def track(self, image):
test_patch = utils.get_subwindow(image, self.pos, self.sz, scale_factor=self.currentScaleFactor)
hog_feature_t = pyhog.features_pedro(test_patch / 255., 1)
hog_feature_t = np.lib.pad(hog_feature_t, ((1, 1), (1, 1), (0, 0)), 'edge')
xt = np.multiply(hog_feature_t, self.cos_window[:, :, None])
xtf = np.fft.fft2(xt, axes=(0, 1))
response = np.real(np.fft.ifft2(np.divide(np.sum(np.multiply(self.x_num, xtf),
axis=2), (self.x_den + self.lamda))))
v_centre, h_centre = np.unravel_index(response.argmax(), response.shape)
vert_delta, horiz_delta = \
[(v_centre - response.shape[0] / 2) * self.currentScaleFactor,
(h_centre - response.shape[1] / 2) * self.currentScaleFactor]
self.pos = [self.pos[0] + vert_delta, self.pos[1] + horiz_delta]
st = utils.get_scale_subwindow(image, self.pos, self.base_target_size,
self.currentScaleFactor * self.scaleSizeFactors, self.scale_window,
self.scale_model_sz)
stf = np.fft.fftn(st, axes=[0])
scale_reponse = np.real(np.fft.ifftn(np.sum(np.divide(np.multiply(self.s_num, stf),
(self.s_den[:, None] + self.lamda_scale)), axis=1)))
recovered_scale = np.argmax(scale_reponse)
self.currentScaleFactor = self.currentScaleFactor * self.scaleFactors[recovered_scale]
if self.currentScaleFactor < self.min_scale_factor:
self.currentScaleFactor = self.min_scale_factor
elif self.currentScaleFactor > self.max_scale_factor:
self.currentScaleFactor = self.max_scale_factor
# update
update_patch = utils.get_subwindow(image, self.pos, self.sz, scale_factor=self.currentScaleFactor)
hog_feature_l = pyhog.features_pedro(update_patch / 255., 1)
hog_feature_l = np.lib.pad(hog_feature_l, ((1, 1), (1, 1), (0, 0)), 'edge')
xl = np.multiply(hog_feature_l, self.cos_window[:, :, None])
xlf = np.fft.fft2(xl, axes=(0, 1))
new_x_num = np.multiply(self.yf[:, :, None], np.conj(xlf))
new_x_den = np.real(np.sum(np.multiply(xlf, np.conj(xlf)), axis=2))
sl = utils.get_scale_subwindow(image, self.pos, self.base_target_size,
self.currentScaleFactor * self.scaleSizeFactors, self.scale_window,
self.scale_model_sz)
slf = np.fft.fftn(sl, axes=[0])
new_s_num = np.multiply(self.ysf[:, None], np.conj(slf))
new_s_den = np.real(np.sum(np.multiply(slf, np.conj(slf)), axis=1))
self.x_num = (1 - self.interp_factor) * self.x_num + self.interp_factor * new_x_num
self.x_den = (1 - self.interp_factor) * self.x_den + self.interp_factor * new_x_den
self.s_num = (1 - self.interp_factor) * self.s_num + self.interp_factor * new_s_num
self.s_den = (1 - self.interp_factor) * self.s_den + self.interp_factor * new_s_den
self.target_size = self.base_target_size * self.currentScaleFactor
return vot.Rectangle(self.pos[1] - self.target_size[1] / 2,
self.pos[0] - self.target_size[0] / 2,
self.target_size[1],
self.target_size[0]
)
def get_subwindow(im, pos, sz, feature='raw'):
"""
Obtain sub-window from image, with replication-padding.
Returns sub-window of image IM centered at POS ([y, x] coordinates),
with size SZ ([height, width]). If any pixels are outside of the image,
they will replicate the values at the borders.
The subwindow is also normalized to range -0.5 .. 0.5, and the given
cosine window COS_WINDOW is applied
(though this part could be omitted to make the function more general).
