def getFeatures(self, image, inithann, scale_adjust=1.0):
extracted_roi = [0,0,0,0] #[int,int,int,int]
roi=list(self._roi)
cx = roi[0] + roi[2]/2 #float
cy = roi[1] + roi[3]/2 #float
if(inithann):
padded_w = roi[2] * self.padding
padded_h = roi[3] * self.padding
if(self.template_size > 1):
if(padded_w >= padded_h):
self._scale = padded_w / float(self.template_size)
else:
self._scale = padded_h / float(self.template_size)
self._tmpl_sz[0] = int(padded_w / self._scale)
self._tmpl_sz[1] = int(padded_h / self._scale)
else:
self._tmpl_sz[0] = int(padded_w)
self._tmpl_sz[1] = int(padded_h)
self._scale = 1.
if(self._hogfeatures):
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // (2*self.cell_size) * 2*self.cell_size + 2*self.cell_size
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // (2*self.cell_size) * 2*self.cell_size + 2*self.cell_size
else:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // 2 * 2
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // 2 * 2
extracted_roi[2] = int(scale_adjust * self._scale * self._tmpl_sz[0])
extracted_roi[3] = int(scale_adjust * self._scale * self._tmpl_sz[1])
extracted_roi[0] = int(cx - extracted_roi[2]//2)
extracted_roi[1] = int(cy - extracted_roi[3]//2)
z = subwindow(image, extracted_roi, cv2.BORDER_REPLICATE)
if(z.shape[1]!=self._tmpl_sz[0] or z.shape[0]!=self._tmpl_sz[1]):
tmplsz=map(int,self._tmpl_sz)
tmplsz=list(tmplsz)
z = cv2.resize(z, tuple(tmplsz))
if(self._hogfeatures):
mapp = {'sizeX':0, 'sizeY':0, 'numFeatures':0, 'map':0}
mapp = fhog.getFeatureMaps(z, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2)
mapp = fhog.PCAFeatureMaps(mapp)
self.size_patch = list(map(int, [mapp['sizeY'], mapp['sizeX'], mapp['numFeatures']]))
FeaturesMap = mapp['map'].reshape((self.size_patch[0]*self.size_patch[1], self.size_patch[2])).T # (size_patch[2], size_patch[0]*size_patch[1])
else:
if(z.ndim==3 and z.shape[2]==3):
FeaturesMap = cv2.cvtColor(z, cv2.COLOR_BGR2GRAY) # z:(size_patch[0], size_patch[1], 3) FeaturesMap:(size_patch[0], size_patch[1]) #np.int8 #0~255
elif(z.ndim==2):
FeaturesMap = z #(size_patch[0], size_patch[1]) #np.int8 #0~255
FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
self.size_patch = [z.shape[0], z.shape[1], 1]
if(inithann):
self.createHanningMats() # createHanningMats need size_patch
FeaturesMap = self.hann * FeaturesMap
return FeaturesMap
def get_scale_sample(self, image):
xsf = None
for i in range(self.n_scales):
# Size of subwindow waiting to be detect
patch_width = self.base_width * self.scaleFactors[i] * self.currentScaleFactor
patch_height = self.base_height * self.scaleFactors[i] * self.currentScaleFactor
cx = self._roi[0] + self._roi[2] / 2.
cy = self._roi[1] + self._roi[3] / 2.
