Opencv tutorial 03: how to track an object in a video

Every frame of the video is a picture , Track an object in the video , Break it down , In fact, it is to find the object in the picture of each frame .

Since you want to find that object , You need to define the target object first , Since it's a picture , Then there are colors , First understand the commonly used 3 Color model .

1. RGB.RGB It is the three primary color light mode , also called RGB Color model or red, green and blue color model , It's an additive model , Will be red 、 green 、 The light of the three primary colors of blue is added in different proportions , To synthesize light of various colors . Every pixel has 3 A value represents the color , This is the most common color model .OpenCV The order in is BGR.

2. grayscale . Simple understanding , It's black and white , Each image Pixel point There can only be A value Color , The range of pixel values is [0~255], Most of the existing mature analysis algorithms are based on gray images .

3. HSV. The parameters of color in this model are hue （H）、 saturation （S） And lightness （V）.HSV It is an intuitive color model for users . We can start with a pure color , That is, specify the hue H, And let V=S=1, Then we can add black and white to it to get the color we need .

The above three models , Use HSV It is more convenient to define a target object .

OpenCV It can be easily from BGR Convert to grayscale , from BGR Convert to HSV：

import cv2 as cv
cv.cvtColor(input_image,  cv.COLOR_BGR2GRAY)
cv.cvtColor(input_image,  cv.COLOR_BGR2HSV)

Now we know how to BGR Image to HSV, We can use it to extract color objects . stay HSV in , Indicates that the color ratio is BGR Easier in color space .

Next , We will track blue objects in the video .

step ：

1. Capture every frame of the video

2. from BGR Convert to HSV Color space

3. We will HSV The image threshold is set to the blue range

4. Extract blue objects individually , We can do whatever we want on this image .

The following is the detailed code ：

import cv2 as cv
import numpy as np

#  Read video from the default camera 
cap = cv.VideoCapture(0)
while True:
    #  Get every frame of the video 
    _, frame = cap.read()
    # BGR  Convert to  HSV
    hsv = cv.cvtColor(frame, cv.COLOR_BGR2HSV)

    #  use  HSV  Define the color range of the target object .
    lower_blue = np.array([110,50,50])
    upper_blue = np.array([130,255,255])
    #  Generate a  HSV  The mask only gets blue objects .
    mask = cv.inRange(hsv, lower_blue, upper_blue)


    # application 
    res = cv.bitwise_and(frame,frame, mask= mask)
    cv.imshow('frame',frame)
    cv.imshow('mask',mask)
    cv.imshow('res',res)
    k = cv.waitKey(5) & 0xFF
    if k == 27:
        break
cv.destroyAllWindows()

The renderings are as follows , You can see a little noise , There are technologies behind it that can eliminate .

Last

If help , Please look at the support .