def plot_matches(im1,
im2,
locs1,
locs2,
match_indices,
show_below=True,
figsize=(6, 4)):
'''対応点を線で結んで画像を表示する'''
im3 = concatenate_img_horiz(im1, im2)
if show_below:
im3 = np.vstack((im3, im3))
mip.show_img(im3, figsize=figsize)
width1 = im1.shape[1]
for i, m in enumerate(match_indices):
if m >= 0:
plt.plot([locs1[i][1], locs2[m][1] + width1],
[locs1[i][0], locs2[m][0]], 'c')
plt.axis('off')
def load_merton_data(data_path='./MertonCollege', show_detail=False):
"""
https://www.robots.ox.ac.uk/~vgg/data/mview/の"Merton College I"データを読み込む
Args:
data_path (str):データを展開したディレクトリのpath。image,2D,3Dがある前提
Returns:
corr:ある717点ある3次元上の点が、2D/001.cornes内で示されたのどの2D上の点に対応しているか。
indexを表す。
"""
# 画像を読み込む
im1 = ip.imread(os.path.join(data_path, 'image/001.jpg'))
im2 = ip.imread(os.path.join(data_path, 'image/002.jpg'))
if (show_detail):
ip.show_img(im1)
ip.show_img(im2)
# 画像上の2Dの点をリストに読み込む
points2D = []
for i in range(3):
file_path = os.path.join(data_path, '2D/{:0=3}.corners'.format(i + 1))
points2D.append(np.loadtxt(file_path).T)
# 3Dの点を読み込む
points3D = np.loadtxt(os.path.join(data_path, '3D/p3d')).T
# 対応関係を読み込む
corr = np.genfromtxt(os.path.join(data_path, '2D/nview-corners'),
dtype='int',
missing_values='*')
# カメラパラメーラをCameraオブジェクトに読み込む
P = []
for i in range(3):
istrct_p = np.loadtxt('MertonCollege/2D/{:0=3}.P'.format(i + 1))
P.append(camera.Camera(istrct_p))
return im1, im2, points2D, points3D, corr, P
p.append([c[0] - wid, c[1] + wid, c[2] + wid])
p.append([c[0] - wid, c[1] + wid, c[2] - wid])
p.append([c[0] + wid, c[1] + wid, c[2] - wid])
p.append([c[0] + wid, c[1] + wid, c[2] + wid])
p.append([c[0] + wid, c[1] - wid, c[2] + wid])
p.append([c[0] + wid, c[1] - wid, c[2] - wid])
return np.array(p).T
cap_file.get(cv2.CAP_PROP_FRAME_COUNT)
# +
img_query = ip.imread('CalibrationImage/query.JPG')
#画像を縮小
img_query = img_query[900:2400, 2200:3700, :]
ip.show_img(img_query, show_axis=True)
K = camera.calculate_camera_matrix_w_sz(sz=(6000, 4000), lens='PZ')
K[0, 2] = img_query.shape[1] / 2
K[1, 2] = img_query.shape[0] / 2
cap_file = cv2.VideoCapture('./CalibrationImage/material.MP4')
print(type(cap_file))
print("suceed open video:", cap_file.isOpened())
width = int(cap_file.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap_file.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = cap_file.get(cv2.CAP_PROP_FPS)
total_frame = int(cap_file.get(cv2.CAP_PROP_FRAME_COUNT))
print(cap_file.set(cv2.CAP_PROP_POS_FRAMES, 0))
# VideoWriter を作成する。
import image_processing as ip import desc_val as dv import homography import sfm im1, im2, points2D, points3D, corr, P = sfm.load_merton_data(show_detail=True) #3Dの点を同次座標系にして射影する X = homography.make_homog(points3D) x = P[0].project(X) points2D[0] # + #画像1の上に点を描写する ip.show_img(im1, figsize=(6, 4)) plt.plot(points2D[0][0], points2D[0][1], '*') plt.show() ip.show_img(im1, figsize=(6, 4)) plt.plot(x[0], x[1], '*') plt.show() # - from mpl_toolkits.mplot3d import Axes3D from mpl_toolkits.mplot3d import axes3d import matplotlib print(matplotlib.__version__) # +
import sfm
# +
# %matplotlib inline
show_detail = True
ratio = 0.5
#画像を読み込み特徴点を計算する
