mirror of
https://github.com/opencv/opencv.git
synced 2024-12-14 08:59:11 +08:00
1401 lines
37 KiB
C++
1401 lines
37 KiB
C++
/*********************************************************************
|
|
* Software License Agreement (BSD License)
|
|
*
|
|
* Copyright (c) 2008-2010, Willow Garage, Inc.
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
*
|
|
* * Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* * Redistributions in binary form must reproduce the above
|
|
* copyright notice, this list of conditions and the following
|
|
* disclaimer in the documentation and/or other materials provided
|
|
* with the distribution.
|
|
* * Neither the name of the Willow Garage nor the names of its
|
|
* contributors may be used to endorse or promote products derived
|
|
* from this software without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
|
|
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
|
|
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
|
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
|
|
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*********************************************************************/
|
|
|
|
//
|
|
// The original code was written by
|
|
// Marius Muja
|
|
// and later modified and prepared
|
|
// for integration into OpenCV by
|
|
// Antonella Cascitelli,
|
|
// Marco Di Stefano and
|
|
// Stefano Fabri
|
|
// from Univ. of Rome
|
|
//
|
|
|
|
#include "precomp.hpp"
|
|
#include "opencv2/highgui/highgui.hpp"
|
|
#include <iostream>
|
|
#include <queue>
|
|
|
|
namespace cv
|
|
{
|
|
|
|
using std::queue;
|
|
|
|
typedef std::pair<int,int> coordinate_t;
|
|
typedef float orientation_t;
|
|
typedef std::vector<coordinate_t> template_coords_t;
|
|
typedef std::vector<orientation_t> template_orientations_t;
|
|
typedef std::pair<Point, float> location_scale_t;
|
|
|
|
class ChamferMatcher
|
|
{
|
|
|
|
private:
|
|
class Matching;
|
|
int max_matches_;
|
|
float min_match_distance_;
|
|
|
|
///////////////////////// Image iterators ////////////////////////////
|
|
|
|
class ImageIterator
|
|
{
|
|
public:
|
|
virtual bool hasNext() const = 0;
|
|
virtual location_scale_t next() = 0;
|
|
};
|
|
|
|
class ImageRange
|
|
{
|
|
public:
|
|
virtual ImageIterator* iterator() const = 0;
|
|
};
|
|
|
|
// Sliding window
|
|
|
|
class SlidingWindowImageRange : public ImageRange
|
|
{
|
|
int width_;
|
|
int height_;
|
|
int x_step_;
|
|
int y_step_;
|
|
int scales_;
|
|
float min_scale_;
|
|
float max_scale_;
|
|
|
|
public:
|
|
SlidingWindowImageRange(int width, int height, int x_step = 3, int y_step = 3, int scales = 5, float min_scale = 0.6, float max_scale = 1.6) :
|
|
width_(width), height_(height), x_step_(x_step),y_step_(y_step), scales_(scales), min_scale_(min_scale), max_scale_(max_scale)
|
|
{
|
|
}
|
|
|
|
|
|
ImageIterator* iterator() const;
|
|
};
|
|
|
|
class LocationImageRange : public ImageRange
|
|
{
|
|
const std::vector<Point>& locations_;
|
|
|
|
int scales_;
|
|
float min_scale_;
|
|
float max_scale_;
|
|
|
|
LocationImageRange(const LocationImageRange&);
|
|
LocationImageRange& operator=(const LocationImageRange&);
|
|
|
|
public:
|
|
LocationImageRange(const std::vector<Point>& locations, int scales = 5, float min_scale = 0.6, float max_scale = 1.6) :
|
|
locations_(locations), scales_(scales), min_scale_(min_scale), max_scale_(max_scale)
|
|
{
|
|
}
|
|
|
|
ImageIterator* iterator() const
|
|
{
|
|
return new LocationImageIterator(locations_, scales_, min_scale_, max_scale_);
|
|
}
|
|
};
|
|
|
|
|
|
class LocationScaleImageRange : public ImageRange
|
|
{
|
|
const std::vector<Point>& locations_;
|
|
const std::vector<float>& scales_;
|
|
|
|
LocationScaleImageRange(const LocationScaleImageRange&);
|
|
LocationScaleImageRange& operator=(const LocationScaleImageRange&);
|
|
public:
|
|
LocationScaleImageRange(const std::vector<Point>& locations, const std::vector<float>& scales) :
|
|
locations_(locations), scales_(scales)
|
|
{
|
|
assert(locations.size()==scales.size());
|
|
}
|
|
|
|
ImageIterator* iterator() const
|
|
{
|
|
return new LocationScaleImageIterator(locations_, scales_);
|
|
}
|
|
};
|
|
|
|
|
|
|
|
|
|
public:
|
|
/**
|
|
* Class that represents a template for chamfer matching.
|
|
*/
|
|
class Template
|
|
{
|
|
friend class ChamferMatcher::Matching;
|
|
friend class ChamferMatcher;
|
|
|
|
|
|
public:
|
|
std::vector<Template*> scaled_templates;
|
|
std::vector<int> addr;
|
|
int addr_width;
|
|
float scale;
|
|
template_coords_t coords;
|
|
|
|
template_orientations_t orientations;
|
|
Size size;
|
|
Point center;
|
|
|
|
public:
|
|
Template() : addr_width(-1)
|
|
{
|
|
}
|
|
|
|
Template(Mat& edge_image, float scale_ = 1);
|
|
|
|
~Template()
|
|
{
|
|
for (size_t i=0;i<scaled_templates.size();++i) {
|
|
delete scaled_templates[i];
|
|
}
|
|
scaled_templates.clear();
|
|
coords.clear();
|
|
orientations.clear();
|
|
}
|
|
void show() const;
|
|
|
|
|
|
|
|
private:
|
|
/**
|
|
* Resizes a template
|
|
*
|
|
* @param scale Scale to be resized to
|
|
*/
|
|
Template* rescale(float scale);
|
|
|
|
std::vector<int>& getTemplateAddresses(int width);
|
|
};
|
|
|
|
|
|
|
|
/**
|
|
* Used to represent a matching result.