"""
if np.isscalar(sz): # square sub-window
sz = [sz, sz]
ys = np.floor(pos[0]) + np.arange(sz[0], dtype=int) - np.floor(sz[0] / 2)
xs = np.floor(pos[1]) + np.arange(sz[1], dtype=int) - np.floor(sz[1] / 2)
ys = ys.astype(int)
xs = xs.astype(int)
# check for out-of-bounds coordinates and set them to the values at the borders
ys[ys < 0] = 0
ys[ys >= im.shape[0]] = im.shape[0] - 1
xs[xs < 0] = 0
xs[xs >= im.shape[1]] = im.shape[1] - 1
out = im[np.ix_(ys, xs)]
if feature == 'hog':
hog_feature = pyhog.features_pedro(out / 255., 1)
out = np.lib.pad(hog_feature, ((1, 1), (1, 1), (0, 0)), 'edge')
return out
def track(self, image):
# ---------------------------------------track--------------------------------- #
test_patch = utils.get_subwindow(image, self.pos, self.sz)
hog_feature_t = pyhog.features_pedro(test_patch / 255., 1)
hog_feature_t = np.lib.pad(hog_feature_t, ((1, 1), (1, 1), (0, 0)),
'edge')
xt = np.multiply(hog_feature_t, self.cos_window[:, :, None])
xtf = np.fft.fft2(xt, axes=(0, 1))
#计算响应，直接多通道叠加
response = np.real(
np.fft.ifft2(
np.divide(np.sum(np.multiply(self.x_num, xtf), axis=2),
(self.x_den + self.lamda))))
#找响应最大值
v_centre, h_centre = np.unravel_index(response.argmax(),
response.shape)
vert_delta, horiz_delta = \
[(v_centre - response.shape[0] / 2),
(h_centre - response.shape[1] / 2)]
#新的位置
self.pos = [self.pos[0] + vert_delta, self.pos[1] + horiz_delta]
# ---------------------------------------update--------------------------------- #
update_patch = utils.get_subwindow(image, self.pos, self.sz)
hog_feature_l = pyhog.features_pedro(update_patch / 255., 1)
hog_feature_l = np.lib.pad(hog_feature_l, ((1, 1), (1, 1), (0, 0)),
'edge')
xl = np.multiply(hog_feature_l, self.cos_window[:, :, None])
xlf = np.fft.fft2(xl, axes=(0, 1))
#更新位置滤波器
new_x_num = np.multiply(self.yf[:, :, None], np.conj(xlf))
new_x_den = np.real(np.sum(np.multiply(xlf, np.conj(xlf)), axis=2))
#滤波器学习
self.x_num = (1 - self.interp_factor
) * self.x_num + self.interp_factor * new_x_num
self.x_den = (1 - self.interp_factor
) * self.x_den + self.interp_factor * new_x_den
self.target_size = self.base_target_size
return vot.Rectangle(self.pos[1] - self.target_size[1] / 2,
self.pos[0] - self.target_size[0] / 2,
self.target_size[1], self.target_size[0])
def get_scale_subwindow(im, pos, base_target_size, scaleFactors, scale_window,
scale_model_sz):
from pyhog import pyhog
nScales = len(scaleFactors)
out = []
for i in range(nScales):
patch_sz = np.floor(base_target_size * scaleFactors[i])
scale_patch = get_subwindow(im, pos, patch_sz)
im_patch_resized = transform.resize(scale_patch,
scale_model_sz,
mode='reflect')
temp_hog = pyhog.features_pedro(im_patch_resized, 4)
out.append(np.multiply(temp_hog.flatten(), scale_window[i]))
return np.asarray(out)
def get_scale_sample(self, im, scaleFactors):
from pyhog import pyhog
resized_im_array = []
for i, s in enumerate(scaleFactors):
patch_sz = np.floor(self.first_target_sz * s)
im_patch = self.get_subwindow(im, self.pos, patch_sz) # extract image
# because the hog output is (dim/4)-2>1:
if self.first_target_sz.min() < 12:
scale_up_factor = 12. / np.min(self.first_target_sz)
im_patch_resized = imresize(im_patch, np.asarray(self.first_target_sz * scale_up_factor).astype('int'))
else:
im_patch_resized = imresize(im_patch, self.first_target_sz) #resize image to model size
features_hog = pyhog.features_pedro(im_patch_resized.astype(np.float64)/255.0, 4)
resized_im_array.append(np.multiply(features_hog.flatten(), self.scale_window[i]))
return np.asarray(resized_im_array)