# Get the subwindow
im_patch = extractImage(image, cx, cy, patch_width, patch_height)
if self.scale_model_width > im_patch.shape[1]:
im_patch_resized = cv2.resize(im_patch, (self.scale_model_width, self.scale_model_height), None, 0, 0, 1)
else:
im_patch_resized = cv2.resize(im_patch, (self.scale_model_width, self.scale_model_height), None, 0, 0, 3)
mapp = {'sizeX': 0, 'sizeY': 0, 'numFeatures': 0, 'map': 0}
mapp = fhog.getFeatureMaps(im_patch_resized, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2)
mapp = fhog.PCAFeatureMaps(mapp)
if i == 0:
totalSize = mapp['numFeatures'] * mapp['sizeX'] * mapp['sizeY']
xsf = np.zeros((totalSize, self.n_scales))
# Multiply the FHOG results by hanning window and copy to the output
FeaturesMap = mapp['map'].reshape((totalSize, 1))
FeaturesMap = self.s_hann[0][i] * FeaturesMap
xsf[:, i] = FeaturesMap[:, 0]
return fftd(xsf, False, True)
def getFeatures(self, image, inithann, scale_adjust=1.):
#print('getFeatures')
extracted_roi = [0, 0, 0, 0]
cx = self._roi[0] + self._roi[2] / 2
cy = self._roi[1] + self._roi[3] / 2
if inithann:
padded_w = self._roi[2] * self.padding
padded_h = self._roi[3] * self.padding
if self.template_size > 1:
# 把最大的边缩小到96，_scale是缩小比例
# _tmpl_sz是滤波模板的大小也是裁剪下的PATCH大小
if padded_w >= padded_h:
self._scale = padded_w / float(self.template_size)
else:
self._scale = padded_h / float(self.template_size)
self._tmpl_sz[0] = int(padded_w / self._scale)
self._tmpl_sz[1] = int(padded_h / self._scale)
else:
self._tmpl_sz[0] = int(padded_w)
self._tmpl_sz[1] = int(padded_h)
self._scale = 1.
if self._hogfeatures:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // (2 * self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // (2 * self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
else:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // 2 * 2
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // 2 * 2
# 选取从原图中扣下的图片位置大小
extracted_roi[2] = int(scale_adjust * self._scale * self._tmpl_sz[0] * self.currentScaleFactor)
extracted_roi[3] = int(scale_adjust * self._scale * self._tmpl_sz[1] * self.currentScaleFactor)
extracted_roi[0] = int(cx - extracted_roi[2] / 2)
extracted_roi[1] = int(cy - extracted_roi[3] / 2)
#print('extracted_roi: ', extracted_roi)
# z是当前帧被裁剪下的搜索区域, 基于extracted_roi获取(预测区域)
z = subwindow(image, extracted_roi, cv2.BORDER_REPLICATE)
# print('subwindow, select in z: ', z.shape)
if self.cnn_multilayer == 0: # use frame_diff in multi_KCF
if type(self.pre_frame) == type(z): # get inter-frame difference
self.pre_frame = subwindow(self.pre_frame, extracted_roi, cv2.BORDER_REPLICATE)
self.frame_diff = np.sum(abs(z - self.pre_frame)) / (z.shape[0]*z.shape[1])
self.pre_frame = image
if z.shape[1] != self._tmpl_sz[0] or z.shape[0] != self._tmpl_sz[1]: # 缩小到tmpl_sz
z = cv2.resize(z, (self._tmpl_sz[0],self._tmpl_sz[1]))
# print('resize z to tmpl_sz: ', z.shape)
if self._hogfeatures:
mapp = {'sizeX': 0, 'sizeY': 0, 'numFeatures': 0, 'map': 0}
mapp = fhog.getFeatureMaps(z, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2)
mapp = fhog.PCAFeatureMaps(mapp)
# size_patch为列表，保存裁剪下来的特征图的【长，宽，通道】
self.size_patch = list(map(int, [mapp['sizeY'], mapp['sizeX'], mapp['numFeatures']]))
FeaturesMap = mapp['map'].reshape((self.size_patch[0] * self.size_patch[1], self.size_patch[2])).T # (size_patch[2], size_patch[0]*size_patch[1])
elif self._cnnfeatures:
FeaturesMap = self.get_feat(z)
if self.multi_gray: # test multi_channel features
FeaturesMap = FeaturesMap / 255.0 - 0.5
self.size_patch = [FeaturesMap.shape[0], FeaturesMap.shape[1], FeaturesMap.shape[2]]
else:
if type(FeaturesMap) == type([]): # fill self.size_patch
self.size_patch = [FeaturesMap[0].shape[0], FeaturesMap[0].shape[1], FeaturesMap[0].shape[2]]
self.cnn_multidepth = [FeaturesMap[0].shape[2],FeaturesMap[1].shape[2],FeaturesMap[2].shape[2]]
#print('get multi cnn featMap: ', FeaturesMap[0].shape, self.cnn_multidepth)
else:
#print('get single cnn featMap: ', FeaturesMap.shape)
# write self.size_patch by FeaturesMap.shape
self.size_patch = [FeaturesMap.shape[0], FeaturesMap.shape[1], FeaturesMap.shape[2]]
else: # 将RGB图变为单通道灰度图
if z.ndim == 3 and z.shape[2] == 3:
FeaturesMap = cv2.cvtColor(z, cv2.COLOR_BGR2GRAY)
if self.SAVE_feat:
path_gray = save_path + 'gray' + str(self.count) + '.jpg'
cv2.imwrite(path_gray, FeaturesMap)
elif z.ndim == 2:
FeaturesMap = z
#print(FeaturesMap.shape)
# 从此FeatureMap从-0.5到0.5
FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
# size_patch为列表，保存裁剪下来的特征图的[长，宽，1]
self.size_patch = [z.shape[0], z.shape[1], 1]
self.count += 1
if inithann:
self.createHanningMats()
#FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
if type(FeaturesMap) == type([]):
for i in range(len(FeaturesMap)):
FeaturesMap[i] = self.hann[i] * FeaturesMap[i]
else:
# print('add hann: ', self.hann.shape, '\n')
FeaturesMap = self.hann * FeaturesMap # 加汉宁窗(cosine)减少频谱泄露
return FeaturesMap
def getFeatures(self, image, inithann, scale_adjust=1.):
extracted_roi = [0, 0, 0, 0]
cx = self._roi[0] + self._roi[2] / 2
cy = self._roi[1] + self._roi[3] / 2
if inithann:
padded_w = self._roi[2] * self.padding
padded_h = self._roi[3] * self.padding
if self.template_size > 1:
# 把最大的边缩小到96，_scale是缩小比例
# _tmpl_sz是滤波模板的大小也是裁剪下的PATCH大小
if padded_w >= padded_h:
self._scale = padded_w / float(self.template_size)
else:
self._scale = padded_h / float(self.template_size)
self._tmpl_sz[0] = int(padded_w / self._scale)
self._tmpl_sz[1] = int(padded_h / self._scale)
else:
self._tmpl_sz[0] = int(padded_w)
self._tmpl_sz[1] = int(padded_h)
self._scale = 1.
if self._hogfeatures:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // (2 * self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // (2 * self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
else:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // 2 * 2
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // 2 * 2
# 选取从原图中扣下的图片位置大小
extracted_roi[2] = int(scale_adjust * self._scale * self._tmpl_sz[0] * self.currentScaleFactor)
extracted_roi[3] = int(scale_adjust * self._scale * self._tmpl_sz[1] * self.currentScaleFactor)
extracted_roi[0] = int(cx - extracted_roi[2] / 2)
extracted_roi[1] = int(cy - extracted_roi[3] / 2)
# z是当前帧被裁剪下的搜索区域
z = subwindow(image, extracted_roi, cv2.BORDER_REPLICATE)
if z.shape[1] != self._tmpl_sz[0] or z.shape[0] != self._tmpl_sz[1]: # 缩小到96
z = cv2.resize(z, tuple(self._tmpl_sz))
if self._hogfeatures:
mapp = {'sizeX': 0, 'sizeY': 0, 'numFeatures': 0, 'map': 0}
mapp = fhog.getFeatureMaps(z, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2)
mapp = fhog.PCAFeatureMaps(mapp)
# size_patch为列表，保存裁剪下来的特征图的【长，宽，通道】
self.size_patch = list(map(int, [mapp['sizeY'], mapp['sizeX'], mapp['numFeatures']]))
FeaturesMap = mapp['map'].reshape((self.size_patch[0] * self.size_patch[1], self.size_patch[2])).T # (size_patch[2], size_patch[0]*size_patch[1])
else: # 将RGB图变为单通道灰度图
if z.ndim == 3 and z.shape[2] == 3:
FeaturesMap = cv2.cvtColor(z, cv2.COLOR_BGR2GRAY)
elif z.ndim == 2:
FeaturesMap = z
# 从此FeatureMap从-0.5到0.5
FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
# size_patch为列表，保存裁剪下来的特征图的【长，宽，1】
self.size_patch = [z.shape[0], z.shape[1], 1]
if inithann:
self.createHanningMats()
