如何使用 opencv.omnidir 模块去扭曲鱼眼图像

Posted 2023-02-16

【中文标题】如何使用 opencv.omnidir 模块去扭曲鱼眼图像

【英文标题】：How to use opencv.omnidir module for dewarping fisheye images

【发布时间】：2017-06-13 19:42:44

【问题描述】：

我正在尝试使用 omnidirectional module 在 Python 中对鱼眼图像进行去扭曲。我正在尝试在 Python 中调整这个C++ tutorial，但遇到了问题。这是我的代码：

import numpy as np
import cv2
import glob

nx = 9
ny = 6

# termination criteria
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 30, 0.001)

chessboard_model = np.zeros((1, nx*ny, 3), np.float32)
chessboard_model[0, :, :2] = np.mgrid[0:nx, 0:ny].T.reshape(-1, 2)

# prepare object points, like (0,0,0), (1,0,0), (2,0,0) ....,(6,5,0)
objp = np.zeros((nx*ny, 3), np.float32)
objp[:,:2] = np.mgrid[0:nx, 0:ny].T.reshape(-1,2)

# Arrays to store object points and image points from all the images.
objpoints = [] # 3d point in real world space
imgpoints = [] # 2d points in image plane.

filenames = glob.glob('/home/work/Downloads/fisheye_calibration/*.jpg')
images = []
i = 0
for fname in filenames:
    i += 1
    print("Processing image: ".format(i))
    img = cv2.imread(fname)
    gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
    images.append(img)

    # Find the chess board corners
    ret, corners = cv2.findChessboardCorners(gray, (nx,ny),None)
    # If found, add object points, image points (after refining them)
    if ret == True:
        objpoints.append(objp)
        corners2 = cv2.cornerSubPix(gray,corners,(11,11),(-1,-1),criteria)
        #imgpoints.append(corners2)
        imgpoints.append(corners2.reshape(1, -1, 2))

        # Draw and display the corners
        img = cv2.drawChessboardCorners(img, (nx,ny), corners2,ret)
        cv2.imshow('img',img)
        cv2.waitKey(500)

num_points = len(imgpoints)
K = np.zeros((3, 3))
xi = np.array([])
idx = np.array([])
D = np.zeros((4, 1))
h, w = images[0].shape[:2]
calibration_flags = cv2.omnidir.CALIB_USE_GUESS | cv2.omnidir.CALIB_FIX_SKEW
obj_points_arr = np.array([chessboard_model]*num_points)
img_points_arr = np.array(imgpoints).squeeze(axis=1)
rms = cv2.omnidir.calibrate(obj_points_arr, img_points_arr, (w,h), K, xi, D, calibration_flags, criteria)

我收到以下错误：

cv2.error:/home/work/opencv_contrib/modules/ccalib/src/omnidir.cpp:1065: error: (-215) !patternPoints.empty() && !imagePoints.empty() && patternPoints.total() == imagePoints.total() in function calibrate

这里的obj_points_arr 和img_points_arr 分别是num_points x 1 x (nx * ny) x 3 和num_points x 1 x (nx * ny) x 2。如果我使用

将 obj_points_arr 和 img_points_arr 重塑为 num_points x (nx * ny) x 3 和 num_points x (nx * ny) x 2

obj_points_arr = np.array([chessboard_model]*num_points).squeeze(axis=1)
img_points_arr = np.array(imgpoints).squeeze(axis=1)

然后我得到以下错误

cv2.error: /home/work/opencv/modules/core/src/matrix.cpp:469: error: (-215) 0 <= _rowRange.start && _rowRange.start <= _rowRange.end && _rowRange.end <= m.rows in function Mat

这些数组的形状应该是什么？有人可以分享他们用于去扭曲鱼眼图像的 python 代码吗？

【参考方案1】：

首先你没有得到图像角落，最好写到else并打印出来。

else: print "not found: "+fname

您可以使用 pyfisheye 的模块来完成您的任务。 https://bitbucket.org/amitibo/pyfisheye

【讨论】：

我相信这只有在您的 FOV 小于 180 时才有效