Task02 Михаил Митрофанов ИТМО

src/phg/matching/descriptor_matcher.cpp

@@ -1,5 +1,6 @@
 #include "descriptor_matcher.h"
 
+#include <unordered_set>
 #include <opencv2/flann/miniflann.hpp>
 #include "flann_factory.h"
 
@@ -8,7 +9,10 @@ void phg::DescriptorMatcher::filterMatchesRatioTest(const std::vector<std::vecto
 {
     filtered_matches.clear();
 
-    throw std::runtime_error("not implemented yet");
+    for (auto &keypoint_matches : matches) {
+        if (keypoint_matches.size() < 2 || keypoint_matches[0].distance > keypoint_matches[1].distance * 0.75) continue;
+        filtered_matches.push_back(keypoint_matches[0]);
+    }
 }
 
 
@@ -35,42 +39,73 @@ void phg::DescriptorMatcher::filterMatchesClusters(const std::vector<cv::DMatch>
         points_query.at<cv::Point2f>(i) = keypoints_query[matches[i].queryIdx].pt;
         points_train.at<cv::Point2f>(i) = keypoints_train[matches[i].trainIdx].pt;
     }
-//
-//    // размерность всего 2, так что точное KD-дерево
-//    std::shared_ptr<cv::flann::IndexParams> index_params = flannKdTreeIndexParams(TODO);
-//    std::shared_ptr<cv::flann::SearchParams> search_params = flannKsTreeSearchParams(TODO);
-//
-//    std::shared_ptr<cv::flann::Index> index_query = flannKdTreeIndex(points_query, index_params);
-//    std::shared_ptr<cv::flann::Index> index_train = flannKdTreeIndex(points_train, index_params);
-//
-//    // для каждой точки найти total neighbors ближайших соседей
-//    cv::Mat indices_query(n_matches, total_neighbours, CV_32SC1);
-//    cv::Mat distances2_query(n_matches, total_neighbours, CV_32FC1);
-//    cv::Mat indices_train(n_matches, total_neighbours, CV_32SC1);
-//    cv::Mat distances2_train(n_matches, total_neighbours, CV_32FC1);
-//
-//    index_query->knnSearch(points_query, indices_query, distances2_query, total_neighbours, *search_params);
-//    index_train->knnSearch(points_train, indices_train, distances2_train, total_neighbours, *search_params);
-//
-//    // оценить радиус поиска для каждой картинки
-//    // NB: radius2_query, radius2_train: квадраты радиуса!
-//    float radius2_query, radius2_train;
-//    {
-//        std::vector<double> max_dists2_query(n_matches);
-//        std::vector<double> max_dists2_train(n_matches);
-//        for (int i = 0; i < n_matches; ++i) {
-//            max_dists2_query[i] = distances2_query.at<float>(i, total_neighbours - 1);
-//            max_dists2_train[i] = distances2_train.at<float>(i, total_neighbours - 1);
-//        }
-//
-//        int median_pos = n_matches / 2;
-//        std::nth_element(max_dists2_query.begin(), max_dists2_query.begin() + median_pos, max_dists2_query.end());
-//        std::nth_element(max_dists2_train.begin(), max_dists2_train.begin() + median_pos, max_dists2_train.end());
-//
-//        radius2_query = max_dists2_query[median_pos] * radius_limit_scale * radius_limit_scale;
-//        radius2_train = max_dists2_train[median_pos] * radius_limit_scale * radius_limit_scale;
-//    }
-//
-//    метч остается, если левое и правое множества первых total_neighbors соседей в радиусах поиска(radius2_query, radius2_train) имеют как минимум consistent_matches общих элементов
-//    // TODO заполнить filtered_matches
+
+    // размерность всего 2, так что точное KD-дерево
+    std::shared_ptr<cv::flann::IndexParams> index_params = flannKdTreeIndexParams(1);
+    std::shared_ptr<cv::flann::SearchParams> search_params = flannKsTreeSearchParams(128);
+
+    std::shared_ptr<cv::flann::Index> index_query = flannKdTreeIndex(points_query, index_params);
+    std::shared_ptr<cv::flann::Index> index_train = flannKdTreeIndex(points_train, index_params);
+
+    // для каждой точки найти total neighbors ближайших соседей
+    cv::Mat indices_query(n_matches, total_neighbours, CV_32SC1);
+    cv::Mat distances2_query(n_matches, total_neighbours, CV_32FC1);
+    cv::Mat indices_train(n_matches, total_neighbours, CV_32SC1);
+    cv::Mat distances2_train(n_matches, total_neighbours, CV_32FC1);
+
+    index_query->knnSearch(points_query, indices_query, distances2_query, total_neighbours, *search_params);
+    index_train->knnSearch(points_train, indices_train, distances2_train, total_neighbours, *search_params);
+
+    // оценить радиус поиска для каждой картинки
+    // NB: radius2_query, radius2_train: квадраты радиуса!
+    float radius2_query, radius2_train;
+    {
+        std::vector<double> max_dists2_query(n_matches);
+        std::vector<double> max_dists2_train(n_matches);
+        for (int i = 0; i < n_matches; ++i) {
+            max_dists2_query[i] = distances2_query.at<float>(i, total_neighbours - 1);
+            max_dists2_train[i] = distances2_train.at<float>(i, total_neighbours - 1);
+        }
+
+        int median_pos = n_matches / 2;
+        std::nth_element(max_dists2_query.begin(), max_dists2_query.begin() + median_pos, max_dists2_query.end());
+        std::nth_element(max_dists2_train.begin(), max_dists2_train.begin() + median_pos, max_dists2_train.end());
+
+        radius2_query = max_dists2_query[median_pos] * radius_limit_scale * radius_limit_scale;
+        radius2_train = max_dists2_train[median_pos] * radius_limit_scale * radius_limit_scale;
+    }
+
+    // метч остается, если левое и правое множества первых total_neighbors соседей в радиусах поиска(radius2_query, radius2_train) имеют как минимум consistent_matches общих элементов
+    for (int i = 0; i < n_matches; ++i) {
+
+        std::unordered_set<int> expected_neighs;
+        expected_neighs.insert(i);
+        for (int j = 0; j < total_neighbours; ++j) {
+            if (radius2_query > 1e-4 && distances2_query.at<float>(i, j) > radius2_query) continue;
+            int neigh_query_index = indices_query.at<int>(i, j);
+            expected_neighs.insert(neigh_query_index);
+        }
+
+        std::unordered_set<int> actual_neighs;
+        actual_neighs.insert(i);
+        for (int j = 0; j < total_neighbours; ++j) {
+            if (radius2_train > 1e-4 && distances2_train.at<float>(i, j) > radius2_train) continue;
+            int neigh_train_index = indices_train.at<int>(i, j);
+            actual_neighs.insert(neigh_train_index);
+        }
+
+        int common = 0;
+        if (expected_neighs.size() > actual_neighs.size()) {
+            std::swap(expected_neighs, actual_neighs);
+        }
+        for (int idx : expected_neighs) {
+            if (actual_neighs.count(idx)) {
+                common++;
+            }
+        }
+
+        if (common >= consistent_matches) {
+            filtered_matches.push_back(matches[i]);
+        }
+    }
 }