|
|
*/
|
|
|
|
class Match
|
|
{
|
|
public:
|
|
float cost;
|
|
Point offset;
|
|
const Template* tpl;
|
|
};
|
|
|
|
typedef std::vector<Match> Matches;
|
|
|
|
private:
|
|
/**
|
|
* Implements the chamfer matching algorithm on images taking into account both distance from
|
|
* the template pixels to the nearest pixels and orientation alignment between template and image
|
|
* contours.
|
|
*/
|
|
class Matching
|
|
{
|
|
float truncate_;
|
|
bool use_orientation_;
|
|
|
|
std::vector<Template*> templates;
|
|
public:
|
|
Matching(bool use_orientation = true, float truncate = 10) : truncate_(truncate), use_orientation_(use_orientation)
|
|
{
|
|
}
|
|
|
|
~Matching()
|
|
{
|
|
for (size_t i = 0; i<templates.size(); i++) {
|
|
delete templates[i];
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Add a template to the detector from an edge image.
|
|
* @param templ An edge image
|
|
*/
|
|
void addTemplateFromImage(Mat& templ, float scale = 1.0);
|
|
|
|
/**
|
|
* Run matching using an edge image.
|
|
* @param edge_img Edge image
|
|
* @return a match object
|
|
*/
|
|
ChamferMatcher::Matches* matchEdgeImage(Mat& edge_img, const ImageRange& range, float orientation_weight = 0.5, int max_matches = 20, float min_match_distance = 10.0);
|
|
|
|
void addTemplate(Template& template_);
|
|
|
|
private:
|
|
|
|
float orientation_diff(float o1, float o2)
|
|
{
|
|
return fabs(o1-o2);
|
|
}
|
|
|
|
/**
|
|
* Computes the chamfer matching cost for one position in the target image.
|
|
* @param offset Offset where to compute cost
|
|
* @param dist_img Distance transform image.
|
|
* @param orientation_img Orientation image.
|
|
* @param tpl Template
|
|
* @param templ_orientations Orientations of the target points.
|
|
* @return matching result
|
|
*/
|
|
ChamferMatcher::Match* localChamferDistance(Point offset, Mat& dist_img, Mat& orientation_img, Template* tpl, float orientation_weight);
|
|
|
|
private:
|
|
/**
|
|
* Matches all templates.
|
|
* @param dist_img Distance transform image.
|
|
* @param orientation_img Orientation image.
|
|
*/
|
|
ChamferMatcher::Matches* matchTemplates(Mat& dist_img, Mat& orientation_img, const ImageRange& range, float orientation_weight);
|
|
|
|
void computeDistanceTransform(Mat& edges_img, Mat& dist_img, Mat& annotate_img, float truncate_dt, float a, float b);
|
|
void computeEdgeOrientations(Mat& edge_img, Mat& orientation_img);
|
|
void fillNonContourOrientations(Mat& annotated_img, Mat& orientation_img);
|
|
|
|
|
|
public:
|
|
/**
|
|
* Finds a contour in an edge image. The original image is altered by removing the found contour.
|
|
* @param templ_img Edge image
|
|
* @param coords Coordinates forming the contour.
|
|
* @return True while a contour is still found in the image.
|
|
*/
|
|
static bool findContour(Mat& templ_img, template_coords_t& coords);
|
|
|
|
/**
|
|
* Computes contour points orientations using the approach from:
|
|
*
|
|
* Matas, Shao and Kittler - Estimation of Curvature and Tangent Direction by
|
|
* Median Filtered Differencing
|
|
*
|
|
* @param coords Contour points
|
|
* @param orientations Contour points orientations
|
|
*/
|
|
static void findContourOrientations(const template_coords_t& coords, template_orientations_t& orientations);
|
|
|
|
|
|
/**
|
|
* Computes the angle of a line segment.
|
|
*
|
|
* @param a One end of the line segment
|
|
* @param b The other end.
|
|
* @param dx
|
|
* @param dy
|
|
* @return Angle in radians.
|
|
*/
|
|
static float getAngle(coordinate_t a, coordinate_t b, int& dx, int& dy);
|
|
|
|
/**
|
|
* Finds a point in the image from which to start contour following.
|
|
* @param templ_img
|
|
* @param p
|
|
* @return
|
|
*/
|
|
|
|
static bool findFirstContourPoint(Mat& templ_img, coordinate_t& p);
|
|
/**
|
|
* Method that extracts a single continuous contour from an image given a starting point.
|
|
* When it extracts the contour it tries to maintain the same direction (at a T-join for example).
