Python cv2.cornerHarris() Examples

def display_harris_corners(input_img):
    """
    computes corners in colored image and plot it.
    """
    # first convert to grayscale with float32 values
    gray = cv2.cvtColor(input_img,cv2.COLOR_BGR2GRAY)
    gray = np.float32(gray)
    
    # using opencv harris corner implementation
    corners = cv2.cornerHarris(gray,2,7,0.04)
    
#     # result is dilated for marking the corners, not important
#     dst = cv2.dilate(dst,None)
    
    # additional thresholding and marking corners for plotting
    input_img[corners>0.01*corners.max()]=[255,0,0]
    
    return input_img
    # # plot image
    # plt.figure(figsize=(12, 8))
    # plt.imshow(cv2.cvtColor(input_img, cv2.COLOR_BGR2RGB))
    # plt.axis('off') 

def harris_loader(image, name, **config):
    num_features = config.get('num_features', 0)
    do_nms = config.get('do_nms', False)
    nms_thresh = config.get('nms_thresh', 4)

    detect_map = cv2.cornerHarris(image.astype(np.uint8), 4, 3, 0.04)
    kpts = np.where(detect_map > 1e-6)
    scores = detect_map[kpts]
    kpts = np.stack([kpts[1], kpts[0]], axis=-1)
    if do_nms:
        keep = nms_fast(kpts, scores, image.shape[:2], nms_thresh)
        kpts, scores = kpts[keep], scores[keep]
    if num_features:
        keep_indices = np.argsort(scores)[::-1][:num_features]
        kpts, scores = [i[keep_indices] for i in [kpts, scores]]
    return {'keypoints': kpts, 'scores': scores} 

def detect_and_compute_harris_np(img,num):
    smooth_img = cv2.GaussianBlur(img, (5, 5), 1.5)
    if len(smooth_img.shape)==3:
        smooth_img=cv2.cvtColor(smooth_img, cv2.COLOR_RGB2GRAY)
    harris_img = cv2.cornerHarris(smooth_img.astype(np.float32), 2, 3, 0.04)
    element=np.sort(harris_img.flatten())[-num]
    mask=harris_img>=element
    hs,ws=np.nonzero(mask)
    kps=np.concatenate([ws[:,None],hs[:,None]],1)
    des=np.zeros([kps.shape[0],128],np.float32)
    return kps, des 

def detect(self, img):
        # convert our input image to a floating point data type and then
        # compute the Harris corner matrix
        gray = np.float32(img)
        H = cv2.cornerHarris(gray, self.blockSize, self.apertureSize, self.k)

        # for every (x, y)-coordinate where the Harris value is above the
        # threshold, create a keypoint (the Harris detector returns
        # keypoint size a 3-pixel radius)
        kps = np.argwhere(H > self.T * H.max())
        kps = [cv2.KeyPoint(pt[1], pt[0], 3) for pt in kps]

        # return the Harris keypoints
        return kps 

def get_harris_corners (self, image):
		"""
			Function: get_harris_corners
			----------------------------
			given an image, returns a list of cv2.KeyPoints representing
			the harris corners
		"""
		return cv2.cornerHarris(image,2,3,0.04) 

def compute_harris_corners(input):
    gray = cv2.cvtColor(input,cv2.COLOR_BGR2GRAY)
    gray = np.float32(gray)
    dst = cv2.cornerHarris(gray,2,5,0.04)
    #result is dilated for marking the corners, not important
    dst = cv2.dilate(dst,None)
    # Threshold for an optimal value, it may vary depending on the image.
    input[dst>0.01*dst.max()]=[0,255,0]
    plt.figure(figsize=(12, 8))
    plt.imshow(cv2.cvtColor(input, cv2.COLOR_BGR2RGB))
    plt.axis('off')
    plt.show() 

def cornerDetection(inputImg_edge):
    imgCorners=cv2.imread(inputImg_edge)    
    imgGray=cv2.cvtColor(imgCorners,cv2.COLOR_BGR2GRAY) #将图像转换为灰度，为每一像素位置1个值，可理解为图像的强度(颜色，易受光照影响，难以提供关键信息，故将图像进行灰度化，同时也可以加快特征提取的速度。)
    imgGray=np.float32(imgGray) #强制转换为浮点值，用于棱角检测
    imgHarris=cv2.cornerHarris(imgGray,7,5,0.04) #哈里斯角检测器 Harris corner detector
    print(imgHarris.max(),imgHarris.shape)
    imgHarris=cv2.dilate(imgHarris,np.ones((1,1))) #放大棱角标记
    print(imgCorners[300:500,])
    imgCorners[imgHarris>0.01*imgHarris.max()]=[40,75,236] #定义阈值，显示重要的棱角
    cv2.imshow('harris corners',imgCorners)
    cv2.imwrite(os.path.join(rootDirectory,'harris corners.jpg'),imgCorners)
    cv2.waitKey()

#SIFT(scale invariant feature transform 尺度不变特征变换)特征点检测