src/phg/matching/flann_matcher.cpp

@@ -6,8 +6,8 @@
 phg::FlannMatcher::FlannMatcher()
 {
     // параметры для приближенного поиска
-//    index_params = flannKdTreeIndexParams(TODO);
-//    search_params = flannKsTreeSearchParams(TODO);
+   index_params = flannKdTreeIndexParams(4);
+   search_params = flannKsTreeSearchParams(40);
 }
 
 void phg::FlannMatcher::train(const cv::Mat &train_desc)
@@ -17,5 +17,19 @@ void phg::FlannMatcher::train(const cv::Mat &train_desc)
 
 void phg::FlannMatcher::knnMatch(const cv::Mat &query_desc, std::vector<std::vector<cv::DMatch>> &matches, int k) const
 {
-    throw std::runtime_error("not implemented yet");
+    int descs = query_desc.rows;
+    matches.clear();
+    matches.resize(descs);
+
+    cv::Mat indices(descs, k, CV_32S);
+    cv::Mat distances2(descs, k, CV_32F);
+
+    flann_index->knnSearch(query_desc, indices, distances2, k, *search_params);
+
+    for (int i = 0; i < descs; i++) {
+        matches[i].reserve(k);
+        for (int j = 0; j < k; j++) {
+            matches[i].emplace_back(i, indices.at<int>(i, j), std::sqrt(distances2.at<float>(i, j)));
+        }
+    }
 }

src/phg/sfm/homography.cpp

@@ -2,6 +2,7 @@
 
 #include <opencv2/calib3d/calib3d.hpp>
 #include <iostream>
+#include <opencv2/imgproc.hpp>
 
 namespace {
 
@@ -84,8 +85,16 @@ namespace {
             double w1 = ws1[i];
 