|
|
*
|
|
* @param templ_
|
|
* @param coords
|
|
* @param direction
|
|
*/
|
|
static void followContour(Mat& templ_img, template_coords_t& coords, int direction);
|
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
class LocationImageIterator : public ImageIterator
|
|
{
|
|
const std::vector<Point>& locations_;
|
|
|
|
size_t iter_;
|
|
|
|
int scales_;
|
|
float min_scale_;
|
|
float max_scale_;
|
|
|
|
float scale_;
|
|
float scale_step_;
|
|
int scale_cnt_;
|
|
|
|
bool has_next_;
|
|
|
|
LocationImageIterator(const LocationImageIterator&);
|
|
LocationImageIterator& operator=(const LocationImageIterator&);
|
|
|
|
public:
|
|
LocationImageIterator(const std::vector<Point>& locations, int scales, float min_scale, float max_scale);
|
|
|
|
bool hasNext() const {
|
|
return has_next_;
|
|
}
|
|
|
|
location_scale_t next();
|
|
};
|
|
|
|
class LocationScaleImageIterator : public ImageIterator
|
|
{
|
|
const std::vector<Point>& locations_;
|
|
const std::vector<float>& scales_;
|
|
|
|
size_t iter_;
|
|
|
|
bool has_next_;
|
|
|
|
LocationScaleImageIterator(const LocationScaleImageIterator&);
|
|
LocationScaleImageIterator& operator=(const LocationScaleImageIterator&);
|
|
|
|
public:
|
|
LocationScaleImageIterator(const std::vector<Point>& locations, const std::vector<float>& scales) :
|
|
locations_(locations), scales_(scales)
|
|
{
|
|
assert(locations.size()==scales.size());
|
|
reset();
|
|
}
|
|
|
|
void reset()
|
|
{
|
|
iter_ = 0;
|
|
has_next_ = (locations_.size()==0 ? false : true);
|
|
}
|
|
|
|
bool hasNext() const {
|
|
return has_next_;
|
|
}
|
|
|
|
location_scale_t next();
|
|
};
|
|
|
|
class SlidingWindowImageIterator : public ImageIterator
|
|
{
|
|
int x_;
|
|
int y_;
|
|
float scale_;
|
|
float scale_step_;
|
|
int scale_cnt_;
|
|
|
|
bool has_next_;
|
|
|
|
int width_;
|
|
int height_;
|
|
int x_step_;
|
|
int y_step_;
|
|
int scales_;
|
|
float min_scale_;
|
|
float max_scale_;
|
|
|
|
|
|
public:
|
|
|
|
SlidingWindowImageIterator(int width, int height, int x_step, int y_step, int scales, float min_scale, float max_scale);
|
|
|
|
bool hasNext() const {
|
|
return has_next_;
|
|
}
|
|
|
|
location_scale_t next();
|
|
};
|
|
|
|
|
|
|
|
|
|
int count;
|
|
Matches matches;
|
|
int pad_x;
|
|
int pad_y;
|
|
int scales;
|
|
float minScale;
|
|
float maxScale;
|
|
float orientation_weight;
|
|
float truncate;
|
|
Matching * chamfer_;
|
|
|
|
public:
|
|
ChamferMatcher(int _max_matches = 20, float _min_match_distance = 1.0, int _pad_x = 3,
|
|
int _pad_y = 3, int _scales = 5, float _minScale = 0.6, float _maxScale = 1.6,
|
|
float _orientation_weight = 0.5, float _truncate = 20)
|
|
{
|
|
max_matches_ = _max_matches;
|
|
min_match_distance_ = _min_match_distance;
|
|
pad_x = _pad_x;
|
|
pad_y = _pad_y;
|
|
scales = _scales;
|
|
minScale = _minScale;
|
|
maxScale = _maxScale;
|
|
orientation_weight = _orientation_weight;
|
|
truncate = _truncate;
|
|
count = 0;
|
|
|
|
matches.resize(max_matches_);
|
|
chamfer_ = new Matching(true);
|
|
}
|
|
|
|
void showMatch(Mat& img, int index = 0);
|
|
void showMatch(Mat& img, Match match_);
|
|
|
|
const Matches& matching(Template&, Mat&);
|
|
|
|
private:
|
|
void addMatch(float cost, Point offset, const Template* tpl);
|
|
|
|
|
|
};
|
|
|
|
|
|
///////////////////// implementation ///////////////////////////
|
|
|
|
ChamferMatcher::SlidingWindowImageIterator::SlidingWindowImageIterator( int width,
|
|
int height,
|
|
int x_step = 3,
|
|
int y_step = 3,
|
|
int scales = 5,
|
|
float min_scale = 0.6,
|
|
float max_scale = 1.6) :
|
|
|
|
width_(width),
|
|
height_(height),
|
|
x_step_(x_step),
|
|
y_step_(y_step),
|
|
scales_(scales),
|
|
min_scale_(min_scale),
|
|
max_scale_(max_scale)
|
|
{
|
|
x_ = 0;
|
|
y_ = 0;
|
|
scale_cnt_ = 0;
|
|
scale_ = min_scale_;
|
|
has_next_ = true;
|
|
scale_step_ = (max_scale_-min_scale_)/scales_;
|
|
}
|
|
|
|
location_scale_t ChamferMatcher::SlidingWindowImageIterator::next()
|
|
{
|
|
location_scale_t next_val = std::make_pair(Point(x_,y_),scale_);
|
|
|
|
x_ += x_step_;
|
|
|
|
if (x_ >= width_) {
|
|
x_ = 0;
|
|
y_ += y_step_;
|
|
|
|
if (y_ >= height_) {
|
|
y_ = 0;
|
|
scale_ += scale_step_;
|
|
scale_cnt_++;
|
|
|
|
if (scale_cnt_ == scales_) {
|
|
has_next_ = false;
|
|
scale_cnt_ = 0;
|
|
scale_ = min_scale_;
|
|
}
|
|
}
|
|
}
|
|
|
|
return next_val;
|
|
}
|
|
|
|
|
|
|
|
ChamferMatcher::ImageIterator* ChamferMatcher::SlidingWindowImageRange::iterator() const
|
|
{
|
|
return new SlidingWindowImageIterator(width_, height_, x_step_, y_step_, scales_, min_scale_, max_scale_);
|
|
}
|
|
|
|
|
|
|
|
ChamferMatcher::LocationImageIterator::LocationImageIterator(const std::vector<Point>& locations,
|
|
int scales = 5,
|
|
float min_scale = 0.6,
|
|
float max_scale = 1.6) :
|
|
locations_(locations),
|
|
scales_(scales),
|
|
min_scale_(min_scale),
|
|
max_scale_(max_scale)
|
|
{
|
|
iter_ = 0;
|
|
scale_cnt_ = 0;
|
|
scale_ = min_scale_;
|
|
has_next_ = (locations_.size()==0 ? false : true);
|
|
scale_step_ = (max_scale_-min_scale_)/scales_;
|
|
}
|
|
|
|
location_scale_t ChamferMatcher::LocationImageIterator:: next()