def get_scale_window(image, position, target_size, sfs, scale_window,
scale_model_sz):
# pos = [position[1],position[0]]
# ts = np.array([target_size[1],target_size[0]])
out = []
for i in range(len(sfs)):
patch_sz = np.floor(target_size * sfs[i])
scale_patch = get_ori(image, position, patch_sz)
im_patch_resized = transform.resize(scale_patch,
scale_model_sz,
mode='reflect')
temp_hog = pyhog.features_pedro(im_patch_resized, 4)
out.append(np.multiply(temp_hog.flatten(), scale_window[i]))
return np.asarray(out)
def get_feature_map(im_patch, feature_list, num_feature_ch, out_size, w2c,
cell_size, projection_matrix):
#im_patch_size = im_patch.shape
#if 'fhog' in feature_list:
## out_size = np.round(np.array([im_patch_size[0],im_patch_size[1]])/cell_size)
#if 'fhog' not in feature_list:
# out_size = np.round(np.array([im_patch_size[0],im_patch_size[1]]))
out = np.zeros((np.int(out_size[0]), np.int(out_size[1]), num_feature_ch))
print('out_size')
print(im_patch.shape)
channel_id = 0
if 'gray' in feature_list:
print('using gray feature')
im_patch_255 = np.floor(im_patch * 255)
im_patch_255 = im_patch_255.astype(np.uint8)
dstimg = cv2.cvtColor(im_patch_255, cv2.COLOR_BGR2GRAY)
out[:, :, channel_id] = transform.resize(dstimg,
out_size,
mode='reflect')
channel_id = channel_id + 1
if 'fhog' in feature_list:
print('using fhog feature')
hog_feature_t = pyhog.features_pedro(im_patch / 255., cell_size)
img_crop = np.lib.pad(hog_feature_t, ((1, 1), (1, 1), (0, 0)), 'edge')
img_crop = img_crop[:, :, 18:27]
out[:, :, channel_id:channel_id + 9] = img_crop
channel_id = channel_id + 9
if 'cn' in feature_list:
print('using cn feature')
im_patch_255 = np.floor(im_patch * 255)
cn_out = im2c(im_patch_255, w2c,
(np.int(out_size[0]), np.int(out_size[1])))
cn_out = np.dot(cn_out, projection_matrix)
out[:, :, channel_id:channel_id + 2] = cn_out
if 'raw' in feature_list:
print('using raw feature')
im_patch_255 = np.floor(im_patch * 255)
img_colour = im_patch_255 - im_patch_255.mean()
out[:, :, channel_id:channel_id + 3] = transform.resize(img_colour,
out_size,
mode='reflect')
print(out.shape)
return out
def get_subwindow(im, pos, sz, scale_factor=None, feature='raw'):
"""
Obtain sub-window from image, with replication-padding.
Returns sub-window of image IM centered at POS ([y, x] coordinates),
with size SZ ([height, width]). If any pixels are outside of the image,
they will replicate the values at the borders.
The subwindow is also normalized to range -0.5 .. 0.5, and the given
cosine window COS_WINDOW is applied
(though this part could be omitted to make the function more general).
"""
#测试sz是不是标量，如果是，就是方形窗口
if np.isscalar(sz): # square sub-window
sz = [sz, sz]
sz_ori = sz
if scale_factor != None:
sz = np.floor(sz * scale_factor)
#获取有效区域位置的坐标序号，注意数列/矩阵的运用
ys = np.floor(pos[0]) + np.arange(sz[0], dtype=int) - np.floor(sz[0] / 2)
xs = np.floor(pos[1]) + np.arange(sz[1], dtype=int) - np.floor(sz[1] / 2)
ys = ys.astype(int)
xs = xs.astype(int)
# check for out-of-bounds coordinates and set them to the values at the borders
# 有可能这个取特征的区域已经超出了图像的范围，将超出的范围置为边界数值
ys[ys < 0] = 0
ys[ys >= im.shape[0]] = im.shape[0] - 1
xs[xs < 0] = 0
xs[xs >= im.shape[1]] = im.shape[1] - 1
out = im[np.ix_(ys, xs)]
#misc.imresize函数实现图像的大小调整，可以给数字对表示调整到固定大小，小于1的数字表示按照比例调整，大于1就按照百分比调整
if scale_factor != None:
out = misc.imresize(out, sz_ori.astype(int))
if feature == 'hog':
from pyhog import pyhog
hog_feature = pyhog.features_pedro(out / 255., 1)
out = np.lib.pad(hog_feature, ((1, 1), (1, 1), (0, 0)), 'edge')
print('getting hog features in get_subwindow')
return out