             // 8 elements of matrix + free term as needed by gauss routine
-//            A.push_back({TODO});
-//            A.push_back({TODO});
+            A.push_back({
+                x0, y0, 1,
+                0, 0, 0,
+                -x1*x0, -x1*y0, x1
+            });
+            A.push_back({
+                0, 0, 0,
+                x0, y0, 1,
+                -y1*x0, -y1*y0, y1
+            });
         }
 
         int res = gauss(A, H);
@@ -168,57 +177,61 @@ namespace {
         // * (простое описание для понимания)
         // * [3] http://ikrisoft.blogspot.com/2015/01/ransac-with-contrario-approach.html
 
-//        const int n_matches = points_lhs.size();
-//
-//        // https://en.wikipedia.org/wiki/Random_sample_consensus#Parameters
-//        const int n_trials = TODO;
-//
-//        const int n_samples = TODO;
-//        uint64_t seed = 1;
-//        const double reprojection_error_threshold_px = 2;
-//
-//        int best_support = 0;
-//        cv::Mat best_H;
-//
-//        std::vector<int> sample;
-//        for (int i_trial = 0; i_trial < n_trials; ++i_trial) {
-//            randomSample(sample, n_matches, n_samples, &seed);
-//
-//            cv::Mat H = estimateHomography4Points(points_lhs[sample[0]], points_lhs[sample[1]], points_lhs[sample[2]], points_lhs[sample[3]],
-//                                                  points_rhs[sample[0]], points_rhs[sample[1]], points_rhs[sample[2]], points_rhs[sample[3]]);
-//
-//            int support = 0;
-//            for (int i_point = 0; i_point < n_matches; ++i_point) {
-//                try {
-//                    cv::Point2d proj = phg::transformPoint(points_lhs[i_point], H);
-//                    if (cv::norm(proj - cv::Point2d(points_rhs[i_point])) < reprojection_error_threshold_px) {
-//                        ++support;
-//                    }
-//                } catch (const std::exception &e)
-//                {
-//                    std::cerr << e.what() << std::endl;
-//                }
-//            }
-//
-//            if (support > best_support) {
-//                best_support = support;
-//                best_H = H;
-//
-//                std::cout << "estimateHomographyRANSAC : support: " << best_support << "/" << n_matches << std::endl;
-//
-//                if (best_support == n_matches) {
-//                    break;
-//                }
-//            }
-//        }
-//
-//        std::cout << "estimateHomographyRANSAC : best support: " << best_support << "/" << n_matches << std::endl;
-//
-//        if (best_support == 0) {
-//            throw std::runtime_error("estimateHomographyRANSAC : failed to estimate homography");
-//        }
-//
-//        return best_H;
+        const int n_matches = points_lhs.size();
+
+        const int n_samples = 4;
+        uint64_t seed = 1;
+        const double reprojection_error_threshold_px = 2;
+
+        // https://en.wikipedia.org/wiki/Random_sample_consensus#Parameters
+        const double desired_p = 0.95;
+        const double filtered_matches_ratio = 0.35;
+        const double actual_p = std::pow(filtered_matches_ratio, n_samples);
+        const double standard_deviation = std::sqrt(1 - actual_p) / actual_p;
+        const int n_trials = std::log(1 - desired_p) / std::log(1 - actual_p) + standard_deviation;
+
+        int best_support = 0;
+        cv::Mat best_H;
+
+        std::vector<int> sample;
+        for (int i_trial = 0; i_trial < n_trials; ++i_trial) {
+            randomSample(sample, n_matches, n_samples, &seed);
+
+            cv::Mat H = estimateHomography4Points(points_lhs[sample[0]], points_lhs[sample[1]], points_lhs[sample[2]], points_lhs[sample[3]],
+                                                  points_rhs[sample[0]], points_rhs[sample[1]], points_rhs[sample[2]], points_rhs[sample[3]]);
+
+            int support = 0;
+            for (int i_point = 0; i_point < n_matches; ++i_point) {
+                try {
+                    cv::Point2d proj = phg::transformPoint(points_lhs[i_point], H);
+                    if (cv::norm(proj - cv::Point2d(points_rhs[i_point])) < reprojection_error_threshold_px) {
+                        ++support;
+                    }
+                } catch (const std::exception &e)
+                {
+                    std::cerr << e.what() << std::endl;
+                }
+            }
+
+            if (support > best_support) {
+                best_support = support;
+                best_H = H;
+
+                std::cout << "estimateHomographyRANSAC : support: " << best_support << "/" << n_matches << std::endl;
+
+                if (best_support == n_matches) {
+                    break;
+                }
+            }
+        }
+
+        std::cout << "estimateHomographyRANSAC : best support: " << best_support << "/" << n_matches << std::endl;
+
+        if (best_support == 0) {
+            throw std::runtime_error("estimateHomographyRANSAC : failed to estimate homography");
+        }
+
+        return best_H;
     }
 
