void TrickEstimater::labelContour(unsigned int contour_id, unsigned int rv_size)
{
LabelingBS labeling;
bool at_is_sort_size_order = false;//óÃàÊÇÃëÂÇ´Ç≥èáÇ…É\Å[ÉgǵǻǢ
int at_delete_size = 0; //è¡ãéÇ∑ÇÈè¨óÃàÊÇÃç≈ëÂÉTÉCÉY(int) - DZÇÍà»â∫ÇÃÉTÉCÉYÇÃóÃàÊÇè¡ãéÇ∑ÇÈ
unsigned char *src = new unsigned char[rv_size];
short *result = new short[rv_size];
for (int i = 0; i < rv_size; i++){
src[i] = (unsigned char)is_sharp_angle[contour_id][i];
}
labeling.Exec(src, result, rv_size, 1, at_is_sort_size_order, at_delete_size);
label_index_max = labeling.GetNumOfResultRegions();
for (int i = 0; i < rv_size; i++){
is_sharp_angle[contour_id][i] = (unsigned short)result[i];
}
for (int i = 0; i < label_index_max; i++){
label_index[contour_id][i] = (labeling.GetResultRegionInfo(i))->GetResult();
}
delete[] src;
delete[] result;
}
int main(int argc, char* argv[])
{
// カメラ映像の幅と高さ
int w = 640;
int h = 480;
Mat im, hsv, mask; // 画像オブジェクトの作成
LabelingBS labeling; // ラベリング関連の変数
RegionInfoBS *ri;
short *mask2 = new short[w * h];//ラベリング出力先
VideoCapture cap(0); // カメラのキャプチャ
if (!cap.isOpened()) return -1; // キャプチャ失敗時のエラー処理
while (1){
cap >> im; // カメラから画像を取得
cvtColor(im, hsv, CV_BGR2HSV); // 画像をRGBからHSVに変換
inRange(hsv, Scalar(150, 70, 70), Scalar(360, 255, 255), mask); // 色検出でマスク画像の作成
// 白領域が無い場合のエラー処理
rectangle(mask, Point(0, 0), Point(1, 1), Scalar(255),-1);
//ラベリング処理
labeling.Exec((uchar *)mask.data, mask2, w, h, true, 30);
//最大の領域を四角で囲む
ri = labeling.GetResultRegionInfo(0);
int x1, y1, x2, y2;
ri->GetMin(x1, y1);
ri->GetMax(x2, y2);
rectangle(im, Point(x1, y1), Point(x2, y2), Scalar(255, 0, 0), 3);
imshow("Camera", im); // カメラ映像の表示
imshow("Mask", mask); // マスク画像の作成
if (waitKey(30) >= 0){ // 任意のキー入力があれば終了
break;
}
}
return 0;
}
void OpenCVAdapter::LabelFrame(){
LabelingBS labeling;
labeling.Exec(label,labeled,xnum,ynum,true,trackbarPosition2);
int n = labeling.GetNumOfResultRegions();
printf("%d region found\n",n);
for(int i=0;i<n;i++){
RegionInfoBS *ri;
ri=labeling.GetResultRegionInfo(i);
float x,y;
ri->GetCenterOfGravity(x,y);
int px=(int)x*winx/width;
int py=(int)y*winy/height;
plist.push_back(make_pair((int)px,(int)py));
}
}
void BlobDetector::detect(
const sensor_msgs::Image::ConstPtr& image_msg)
{
vital_checker_->poke();
boost::mutex::scoped_lock lock(mutex_);
cv::Mat image = cv_bridge::toCvShare(image_msg, image_msg->encoding)->image;
cv::Mat label(image.size(), CV_16SC1);
LabelingBS labeling;
labeling.Exec(image.data, (short*)label.data, image.cols, image.rows,
true, min_area_);
cv::Mat label_int(image.size(), CV_32SC1);
for (int j = 0; j < label.rows; j++) {
for (int i = 0; i < label.cols; i++) {
label_int.at<int>(j, i) = label.at<short>(j, i);
}
}
pub_.publish(
cv_bridge::CvImage(image_msg->header,
sensor_msgs::image_encodings::TYPE_32SC1,
label_int).toImageMsg());
}
void ImgProcQt::detectCard(cv::Mat& bgrImg){
// 画面の中の明るい部分(カードは明るいはず)を抽出するためにHSVへ変換
cv::Mat hsvImg;
cv::vector<cv::Mat> splited(3);
cv::cvtColor(bgrImg, hsvImg, CV_BGR2HSV);
cv::split(hsvImg, splited);
// 2値化
cv::Mat binImg;
cv::threshold(splited[2], binImg, 128, 255, CV_THRESH_BINARY);
//Labeling
cv::Mat rImg(binImg.size(), CV_16SC1);
LabelingBS labeling;
labeling.Exec(binImg.data, (short *)rImg.data, binImg.cols, binImg.rows, true, 10);
// カードのエリアを取得
cv::Mat cardImg;
cv::compare(rImg, 1, cardImg, CV_CMP_EQ);
// 1箇所だけの輪郭を取得
cv::vector<cv::vector<cv::Point2i> > contours;
cv::findContours(cardImg, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_TC89_L1);
