Learning about opencv (3)

This time, I will study contour detection and template matching , Because of some problems , It takes a lot of time .（ It may be the problem of image selection ）

Contour detection ：

mode: Contour retrieval mode 
RETR_EXTERNAL: Retrieve only the outermost contour 
RETR_LIST: Retrieve all the contours , And save it in a linked list 
RETR_CCOMP: Retrieve all the contours , And organize them into two layers ： The top layer is the outer boundary of each part , The second layer is the boundary of the void 
RETR_TREE: Check all contours , And reconstruct the entire hierarchy of nested profiles 

method Contour approximation method 
CHAIN_APPROX_NONE: With Freeman Chain code output outline , All other methods output polygons （ The sequence of vertices ）
CHAIN_APPROX_SIMPLE: Compressed horizontal , The vertical and oblique parts , Functions keep only the end part of them 

The first step is to introduce images ：（ For higher accuracy , Use binary images ） 

img=cv2.imread('people.jpg')
print(img.shape)
gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)
print(thresh.shape)
binary,contours,hierarchy=cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

Next, draw the border ：

# Pass in the drawing image , outline , Outline index , Color mode , Line thickness 
# Be careful copy, Or the original picture will change 
draw_ = img.copy()
res=cv2.drawContours(img,contours,-1,(0,0,255),2)      #2 Represents the line width ,-1 Represent all 
cv_show('b',res)

res =cv2.drawContours(draw_,contours,0,(0,0,255),2)
cv_show('c',res)
# Contour feature 
cnt = contours[0]
# area 
s=cv2.contourArea(cnt)
# Perimeter ,Ture It means closed 
c=cv2.arcLength(cnt,True)

# The outline is approximate ：
epsilon = 0.1*cv2.arcLength(cnt,True)       # You can also set it yourself , Usually the percentage of installation perimeter 
approx = cv2.approxPolyDP(cnt,epsilon,True)

drwaContours（ Images , Outline index ,-1（ Represents the whole outline , Other numbers represent the corresponding contour ）, Color , The thickness of the line ）

This is the complete code of the outline ：

img=cv2.imread('people.jpg')
gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)
binary,contours,hierarchy=cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

cnt = contours[0]
s=cv2.contourArea(cnt)
c=cv2.arcLength(cnt,True)

draw_ = img.copy()
res=cv2.drawContours(img,[cnt],-1,(0,0,255),2)


# Approximate image contour 
epsilon = 0.1*c
approx = cv2.approxPolyDP(cnt,epsilon,True)
draw_ = img.copy()
res=cv2.drawContours(draw_,[approx],-1,(0,0,255),2)

Border rectangle ：

Frame the contour on the basis of contour detection ：

# Border rectangle 
img=cv2.imread('people.jpg')
gray=cv2.cvtColor(img,cv2.COLOR_BGR2RGBA)
ret,thresh = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)
cnt = contours[0]

x,y,w,h=cv2.boundingRect(cnt)         #B  G  R
img=cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)   #2 Represents the line width 
cv_show('e',img)

  Ratio of contour area to boundary rectangle ：（ At present, it is not clear what use it has ）

rect_ =w * h
area = cv2.contourArea(cnt)
extent = float(area)/rect_        # Ratio of contour area to boundary rectangle 
print(extent)

Learning outline cv2.findContours(img.mode,method) Errors are always reported here , Some versions have two values , There are three values , It took me a long time to solve this problem .

Next is template matching , Single matching has always had problems in the display of results （ Multiple matches can be used instead of a single match ）

 TM_SQDIFF:  The square is different , The smaller the calculated value is , The more relevant  TM_CCORR:  Calculate correlation , The larger the calculated value , The more relevant  TM_CCOEFF: Calculate the correlation coefficient , The larger the calculated value , The more relevant   Normalization is more reliable  TM_SQDIFF_NORMED : Calculating the normalized square is different , The closer the calculated value is to 0, The more relevant  TM_CCORR_NORMED : Calculate the normalized correlation , The closer the calculated value is to 1, The more relevant  TM_CCOEFF_NORMED: Calculate the normalized correlation coefficient , The closer the calculated value is to 1, The more relevant 

altogether 6 function , Below 3 This function is more reliable .

methods =['cv2.TM_SQDIFF','cv2.TM_CCORR','cv2.TM_CCOEFF','cv2.TM_SQDIFF_NORMED','cv2.TM_CCORR_NORMED','cv2.TM_CCOEFF_NORMED']
img = cv2.imread('people.jpg',0)
print(img.shape)
img1 = cv2.imread('people1.jpg',0)
cv_show('a',img1)
print(img1.shape)

for meth in methods:
    img2=img.copy()
    method=eval(meth)
    print(method)


    res=cv2.matchTemplate(img,img1,method)      # Small value 
    cv_show('a',res)
    print(res.shape)        # result =A-a+1,B-b+1
    min_val ,max_val , min_joc,max_joc= cv2.minMaxLoc(res)
    # minimum value , Maximum , Minimum coordinate position , Maximum coordinate position 




    #cv2.TM_SQDIFF  cv2.TM_SQDIFF_NORMED  Minimum value 
    if method in [cv2.TM_SQDIFF,cv2.TM_SQDIFF_NORMED]:
        top = min_joc
    else:
        top = max_joc
    tmp = cv2.imread('people1.jpg')
    h, w = tmp.shape[:2]
    bottom = (top[0]+w,top[1]+h)
    print('bottom')
    print(bottom)

    # Draw a rectangular 
    cv2.rectangle(img1,top,bottom,255,2)
    plt.subplots(221)              # The first parameter represents the number of rows in the subgraph ; The second parameter represents the number of columns in the row of images ;  The third parameter represents the number of images in each row .
    plt.imshow(res,cmap='gray')
    plt.xticks([]),plt.yticks([])# Hide axis 
    plt.subplots(222)
    plt.imshow(img2,cmap='gray')
    plt.xticks([]),plt.yticks([])
    plt.suptitle(meth)
    plt.show()

This is going to be 6 Put the functions together , You can observe their differences .（ I don't know why there has been a problem here , So I will replace a single match with multiple matches ）

img = cv2.imread('people.jpg')
gary=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
tmp = cv2.imread('people1.jpg')
h,w=tmp.shape[:2]

res =cv2.matchTemplate(img,tmp,cv2.TM_CCOEFF_NORMED)
threshold =0.8
loc =np.where(res>=threshold)
for pt in zip(*loc[::-1]):    #* Represents an optional parameter 
    bottom =(pt[0]+w,pt[1]+h)
    cv2.rectangle(img,pt,bottom,(0,0,255),2)
cv2.imshow('sss',img)
cv2.waitKey(0)

Next, I continue to look for the cause of the problem ,( It is also possible to match only multiple results )  Complete the goals set for yourself in the summer vacation at your own pace .