[opencv learning] [contour detection]

Today, learn the contour detection method

import cv2
import numpy as np


#  Show the image , Encapsulate as a function 
def cv_show_image(name, img):
    cv2.imshow(name, img)
    cv2.waitKey(0)  #  Waiting time , In milliseconds ,0 Represents any key termination 
    cv2.destroyAllWindows()

#  Function of edge detection 
# cv2.findContours(img, mode, method)
# img:  The input image , For high accuracy , It is recommended to use binary image 
# mode: RETR_EXTERNAL( Only the outer contour is detected )、
# RETR_LIST( Detect all contours and keep the results to a List in )
# RETR_CCOMP, Detect all contours , And wrap the results in two layers , The first layer is the external outline of each part , The second layer is the inner boundary of the cavity 
# RETR_TREE,  Detect all contours , And reconstruct the hierarchy of nested contours 
# RETR_LIST  From an explanatory point of view , This should be the simplest . It just extracts all the contours , Without creating any parent-child relationship .
# RETR_EXTERNAL  If you choose this mode , Only the outermost outline will be returned , All sub contours are ignored .
# RETR_CCOMP  In this mode, all contours will be returned and divided into two levels of organizational structure .
# RETR_TREE  In this mode, all contours are returned , And create a complete list of organizational structures . It will even tell you who is Grandpa , Dad , son , Grandson, etc .
# methord:  Method of contour detection 
# CHAIN_APPROX_NONE:  With freeman Chain code output outline ,
# CHAIN_APPROX_SIMPLE:  Keep only the end part , That is, only the end points are saved for horizontal, vertical and inclined lines . Less and more concise data .


img = cv2.imread('images/contour.png')  #  Turn to grayscale 
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret, threshold = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)  #  Convert to binary image 
cv_show_image('binary_src_img', threshold)

# contours, All contour information , It's a list structure 
# hierarchy  It's a hierarchy , It is also a structure that retains all results 
contours, hierarchy = cv2.findContours(threshold, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
print(len(contours))
print(type(contours))

draw_img = img.copy()  #  Copy an image , Keep the original image 
# -1  It means that all the outlines are drawn , The last two parameters are the color and thickness of the outline ,drawContours It will be modified in place on the input image 
res = cv2.drawContours(draw_img, contours, contourIdx=-1, color=(0, 255, 0), thickness=1)
cv_show_image('draw_contours_img', res)


#  Features of contour 
for i in range(len(contours)):
    cont = contours[i]
    print(' The first {} The area of the outline is {}, The circumference is {}', i, cv2.contourArea(cont), cv2.arcLength(cont, True))  # true Closed 


#  The outline is approximate 
#  There is a curve from point A point-to-point B, There is a connection AB, There is a point on the curve C,C The straight line AB Distance of d Maximum . If d <=  A certain threshold , Then we can use AB Straight lines directly replace curves AB
#  If d >  A certain threshold , So you can't use another curve AB Straight line substituted , Then at this time, we have to put the curve AB Continue to divide into two sections , They are curves AC Sum curve CB. Try further 
#  Try further , See if this curve can be straight AC And straight lines CB Instead of . This is a divide and conquer / Recursively solve .
draw_img = img.copy()
for i in range(len(contours)):
    cont = contours[i]
    epsilon = 0.03 * cv2.arcLength(cont, True)
    approx = cv2.approxPolyDP(cont, epsilon, True)  #  Find the approximation of the contour ,True It means closed 

    #  Draw an approximate outline , Pass it on to drawContours The contour type of must be list
    draw_img = cv2.drawContours(draw_img, [approx], contourIdx=-1, color=(0, 255, 0), thickness=1)

cv_show_image('approx_contours_img', draw_img)



#  Outer rectangle 
draw_img = img.copy()
for i in range(len(contours)):
    cont = contours[i]
    x,y,w,h = cv2.boundingRect(cont)  #  According to the contour, calculate the rectangular coordinates with the farthest circumscribed .

    draw_img = cv2.rectangle(draw_img, pt1=(x,y), pt2=(x+w, y+h), color=(0, 255, 0), thickness=2)  #  Draw this rectangle , Will draw on the original picture 

cv_show_image('rectangle_contours_img', draw_img)



#  Circumcircle 
draw_img = img.copy()
for i in range(len(contours)):
    cont = contours[i]
    (x,y), radius = cv2.minEnclosingCircle(cont)  #  Calculate the center and radius of the outermost circle according to the contour .
    center = (int(x), int(y))
    radius = int(radius)

    draw_img = cv2.circle(draw_img, center=center, radius=radius, color=(0, 255, 0), thickness=2)  #  Draw this rectangle , Will draw on the original picture 

cv_show_image('circle_contours_img', draw_img)

The original image is as follows ：

Draw all the outlines ：
The outline has internal outline and external outline , If you use mode = RETR_EXTERNAL Can only draw the outline

Draw all approximate outlines ：

Draw the circumscribed largest rectangle of all contours

Draw the circumscribed maximum circle of all contours