济南做网络安全的公司seo学校培训班

利用opencv，使用边缘检测、全局变化梯度阈值过滤、算子角度过滤、HLS阈值过滤的方法进行车道线分割检测，综合多种阈值过滤进行检测提高检测精度。

1.利用cv2.Sobel()计算图像梯度(边缘检测)

import cv2
import numpy as np
import matplotlib.pyplot as plt
import ossrc = np.float32([[200, 720], [1100, 720], [595, 450], [685, 450]]) #src 输入图像
dst = np.float32([[300, 720], [980, 720], [300, 0], [980, 0]]) #dst 输出图像
m_inv = cv2.getPerspectiveTransform(dst, src)
m = cv2.getPerspectiveTransform(src, dst)# 利用cv2.Sobel()计算图像梯度(边缘检测)
def abs_sobel_threshold(img, orient='x', thresh_min=40, thresh_max=255):###利用X，y方向上sobel，二值化图像######gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)if orient == 'x':abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 1, 0))if orient == 'y':########参考求x方向的sobel算子，计算y方向上sobel算子####################填空1 （1行代码）########################abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 0, 1))#############填空1 （1行代码）########################scaled_sobel = np.uint8(255 * abs_sobel / np.max(abs_sobel))binary_output = np.zeros_like(scaled_sobel)#############二值图像，大于最小阈值并且小于最大阈值的区间置为255， 其余为0，可通过修改最大最小值查看差异####################填空2（1行代码）########################binary_output[(scaled_sobel >= thresh_min) & (scaled_sobel <= thresh_max)] = 255#############填空2 （1行代码）########################return binary_output

使用检测纵向边缘（x方向的梯度）

path = r"d:\Users\WYN\Desktop\temp\week1HomeWork\testImage"
path_list = os.listdir(path)
plt.figure(figsize=(16, 9))
for i in range(len(path_list)):path_now = "\\".join([str(path), str(path_list[i])])img = cv2.imread(path_now)img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)plt.subplot(2, 3, i + 1)plt.imshow(abs_sobel_threshold(img, orient='x', thresh_min=40, thresh_max=255))plt.xticks([])plt.yticks([])plt.subplots_adjust(wspace=0, hspace=0)
plt.show()

使用检测纵向边缘（y方向的梯度）

# 检测纵向边缘（y方向的梯度）
plt.figure(figsize=(16, 9))
for i in range(len(path_list)):path_now = "\\".join([str(path), str(path_list[i])])img = cv2.imread(path_now)img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)plt.subplot(2, 3, i + 1)plt.imshow(abs_sobel_threshold(img, orient='y', thresh_min=40, thresh_max=255))plt.xticks([])plt.yticks([])plt.subplots_adjust(wspace=0, hspace=0)
plt.show()

2.使用全局的变化梯度来进行阈值过滤

def mag_threshold(img, sobel_kernel=3, mag_threshold=(50, 255)):gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)########根据x方向的sobel算子和y方向上sobel算子，计算梯度，公式为sqrt（x^2 + y ^2）####################填空3 （1行代码）########################gradmag = np.sqrt(sobelx ** 2 + sobely ** 2)#############填空3 （1行代码）########################scale_factor = np.max(gradmag) / 255gradmag = (gradmag / scale_factor).astype(np.uint8)binary_out = np.zeros_like(gradmag)########转换为二值图，最大最小值可调，kernel_size也可以调整看看差异####################填空4 （1行代码）########################binary_out[(gradmag >= mag_threshold[0]) & (gradmag <= mag_threshold[1])] = 255#############填空4 （1行代码）########################return binary_out

通过全局阈值过滤来检测车道线

path = r"d:\Users\WYN\Desktop\temp\week1HomeWork\testImage"
path_list = os.listdir(path)
plt.figure(figsize=(16, 9))
plt.figure(1)
for i in range(len(path_list)):path_now = "\\".join([str(path), str(path_list[i])])img = cv2.imread(path_now)img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)plt.subplot(2, 3, i + 1)plt.imshow(mag_threshold(img, sobel_kernel=3, mag_threshold=(50, 255)))plt.xticks([])plt.yticks([])plt.subplots_adjust(wspace=0, hspace=0)
plt.show()

3.通过算子角度来进行阈值过滤

def dir_threshold(img, sobel_kernel=5, thresh=(0.7, 1.3)):gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)########根据x方向的sobel算子和y方向上sobel算子，计算角度，公式为arctan（y/x），将倾斜角度过大的过滤掉####################填空5 （1行代码）########################absgraddir = np.arctan(sobely / sobelx)#############填空5 （1行代码）########################binary_output = np.zeros_like(absgraddir)########转换为二值图，最大最小值可调，kernel_size也可以调整看看差异####################填空6 （1行代码）########################binary_output[((absgraddir >= thresh[0]) & (absgraddir <= thresh[1]))] = 255#############填空6 （1行代码）########################return binary_output

