class SURFDemo(ImageProcessDemo):
TITLE = "SURF Demo"
DEFAULT_IMAGE = "lena.jpg"
SETTINGS = ["m_perspective", "hessian_threshold", "n_octaves"]
m_perspective = Array(np.float, (3, 3))
m_perspective2 = Array(np.float, (3, 3))
hessian_threshold = Int(2000)
n_octaves = Int(2)
poly = Instance(PolygonWidget)
def control_panel(self):
return VGroup(
Item("m_perspective",
label="变换矩阵",
editor=ArrayEditor(format_str="%g")),
Item("m_perspective2",
label="变换矩阵",
editor=ArrayEditor(format_str="%g")),
Item("hessian_threshold", label="hessianThreshold"),
Item("n_octaves", label="nOctaves"))
def __init__(self, **kwargs):
super(SURFDemo, self).__init__(**kwargs)
self.poly = None
self.init_points = None
self.lines = LineCollection([], linewidths=1, alpha=0.6, color="red")
self.axe.add_collection(self.lines)
self.connect_dirty("poly.changed,hessian_threshold,n_octaves")
def init_poly(self):
if self.poly is None:
return
h, w, _ = self.img_color.shape
self.init_points = np.array([(w, 0), (2 * w, 0), (2 * w, h), (w, h)],
np.float32)
self.poly.set_points(self.init_points)
self.poly.update()
def init_draw(self):
style = {"marker": "o"}
self.poly = PolygonWidget(axe=self.axe,
points=np.zeros((3, 2)),
style=style)
self.init_poly()
@on_trait_change("hessian_threshold, n_octaves")
def calc_surf1(self):
self.surf = cv2.SURF(self.hessian_threshold, self.n_octaves)
self.key_points1, self.features1 = self.surf.detectAndCompute(
self.img_gray, None)
self.key_positions1 = np.array([kp.pt for kp in self.key_points1])
def _img_changed(self):
self.img_gray = cv2.cvtColor(self.img, cv2.COLOR_BGR2GRAY)
self.img_color = cv2.cvtColor(self.img_gray, cv2.COLOR_GRAY2RGB)
self.img_show = np.concatenate([self.img_color, self.img_color],
axis=1)
self.size = self.img_color.shape[1], self.img_color.shape[0]
self.calc_surf1()
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=100)
self.matcher = cv2.FlannBasedMatcher(index_params, search_params)
self.init_poly()
def settings_loaded(self):
src = self.init_points.copy()
w, h = self.size
src[:, 0] -= w
dst = cv2.perspectiveTransform(src[None, :, :], self.m_perspective)
dst = dst.squeeze()
dst[:, 0] += w
self.poly.set_points(dst)
self.poly.update()
def draw(self):
if self.poly is None:
return
w, h = self.size
src = self.init_points.copy()
dst = self.poly.points.copy().astype(np.float32)
src[:, 0] -= w
dst[:, 0] -= w
m = cv2.getPerspectiveTransform(src, dst)
self.m_perspective = m
img2 = cv2.warpPerspective(self.img_gray,
m,
self.size,
borderValue=[255] * 4)
self.img_show[:, w:, :] = img2[:, :, None]
key_points2, features2 = self.surf.detectAndCompute(img2, None)
key_positions2 = np.array([kp.pt for kp in key_points2])
match_list = self.matcher.knnMatch(self.features1, features2, k=1)
index1 = np.array([m[0].queryIdx for m in match_list])
index2 = np.array([m[0].trainIdx for m in match_list])
distances = np.array([m[0].distance for m in match_list])
n = min(50, len(distances))
best_index = np.argsort(distances)[:n]
matched_positions1 = self.key_positions1[index1[best_index]]
matched_positions2 = key_positions2[index2[best_index]]
self.m_perspective2, mask = cv2.findHomography(matched_positions1,
matched_positions2,
cv2.RANSAC)
lines = np.concatenate([matched_positions1, matched_positions2],
axis=1)
lines[:, 2] += w
line_colors = COLORS[mask.ravel()]
self.lines.set_segments(lines.reshape(-1, 2, 2))
self.lines.set_color(line_colors)
self.draw_image(self.img_show)