 }
@@ -238,7 +251,13 @@ cv::Mat phg::findHomographyCV(const std::vector<cv::Point2f> &points_lhs, const
 // таким преобразованием внутри занимается функции cv::perspectiveTransform и cv::warpPerspective
 cv::Point2d phg::transformPoint(const cv::Point2d &pt, const cv::Mat &T)
 {
-    throw std::runtime_error("not implemented yet");
+    double x = T.at<double>(0, 0) * pt.x + T.at<double>(0, 1) * pt.y + T.at<double>(0, 2) * 1;
+    double y = T.at<double>(1, 0) * pt.x + T.at<double>(1, 1) * pt.y + T.at<double>(1, 2) * 1;
+    double w = T.at<double>(2, 0) * pt.x + T.at<double>(2, 1) * pt.y + T.at<double>(2, 2) * 1;
+
+    x /= w;
+    y /= w;
+    return {x, y};
 }
 
 cv::Point2d phg::transformPointCV(const cv::Point2d &pt, const cv::Mat &T) {

src/phg/sfm/panorama_stitcher.cpp

@@ -3,6 +3,7 @@
 
 #include <libutils/bbox2.h>
 #include <iostream>
+#include <queue>
 
 /*
  * imgs - список картинок
@@ -23,7 +24,30 @@ cv::Mat phg::stitchPanorama(const std::vector<cv::Mat> &imgs,
     {
         // здесь надо посчитать вектор Hs
         // при этом можно обойтись n_images - 1 вызовами функтора homography_builder
-        throw std::runtime_error("not implemented yet");
+        std::vector<std::vector<int>> children(n_images);
+        int root = -1;
+        for (int i = 0; i < n_images; ++i) {
+            if (parent[i] == -1) {
+                root = i;
+            } else {
+                children[parent[i]].push_back(i);
+            }
+        }
+        std::queue<int> q;
+        q.push(root);
+        Hs[root] = cv::Mat::eye(3, 3, CV_64FC1);
+        while (!q.empty()) {
+            int par = q.front();
+            q.pop();
+            for (auto child : children[par]) {
+                if (parent[child] == par) {
+                    cv::Mat H_temp = homography_builder(imgs[child], imgs[par]);
+
+                    Hs[child] = Hs[par] * H_temp;
+                    q.push(child);
+                }
+            }
+        }
     }
 
     bbox2<double, cv::Point2d> bbox;
@@ -43,6 +67,8 @@ cv::Mat phg::stitchPanorama(const std::vector<cv::Mat> &imgs,
 
     cv::Mat result = cv::Mat::zeros(result_height, result_width, CV_8UC3);
 
+    // я не понял надо ли это раскомментить или нет, но результат с ним хуже)
+
     // из-за растяжения пикселей при использовании прямой матрицы гомографии после отображения между пикселями остается пустое пространство
     // лучше использовать обратную и для каждого пикселя на итоговвой картинке проверять, с какой картинки он может получить цвет
     // тогда в некоторых пикселях цвет будет дублироваться, но изображение будет непрерывным