Object detection with OpenCV

1. Pre knowledge points

（1）Opencv Global binarization and local binarization

https://mydreamambitious.blog.csdn.net/article/details/125249121

（2）Opencv Medium erode and dilate

https://mydreamambitious.blog.csdn.net/article/details/125265431

（3）Opencv Medium findContours（） and drawContours（）

https://mydreamambitious.blog.csdn.net/article/details/125339604

（4）Opencv Draw a straight line in , rectangular , round , The ellipse , polygon （ Include polygon fill ）, Draw text

https://mydreamambitious.blog.csdn.net/article/details/125392536

（5）Opencv Calculate the perimeter and area of the contour in

https://mydreamambitious.blog.csdn.net/article/details/125340081

（6）Opencv Basic knowledge points in

https://mydreamambitious.blog.csdn.net/article/details/125351256

---

2. Code combat

（1） First, the read frame is scaled , Extracting regions of interest and binarization

frame=cv2.resize(src=frame,dsize=(750,500))
    # Extract the region of interest 
    roi=frame[200:400,450:700]
    mask=bgsegment.apply(roi)

    # Remove the video mask Medium gray shadow 
    OK,mask=cv2.threshold(src=mask,thresh=245,maxval=255,type=cv2.THRESH_BINARY)

（2） Find outline and draw rectangle

# Find the outline 
    binary,contours,hierarchy=cv2.findContours(image=mask,mode=cv2.RETR_TREE,method=cv2.CHAIN_APPROX_SIMPLE)

    # Record the position of the car 
    box=[]

    # Perform traversal filtering 
    for step,cont in enumerate(contours):
        # Calculated area 
        area=cv2.contourArea(cont)
        if area>=150:
            # cv2.drawContours(image=roi,contours=[cont],contourIdx=-1,
            # color=(0,255,0),thickness=2)
            # Get the coordinates, height and width of the object in the video 
            (x,y,w,h)=cv2.boundingRect(cont)
            # # Drawing a matrix （ Frame column object ）
            cv2.rectangle(img=roi,pt1=(x,y),pt2=(x+w,y+h),color=(0,255,0),thickness=2)

（3） Draw the number of frames read by the video

 #  Show FPS
    endtime = time.time()
    FPS = 1 / (endtime - startime)
    cv2.putText(img=frame, text='FPS: ' + str(int(FPS)), org=(30, 450),
                fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=1.0,
                color=(0, 255, 0), thickness=2)

（4） The overall code

import os
import cv2
import time
import numpy as np

# frame=np.zeros(shape=(450,450,3),dtype=np.uint8)
cap=cv2.VideoCapture('Video/DJ_1.mp4')
# Remove the background 
bgsegment=cv2.createBackgroundSubtractorMOG2(history=100,varThreshold=40)

# Instance tracker 
# trackers=cv2.MultiTracker_create()
# # The instance tracking algorithm used 
# track_algorithm=cv2.TrackerKCF_create

# Track the overall situation of the car ID
class_id=0

while cap.isOpened():
    startime=time.time()
    OK,frame=cap.read()
    if OK==False:
        break
    frame=cv2.resize(src=frame,dsize=(750,500))
    # Extract the region of interest 
    roi=frame[200:400,450:700]
    mask=bgsegment.apply(roi)

    # Remove the video mask Medium gray shadow 
    OK,mask=cv2.threshold(src=mask,thresh=245,maxval=255,type=cv2.THRESH_BINARY)
    # Do the open operation , take mask Remove all small objects in the 
    # kernel=cv2.getStructuringElement(shape=cv2.MORPH_RECT,ksize=(3,3))
    # mask=cv2.morphologyEx(src=mask,op=cv2.MORPH_OPEN,kernel=kernel,iterations=1)
    # Find the outline 
    binary,contours,hierarchy=cv2.findContours(image=mask,mode=cv2.RETR_TREE,method=cv2.CHAIN_APPROX_SIMPLE)

    # Record the position of the car 
    box=[]

    # Perform traversal filtering 
    for step,cont in enumerate(contours):
        # Calculated area 
        area=cv2.contourArea(cont)
        if area>=150:
            # cv2.drawContours(image=roi,contours=[cont],contourIdx=-1,
            # color=(0,255,0),thickness=2)
            # Get the coordinates, height and width of the object in the video 
            (x,y,w,h)=cv2.boundingRect(cont)
            # # Drawing a matrix （ Frame column object ）
            cv2.rectangle(img=roi,pt1=(x,y),pt2=(x+w,y+h),color=(0,255,0),thickness=2)

    #  Show FPS
    endtime = time.time()
    FPS = 1 / (endtime - startime)
    cv2.putText(img=frame, text='FPS: ' + str(int(FPS)), org=(30, 450),
                fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=1.0,
                color=(0, 255, 0), thickness=2)

    cv2.imshow('image', frame)
    cv2.imshow('roi', roi)
    cv2.imshow('mask', mask)

    key=cv2.waitKey(30)
    if key==27:
        break


cap.release()
cv2.destroyAllWindows()

if __name__ == '__main__':
    print('Pycharm')

Indicate the ： When the reader is extracting the region of interest , You can use your own video to , Because I extracted it according to the moving position of the car in my video .