opencv/modules/objdetect/src/softcascade.cpp
2012-11-07 03:19:04 +04:00

420 lines
13 KiB
C++

/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// 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.
//
// * The name of the copyright holders may not 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 Intel Corporation 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.
//M*/
#include <precomp.hpp>
#include <opencv2/objdetect/objdetect.hpp>
#include <opencv2/core/core.hpp>
#include <vector>
#include <string>
#include <stdio.h>
namespace {
static const char* SC_OCT_SCALE = "scale";
static const char* SC_OCT_STAGES = "stageNum";
struct Octave
{
float scale;
int stages;
Octave(){}
Octave(const cv::FileNode& fn) : scale((float)fn[SC_OCT_SCALE]), stages((int)fn[SC_OCT_STAGES])
{/*printf("octave: %f %d\n", scale, stages);*/}
};
static const char *SC_STAGE_THRESHOLD = "stageThreshold";
static const char *SC_STAGE_WEIGHT = "weight";
struct Stage
{
float threshold;
float weight;
Stage(){}
Stage(const cv::FileNode& fn) : threshold((float)fn[SC_STAGE_THRESHOLD]), weight((float)fn[SC_STAGE_WEIGHT])
{/*printf(" stage: %f %f\n",threshold, weight);*/}
};
// according to R. Benenson, M. Mathias, R. Timofte and L. Van Gool paper
struct CascadeIntrinsics
{
static const float lambda = 1.099f, a = 0.89f;
static const float intrinsics[10][4];
static float getFor(int channel, float scaling)
{
CV_Assert(channel < 10);
if ((scaling - 1.f) < FLT_EPSILON)
return 1.f;
int ud = (int)(scaling < 1.f);
return intrinsics[channel][(ud << 1)] * pow(scaling, intrinsics[channel][(ud << 1) + 1]);
}
};
const float CascadeIntrinsics::intrinsics[10][4] =
{ //da, db, ua, ub
// hog-like orientation bins
{a, lambda / log(2), 1, 2},
{a, lambda / log(2), 1, 2},
{a, lambda / log(2), 1, 2},
{a, lambda / log(2), 1, 2},
{a, lambda / log(2), 1, 2},
{a, lambda / log(2), 1, 2},
// gradient magnitude
{a, lambda / log(2), 1, 2},
// luv color channels
{1, 2, 1, 2},
{1, 2, 1, 2},
{1, 2, 1, 2}
};
static const char *SC_F_THRESHOLD = "threshold";
static const char *SC_F_DIRECTION = "direction";
static const char *SC_F_CHANNEL = "channel";
static const char *SC_F_RECT = "rect";
struct Feature
{
float threshold;
int direction;
int channel;
cv::Rect rect;
Feature() {}
Feature(const cv::FileNode& fn)
: threshold((float)fn[SC_F_THRESHOLD]), direction((int)fn[SC_F_DIRECTION]),
channel((int)fn[SC_F_CHANNEL])
{
cv::FileNode rn = fn[SC_F_RECT];
cv::FileNodeIterator r_it = rn.begin();
rect = cv::Rect(*(r_it++), *(r_it++), *(r_it++), *(r_it++));
// printf(" feature: %f %d %d [%d %d %d %d]\n",threshold, direction, channel, rect.x, rect.y, rect.width, rect.height);
}
Feature rescale(float relScale)
{
Feature res(*this);
res.rect = cv::Rect (cvRound(rect.x * relScale), cvRound(rect.y * relScale),
cvRound(rect.width * relScale), cvRound(rect.height * relScale));
res.threshold = threshold * CascadeIntrinsics::getFor(channel, relScale);
return res;
}
};
struct Level
{
int index;
float factor;
float logFactor;
int width;
int height;
float octave;
cv::Size objSize;