通过全局阈值过滤来检测车道线

path = r"d:\Users\WYN\Desktop\temp\week1HomeWork\testImage"
path_list = os.listdir(path)
plt.figure(figsize=(16, 9))
plt.figure(1)
for i in range(len(path_list)):path_now = "\\".join([str(path), str(path_list[i])])img = cv2.imread(path_now)img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)plt.subplot(2, 3, i + 1)plt.imshow(dir_threshold(img, sobel_kernel=5, thresh=(np.pi/4, np.pi/3)))plt.xticks([])plt.yticks([])plt.subplots_adjust(wspace=0, hspace=0)
plt.show()

4.使用HLS进行阈值过滤

def hls_thresh(img, thresh=(100, 255)):hls = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)########分离出s通道s_channel####################填空7 （1行代码）########################_, _, s_channel = cv2.split(hls)#############填空7 （1行代码）########################binary_output = np.zeros_like(s_channel)########转换为二值图，最大最小值可调##################################填空8 （1行代码）########################binary_output[(s_channel > thresh[0]) & (s_channel < thresh[1])] = 255#############填空8 （1行代码）########################return binary_output

5.综合多种阈值过滤方法

def combined_threshold(img):abs_bin = abs_sobel_threshold(img, orient='x', thresh_min=50, thresh_max=255)mag_bin = mag_threshold(img, sobel_kernel=3, mag_threshold=(50, 255))dir_bin = dir_threshold(img, sobel_kernel=15, thresh=(0.7, 1.3))hls_bin = hls_thresh(img, thresh=(170, 255))combined = np.zeros_like(dir_bin)#############组合四个阈值结果，判定车道线，###################例如(abs_bin == 255 | ((mag_bin == 255) & (dir_bin == 255))) | hls_bin == 25）###########可以尝试不同的组合###################################填空9（1行代码）######################### combined[(abs_bin == 255 | ((mag_bin == 255) & (dir_bin == 255))) | hls_bin == 255] = 255combined[(abs_bin == 255) & (mag_bin == 255) & (dir_bin == 255) | (hls_bin == 255)] = 255#############填空9 （1行代码）########################return combined, abs_bin, mag_bin, dir_bin, hls_bin

