minY) {
int arcStep = ModifiableConst.getSolpeArcStep();
if (matches.size() <= 1) return matches;
int[] ms = new int[90 / arcStep]; // 用数据的索引拂过每个度数的key，不使用map来保存，性能优化
for (SurfMatch m : matches) {
double r = Math.atan((m.sp1.getY() - (m.sp2.getY() + minY)) / (m.sp1.getX() + minX - m.sp2.getX())) * 360
/ (2 * Math.PI);
m.slopeArc = ((int) r / arcStep * arcStep); // 第一次计算就把 match的斜率保存起来。
if (m.slopeArc < 0) m.slopeArc += 90;
ms[m.slopeArc / arcStep] = ms[m.slopeArc / arcStep] + 1;
}
int count = 0;
int idx = 0;
for (int i = 0; i < ms.length; i++) {// 找到斜率相同的最多的一个度数
if (ms[i] > count) {
count = ms[i];
idx = i;
}
}
idx = idx * arcStep;
ArrayList<SurfMatch> survivors = new ArrayList<SurfMatch>();
for (SurfMatch m : matches) {
if (m.slopeArc == idx) survivors.add(m);
}
return survivors;
}
}