图片网站模版/随州seo

# 导入相关包

# 加载图片
car_filename = glob.glob('./vehicles/*/*/*.png')
not_car_filename = glob.glob('./non-vehicles/*/*/*.png')
num_car_image = len(car_filename)
not_car_image = len(not_car_filename)
car_image = mping.imread(car_filename[0])
print('car images: ', num_car_image)
print('not car images: ', not_car_image)
print('Image shape{} and type {}'.format(car_image.shape,car_image.dtype))

car images

def 

# 提取HOG特征(方向梯度直方图)
def get_hog_features(img, orient,pix_per_cell, cell_per_block,vis=False, feature_vec =True):if vis == True:features, hog_image = hog(img, orientations=orient, pixels_per_cell=(pix_per_cell, pix_per_cell),cells_per_block=(cell_per_block, cell_per_block), block_norm= 'L2-Hys',transform_sqrt=True, visualise=vis, feature_vector=feature_vec)return features, hog_imageelse:features = hog(img, orientations=orient, pixels_per_cell=(pix_per_cell, pix_per_cell),cells_per_block=(cell_per_block, cell_per_block), block_norm= 'L2-Hys',transform_sqrt=True, visualise=vis, feature_vector=feature_vec)return features

gray = convert_color(car_image,'Gray')
features,hog_image=get_hog_features(gray, orient = 9, pix_per_cell =8,cell_per_block = 1,vis = True, feature_vec = False)gray = convert_color(noncar_image, 'Gray')
features, hog_image = get_hog_features(gray, orient= 9, pix_per_cell= 8, cell_per_block= 1, vis=True, feature_vec=False)

# 计算颜色histogram
def color_hist(img, nbins=8, bins_range=(0.1,0.5),visualize = False):channel1_hist = np.histogram(img[:,:,0],bins = nbins, range = bins_range)channel2_hist = np.histogram(img[:,:,1],bins = nbins, range = bins_range)channel3_hist = np.histogram(img[:,:,2],bins = nbins, range = bins_range)hist_features = np.concatenate((channel1_hist[0],channel2_hist[0],channel3_hist[0]))    if visualize:bin_edges = channel1_hist[1]bin_centers = (bin_edges[1:]+bin_edges[0:len(bin_edges)-1])/2return hist_features, bin_centersreturn hist_features

# 原始像素信息

def extract_features(image,params,spatial_feat=True, hist_feat=True,hog_feat=True):# 导入参数cspace = params.cspaceorient = params.orientpix_per_cell = params.pix_per_cellcell_per_block = params.cell_per_blockhog_channel = params.hog_channelsize = params.sizehist_bins = params.hist_binshist_range = params.hist_range    # 颜色转换feature_image = convert_color(image, cspace)   img_features = []   if hog_feat == True:if hog_channel == 'ALL':hog_features = []for channel in range(feature_image.shape[2]):hog_features.append(get_hog_features(feature_image[:,:,channel], orient, pix_per_cell, cell_per_block, vis=False, feature_vec=True))hog_features = np.ravel(hog_features) else:hog_features = get_hog_features(feature_image[:,:,hog_channel],orient,pix_per_cell,cell_per_block,vis = False, feature_vec=True)img_features.append(hog_features)if spatial_feat == True:spatial_features = bin_spatial(feature_image,size)img_features.append(spatial_features)if hist_feat == True:hist_features = color_hist(feature_image,nbins = hist_bins, bins_range = hist_range)img_features.append(hist_features)return np.concatenate(img_features)

# 参数类
class FeatureParams():def __init__(self):# HOG parametersself.cspace = 'YCrCb'self.orient = 9self.pix_per_cell = 8self.cell_per_block = 1self.hog_channel = 'ALL'self.size = (8,8)self.hist_bins = 8self.hist_range = (0.1,0.5)
params = FeatureParams()

# 颜色转换函数
def convert_color(image, color_space = 'YCrCb'):if color_space != 'RGB':if color_space == 'HSV':return cv2.cvtColor(image, cv2.COLOR_RGB2HSV)if color_space == 'LUV':return cv2.cvtColor(image, cv2.COLOR_RGB2LUV)elif color_space == 'HLS':return cv2.cvtColor(image, cv2.COLOR_RGB2HLS)elif color_space == 'YUV':return cv2.cvtColor(image, cv2.COLOR_RGB2YUV)elif color_space == 'YCrCb':return cv2.cvtColor(image, cv2.COLOR_RGB2YCrCb)elif color_space == 'Gray':return cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)else: return np.copy(image)

def split_train_test(car_features, noncar_features):x = np.vstack((car_features, noncar_features)).astype(np.float64)
# car的标签命名为1，not car的标签命名为0y = np.hstack((np.ones(len(car_features)), np.zeros(len(noncar_features))))rand_state = 43x_train,x_test,y_train,y_test = train_test_split(x,y,test_size = 0.2, random_state=rand_state)return x_train,x_test, y_train, y_test