im1 = ip.imread('./CalibrationImage/sfm_005.JPG')
#画像を縮小
im1 = cv2.resize(im1, None, fx=ratio, fy=ratio, interpolation=cv2.INTER_AREA)
im2 = ip.imread('./CalibrationImage/sfm_006.JPG')
#画像を縮小
im2 = cv2.resize(im2, None, fx=ratio, fy=ratio, interpolation=cv2.INTER_AREA)
ip.show_img(im1, show_axis=True)
ip.show_img(im2, show_axis=True)
plt.show()
# +
# K = np.array([[2394,0,932],[0,2398,628],[0,0,1]])
K = camera.calculate_camera_matrix_w_sz(im1.shape[1::-1], lens='PZ')
# Initiate AKAZE detector
akaze = cv2.AKAZE_create()
# key pointとdescriptorを計算
kp1, des1 = akaze.detectAndCompute(im1, None)
kp2, des2 = akaze.detectAndCompute(im2, None)
# matcherとしてflannを使用。
import matplotlib.pyplot as plt
#自作モジュール
import image_processing as ip
# import desc_val as dv
# import camera
# import feature_detection as fd
# import homography
# +
ratio=0.3
img_query=ip.imread('CalibrationImage/FeatureDetection00001.JPG')
#画像を縮小
img_query=cv2.resize(img_query,None,fx=ratio,fy=ratio,interpolation=cv2.INTER_AREA)
ip.show_img(img_query,show_axis=True)
img_train=ip.imread('CalibrationImage/FeatureDetection00003.JPG')
#画像を縮小
img_train=cv2.resize(img_train,None,fx=ratio,fy=ratio,interpolation=cv2.INTER_AREA)
ip.show_img(img_train,show_axis=True)
# +
# Initiate AKAZE detector
akaze = cv2.AKAZE_create()
# key pointとdescriptorを計算
kp1, des1 = akaze.detectAndCompute(img_query, None)
kp2, des2 = akaze.detectAndCompute(img_train, None)
#matcherとしてflannを使用。
def expand(image, ratio):
h, w = image.shape[:2]
src = np.array([[0.0, 0.0],[0.0, 1.0],[1.0, 0.0]], np.float32)
dest = src * ratio
affine = cv2.getAffineTransform(src, dest)
return cv2.warpAffine(image, affine, (2*w, 2*h), cv2.INTER_LANCZOS4) # 補間法も指定できる
# +
ratio=0.3
img_for_camera_calibration=ip.imread('CalibrationImage/FeatureDetection00001.JPG')
im1=expand(img_for_camera_calibration,ratio)
im1=im1[:int(4000*ratio),:int(6000*ratio)]
ip.show_img(im1,show_axis=True)
img_math_test=ip.imread('CalibrationImage/FeatureDetection00003.JPG')
im2=expand(img_math_test,ratio)
im2=im2[:int(4000*ratio),:int(6000*ratio)]
ip.show_img(im2,show_axis=True)
# + {}
akaze = cv2.AKAZE_create()
kp1, des1 = akaze.detectAndCompute(im1,None)
kp2, des2 = akaze.detectAndCompute(im2,None)
# Brute-Force Matcher生成
# create BFMatcher object
bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
matches = bf.match(des1, des2)
def compute_rasac_homology(img_query_orig,
img_train_orig,
MIN_MATCH_COUNT=10,
show_detail=False,
save_result=False):
"""
query画像とtrain画像についてakazeでマッチングし、
ransacによって外れ値を除去してHomology行列を算出する。
Args:
MIN_MATCH_COUNT (int):mathesの数の最小値。これ以下だとHomologyを計算しない
"""
img_query = img_query_orig.copy()
img_train = img_train_orig.copy()
# Initiate AKAZE detector
akaze = cv2.AKAZE_create()
# key pointとdescriptorを計算
kp1, des1 = akaze.detectAndCompute(img_query, None)
kp2, des2 = akaze.detectAndCompute(img_train, None)
# matcherとしてflannを使用。
# FLANN parameters
FLANN_INDEX_LSH = 6
index_params = dict(algorithm=FLANN_INDEX_LSH,
table_number=6,
key_size=12,
multi_probe_level=1)
search_params = dict(checks=50)
# ANNで近傍２位までを出力
flann = cv2.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(des1, des2, k=2)
