[SOLVED] Error with cv2.minAreaRect , gives me Error related to convex hull function which i haven't used

Error with cv2.minAreaRect , gives me Error related to convex hull function which i haven't used

In the code below I have tried to find out if an object is green in colour and draw a contour around it. With that information I would also like to find the corners of the shape and crop it, but I get an error related to the ConvexHull function which I have not used in the code. I have used python 3 and Opencv 3.2.0

The code that I have written gives me this error

Traceback (most recent call last):
  File "/home/pi/Desktop/SelfDrivingCar/code/picamsense2.py", line 29, in <module>
    points = cv2.minAreaRect(contour)
error: /home/pi/opencv/opencv-3.2.0/modules/imgproc/src/convhull.cpp:136: error: (-215) total >= 0 && (depth == CV_32F || depth == CV_32S) in function convexHull

Here is my code

from picamera import PiCamera
import time
import cv2
import numpy as np 
camera = PiCamera()
camera.start_preview()
time.sleep(5)
camera.capture('/home/pi/Desktop/piImage/image.jpg')
camera.stop_preview()
img = cv2.imread('/home/pi/Desktop/piImage/image.jpg', 1)
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
lower = np.array([46, 100, 100])
upper = np.array([86, 255, 255])
mask = cv2.inRange (hsv, lower, upper)
blur = cv2.GaussianBlur(mask, (7,5),0 )
erosion =cv2.erode(blur, (5,5), iterations = 3)
contour =np.array( cv2.findContours(erosion, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)[1])
cv2.drawContours(img, contour, -1, (0,255,0), 3)
for cnt in contour:
    approx = cv2.approxPolyDP(cnt, 0.04*cv2.arcLength(cnt, True), True)
    if len (approx)== 3:
        print "triangle"
    elif len (approx) == 4 :
        print "quadrilateral"
    elif len(approx) > 4:
        print "circle"        
points = cv2.minAreaRect(contour)
points = cv2.boxPoints(points)
points = np.int0(points)
for p in points :
    pt = (p[0],p[1])
    print pt
print points
cv2.imshow('eroded', blur)
cv2.imshow('original', img)
while (1):
    k = cv2.waitKey(0)
    if(k == 27):
     break
cv2.destroyAllWindows()

I have been trying to look for a fix for quiet some time now

Any help would be appreciated

thanks in advance

-Sam

Solution

The mistake you are making is that you are passing an array of contours to a function that expects a contour.

Here is a piece of code as illustration. It computes the bounding boxes of all the separate contours in the contour array and stores them in a list.

rectangles = []
for cont in contour:
    rectangles.append(cv2.minAreaRect(cont))