FeaturesMap = self.hann * FeaturesMap # 加汉宁（余弦）窗减少频谱泄露
return FeaturesMap
def getFeatures(self,
image,
inithann,
scale_adjustx=1.0,
scale_adjusty=1.0):
extracted_roi = [0, 0, 0, 0] # [int,int,int,int]
cx = self._roi[0] + self._roi[2] / 2 # float
cy = self._roi[1] + self._roi[3] / 2 # float
if inithann:
padded_w = self._roi[2] * self.padding
padded_h = self._roi[3] * self.padding
if (self.template_size > 1):
if (padded_w >= padded_h):
self._scalex = padded_w / float(self.template_size)
else:
self._scaley = padded_h / float(self.template_size)
self._tmpl_sz[0] = int(padded_w / self._scalex)
self._tmpl_sz[1] = int(padded_h / self._scaley)
else:
self._tmpl_sz[0] = int(padded_w)
self._tmpl_sz[1] = int(padded_h)
self._scalex = 1.
self._scaley = 1.
if (self._hogfeatures):
self._tmpl_sz[0] = int(self._tmpl_sz[0]) / (
2 *
self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
self._tmpl_sz[1] = int(self._tmpl_sz[1]) / (
2 *
self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
else:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) / 2 * 2
self._tmpl_sz[1] = int(self._tmpl_sz[1]) / 2 * 2
extracted_roi[2] = int(scale_adjustx * self._scalex * self._tmpl_sz[0])
extracted_roi[3] = int(scale_adjusty * self._scaley * self._tmpl_sz[1])
extracted_roi[0] = int(cx - extracted_roi[2] / 2)
extracted_roi[1] = int(cy - extracted_roi[3] / 2)
z = subwindow(image, extracted_roi, cv2.BORDER_REPLICATE)
if (z.shape[1] != self._tmpl_sz[0] or z.shape[0] != self._tmpl_sz[1]):
z = cv2.resize(z, tuple(np.array(self._tmpl_sz).astype(np.uint8)))
if (self._hogfeatures):
mapp = {'sizeX': 0, 'sizeY': 0, 'numFeatures': 0, 'map': 0}
mapp = fhog.getFeatureMaps(z, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2)
mapp = fhog.PCAFeatureMaps(mapp)
self.size_patch = list(
map(int, [mapp['sizeY'], mapp['sizeX'], mapp['numFeatures']]))
FeaturesMap = mapp['map'].reshape(
(self.size_patch[0] * self.size_patch[1], self.size_patch[2]
)).T # (size_patch[2], size_patch[0]*size_patch[1])
# t1 = time.time()
# colormap = im2color(z, 1, self.size_patch[1], self.size_patch[0])
# FeaturesMap = np.array(FeaturesMap.tolist() + colormap.tolist())
# t2 = time.time()
# self.size_patch[2] += 11
else:
if (z.ndim == 3 and z.shape[2] == 3):
# z:(size_patch[0], size_patch[1], 3) FeaturesMap:(size_patch[0], size_patch[1]) #np.int8 #0~255
FeaturesMap = cv2.cvtColor(z, cv2.COLOR_BGR2HSV)[:, :, 0]
elif (z.ndim == 2):
FeaturesMap = z # (size_patch[0], size_patch[1]) #np.int8 #0~255
FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
self.size_patch = [z.shape[0], z.shape[1], 1]
if (inithann):
self.createHanningMats() # createHanningMats need size_patch
FeaturesMap = self.hann * FeaturesMap
return FeaturesMap
def getFeatures(self, image, inithann, scale_adjust=1.0):
extracted_roi = [0, 0, 0, 0] # [int,int,int,int]
print("11111111111111 = ", self._roi)
# _roi = [0,0,0,0]
# for item in self._roi:
# print("item = ", item)
# print("11111111111111111 ",self._roi)
cx = self._roi[0] + self._roi[2] / 2
# float 0123分别是左上角的坐标Xy,以及矩形的长和宽
cy = self._roi[1] + self._roi[3] / 2 # float
if (inithann):
padded_w = self._roi[2] * self.padding
padded_h = self._roi[3] * self.padding
print("self.padding = ", self.padding)
print("self._roi[ = ", self._roi)
if (self.template_size >
1): # template_size?? #把最大的边缩小到96，_scale是缩小比例
# _tmpl_sz是滤波模板的大小也是裁剪下的PATCH大小
print("padded_w = ", padded_w)
print("padded_h = ", padded_h)
if (padded_w >= padded_h):
self._scale = padded_w / float(self.template_size)
else:
self._scale = padded_h / float(self.template_size)
print("self._scale = ", self._scale)
self._tmpl_sz[0] = int(padded_w / self._scale) # ??