6.滑动窗口扫描

def line_fit_and_draw_line(binary_warped):# "查找拟合直线"# 对图像对下半部分查找直方图#############填空10（1行代码）截取图像高度的下方1/2处########################histogram = np.sum(binary_warped[int(binary_warped.shape[0]//2):, :], axis=0)#############填空10（1行代码）截取图像高度的下方1/2处########################out_img = (np.dstack((binary_warped, binary_warped, binary_warped)) * 255).astype('uint8')#查找直方图中左右两侧对峰值midpoint = np.int(histogram.shape[0] / 2)#左侧从100到 midpoint的最大值，转换成图像坐标还要加上100哦～#############右侧从midpoint到图像宽度减100的最大值，转换成图像坐标还要加上midpoint哦～################也就是图像左右边缘100像素内不查找车道线###############################填空11（2行代码）查找左侧右侧最大值基本点########################leftx_base = np.argmax(histogram[100:midpoint])+100rightx_base = np.argmax(histogram[midpoint:binary_warped.shape[1]-100])+midpoint#############填空11（2行代码）查找左侧右侧最大值基本点##################################以下是关于滑动窗口查找车道线的代码#####################nwindows = 9  # 将窗口划分为9行window_height = np.int(binary_warped.shape[0] / nwindows)nonzero = binary_warped.nonzero()nonzeroy = np.array(nonzero[0])  # 非零元素所在行nonzerox = np.array(nonzero[1])  # 非零元素所在列leftx_current = leftx_baserightx_current = rightx_basemargin = 100minpix = 10left_lane_inds = []right_lane_inds = []for window in range(nwindows):win_y_low = binary_warped.shape[0] - (window + 1) * window_heightwin_y_high = binary_warped.shape[0] - window * window_heightwin_xleft_low = leftx_current - marginwin_xleft_high = leftx_current + marginwin_xright_low = rightx_current - marginwin_xright_high = rightx_current + margingood_left_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) & (nonzerox > win_xleft_low) & (nonzerox < win_xleft_high)).nonzero()[0]good_right_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) & (nonzerox >= win_xright_low) & (nonzerox < win_xright_high)).nonzero()[0]left_lane_inds.append(good_left_inds)right_lane_inds.append(good_right_inds)if len(good_left_inds) > minpix:leftx_current = np.int(np.mean(nonzerox[good_left_inds]))if len(good_right_inds) > minpix:rightx_current = np.int(np.mean(nonzerox[good_right_inds]))left_lane_inds = np.concatenate(left_lane_inds)right_lane_inds = np.concatenate(right_lane_inds)leftx = nonzerox[left_lane_inds]lefty = nonzeroy[left_lane_inds]rightx = nonzerox[right_lane_inds]righty = nonzeroy[right_lane_inds]##########以上是关于滑动窗口查找车道线的代码######################将左侧，右侧车道线3次拟合，用函数np.polyfit#######################填空12（2行代码）左侧、右侧车道线拟合#######################para_l = np.polyfit(lefty, leftx, 3) #得到曲线参数para_r = np.polyfit(righty, rightx, 3)#############填空12（2行代码）左侧、右侧车道线拟合#######################################在图上画出拟合的线########################ploty = np.linspace(0, undist.shape[0]-1, undist.shape[0])#########对y进行拟合，x = a * y ^ 2 + b * y + C#############填空13（2行代码）左侧、右侧车道线方程坐标#######################left_fitx = para_l[0] * ploty**3 + para_l[1] * ploty**2 + para_l[2] * ploty + para_l[3]right_fitx = para_r[0] * ploty**3 + para_r[1] * ploty**2 + para_r[2] * ploty + para_r[3]#############填空13（2行代码）左侧、右侧车道线方程坐标#############################生成一张黑图，做mask，将车道线区域标注出来##########color_warp = np.zeros((720, 1280, 3), dtype='uint8')pts_left = np.array([np.transpose(np.vstack([left_fitx, ploty]))])pts_right = np.array([np.flipud(np.transpose(np.vstack([right_fitx, ploty])))])pts = np.hstack((pts_left, pts_right))# 在透射变换后的图上画出车道线cv2.fillPoly(color_warp, np.int_([pts]), (0, 255, 0))# 将画出的车道线的图，逆变换到原来的图上，将color_warp逆变换为newwarp#############填空14（1行代码）#######################newwarp = cv2.warpPerspective(color_warp, m_inv, imgOut_size, flags=cv2.INTER_LINEAR)#############填空14（1行代码）######################## 将原来的图和标注好车道线的图叠加，用cv2.addWeighted，可画成半透明，最终图为result#############填空15（1行代码）#######################result = cv2.addWeighted(img, 0.5, newwarp, 0.5, 0)#############填空15（1行代码）#######################plt.figure(figsize = (30, 30))plt.title('lane')plt.subplot(1, 1, 1)plt.imshow(result)plt.axis('off')

7.图像显示

img = cv2.imread("./testImage/test6.jpg")
out1 = mag_threshold(img)
out2 = abs_sobel_threshold(img)
out3 = dir_threshold(img)
out4 = hls_thresh(img)
imgOut, abs_bin, mab_bin, dir_bin, hls_bin = combined_threshold(img)
plt.figure(figsize = (30, 30))
plt.title('calibration')
plt.subplot(1, 5, 1)
plt.imshow(out1)
plt.title("mag_threshold")
plt.subplot(1, 5, 2)
plt.imshow(out2)
plt.title("abs_sobel_threshold")
plt.subplot(1, 5, 3)
plt.imshow(out3)
plt.title("dir_threshold")
plt.subplot(1, 5, 4)
plt.imshow(out4)
plt.title("hls_thresh")
plt.subplot(1, 5, 5)
plt.imshow(imgOut)
plt.title("combined_threshold")
plt.axis('off')
imgOut_size = (imgOut.shape[1], imgOut.shape[0])
binary_warped = cv2.warpPerspective(imgOut, m, imgOut_size, flags=cv2.INTER_LINEAR)
undist = cv2.imread("./testImage/test6.jpg")
line_fit_and_draw_line(binary_warped)
plt.show()

8.创建空白画布，并绘制指定点

import numpy as np
import cv2
import matplotlib.pyplot as plt
###在图中把点标记出来
plt.figure(figsize=(30, 30))
img = np.zeros((1000,1000,3),dtype=np.uint8)
point_list = [(200, 720), (1100, 720), (595, 450), (685, 450)]  # src
point_list2 =[(300, 720), (980, 720), (300, 0), (980, 0)]  # dst
for point in point_list:cv2.circle(img, point, 10, (255, 0, 0), 3)
for point in point_list2:cv2.circle(img, point, 10, (0, 0, 255), 3)
plt.imshow(img)
plt.show()