Как нанести Ransac на сегменты и построить их - PullRequest
Новая
       43

Как нанести Ransac на сегменты и построить их

0 голосов
/ 25 марта 2019

Я пытаюсь реализовать https://rd.springer.com/chapter/10.1007/978-3-319-68505-2_6, но сталкиваюсь с проблемами, чтобы отфильтровать и построить результаты. Я использую https://docs.opencv.org/3.4.2/d1/de0/tutorial_py_feature_homography.html для фильтрации совпадений. Когда я запускаю cv2.findHomography (sr [i ], de [i], cv2.RANSAC, 5.0), каждый раз происходит сбой python. 

    from skimage.segmentation import slic
    from skimage.segmentation import mark_boundaries
    from skimage.util import img_as_float
    from skimage import io
    import matplotlib.pyplot as plt
    import cv2
    import numpy as np
    # load the image and convert it to a floating point data type
    image = cv2.imread('C:\\Users\\pragyan.prakash\\Desktop\\p_orb\\001_F.png')
    img1=image, cv2.COLOR_BGR2GRAY


    segments = slic(img_as_float(image), n_segments = 100, sigma = 5)

    # show the output of SLIC
    fig = plt.figure("Superpixels")
    ax = fig.add_subplot(1, 1, 1)
    ax.imshow(mark_boundaries(img_as_float(cv2.cvtColor(image, cv2.COLOR_BGR2RGB)), segments))
    plt.axis("off")
    plt.show()
    feature=[]
    # loop over the unique segment values
    for (i, segVal) in enumerate(np.unique(segments)):

        # construct a mask for the segment
        #print ("[x] inspecting segment %d" % (i))
        mask = np.zeros(image.shape[:2], dtype = "uint8")
        mask[segments == segVal] = 255
        img=cv2.bitwise_and(image, image, mask = mask)
        feature.append(cv2.cvtColor(img, cv2.COLOR_BGR2GRAY))
        cv2.waitKey(0)

    # Initiate ORB detector
    orb = cv2.ORB_create()
    key=[]
    desc=[]
    sel= []

    for i in feature:
        kp,des=orb.detectAndCompute(i,None)
        if des is not None: 
            key.append(kp)
            desc.append(des)
            sel.append(i)

    # create BFMatcher object
    bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)

    match=[]
    pairs = []

    for i in range(len(desc)):
        for j in range(len(desc)):
            #print (i,j)
            if i!=j:
                match.append(bf.match(desc[i],desc[j]))
                pairs.append((i,j))

    sorted_match=[]

    for i in match:
        sorted_match.append(sorted(i,key=lambda x:x.distance))


    list_kp1=[]
    list_kp2=[]
    sr=[]
    de=[]

    for m in sorted_match:
        for e in m:
            img1_idx = e.queryIdx
            img2_idx = e.trainIdx

            # x - columns
            # y - rows
            # Get the coordinates
            i = sorted_match.index(m)
            p,q = pairs[i]


            (x1,y1) = key[p][img1_idx].pt
            (x2,y2) = key[q][img2_idx].pt
            sr.append(np.float32((x1,y1)).reshape(-1,1,2))
            de.append(np.float32((x2,y2)).reshape(-1,1,2))

            # Append to each list
            x1=int(x1)
            x2=int(x2)
            y1=int(y1)
            y2=int(y2)
            list_kp1.append((x1, y1))
            list_kp2.append((x2, y2))

    lst=[]
    ls1=[]
    for i in range(len(sr)):
        lst.append(cv2.findHomography(sr[i], de[i], cv2.RANSAC,5.0))
        ls1.append(mask.ravel().tolist())
...