// ラベリング時点で1つしか輪郭が取れないことが保障されている。はず。
cv::drawContours(bgrImg, contours, 0, CV_RGB(0,255,0), 1);
void DemoInterface::ImgCallBack(const sensor_msgs::ImageConstPtr &msg) {
if (raw_img_.empty())
return;
cv::Rect roi(0, 0, raw_img_.cols, raw_img_.rows);
cv::Mat marker_img = cv::Mat::zeros(raw_img_.size(), CV_8UC1);
//一度平滑化した後，HSV色空間に変換
cv::Mat hsv_img;
cv::medianBlur(raw_img_, hsv_img, 13);
cv::cvtColor(hsv_img, hsv_img, CV_BGR2HSV);
//黄色の領域を抽出
ExtractColor(hsv_img, marker_img);
//ラベリング
cv::Mat label(raw_img_.size(), CV_16SC1);
LabelingBS labeling;
labeling.Exec(marker_img.data, (short *) label.data, raw_img_.cols, raw_img_.rows, false, 0);
cv::Mat out_img(raw_img_.size(), CV_8UC3, cv::Scalar(0, 0, 0));
int num_regions = labeling.GetNumOfRegions();
//ラベリング数が2以上，10未満の時
if (num_regions > 1 && num_regions < 10) {
cv::Point point[2];
cv::Point max_pt[2] = {};
int max_idx[2] = {0, 1};
int max_sum[2] = {};
int num_pt = 0;
//面積最大の要素を2つ計算する
for (int i = 0; i < num_regions; i++) {
int sum = 0;
cv::Point pt(0, 0);
cv::Mat_<uchar> labelarea;
cv::compare(label, i + 1, labelarea, CV_CMP_EQ);
for (int y = 0; y < labelarea.rows; y++) {
const uchar *labelarea_row = labelarea.ptr<uchar>(y);
for (int x = 0; x < labelarea.cols; x++) {
if (labelarea_row[x] > 250) {
sum++;
pt.x += x;
pt.y += y;
}
}
}
for (int j = 0; j < 2; j++) {
if (sum > max_sum[j]) {
if (pt.x == 0 && pt.y == 0)
break;
else if (sum < SUM_MIN)
break;
for (int n = 1; n > j; n--) {
max_idx[n] = max_idx[n - 1];
max_sum[n] = max_sum[n - 1];
max_pt[n] = max_pt[n - 1];
}
max_idx[j] = i;
max_sum[j] = sum;
max_pt[j] = pt;
num_pt++;
break;
}
}
}
if (num_pt != 2)
return;
for (int i = 0; i < 2; i++) {
int sum = max_sum[i];
int idx = max_idx[i];
point[i] = max_pt[i];
cv::Mat_<uchar> labelarea;
cv::compare(label, idx + 1, labelarea, CV_CMP_EQ);
cv::Mat color(raw_img_.size(), CV_8UC3, cv::Scalar(255, 255, 255));
if (sum > 0) {
point[i].x /= sum;
point[i].y /= sum;
color.copyTo(out_img, labelarea);
cv::circle(out_img, point[i], 20, cv::Scalar(0, 200, 0), 8, 8);
}
}
//指定領域の行列のサイズと中点
int row_diff = abs(point[0].y - point[1].y);
int col_diff = abs(point[0].x - point[1].x);
int row_center = (point[0].y + point[1].y) / 2;
int col_center = (point[0].x + point[1].x) / 2;
int length = 0;
//縦幅と横幅の大きい方を正方形の一辺に取る
if (row_diff > col_diff)
length = row_diff;
else
length = col_diff;
roi.x = col_center - length / 2;
roi.y = row_center - length / 2;
roi.width = length;
roi.height = length;
//境界付近の処理
if (roi.x + roi.width > raw_img_.cols)
roi.width = raw_img_.cols - roi.x;
if (roi.y + roi.height > raw_img_.rows)
roi.height = raw_img_.rows - roi.y;
if (roi.x < 0) {
roi.width += roi.x;
roi.x = 0;
}
if (roi.y < 0) {
roi.height += roi.y;
roi.y = 0;
}
if (roi.x || roi.y || roi.width || roi.height) {
point_rec_1_ = cv::Point(col_center - length / 2, row_center - length / 2);
point_rec_2_ = cv::Point(col_center + length / 2, row_center + length / 2);
//受信画像の指定領域をROSメッセージにセット
pinch_classifier::classify srv;
srv.request.image = *cv_bridge::CvImage(std_msgs::Header(), "bgr8", raw_img_(roi)).toImageMsg();
//Caffeサーバを呼び出す
CHECK(client_.call(srv)) << "Failed to call service";
//返ってきた結果を整形
result_ = ConvertResponse(srv.response.label);
point_ = cv::Point(roi.x, roi.y - 20);
draw_ = true;
delay_ = DELAY;
}
}
else {
if (--delay_ == 0) {
draw_ = false;
delay_ = DELAY;
}
}
cv::rectangle(out_img, point_rec_1_, point_rec_2_, cv::Scalar(0, 200, 0), 3, 4);
cv::imshow("Yellow", out_img);
cv::imshow("Marker", marker_img);
}