|
|
{
|
|
location_scale_t next_val = std::make_pair(locations_[iter_],scale_);
|
|
|
|
iter_ ++;
|
|
if (iter_==locations_.size()) {
|
|
iter_ = 0;
|
|
scale_ += scale_step_;
|
|
scale_cnt_++;
|
|
|
|
if (scale_cnt_ == scales_) {
|
|
has_next_ = false;
|
|
scale_cnt_ = 0;
|
|
scale_ = min_scale_;
|
|
}
|
|
}
|
|
|
|
return next_val;
|
|
}
|
|
|
|
|
|
location_scale_t ChamferMatcher::LocationScaleImageIterator::next()
|
|
{
|
|
location_scale_t next_val = std::make_pair(locations_[iter_],scales_[iter_]);
|
|
|
|
iter_ ++;
|
|
if (iter_==locations_.size()) {
|
|
iter_ = 0;
|
|
|
|
has_next_ = false;
|
|
}
|
|
|
|
return next_val;
|
|
}
|
|
|
|
|
|
|
|
bool ChamferMatcher::Matching::findFirstContourPoint(Mat& templ_img, coordinate_t& p)
|
|
{
|
|
for (int y=0;y<templ_img.rows;++y) {
|
|
for (int x=0;x<templ_img.cols;++x) {
|
|
if (templ_img.at<uchar>(y,x)!=0) {
|
|
p.first = x;
|
|
p.second = y;
|
|
return true;
|
|
}
|
|
}
|
|
}
|
|
return false;
|
|
}
|
|
|
|
|
|
|
|
void ChamferMatcher::Matching::followContour(Mat& templ_img, template_coords_t& coords, int direction = -1)
|
|
{
|
|
const int dir[][2] = { {-1,-1}, {-1,0}, {-1,1}, {0,1}, {1,1}, {1,0}, {1,-1}, {0,-1} };
|
|
coordinate_t next;
|
|
coordinate_t next_temp;
|
|
unsigned char ptr;
|
|
|
|
assert (direction==-1 || !coords.empty());
|
|
|
|
coordinate_t crt = coords.back();
|
|
|
|
// mark the current pixel as visited
|
|
templ_img.at<uchar>(crt.second,crt.first) = 0;
|
|
if (direction==-1) {
|
|
for (int j = 0; j<7; ++j) {
|
|
next.first = crt.first + dir[j][1];
|
|
next.second = crt.second + dir[j][0];
|
|
if (next.first >= 0 && next.first < templ_img.cols &&
|
|
next.second >= 0 && next.second < templ_img.rows){
|
|
ptr = templ_img.at<uchar>(next.second, next.first);
|
|
if (ptr!=0) {
|
|
coords.push_back(next);
|
|
followContour(templ_img, coords,j);
|
|
// try to continue contour in the other direction
|
|
reverse(coords.begin(), coords.end());
|
|
followContour(templ_img, coords, (j+4)%8);
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
else {
|
|
int k = direction;
|
|
int k_cost = 3;
|
|
next.first = crt.first + dir[k][1];
|
|
next.second = crt.second + dir[k][0];
|
|
if (next.first >= 0 && next.first < templ_img.cols &&
|
|
next.second >= 0 && next.second < templ_img.rows){
|
|
ptr = templ_img.at<uchar>(next.second, next.first);
|
|
if (ptr!=0) {
|
|
k_cost = std::abs(dir[k][1]) + std::abs(dir[k][0]);
|
|
}
|
|
int p = k;
|
|
int n = k;
|
|
|
|
for (int j = 0 ;j<3; ++j) {
|
|
p = (p + 7) % 8;
|
|
n = (n + 1) % 8;
|
|
next.first = crt.first + dir[p][1];
|
|
next.second = crt.second + dir[p][0];
|
|
if (next.first >= 0 && next.first < templ_img.cols &&
|
|
next.second >= 0 && next.second < templ_img.rows){
|
|
ptr = templ_img.at<uchar>(next.second, next.first);
|
|
if (ptr!=0) {
|
|
int p_cost = std::abs(dir[p][1]) + std::abs(dir[p][0]);
|
|
if (p_cost<k_cost) {
|
|
k_cost = p_cost;
|
|
k = p;
|
|
}
|
|
}
|
|
next.first = crt.first + dir[n][1];
|
|
next.second = crt.second + dir[n][0];
|
|
if (next.first >= 0 && next.first < templ_img.cols &&
|
|
next.second >= 0 && next.second < templ_img.rows){
|
|
ptr = templ_img.at<uchar>(next.second, next.first);
|
|
if (ptr!=0) {
|
|
int n_cost = std::abs(dir[n][1]) + std::abs(dir[n][0]);
|
|
if (n_cost<k_cost) {
|
|
k_cost = n_cost;
|
|
k = n;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if (k_cost!=3) {
|
|
next.first = crt.first + dir[k][1];
|
|
next.second = crt.second + dir[k][0];
|
|
if (next.first >= 0 && next.first < templ_img.cols &&
|
|
next.second >= 0 && next.second < templ_img.rows) {
|
|
coords.push_back(next);
|
|
followContour(templ_img, coords, k);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
bool ChamferMatcher::Matching::findContour(Mat& templ_img, template_coords_t& coords)
|
|
{
|
|
coordinate_t start_point;
|
|
|
|
bool found = findFirstContourPoint(templ_img,start_point);
|
|
if (found) {
|
|
coords.push_back(start_point);
|
|
followContour(templ_img, coords);
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
|
|
float ChamferMatcher::Matching::getAngle(coordinate_t a, coordinate_t b, int& dx, int& dy)
|
|
{
|
|
dx = b.first-a.first;
|
|
dy = -(b.second-a.second); // in image coordinated Y axis points downward
|
|
float angle = atan2((float)dy,(float)dx);
|
|
|
|
if (angle<0) {
|
|
angle+=(float)CV_PI;
|
|
}
|
|
|
|
return angle;
|
|
}
|
|
|
|
|
|
|
|
void ChamferMatcher::Matching::findContourOrientations(const template_coords_t& coords, template_orientations_t& orientations)
|
|
{
|
|
const int M = 5;
|
|
int coords_size = (int)coords.size();
|
|
|
|
std::vector<float> angles(2*M);
|
|
orientations.insert(orientations.begin(), coords_size, float(-3*CV_PI)); // mark as invalid in the beginning
|
|
|
|
if (coords_size<2*M+1) { // if contour not long enough to estimate orientations, abort
|
|
return;
|
|
}
|
|
|
|
for (int i=M;i<coords_size-M;++i) {
|
|
coordinate_t crt = coords[i];
|
|
coordinate_t other;
|
|
int k = 0;
|
|
int dx, dy;
|
|
// compute previous M angles
|
|
for (int j=M;j>0;--j) {
|
|
other = coords[i-j];
|
|
