1. Introduction
To start exploring OpenCV with Python, we’ll start by a simple exercise : Detect a cardboard box on a industrial conveyor (from a average picture took with a smartphone) :
As as first step, we’ll identity the contour of the box and define a “perfect” rotated rectangle of the minimum area enclosing our box. The result will be this :
2. Code details
Load the libraries :
|
1 2 |
import numpy as np import cv2 |
Read the picture (here the picture “box-1.jpg is in the same directory) :
|
1 |
img = cv2.imread('box-1.jpg') |
Blur the picture :
|
1 |
blurred = cv2.GaussianBlur(img, (5, 5), 0) |
Convert to gray scale :
|
1 |
gray = cv2.cvtColor(blurred,cv2.COLOR_BGR2GRAY) |
Calculate the threshold level and we apply it :
|
1 2 |
threshold = cv2.adaptiveThreshold(gray,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,11,2) retval, bin = cv2.threshold(gray, 100, 255, cv2.THRESH_BINARY) |
Detect contours :
|
1 |
bin, contours, hierarchy = cv2.findContours(bin, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE) |
If we draw all the contours :
|
1 |
cv2.drawContours(img, contours, -1, (0, 255, 0), 3 ) |
The result is :
That’s why, to do it quick, we sort out the biggest and convex (close) contour:
|
1 2 3 4 5 6 7 8 9 |
max_area = 0 max_index = -1 index = -1 for i in contours: area = cv2.contourArea(i) index=index+1 if area > max_area : max_area = area max_index = index |
And then we draw it :
|
1 |
cv2.drawContours(img, contours, max_index, (0, 255, 0), 3 ) |
Now we just have the contour that we wanted, but it’s not a rectangle !
Let’s draw a rotated rectangle of the minimum area enclosing our box, in red :
|
1 2 3 4 5 6 7 8 9 10 11 12 13 |
" rect = cv2.minAreaRect(cnt) box = cv2.boxPoints(rect) box = np.int0(box) img = cv2.drawContours(img,[box],0,(0,0,255),2) </pre> <a href= http://www.fontenay-ronan.fr/wp-content/uploads/2015/06/box-1-result.jpg"] cnt=contours[max_index] rect = cv2.minAreaRect(cnt) box = cv2.boxPoints(rect) box = np.int0(box) img = cv2.drawContours(img,[box],0,(0,0,255),2) |
And to finish, just the code to show a picture in a good size windows in Python :
|
1 2 3 4 5 6 7 8 9 10 11 |
cv2.namedWindow('image', cv2.WINDOW_NORMAL) cv2.imshow('image', img) cv2.waitKey(0) cv2.destroyAllWindows() <pre class="lang:python> <pre class="lang:python> cv2.namedWindow('image', cv2.WINDOW_NORMAL) cv2.imshow('image', img) cv2.waitKey(0) cv2.destroyAllWindows() |
3. Code overview
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 |
rect = cv2.minAreaRect(cnt) box = cv2.boxPoints(rect) box = np.int0(box) img = cv2.drawContours(img,[box],0,(0,0,255),2) #Show original picture with contour cv2.namedWindow('image', cv2.WINDOW_NORMAL) cv2.imshow('image', img) cv2.waitKey(0) cv2.destroyAllWindows() </pre] import numpy as np import cv2 #Read image img = cv2.imread('box-1.jpg') #Gaussian blur blurred = cv2.GaussianBlur(img, (5, 5), 0) #Convert to graysscale gray = cv2.cvtColor(blurred,cv2.COLOR_BGR2GRAY) #Autocalculate the thresholding level threshold = cv2.adaptiveThreshold(gray,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,11,2) #Threshold retval, bin = cv2.threshold(gray, 100, 255, cv2.THRESH_BINARY) #Find contours bin, contours, hierarchy = cv2.findContours(bin, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE) #Sort out the biggest contour (biggest area) max_area = 0 max_index = -1 index = -1 for i in contours: area = cv2.contourArea(i) index=index+1 if area > max_area : max_area = area max_index = index #Draw the raw contours cv2.drawContours(img, contours, max_index, (0, 255, 0), 3 ) cv2.imwrite("box-1-biggest-contour.png", img) #Draw a rotated rectangle of the minimum area enclosing our box (red) cnt=contours[max_index] rect = cv2.minAreaRect(cnt) box = cv2.boxPoints(rect) box = np.int0(box) img = cv2.drawContours(img,[box],0,(0,0,255),2) #Show original picture with contour cv2.namedWindow('image', cv2.WINDOW_NORMAL) cv2.imshow('image', img) cv2.waitKey(0) cv2.destroyAllWindows() |