Level(int i,float f, float lf, int w, int h): index(i), factor(f), logFactor(lf), width(w), height(h), octave(0.f) {}
void assign(float o, int detW, int detH)
{
octave = o;
objSize = cv::Size(cv::saturate_cast<int>(detW * o), cv::saturate_cast<int>(detH * o));
}
float relScale() {return (factor / octave); }
};
}
struct cv::SoftCascade::Filds
{
float minScale;
float maxScale;
int origObjWidth;
int origObjHeight;
int noctaves;
std::vector<Octave> octaves;
std::vector<Stage> stages;
std::vector<Feature> features;
std::vector<Level> levels;
// compute levels of full pyramid
void calcLevels(int frameW, int frameH, int scales)
{
CV_Assert(scales > 1);
levels.clear();
float logFactor = (log(maxScale) - log(minScale)) / (scales -1);
float scale = minScale;
for (int sc = 0; sc < scales; ++sc)
{
Level level(sc, scale, log(scale) + logFactor,
std::max(0.0f, frameW - (origObjWidth * scale)), std::max(0.0f, frameH - (origObjHeight * scale)));
if (!level.width || !level.height)
break;
else
levels.push_back(level);
if (fabs(scale - maxScale) < FLT_EPSILON) break;
scale = std::min(maxScale, expf(log(scale) + logFactor));
}
for (std::vector<Level>::iterator level = levels.begin(); level < levels.end(); ++level)
{
float minAbsLog = FLT_MAX;
for (std::vector<Octave>::iterator oct = octaves.begin(); oct < octaves.end(); ++oct)
{
const Octave& octave =*oct;
float logOctave = log(octave.scale);
float logAbsScale = fabs((*level).logFactor - logOctave);
if(logAbsScale < minAbsLog)
(*level).assign(octave.scale, ORIG_OBJECT_WIDTH, ORIG_OBJECT_HEIGHT);
}
}
}
bool fill(const FileNode &root, const float mins, const float maxs)
{
minScale = mins;
maxScale = maxs;
// cascade properties
const char *SC_STAGE_TYPE = "stageType";
const char *SC_BOOST = "BOOST";
const char *SC_FEATURE_TYPE = "featureType";
const char *SC_ICF = "ICF";
const char *SC_TREE_TYPE = "stageTreeType";
const char *SC_STAGE_TH2 = "TH2";
const char *SC_NUM_OCTAVES = "octavesNum";
const char *SC_ORIG_W = "origObjWidth";
const char *SC_ORIG_H = "origObjHeight";
const char* SC_OCTAVES = "octaves";
const char *SC_STAGES = "stages";
const char *SC_FEATURES = "features";
// only boost supported
std::string stageTypeStr = (string)root[SC_STAGE_TYPE];
CV_Assert(stageTypeStr == SC_BOOST);
// only HOG-like integral channel features cupported
string featureTypeStr = (string)root[SC_FEATURE_TYPE];
CV_Assert(featureTypeStr == SC_ICF);
// only trees of height 2
string stageTreeTypeStr = (string)root[SC_TREE_TYPE];
CV_Assert(stageTreeTypeStr == SC_STAGE_TH2);
// not empty
noctaves = (int)root[SC_NUM_OCTAVES];
CV_Assert(noctaves > 0);
origObjWidth = (int)root[SC_ORIG_W];
CV_Assert(origObjWidth == SoftCascade::ORIG_OBJECT_WIDTH);
origObjHeight = (int)root[SC_ORIG_H];
CV_Assert(origObjHeight == SoftCascade::ORIG_OBJECT_HEIGHT);
// for each octave (~ one cascade in classic OpenCV xml)
FileNode fn = root[SC_OCTAVES];
if (fn.empty()) return false;
octaves.reserve(noctaves);
FileNodeIterator it = fn.begin(), it_end = fn.end();
for (; it != it_end; ++it)
{
FileNode fns = *it;
Octave octave = Octave(fns);
CV_Assert(octave.stages > 0);