angles[k++] = getAngle(other,crt, dx, dy);
|
|
}
|
|
// compute next M angles
|
|
for (int j=1;j<=M;++j) {
|
|
other = coords[i+j];
|
|
angles[k++] = getAngle(crt, other, dx, dy);
|
|
}
|
|
|
|
// get the middle two angles
|
|
nth_element(angles.begin(), angles.begin()+M-1, angles.end());
|
|
nth_element(angles.begin()+M-1, angles.begin()+M, angles.end());
|
|
// sort(angles.begin(), angles.end());
|
|
|
|
// average them to compute tangent
|
|
orientations[i] = (angles[M-1]+angles[M])/2;
|
|
}
|
|
}
|
|
|
|
//////////////////////// Template /////////////////////////////////////
|
|
|
|
ChamferMatcher::Template::Template(Mat& edge_image, float scale_) : addr_width(-1), scale(scale_)
|
|
{
|
|
template_coords_t local_coords;
|
|
template_orientations_t local_orientations;
|
|
|
|
while (ChamferMatcher::Matching::findContour(edge_image, local_coords)) {
|
|
ChamferMatcher::Matching::findContourOrientations(local_coords, local_orientations);
|
|
|
|
coords.insert(coords.end(), local_coords.begin(), local_coords.end());
|
|
orientations.insert(orientations.end(), local_orientations.begin(), local_orientations.end());
|
|
local_coords.clear();
|
|
local_orientations.clear();
|
|
}
|
|
|
|
|
|
size = edge_image.size();
|
|
Point min, max;
|
|
min.x = size.width;
|
|
min.y = size.height;
|
|
max.x = 0;
|
|
max.y = 0;
|
|
|
|
center = Point(0,0);
|
|
for (size_t i=0;i<coords.size();++i) {
|
|
center.x += coords[i].first;
|
|
center.y += coords[i].second;
|
|
|
|
if (min.x>coords[i].first) min.x = coords[i].first;
|
|
if (min.y>coords[i].second) min.y = coords[i].second;
|
|
if (max.x<coords[i].first) max.x = coords[i].first;
|
|
if (max.y<coords[i].second) max.y = coords[i].second;
|
|
}
|
|
|
|
size.width = max.x - min.x;
|
|
size.height = max.y - min.y;
|
|
int coords_size = (int)coords.size();
|
|
|
|
center.x /= MAX(coords_size, 1);
|
|
center.y /= MAX(coords_size, 1);
|
|
|
|
for (int i=0;i<coords_size;++i) {
|
|
coords[i].first -= center.x;
|
|
coords[i].second -= center.y;
|
|
}
|
|
}
|
|
|
|
|
|
vector<int>& ChamferMatcher::Template::getTemplateAddresses(int width)
|
|
{
|
|
if (addr_width!=width) {
|
|
addr.resize(coords.size());
|
|
addr_width = width;
|
|
|
|
for (size_t i=0; i<coords.size();++i) {
|
|
addr[i] = coords[i].second*width+coords[i].first;
|
|
}
|
|
}
|
|
return addr;
|
|
}
|
|
|
|
|
|
/**
|
|
* Resizes a template
|
|
*
|
|
* @param scale Scale to be resized to
|
|
*/
|
|
ChamferMatcher::Template* ChamferMatcher::Template::rescale(float new_scale)
|
|
{
|
|
|
|
if (fabs(scale-new_scale)<1e-6) return this;
|
|
|
|
for (size_t i=0;i<scaled_templates.size();++i) {
|
|
if (fabs(scaled_templates[i]->scale-new_scale)<1e-6) {
|
|
return scaled_templates[i];
|
|
}
|
|
}
|
|
|
|
float scale_factor = new_scale/scale;
|
|
|
|
Template* tpl = new Template();
|
|
tpl->scale = new_scale;
|
|
|
|
tpl->center.x = int(center.x*scale_factor+0.5);
|
|
tpl->center.y = int(center.y*scale_factor+0.5);
|
|
|
|
tpl->size.width = int(size.width*scale_factor+0.5);
|
|
tpl->size.height = int(size.height*scale_factor+0.5);
|
|
|
|
tpl->coords.resize(coords.size());
|
|
tpl->orientations.resize(orientations.size());
|
|
for (size_t i=0;i<coords.size();++i) {
|
|
tpl->coords[i].first = int(coords[i].first*scale_factor+0.5);
|
|
tpl->coords[i].second = int(coords[i].second*scale_factor+0.5);
|
|
tpl->orientations[i] = orientations[i];
|
|
}
|
|
scaled_templates.push_back(tpl);
|
|
|
|
return tpl;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ChamferMatcher::Template::show() const
|
|
{
|
|
int pad = 50;
|
|
//Attention size is not correct
|
|
Mat templ_color (Size(size.width+(pad*2), size.height+(pad*2)), CV_8UC3);
|
|
templ_color.setTo(0);
|
|
|
|
for (size_t i=0;i<coords.size();++i) {
|
|
|
|
int x = center.x+coords[i].first+pad;
|
|
int y = center.y+coords[i].second+pad;
|
|
templ_color.at<Vec3b>(y,x)[1]=255;
|
|
//CV_PIXEL(unsigned char, templ_color,x,y)[1] = 255;
|
|
|
|
if (i%3==0) {
|
|
if (orientations[i] < -CV_PI) {
|
|
continue;
|
|
}
|
|
Point p1;
|
|
p1.x = x;
|
|
p1.y = y;
|
|
Point p2;
|
|
p2.x = x + pad*(int)(sin(orientations[i])*100)/100;
|
|
p2.y = y + pad*(int)(cos(orientations[i])*100)/100;
|
|
|
|
line(templ_color, p1,p2, CV_RGB(255,0,0));
|
|
}
|
|
}
|
|
|
|
|
|
circle(templ_color,Point(center.x + pad, center.y + pad),1,CV_RGB(0,255,0));
|
|
|
|
namedWindow("templ",1);
|
|
imshow("templ",templ_color);
|
|
|
|
cvWaitKey(0);
|
|
|
|
templ_color.release();
|
|
}
|
|
|
|
|
|
//////////////////////// Matching /////////////////////////////////////
|
|
|
|
|
|
void ChamferMatcher::Matching::addTemplateFromImage(Mat& templ, float scale)
|
|
{
|
|
Template* cmt = new Template(templ, scale);
|
|
if(templates.size() > 0)
|
|
templates.clear();
|
|
templates.push_back(cmt);
|
|
cmt->show();
|
|
}
|
|
|
|
void ChamferMatcher::Matching::addTemplate(Template& template_){
|
|
if(templates.size() > 0)
|
|
templates.clear();
|
|
templates.push_back(&template_);
|
|
}
|
|
/**
|
|
* Alternative version of computeDistanceTransform, will probably be used to compute distance
|
|
* transform annotated with edge orientation.