class SURFDemo(ImageProcessDemo):
TITLE = "SURF Demo"
DEFAULT_IMAGE = "lena.jpg"
SETTINGS = ["m_perspective", "hessian_threshold", "n_octaves"]
m_perspective = Array(np.float, (3, 3))
m_perspective2 = Array(np.float, (3, 3))
hessian_threshold = Int(2000)
n_octaves = Int(2)
poly = Instance(PolygonWidget)
def control_panel(self):
return VGroup(
Item("m_perspective", label=u"变换矩阵", editor=ArrayEditor(format_str="%g")),
Item("m_perspective2", label=u"变换矩阵", editor=ArrayEditor(format_str="%g")),
Item("hessian_threshold", label=u"hessianThreshold"),
Item("n_octaves", label=u"nOctaves")
)
def __init__(self, **kwargs):
super(SURFDemo, self).__init__(**kwargs)
self.poly = None
self.init_points = None
self.lines = LineCollection([], linewidths=1, alpha=0.6, color="red")
self.axe.add_collection(self.lines)
self.connect_dirty("poly.changed,hessian_threshold,n_octaves")
def init_poly(self):
if self.poly is None:
return
h, w, _ = self.img_color.shape
self.init_points = np.array([(w, 0), (2*w, 0), (2*w, h), (w, h)], np.float32)
self.poly.set_points(self.init_points)
self.poly.update()
def init_draw(self):
style = {"marker": "o"}
self.poly = PolygonWidget(axe=self.axe, points=np.zeros((3, 2)), style=style)
self.init_poly()
@on_trait_change("hessian_threshold, n_octaves")
def calc_surf1(self):
self.surf = cv2.SURF(self.hessian_threshold, self.n_octaves)
self.key_points1, self.features1 = self.surf.detectAndCompute(self.img_gray, None)
self.key_positions1 = np.array([kp.pt for kp in self.key_points1])
def _img_changed(self):
self.img_gray = cv2.cvtColor(self.img, cv2.COLOR_BGR2GRAY)
self.img_color = cv2.cvtColor(self.img_gray, cv2.COLOR_GRAY2RGB)
self.img_show = np.concatenate([self.img_color, self.img_color], axis=1)
self.size = self.img_color.shape[1], self.img_color.shape[0]
self.calc_surf1()
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=100)
self.matcher = cv2.FlannBasedMatcher(index_params, search_params)
self.init_poly()
def settings_loaded(self):
src = self.init_points.copy()
w, h = self.size
src[:, 0] -= w
dst = cv2.perspectiveTransform(src[None, :, :], self.m_perspective)
dst = dst.squeeze()
dst[:, 0] += w
self.poly.set_points(dst)
self.poly.update()
def draw(self):
if self.poly is None:
return
w, h = self.size
src = self.init_points.copy()
dst = self.poly.points.copy().astype(np.float32)
src[:, 0] -= w
dst[:, 0] -= w
m = cv2.getPerspectiveTransform(src, dst)
self.m_perspective = m
img2 = cv2.warpPerspective(self.img_gray, m, self.size, borderValue=[255]*4)
self.img_show[:, w:, :] = img2[:, :, None]
key_points2, features2 = self.surf.detectAndCompute(img2, None)
key_positions2 = np.array([kp.pt for kp in key_points2])
match_list = self.matcher.knnMatch(self.features1, features2, k=1)
index1 = np.array([m[0].queryIdx for m in match_list])
index2 = np.array([m[0].trainIdx for m in match_list])
distances = np.array([m[0].distance for m in match_list])
n = min(50, len(distances))
best_index = np.argsort(distances)[:n]
matched_positions1 = self.key_positions1[index1[best_index]]
matched_positions2 = key_positions2[index2[best_index]]
self.m_perspective2, mask = cv2.findHomography(matched_positions1, matched_positions2, cv2.RANSAC)
lines = np.concatenate([matched_positions1, matched_positions2], axis=1)
lines[:, 2] += w
line_colors = COLORS[mask.ravel()]
self.lines.set_segments(lines.reshape(-1, 2, 2))
self.lines.set_color(line_colors)
self.draw_image(self.img_show)