octaves.push_back(octave);
stages.reserve(stages.size() + octave.stages);
fns = fns[SC_STAGES];
if (fn.empty()) return false;
// for each stage (~ decision tree with H = 2)
FileNodeIterator st = fns.begin(), st_end = fns.end();
for (; st != st_end; ++st )
{
fns = *st;
stages.push_back(Stage(fns));
fns = fns[SC_FEATURES];
// for each feature for tree. features stored in order {root, left, right}
FileNodeIterator ftr = fns.begin(), ft_end = fns.end();
for (; ftr != ft_end; ++ftr)
{
features.push_back(Feature(*ftr));
}
}
}
return true;
}
};
cv::SoftCascade::SoftCascade() : filds(0) {}
cv::SoftCascade::SoftCascade( const string& filename, const float minScale, const float maxScale)
{
filds = new Filds;
load(filename, minScale, maxScale);
}
cv::SoftCascade::~SoftCascade()
{
delete filds;
}
bool cv::SoftCascade::load( const string& filename, const float minScale, const float maxScale)
{
delete filds;
filds = 0;
cv::FileStorage fs(filename, FileStorage::READ);
if (!fs.isOpened()) return false;
filds = new Filds;
Filds& flds = *filds;
if (!flds.fill(fs.getFirstTopLevelNode(), minScale, maxScale)) return false;
flds.calcLevels(FRAME_WIDTH, FRAME_HEIGHT, TOTAL_SCALES);
return true;
}
namespace {
void calcHistBins(const cv::Mat& grey, std::vector<cv::Mat>& histInts, const int bins)
{
CV_Assert( grey.type() == CV_8U);
const int rows = grey.rows + 1;
const int cols = grey.cols + 1;
cv::Size intSumSize(cols, rows);
histInts.clear();
std::vector<cv::Mat> hist;
for (int bin = 0; bin < bins; ++bin)
{
hist.push_back(cv::Mat(rows, cols, CV_32FC1));
}
cv::Mat df_dx, df_dy, mag, angle;
cv::Sobel(grey, df_dx, CV_32F, 1, 0);
cv::Sobel(grey, df_dy, CV_32F, 0, 1);
cv::cartToPolar(df_dx, df_dy, mag, angle, true);
const float magnitudeScaling = 1.0 / sqrt(2);
mag *= magnitudeScaling;
angle /= 60;
for (int h = 0; h < mag.rows; ++h)
{
float* magnitude = mag.ptr<float>(h);
float* ang = angle.ptr<float>(h);
for (int w = 0; w < mag.cols; ++w)
{
hist[(int)ang[w]].ptr<float>(h)[w] = magnitude[w];
}
}
for (int bin = 0; bin < bins; ++bin)
{
cv::Mat sum;
cv::integral(hist[bin], sum);
histInts.push_back(sum);
}
cv::Mat magIntegral;
cv::integral(mag, magIntegral, mag.depth());
}
struct Integrals
{
/* data */
};
}
void cv::SoftCascade::detectInRoi()
{}
void cv::SoftCascade::detectMultiScale(const Mat& image, const std::vector<cv::Rect>& rois, std::vector<cv::Rect>& objects,
const int step, const int rejectfactor)
{
typedef std::vector<cv::Rect>::const_iterator RIter_t;
// only color images are supperted
CV_Assert(image.type() == CV_8UC3);
// only this window size allowed
CV_Assert(image.cols == 640 && image.rows == 480);
objects.clear();
// create integrals
cv::Mat luv;
cv::cvtColor(image, luv, CV_BGR2Luv);
cv::Mat luvIntegral;
cv::integral(luv, luvIntegral);
cv::Mat grey;
cv::cvtColor(image, grey, CV_RGB2GRAY);
std::vector<cv::Mat> hist;
const int bins = 6;
calcHistBins(grey, hist, bins);
for (RIter_t it = rois.begin(); it != rois.end(); ++it)
{
const cv::Rect& roi = *it;
// detectInRoi(roi, objects, step);
}
}
void cv::SoftCascade::detectForOctave(const int octave)
{}