|
|
*/
|
|
void ChamferMatcher::Matching::computeDistanceTransform(Mat& edges_img, Mat& dist_img, Mat& annotate_img, float truncate_dt, float a = 1.0, float b = 1.5)
|
|
{
|
|
int d[][2] = { {-1,-1}, { 0,-1}, { 1,-1},
|
|
{-1,0}, { 1,0},
|
|
{-1,1}, { 0,1}, { 1,1} };
|
|
|
|
|
|
Size s = edges_img.size();
|
|
int w = s.width;
|
|
int h = s.height;
|
|
// set distance to the edge pixels to 0 and put them in the queue
|
|
std::queue<std::pair<int,int> > q;
|
|
|
|
|
|
|
|
for (int y=0;y<h;++y) {
|
|
for (int x=0;x<w;++x) {
|
|
|
|
unsigned char edge_val = edges_img.at<uchar>(y,x);
|
|
if ( (edge_val!=0) ) {
|
|
q.push(std::make_pair(x,y));
|
|
dist_img.at<float>(y,x)= 0;
|
|
|
|
if (&annotate_img!=NULL) {
|
|
annotate_img.at<Vec2i>(y,x)[0]=x;
|
|
annotate_img.at<Vec2i>(y,x)[1]=y;
|
|
}
|
|
}
|
|
else {
|
|
dist_img.at<float>(y,x)=-1;
|
|
}
|
|
}
|
|
}
|
|
|
|
// breadth first computation of distance transform
|
|
std::pair<int,int> crt;
|
|
while (!q.empty()) {
|
|
crt = q.front();
|
|
q.pop();
|
|
|
|
int x = crt.first;
|
|
int y = crt.second;
|
|
|
|
float dist_orig = dist_img.at<float>(y,x);
|
|
float dist;
|
|
|
|
for (size_t i=0;i<sizeof(d)/sizeof(d[0]);++i) {
|
|
int nx = x + d[i][0];
|
|
int ny = y + d[i][1];
|
|
|
|
if (nx<0 || ny<0 || nx>=w || ny>=h) continue;
|
|
|
|
if (std::abs(d[i][0]+d[i][1])==1) {
|
|
dist = (dist_orig)+a;
|
|
}
|
|
else {
|
|
dist = (dist_orig)+b;
|
|
}
|
|
|
|
float dt = dist_img.at<float>(ny,nx);
|
|
|
|
if (dt==-1 || dt>dist) {
|
|
dist_img.at<float>(ny,nx) = dist;
|
|
q.push(std::make_pair(nx,ny));
|
|
|
|
if (&annotate_img!=NULL) {
|
|
annotate_img.at<Vec2i>(ny,nx)[0]=annotate_img.at<Vec2i>(y,x)[0];
|
|
annotate_img.at<Vec2i>(ny,nx)[1]=annotate_img.at<Vec2i>(y,x)[1];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
// truncate dt
|
|
|
|
if (truncate_dt>0) {
|
|
Mat dist_img_thr = dist_img.clone();
|
|
threshold(dist_img, dist_img_thr, truncate_dt,0.0 ,THRESH_TRUNC);
|
|
dist_img_thr.copyTo(dist_img);
|
|
}
|
|
}
|
|
|
|
|
|
void ChamferMatcher::Matching::computeEdgeOrientations(Mat& edge_img, Mat& orientation_img)
|
|
{
|
|
Mat contour_img(edge_img.size(), CV_8UC1);
|
|
|
|
orientation_img.setTo(3*(-CV_PI));
|
|
template_coords_t coords;
|
|
template_orientations_t orientations;
|
|
|
|
while (ChamferMatcher::Matching::findContour(edge_img, coords)) {
|
|
|
|
ChamferMatcher::Matching::findContourOrientations(coords, orientations);
|
|
|
|
// set orientation pixel in orientation image
|
|
for (size_t i = 0; i<coords.size();++i) {
|
|
int x = coords[i].first;
|
|
int y = coords[i].second;
|
|
// if (orientations[i]>-CV_PI)
|
|
// {
|
|
//CV_PIXEL(unsigned char, contour_img, x, y)[0] = 255;
|
|
contour_img.at<uchar>(y,x)=255;
|
|
// }
|
|
//CV_PIXEL(float, orientation_img, x, y)[0] = orientations[i];
|
|
orientation_img.at<float>(y,x)=orientations[i];
|
|
}
|
|
|
|
|
|
coords.clear();
|
|
orientations.clear();
|
|
}
|
|
|
|
//imwrite("contours.pgm", contour_img);
|
|
}
|
|
|
|
|
|
void ChamferMatcher::Matching::fillNonContourOrientations(Mat& annotated_img, Mat& orientation_img)
|
|
{
|
|
int cols = annotated_img.cols;
|
|
int rows = annotated_img.rows;
|
|
|
|
assert(orientation_img.cols==cols && orientation_img.rows==rows);
|
|
|
|
for (int y=0;y<rows;++y) {
|
|
for (int x=0;x<cols;++x) {
|
|
int xorig = annotated_img.at<Vec2i>(y,x)[0];
|
|
int yorig = annotated_img.at<Vec2i>(y,x)[1];
|
|
|
|
if (x!=xorig || y!=yorig) {
|
|
//orientation_img.at<float>(yorig,xorig)=orientation_img.at<float>(y,x);
|
|
orientation_img.at<float>(y,x)=orientation_img.at<float>(yorig,xorig);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
ChamferMatcher::Match* ChamferMatcher::Matching::localChamferDistance(Point offset, Mat& dist_img, Mat& orientation_img,
|
|
ChamferMatcher::Template* tpl, float alpha)
|
|
{
|
|
int x = offset.x;
|
|
int y = offset.y;
|
|
|
|
float beta = 1-alpha;
|
|
|
|
std::vector<int>& addr = tpl->getTemplateAddresses(dist_img.cols);
|
|
|
|
float* ptr = dist_img.ptr<float>(y)+x;
|
|
|
|
|
|
float sum_distance = 0;
|
|
for (size_t i=0; i<addr.size();++i) {
|
|