self._tmpl_sz[1] = int(padded_h / self._scale)
else:
self._tmpl_sz[0] = int(padded_w)
self._tmpl_sz[1] = int(padded_h)
self._scale = 1.
if (self._hogfeatures):
self._tmpl_sz[0] = int(
(self._tmpl_sz[0]) /
(2 *
self.cell_size)) * 2 * self.cell_size + 2 * self.cell_size
self._tmpl_sz[1] = int(
(self._tmpl_sz[1]) /
(2 *
self.cell_size)) * 2 * self.cell_size + 2 * self.cell_size
else:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) / 2 * 2 # ??
self._tmpl_sz[1] = int(self._tmpl_sz[1]) / 2 * 2
extracted_roi[2] = int(scale_adjust * self._scale *
self._tmpl_sz[0]) # 选取从原图中扣下的图片位置大小
extracted_roi[3] = int(scale_adjust * self._scale * self._tmpl_sz[1])
extracted_roi[0] = int(cx - extracted_roi[2] / 2)
extracted_roi[1] = int(cy - extracted_roi[3] / 2)
z = subwindow(image, extracted_roi,
cv2.BORDER_REPLICATE) # z是当前帧被裁剪下的搜索区域
if (z.shape[1] != self._tmpl_sz[0]
or z.shape[0] != self._tmpl_sz[1]): # 缩小到96
self._tmpl_sz[0] = int(self._tmpl_sz[0])
self._tmpl_sz[1] = int(self._tmpl_sz[1])
print("self._tmpl_sz = ", self._tmpl_sz)
z = cv2.resize(z, tuple(self._tmpl_sz))
if (self._hogfeatures):
mapp = {'sizeX': 0, 'sizeY': 0, 'numFeatures': 0, 'map': 0}
mapp = fhog.getFeatureMaps(z, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2)
mapp = fhog.PCAFeatureMaps(mapp)
print("sizeY = ", mapp['sizeY'])
print("sizeX = ", mapp['sizeX'])
print("numFeatures = ", mapp['numFeatures'])
self.size_patch = map(
int, [mapp['sizeY'], mapp['sizeX'], mapp['numFeatures']])
# size_patch为列表，保存裁剪下来的特征图的【长，宽，通道】
self.size_patch = list(self.size_patch)
FeaturesMap = mapp['map'].reshape(
(self.size_patch[0] * self.size_patch[1], self.size_patch[2]
)).T # (size_patch[2], size_patch[0]*size_patch[1])
else:
if (z.ndim == 3 and z.shape[2] == 3): # 三维且第三维长度为3?? 三通道，也就是RGB图像
FeaturesMap = cv2.cvtColor(z, cv2.COLOR_BGR2GRAY)
# z:(size_patch[0], size_patch[1], 3) FeaturesMap:(size_patch[0], size_patch[1])
# #np.int8 #0~255
elif (z.ndim == 2):
FeaturesMap = z # (size_patch[0], size_patch[1]) #np.int8 #0~255
FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
# 从此FeatureMap从-0.5到0.5
self.size_patch = [z.shape[0], z.shape[1],
1] # size_patch为列表，保存裁剪下来的特征图的【长，宽，1】
if (inithann):
self.createHanningMats() # createHanningMats need size_patch
FeaturesMap = self.hann * FeaturesMap # 加汉宁（余弦）窗减少频谱泄露
return FeaturesMap
def getFeatures(self, image, inithann, scale_adjust=1.0):
"""
提取窗口特征
//步骤：
