#include "diploma.hpp"

#include "track.hpp"

#include <opencv2/flann.hpp>

using namespace cv::cuda;

void cpu_orb(char *fileName) {

cv::VideoCapture cap(fileName);

if(!cap.isOpened()) {

std::cout << "[ERROR] bad video file" << std::endl;

return;

}

int nFrames = 50;

int nWinSize = 11;

auto pTracker = std::make_unique<Track::FeatureTracker>(nFrames, nWinSize);

cv::Ptr<cv::ORB> ORBDetector = cv::ORB::create(30);

cv::Mat frame, frameClone, grayed, descriptors, oldDescriptors;

std::vector<cv::KeyPoint> keypoints, oldKeypoints;

cv::FlannBasedMatcher matcher(new cv::flann::LshIndexParams(20, 10, 2));

std::cout << "[INFO] ORB on CPU" << std::endl;

while(cap.read(frame)) {

keypoints.clear();

frameClone = frame.clone();

cv::cvtColor(frame, grayed, cv::COLOR_RGB2GRAY);

uint64_t t0 = cv::getTickCount();

ORBDetector->detectAndCompute(frameClone, cv::noArray(), keypoints, descriptors);

uint64_t t1 = cv::getTickCount();

double timegap = (t1*1.0 - t0) / cv::getTickFrequency();

if(descriptors.empty()) {

std::cout << "[WARNING] no descriptors" << std::endl;

continue;

}

if(oldKeypoints.empty()) {

oldKeypoints = keypoints;

keypoints.clear();

oldDescriptors = descriptors.clone();

continue;

}

std::vector<std::vector<cv::DMatch>> vKnnMatches;

std::vector<cv::DMatch> dmMatches;

matcher.knnMatch(oldDescriptors, descriptors, vKnnMatches, 2 );

//-- Filter matches using the Lowe's ratio test

const float ratio_thresh = 0.7f;

for (size_t i = 0; i < vKnnMatches.size(); i++) {

if (vKnnMatches[i][0].distance < ratio_thresh * vKnnMatches[i][1].distance) {

dmMatches.push_back(vKnnMatches[i][0]);

}

}

std::cout << "elapsed time: " << timegap << "; featured found: " << keypoints.size() << "; features tracked: " << dmMatches.size() << std::endl;

oldKeypoints = keypoints;

oldDescriptors = descriptors.clone();

keypoints.clear();

}

}