if(addr[i] < (dist_img.cols*dist_img.rows) - (offset.y*dist_img.cols + offset.x)){
|
|
sum_distance += *(ptr+addr[i]);
|
|
}
|
|
}
|
|
|
|
float cost = (sum_distance/truncate_)/addr.size();
|
|
|
|
|
|
if (&orientation_img!=NULL) {
|
|
float* optr = orientation_img.ptr<float>(y)+x;
|
|
float sum_orientation = 0;
|
|
int cnt_orientation = 0;
|
|
|
|
for (size_t i=0;i<addr.size();++i) {
|
|
|
|
if(addr[i] < (orientation_img.cols*orientation_img.rows) - (offset.y*orientation_img.cols + offset.x)){
|
|
if (tpl->orientations[i]>=-CV_PI && (*(optr+addr[i]))>=-CV_PI) {
|
|
sum_orientation += orientation_diff(tpl->orientations[i], (*(optr+addr[i])));
|
|
cnt_orientation++;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (cnt_orientation>0) {
|
|
cost = (float)(beta*cost+alpha*(sum_orientation/(2*CV_PI))/cnt_orientation);
|
|
}
|
|
|
|
}
|
|
|
|
if(cost > 0){
|
|
ChamferMatcher::Match* istance(new ChamferMatcher::Match());
|
|
istance->cost = cost;
|
|
istance->offset = offset;
|
|
istance->tpl = tpl;
|
|
|
|
return istance;
|
|
}
|
|
|
|
return NULL;
|
|
}
|
|
|
|
|
|
ChamferMatcher::Matches* ChamferMatcher::Matching::matchTemplates(Mat& dist_img, Mat& orientation_img, const ImageRange& range, float orientation_weight)
|
|
{
|
|
|
|
ChamferMatcher::Matches* matches(new Matches());
|
|
// try each template
|
|
for(size_t i = 0; i < templates.size(); i++) {
|
|
ImageIterator* it = range.iterator();
|
|
while (it->hasNext()) {
|
|
location_scale_t crt = it->next();
|
|
|
|
Point loc = crt.first;
|
|
float scale = crt.second;
|
|
Template* tpl = templates[i]->rescale(scale);
|
|
|
|
|
|
if (loc.x-tpl->center.x<0 || loc.x+tpl->size.width/2>=dist_img.cols) continue;
|
|
if (loc.y-tpl->center.y<0 || loc.y+tpl->size.height/2>=dist_img.rows) continue;
|
|
|
|
ChamferMatcher::Match* is = localChamferDistance(loc, dist_img, orientation_img, tpl, orientation_weight);
|
|
if(is)
|
|
matches->push_back(*is);
|
|
}
|
|
|
|
delete it;
|
|
}
|
|
return matches;
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
* Run matching using an edge image.
|
|
* @param edge_img Edge image
|
|
* @return a match object
|
|
*/
|
|
ChamferMatcher::Matches* ChamferMatcher::Matching::matchEdgeImage(Mat& edge_img, const ImageRange& range, float orientation_weight, int /*max_matches*/, float /*min_match_distance*/)
|
|
{
|
|
CV_Assert(edge_img.channels()==1);
|
|
|
|
Mat dist_img;
|
|
Mat annotated_img;
|
|
Mat orientation_img;
|
|
|
|
annotated_img.create(edge_img.size(), CV_32SC2);
|
|
dist_img.create(edge_img.size(),CV_32FC1);
|
|
dist_img.setTo(0);
|
|
// Computing distance transform
|
|
computeDistanceTransform(edge_img,dist_img, annotated_img, truncate_);
|
|
|
|
|
|
//orientation_img = NULL;
|
|
if (use_orientation_) {
|
|
orientation_img.create(edge_img.size(), CV_32FC1);
|
|
orientation_img.setTo(0);
|
|
Mat edge_clone = edge_img.clone();
|
|
computeEdgeOrientations(edge_clone, orientation_img );
|
|
edge_clone.release();
|
|
fillNonContourOrientations(annotated_img, orientation_img);
|
|
}
|
|
|
|
|
|
// Template matching
|
|
ChamferMatcher::Matches* matches = matchTemplates( dist_img,
|
|
orientation_img,
|
|
range,
|
|
orientation_weight);
|
|
|
|
|
|
if (use_orientation_) {
|
|
orientation_img.release();
|
|
}
|
|
dist_img.release();
|
|
annotated_img.release();
|
|
|
|
return matches;
|
|
}
|
|
|
|
|
|
void ChamferMatcher::addMatch(float cost, Point offset, const Template* tpl)
|
|
{
|
|
bool new_match = true;
|
|
for (int i=0; i<count; ++i) {
|
|
if (std::abs(matches[i].offset.x-offset.x)+std::abs(matches[i].offset.y-offset.y)<min_match_distance_) {
|
|
// too close, not a new match
|
|
new_match = false;
|
|
// if better cost, replace existing match
|
|
if (cost<matches[i].cost) {
|
|
matches[i].cost = cost;
|
|
matches[i].offset = offset;
|
|
matches[i].tpl = tpl;
|
|
}
|
|
// re-bubble to keep ordered
|
|
int k = i;
|
|
while (k>0) {
|
|
if (matches[k-1].cost>matches[k].cost) {
|
|
std::swap(matches[k-1],matches[k]);
|
|
}
|
|
k--;
|
|
}
|
|
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (new_match) {
|
|
// if we don't have enough matches yet, add it to the array
|
|