//1.根据给定的框框找到合适的框框
//2.提取HOG特征
//3.对特征进行归一化和截断
//4.对特征进行PCA降维
//5.创建一个常数阵，对所有特征根据cell的位置进行加权
:param image: 输入图像
:param inithann: 是否使用汉宁窗
:param scale_adjust: 调整因子
:return: 返回提取到的特征
"""
extracted_roi = [0, 0, 0, 0] # 用于提取特征的窗口
cx = self._roi[0] + self._roi[2] / 2
cy = self._roi[1] + self._roi[3] / 2
if inithann: # 初始化汉宁窗，只在第一帧的时候执行一次
padded_w = self._roi[2] * self.padding
padded_h = self._roi[3] * self.padding
if self.template_size > 1: # 按照长宽比修改_tmpl长宽大小，保证长边为96
if padded_w >= padded_h:
self._scale = padded_w / float(self.template_size)
else:
self._scale = padded_h / float(self.template_size)
self._tmpl_sz[0] = int(padded_w / self._scale)
self._tmpl_sz[1] = int(padded_h / self._scale)
else:
self._tmpl_sz[0] = int(padded_w)
self._tmpl_sz[1] = int(padded_h)
self._scale = 1.
if self._hogfeatures: # 设置_tmpl_sz的长宽:向上取原来长宽的最小2*cell_size倍
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // (
2 *
self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // (
2 *
self.cell_size) * 2 * self.cell_size + 2 * self.cell_size
else:
self._tmpl_sz[0] = int(self._tmpl_sz[0]) // 2 * 2
self._tmpl_sz[1] = int(self._tmpl_sz[1]) // 2 * 2
extracted_roi[2] = int(scale_adjust * self._scale *
self._tmpl_sz[0]) # 检测区域大小
extracted_roi[3] = int(scale_adjust * self._scale * self._tmpl_sz[1])
extracted_roi[0] = int(cx - extracted_roi[2] / 2) # 检测区域的左上角坐标
extracted_roi[1] = int(cy - extracted_roi[3] / 2)
z = tools.subwindow(image, extracted_roi,
cv2.BORDER_REPLICATE) # 提取目标区域像素，超边界则做填充
if z.shape[1] != self._tmpl_sz[0] or z.shape[0] != self._tmpl_sz[1]:
z = cv2.resize(z, tuple(self._tmpl_sz)) # 按照比例缩小边界大小
if self._hogfeatures: # 提取HOG特征点
mapp = {'sizeX': 0, 'sizeY': 0, 'numFeatures': 0, 'map': 0}
mapp = fhog.getFeatureMaps(z, self.cell_size, mapp)
mapp = fhog.normalizeAndTruncate(mapp, 0.2) # 对HOG特征归一化和截断
mapp = fhog.PCAFeatureMaps(mapp) # PCA降维
self.size_patch = list(
map(int, [mapp['sizeY'], mapp['sizeX'], mapp['numFeatures']]))
FeaturesMap = mapp['map'].reshape(
(self.size_patch[0] * self.size_patch[1],
self.size_patch[2])).T
else:
if z.ndim == 3 and z.shape[2] == 3:
FeaturesMap = cv2.cvtColor(
z, cv2.COLOR_BGR2GRAY) # 如果图像时三通道则转换为灰度图
elif z.ndim == 2:
FeaturesMap = z
FeaturesMap = FeaturesMap.astype(np.float32) / 255.0 - 0.5
self.size_patch = [z.shape[0], z.shape[1], 1]
if inithann:
self.createHanningMats()
FeaturesMap = self.hann * FeaturesMap
return FeaturesMap