# store all the good matches as per Lowe's ratio test.
# 2番めに近かったkey pointと差があるものをいいkey pointとする。
good_matches = []
for i in range(len(matches)):
if (len(matches[i]) < 2):
continue
m, n = matches[i]
if m.distance < 0.7 * n.distance:
good_matches.append(m)
# descriptorの距離が近かったもの順に並び替え
good_matches = sorted(good_matches, key=lambda x: x.distance)
if (show_detail):
# 結果を描写
img_result = cv2.drawMatches(img_query,
kp1,
img_train,
kp2,
good_matches[:10],
None,
flags=2)
ip.show_img(img_result, figsize=(20, 30))
print('queryのkp:{}個、trainのkp:{}個、good matchesは:{}個'.format(
len(kp1), len(kp2), len(good_matches)))
# ransacによって外れ値を除去してHomology行列を算出する。
# opencvの座標は3次元のarrayで表さなければならないのに注意
if len(good_matches) > MIN_MATCH_COUNT:
# matching点の座標を取り出す
src_pts = np.float32([kp1[m.queryIdx].pt
for m in good_matches]).reshape(-1, 1, 2)
dst_pts = np.float32([kp2[m.trainIdx].pt
for m in good_matches]).reshape(-1, 1, 2)
# ransacによって外れ値を除去
homology_matrix, mask = cv2.findHomography(src_pts, dst_pts,
cv2.RANSAC, 5.0)
else:
print("Not enough matches are found - %d/%d" %
(len(good_matches), MIN_MATCH_COUNT))
matchesMask = None
return None, None
if (show_detail or save_result):
# 結果を描写
matchesMask = mask.ravel().tolist()
# query画像の高さ、幅を取得し、query画像を囲う長方形の座標を取得し、
# それを算出された変換行列homology_matrixで変換する
# 変換した長方形をtrain画像に描写
h, w = img_query.shape[:2]
pts = np.float32([[0, 0], [0, h - 1], [w - 1, h - 1],
[w - 1, 0]]).reshape(-1, 1, 2)
dst = cv2.perspectiveTransform(pts, homology_matrix)
cv2.polylines(img_train, [np.int32(dst)], True, (255, 100, 0), 3,
cv2.LINE_AA)
num_draw = 50
draw_params = dict(
# matchColor = (0,255,0), # draw matches in green color
singlePointColor=None,
matchesMask=matchesMask[:num_draw], # draw only inliers
flags=2)
img_result_2 = cv2.drawMatches(img_query, kp1, img_train, kp2,
good_matches[:num_draw], None,
**draw_params)
if (show_detail):
ip.show_img(img_result_2, figsize=(20, 30))
num_inlier = (mask == 1).sum()
print('inlier:%d個' % num_inlier)
if (save_result):
ip.imwrite('ransac_match.jpg', img_result_2)
return homology_matrix, mask
import camera
import feature_detection as fd
import homography
# +
ratio = 0.3
img_for_camera_calibration = ip.imread(
'CalibrationImage/ImageforCameraCalibration.jpg')
#画像を縮小
im1 = cv2.resize(img_for_camera_calibration,
None,
fx=ratio,
fy=ratio,
interpolation=cv2.INTER_AREA)
ip.show_img(im1, show_axis=True)
# -
#本の左上にプロットしてみる
img_for_edit = img_for_camera_calibration.copy()
cv2.circle(img_for_edit, (2789, 1140), 50, (255, 255, 255), thickness=-1)
ip.show_img(img_for_edit, show_axis=True)
# 写真上での四隅の座標
# (左上の隅からのx座標、y座標)
# 左上(2789,1140)
# 右上(3910,1135)
# 左下(2782,2685)
# 右下(3902,2693)
#左上-左下→縦の長さ