def 

def 

13.57 Seconds to train SVC...
Train Accuracy: 0.9665
Test Accuracy: 0.9406
My SVC predicts:  [ 0.  0.  0.  1.  0.  1.  0.  0.  1.  1.]
For these 10 labels:  [ 1.  0.  1.  1.  0.  1.  0.  0.  1.  1.]
0.00363 Seconds to predict 10 labels with SVC

def slide_window(img, x_start_stop=[None, None], y_start_stop=[None, None], xy_window=(64, 64), xy_overlap=(0.85, 0.85)):# 目前定义的相邻两个滑移窗的重合度为85%# 确定滑移窗的边界，如果没有定义边界就以图形的尺寸为最终边界；if x_start_stop[0] == None:x_start_stop[0] = 0if x_start_stop[1] == None:x_start_stop[1] = img.shape[1]if y_start_stop[0] == None:y_start_stop[0] = 0if y_start_stop[1] == None:y_start_stop[1] = img.shape[0]# 计算整个滑移窗边界的大小；   xspan = x_start_stop[1] - x_start_stop[0]yspan = y_start_stop[1] - y_start_stop[0]# 计算每移动一步需要移动多少个pix，窗口的大小为xy_windownx_pix_per_step = np.int(xy_window[0]*(1 - xy_overlap[0]))ny_pix_per_step = np.int(xy_window[1]*(1 - xy_overlap[1]))# 计算x和y方向窗口的数量nx_buffer = np.int(xy_window[0]*(xy_overlap[0]))ny_buffer = np.int(xy_window[1]*(xy_overlap[1]))nx_windows = np.int((xspan-nx_buffer)/nx_pix_per_step) ny_windows = np.int((yspan-ny_buffer)/ny_pix_per_step) # I记录窗口的位置window_list = []for ys in range(ny_windows):for xs in range(nx_windows):# Calculate window positionstartx = xs*nx_pix_per_step + x_start_stop[0]endx = startx + xy_window[0]starty = ys*ny_pix_per_step + y_start_stop[0]endy = starty + xy_window[1]window_list.append(((startx, starty), (endx, endy)))# 返回窗口的位置return window_list

def 

def multi_scale_find_cars(img, svc, scaler, params, return_all=False):# 4个区域分别采用不同大小的比例缩放因子y_start_stops = [[400,464],[400,496],[432,560],[448,644]]x_start_stops = [[620,1000],[560,1280],[500,1280],[380,1280]]car_windows = []for i in range(len(scales)):scale = scales[i]for j in range(2):y_offset = j*16windows = search_windows(img, clf, scaler, params, y_start_stop=[y_start_stops[i][0],y_start_stops[i][1]+y_offset], xy_window=(64, 64), xy_overlap=(0.8, 0.8))car_windows.extend(windows)return np.array(car_windows)

def add_heat(heatmap, bbox_list):for box in bbox_list:# 每出现一个框，热度图加1heatmap[box[0][1]:box[1][1], box[0][0]:box[1][0]] += 1return heatmapdef apply_threshold(heatmap, threshold):# 小于阈值的热度图置0heatmap[heatmap <= threshold] = 0# 返回热度图return heatmap
# 画框
def draw_labeled_bboxes(img, labels):for car_number in range(1, labels[1]+1):nonzero = (labels[0] == car_number).nonzero()nonzeroy = np.array(nonzero[0])nonzerox = np.array(nonzero[1])bbox = ((np.min(nonzerox), np.min(nonzeroy)), (np.max(nonzerox), np.max(nonzeroy)))cv2.rectangle(img, bbox[0], bbox[1], (0,0,255), 6)return imgdef draw_labeled_windows(image, boxes, threshold=2):heat = np.zeros_like(image[:,:,0]).astype(np.float)heat = add_heat(heat,boxes) heat = apply_threshold(heat,threshold) heatmap = np.clip(heat, 0, 255) 
# 给剩下的热图打标签，相当于统计车的数量  labels = label(heatmap)draw_img = draw_labeled_bboxes(np.copy(image), labels)    return heatmap, labels, draw_img

from collections import deque
def pipeline(image, svc=svc, scaler=scaler, params=params):car_windows = multi_scale_find_cars(image, svc, scaler, params)  _,_,draw_image = draw_labeled_windows(image, car_windows, threshold=2)return draw_image

def test_video(src_path, dst_path):project_output = dst_path        clip1 = VideoFileClip(src_path)white_clip = clip1.fl_image(pipeline)test_video('test_video.mp4','test_videos_output/test_video.mp4')