Tip: Actual Handling of Real Conditions
Article Directory
foreword
Hough transform is an algorithm that image processing must come into contact with. It detects objects with a specific shape through a voting algorithm. This process obtains a set that conforms to the specific shape by calculating the local maximum of the accumulated results in a parameter space as Hough transform results, this method can detect circles, straight lines, ellipses and other shapes. In lane line detection, one of the solutions considered at the beginning is to use Hough transform to detect straight lines for lane line extraction.
1. Image processing
Example: For image feature detection, the definition of the image is relatively high, and for the blurred image, it needs to be processed by the device:
- Gaussian filter
- mean filtering
- median filter
- bilateral filtering
- Histogram filtering
, etc. are more commonly used methods for blurred image processing;
# 高斯滤波
img = cv2.GaussianBlur(img, (5,5),1)
# cv2.imshow('gaosi', img)
# # 均值滤波
img = cv2.blur(img, (3,3))# 5,5
# cv2.imshow('junzhi', img)
# 中值滤波
img = cv2.medianBlur(img, 5)
# 双边滤波
"""
src：输入图像
d：过滤时周围每个像素领域的直径
sigmaColor：Sigma_color较大，则在邻域中的像素值相差较大的像素点也会用来平均。
sigmaSpace：Sigma_space较大，则虽然离得较远，但是，只要值相近，就会互相影响。
将sigma_space设置较大，sigma_color设置较小，可获得较好的效果（椒盐噪声）。
"""
# img = cv2.bilateralFilter(img, 10, 75,75)#10,75,75
#直方滤波
(b, g, r) = cv2.split(img)
bH = cv2.equalizeHist(b)
gH = cv2.equalizeHist(g)
rH = cv2.equalizeHist(r)
The image after image enhancement has more obvious contour information and better feature extraction;
2. Line detection
code analysis
Fragment Analysis 1
edges = cv2.Canny(img, 10, 200, apertureSize=3) # 边缘检测 #10
lines = cv2.HoughLines(edges, 1, np.pi / 180, 20) # 直线检测
HoughLinesAfter Hough transform, return rho (the distance from the origin (0, 0)) and theta (the angle of the straight line), and finally calculate according to the triangle theorem, you can get the end coordinates of the longest line segment.
[[0. 3.0019662]] 0.0 3.0019662 0.7853981633974483 2.356194490192345 ----------------------------------
[[-2. 3.0194197]] -2.0 3.0194197 0.7853981633974483 2.356194490192345 ----------------------------------
[[0. 3.0019662]] 0.0 3.0019662 0.7853981633974483 2.356194490192345 ----------------------------------
[[0. 3.0019662]] 0.0 3.0019662 0.7853981633974483 2.356194490192345 ----------------------------------
[[0. 3.0019662]] 0.0 3.0019662 0.7853981633974483 2.356194490192345 ----------------------------------
Fragment Analysis 2
if (theta < (np.pi / 4.)) or (theta > (3. * np.pi / 4.0)): # 垂直直线
pt1 = (int(rho / np.cos(theta)), 0) # 该直线与第一行的交点
# 该直线与最后一行的焦点
pt2 = (int((rho - img.shape[0] * np.sin(theta)) / np.cos(theta)), img.shape[0])
else: # 水平直线
pt1 = (0, int(rho / np.sin(theta))) # 该直线与第一列的交点
# 该直线与最后一列的交点
pt2 = (img.shape[1], int((rho - img.shape[1] * np.cos(theta)) / np.sin(theta)))
这里我计算也没有理解过来，它是怎么转化过来的，有大佬知道，希望能指点一下
full code
import time
import cv2
import numpy as np
def opencv_img(img):# img 必须为灰度图
img2 = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
cv2.imshow('ded',img2)
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 转灰度图
cv2.imshow('huidutu',img)
# 高斯滤波
img = cv2.GaussianBlur(img, (5,5),1)
# cv2.imshow('gaosi', img)
# # 均值滤波
img = cv2.blur(img, (3,3))# 5,5
# cv2.imshow('junzhi', img)
# 中值滤波
img = cv2.medianBlur(img, 5)
# 双边滤波
"""
src：输入图像
d：过滤时周围每个像素领域的直径
sigmaColor：Sigma_color较大，则在邻域中的像素值相差较大的像素点也会用来平均。
sigmaSpace：Sigma_space较大，则虽然离得较远，但是，只要值相近，就会互相影响。
将sigma_space设置较大，sigma_color设置较小，可获得较好的效果（椒盐噪声）。
"""
# img = cv2.bilateralFilter(img, 10, 75,75)#10,75,75
edges = cv2.Canny(img, 10, 200, apertureSize=3) # 边缘检测 #10
cv2.imshow('Canny',edges)
lines = cv2.HoughLines(edges, 1, np.pi / 180, 20) # 直线检测
print(lines,'\n')
if lines is None:
return ((0,1),(1,0))
for line in lines:
rho = line[0][0] # 第一个元素是距离rho
theta = line[0][1] # 第二个元素是角度theta
# print(line,rho,theta,np.pi / 4.,3. * np.pi / 4.0,'----------------------------------')
if (theta < (np.pi / 4.)) or (theta > (3. * np.pi / 4.0)): # 垂直直线
"""
1
"""
pt1 = (int(rho / np.cos(theta))+490, 0+372) # 该直线与第一行的交点
# 该直线与最后一行的焦点
pt2 = (int((rho - img.shape[0] * np.sin(theta)) / np.cos(theta))+490, img.shape[0]+372)
"""
2
"""
# pt1 = (int(rho / np.cos(theta)) + 66, 0 + 487) # 该直线与第一行的交点
# # 该直线与最后一行的焦点
# pt2 = (int((rho - img.shape[0] * np.sin(theta)) / np.cos(theta)) + 66, img.shape[0] + 487)
# print(pt1,pt2)
else: # 水平直线
"""
1
"""
pt1 = (0, int(rho / np.sin(theta))) # 该直线与第一列的交点
# 该直线与最后一列的交点
pt2 = (img.shape[1], int((rho - img.shape[1] * np.cos(theta)) / np.sin(theta)))
"""
2
"""
# pt1 = (int(rho / np.cos(theta)) + 66, 0 + 487) # 该直线与第一行的交点
# # 该直线与最后一行的焦点
# pt2 = (int((rho - img.shape[0] * np.sin(theta)) / np.cos(theta)) + 66, img.shape[0] + 487)
# print(pt1, pt2)
# print(pt1,pt2)
return (pt1,pt2)
pass
cap = cv2.VideoCapture(r'E:\1\pidai\pinayi_1.mp4') # 检测数据地址
while(True):
ret, frame = cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
t0 = time.time()
# print(a)
img = frame[372:560,490:518] # 切片后图像 这是为了能够更好的检测，选取图像中的某一位置的图像经行检测
#直方滤波
(b, g, r) = cv2.split(img)
bH = cv2.equalizeHist(b)
gH = cv2.equalizeHist(g)
rH = cv2.equalizeHist(r)
# 合并每一个通道
img = cv2.merge((bH, gH, rH))
cv2.imshow('a',img)
# Display the resulting frame
pt1, pt2 = opencv_img(img)
cv2.line(frame, pt1, pt2,(0,0,255) , 2) # 绘制一条蓝线
t1 = time.time()
print("Total time running seconds:" , str(t1 - t0))
cv2.imshow('frame',frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
Hope this works for you!
Thank you for your likes and comments~~~