openCV doc

高速輪郭抽出

Gallery

src 元画像

Sobal

Canny

dst 高速輪郭抽出

what process?

輪郭抽出する有名なアルゴリズムとしてSobalやCannyなどがあります。ここで使われているものは、ひとが手動で抽出した輪郭・物体の境界線に近い線が抽出されます。 ※ 輪郭抽出の精度評価に使われるデータセットとして、Berkeley Segmentation Dataset (BSDS500) Benchmark が使われています。 https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html ** model.yml は下記のリンクからダウンロードして実行時に読み込んでください。 https://github.com/opencv/opencv_extra/blob/master/testdata/cv/ximgproc/model.yml.gz

cook code C++

#include <iostream>
#include <opencv2/core.hpp>
#include <opencv2/core/utility.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/videoio.hpp>
#include <opencv2/ximgproc.hpp>

int main(int argc, const char* argv[])
{
    // ./out -src=[image path]
    cv::String keys = "{src||}";
    cv::CommandLineParser parser(argc, argv, keys);
    cv::String src_path = parser.get<cv::String>("src");

cv::VideoCapture capture(src_path);
    if (!capture.isOpened()){
        //error in opening the video input
        std::cerr << "Unable to open file!" << std::endl;
        return 0;
    }

/* 初期設定 */
    int cap_width = int(capture.get(cv::CAP_PROP_FRAME_WIDTH));
    int cap_height = int(capture.get(cv::CAP_PROP_FRAME_HEIGHT));
    int fps = int(capture.get(cv::CAP_PROP_FPS));
    int fourcc = cv::VideoWriter::fourcc('m','p','4','v');
    cv::Size video_size = cv::Size(cap_width,cap_height);

cv::VideoWriter writer("output.mp4", fourcc, fps, video_size);

/* read model */
    std::string modelFileName = "model.yml";
    cv::Ptr<cv::ximgproc::StructuredEdgeDetection> obj 
        = cv::ximgproc::createStructuredEdgeDetection(modelFileName);

while(true){
        cv::Mat frame, dst, edges;

capture >> frame;
        if (frame.empty())
            break;

frame.convertTo(dst, CV_32F, 1/255.0);
        obj->detectEdges(dst, edges);

edges.convertTo(edges, CV_8U, 255);
        cv::cvtColor(edges, edges, cv::COLOR_GRAY2BGR);
        writer << edges;
    }
    capture.release();
    writer.release();
    return 0;
}

cook code Python

import numpy as np
import cv2
import argparse

def main():

parser = argparse.ArgumentParser()
    parser.add_argument('--src')
    args = parser.parse_args()

cap = cv2.VideoCapture(args.src)

cap_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    cap_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
    fps = cap.get(cv2.CAP_PROP_FPS)

fmt = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')
    writer = cv2.VideoWriter('output.mp4', fmt, fps, (cap_width,cap_height))

#read model
    model='model.yml'
    retval=cv2.ximgproc.createStructuredEdgeDetection(model)

while cap.isOpened():
        _ret, img = cap.read()
        
        if img is None:
            break

imgrgb=cv2.cvtColor(img,cv2.COLOR_BGR2RGB)/255
        imgrgb=imgrgb.astype(np.float32)
        edges=retval.detectEdges(imgrgb)
        edges=(edges*255).astype(np.uint8)
        dst = cv2.cvtColor(edges, cv2.COLOR_GRAY2BGR)

writer.write(dst)

cap.release()
    writer.release()

if __name__ == '__main__':
    main()

Reference

https://docs.opencv.org/4.3.0/d0/da5/tutorial_ximgproc_prediction.html
https://stackoverflow.com/questions/33317152/model-file-for-opencvs-structured-edge-detector
Thesis : Piotr Dollár and C Lawrence Zitnick. Structured forests for fast edge detection. In Computer Vision (ICCV), 2013 IEEE International Conference on, pages 1841–1848. IEEE, 2013.
アキラ100%の城本クリニック∞