Actual combat | Low-contrast target extraction implemented by OpenCV (steps + source code)

guide

    This article mainly introduces a case of low-contrast target extraction implemented by OpenCV (steps + source code).

background introduction

    The example comes from the network, the following is the test picture, the goal is to extract the outline of the object:

    The effect is as follows:   

Implementation steps

    Initially, the contrast between the background and the target is not obvious, and it is difficult to distinguish them. But we can still extract the target outline through the traditional Blob analysis method, the steps are as follows:

[1] Convert to grayscale image + Gaussian filter

gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)blur = cv2.GaussianBlur(gray, (11, 11), 7)cv2.imshow('blur', blur)

【2】Canny edge detection​​​​​​

canny = cv2.Canny(blur, 0, 42)cv2.imshow('canny', canny)

【3】Closing operation (connecting the separated contours)​​​​​​

kernel = cv2.getStructuringElement(cv2.MORPH_RECT,(7,7))closing = cv2.morphologyEx(canny, cv2.MORPH_CLOSE, kernel, iterations=2)#闭运算cv2.imshow('closing', closing)

[4] Find the contour + find the maximum area contour (removing interference), you can get the overall contour​​​​​​​​

contours, hierarchies = cv2.findContours(closing, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)cnt = max(contours, key=cv2.contourArea)cv2.drawContours(img, cnt, -1, (0, 0, 255), 2)

[5] Perform an open operation on the result of step [3], and wait to get the ontology part:​​​​​​​​

opening = cv2.morphologyEx(closing, cv2.MORPH_OPEN, kernel, iterations=1)#闭运算cv2.imshow('opening', opening)

[6] Find the contour + find the maximum area contour (removing interference):​​​​​​

contours, hierarchies = cv2.findContours(opening, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)cnt = max(contours, key=cv2.contourArea)cv2.drawContours(img, cnt, -1, (0, 255, 0), 2)

Complete source code:

# 公众号：计算机视觉之家import numpy as npimport cv2
img = cv2.imread('1.jpg')cv2.imshow('src', img)
h,w,c = img.shapeimg = cv2.resize(img, (w // 2, h // 2))
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)blur = cv2.GaussianBlur(gray, (11, 11), 7)cv2.imshow('blur', blur)canny = cv2.Canny(blur, 0, 42)cv2.imshow('canny', canny)
kernel = cv2.getStructuringElement(cv2.MORPH_RECT,(7,7))closing = cv2.morphologyEx(canny, cv2.MORPH_CLOSE, kernel, iterations=2)#闭运算cv2.imshow('closing', closing)
##contours, hierarchies = cv2.findContours(closing, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)##cnt = max(contours, key=cv2.contourArea)##cv2.drawContours(img, cnt, -1, (0, 0, 255), 2)
opening = cv2.morphologyEx(closing, cv2.MORPH_OPEN, kernel, iterations=1)#闭运算cv2.imshow('opening', opening)
contours, hierarchies = cv2.findContours(opening, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)cnt = max(contours, key=cv2.contourArea)cv2.drawContours(img, cnt, -1, (0, 255, 0), 2)
cv2.imshow('result', img)cv2.waitKey(0)cv2.destroyAllWindows()