if (count<max_matches_) {
|
|
matches[count].cost = cost;
|
|
matches[count].offset = offset;
|
|
matches[count].tpl = tpl;
|
|
count++;
|
|
}
|
|
// otherwise find the right position to insert it
|
|
else {
|
|
// if higher cost than the worst current match, just ignore it
|
|
if (matches[count-1].cost<cost) {
|
|
return;
|
|
}
|
|
|
|
int j = 0;
|
|
// skip all matches better than current one
|
|
while (matches[j].cost<cost) j++;
|
|
|
|
// shift matches one position
|
|
int k = count-2;
|
|
while (k>=j) {
|
|
matches[k+1] = matches[k];
|
|
k--;
|
|
}
|
|
|
|
matches[j].cost = cost;
|
|
matches[j].offset = offset;
|
|
matches[j].tpl = tpl;
|
|
}
|
|
}
|
|
}
|
|
|
|
void ChamferMatcher::showMatch(Mat& img, int index)
|
|
{
|
|
if (index>=count) {
|
|
std::cout << "Index too big.\n" << std::endl;
|
|
}
|
|
|
|
assert(img.channels()==3);
|
|
|
|
Match match = matches[index];
|
|
|
|
const template_coords_t& templ_coords = match.tpl->coords;
|
|
int x, y;
|
|
for (size_t i=0;i<templ_coords.size();++i) {
|
|
x = match.offset.x + templ_coords[i].first;
|
|
y = match.offset.y + templ_coords[i].second;
|
|
|
|
if ( x > img.cols-1 || x < 0 || y > img.rows-1 || y < 0) continue;
|
|
img.at<Vec3b>(y,x)[0]=0;
|
|
img.at<Vec3b>(y,x)[2]=0;
|
|
img.at<Vec3b>(y,x)[1]=255;
|
|
}
|
|
}
|
|
|
|
void ChamferMatcher::showMatch(Mat& img, Match match)
|
|
{
|
|
assert(img.channels()==3);
|
|
|
|
const template_coords_t& templ_coords = match.tpl->coords;
|
|
for (size_t i=0;i<templ_coords.size();++i) {
|
|
int x = match.offset.x + templ_coords[i].first;
|
|
int y = match.offset.y + templ_coords[i].second;
|
|
if ( x > img.cols-1 || x < 0 || y > img.rows-1 || y < 0) continue;
|
|
img.at<Vec3b>(y,x)[0]=0;
|
|
img.at<Vec3b>(y,x)[2]=0;
|
|
img.at<Vec3b>(y,x)[1]=255;
|
|
}
|
|
match.tpl->show();
|
|
}
|
|
|
|
const ChamferMatcher::Matches& ChamferMatcher::matching(Template& tpl, Mat& image_){
|
|
chamfer_->addTemplate(tpl);
|
|
|
|
matches.clear();
|
|
matches.resize(max_matches_);
|
|
count = 0;
|
|
|
|
|
|
Matches* matches_ = chamfer_->matchEdgeImage( image_,
|
|
ChamferMatcher::
|
|
SlidingWindowImageRange(image_.cols,
|
|
image_.rows,
|
|
pad_x,
|
|
pad_y,
|
|
scales,
|
|
minScale,
|
|
maxScale),
|
|
orientation_weight,
|
|
max_matches_,
|
|
min_match_distance_);
|
|
|
|
|
|
|
|
for(int i = 0; i < (int)matches_->size(); i++){
|
|
addMatch(matches_->at(i).cost, matches_->at(i).offset, matches_->at(i).tpl);
|
|
}
|
|
|
|
matches_->clear();
|
|
delete matches_;
|
|
matches_ = NULL;
|
|
|
|
matches.resize(count);
|
|
|
|
|
|
return matches;
|
|
|
|
}
|
|
|
|
|
|
int chamerMatching( Mat& img, Mat& templ,
|
|
std::vector<std::vector<Point> >& results, std::vector<float>& costs,
|
|
double templScale, int maxMatches, double minMatchDistance, int padX,
|
|
int padY, int scales, double minScale, double maxScale,
|
|
double orientationWeight, double truncate )
|
|
{
|
|
CV_Assert(img.type() == CV_8UC1 && templ.type() == CV_8UC1);
|
|
|
|
ChamferMatcher matcher_(maxMatches, (float)minMatchDistance, padX, padY, scales,
|
|
(float)minScale, (float)maxScale,
|
|
(float)orientationWeight, (float)truncate);
|
|
|
|
ChamferMatcher::Template template_(templ, (float)templScale);
|
|
ChamferMatcher::Matches match_instances = matcher_.matching(template_, img);
|
|
|
|
size_t i, nmatches = match_instances.size();
|
|
|
|
results.resize(nmatches);
|
|
costs.resize(nmatches);
|
|
|
|
int bestIdx = -1;
|
|
double minCost = DBL_MAX;
|
|
|
|
for( i = 0; i < nmatches; i++ )
|
|
{
|
|
const ChamferMatcher::Match& match = match_instances[i];
|
|
double cval = match.cost;
|
|
if( cval < minCost)
|
|
{
|
|
minCost = cval;
|
|
bestIdx = (int)i;
|
|
}
|
|
costs[i] = (float)cval;
|
|
|
|
const template_coords_t& templ_coords = match.tpl->coords;
|
|
std::vector<Point>& templPoints = results[i];
|
|
size_t j, npoints = templ_coords.size();
|
|
templPoints.resize(npoints);
|
|
|
|
for (j = 0; j < npoints; j++ )
|
|
{
|
|
int x = match.offset.x + templ_coords[j].first;
|
|
int y = match.offset.y + templ_coords[j].second;
|
|
templPoints[j] = Point(x,y);
|
|
}
|
|
}
|
|
|
|
return bestIdx;
|
|
}
|
|
|
|
}
|