Opencv Python 如何检查齿轮的方向？

Question

I'm checking the direction of the gear, whether it is "up or down", "and print the result to the screen". The background color is always black.

This gear is in the "up" position:

https://imgur.com/a/DON8GJs

This gear is in the "down" position:

https://imgur.com/a/4ODZQAt

I did try to binary the two images and canny edge detection but all I've found is the result of those algorithms, nothing more.! I'm wondering what should I need to do to check the direction of the gear? Your help would be greatly appreciated.!

Answer 1

The down position has a distinct shape, you can use shapematch to detect the presence of this shape.
To do that, you'll need a reference shape. I created this by detecting the edges, saved that image and used MS paint to only leave the shape needed.
The code below shows you how to detect the shape. It prints the position of the gear to the terminal and draws the shape if it is in the down position.

Shapematch is able to handle rotation, but you might need to test and tweak some of the settings if you want the use this in some sort of automation.

Result:

Image of reference gear: https://i.stack.imgur.com/s6E9C.jpg

Code:

import numpy as np 
import cv2

# load image of reference shape
image_reference = cv2.imread("ReferenceGear.jpg",0)
# threshold to remove artefacts
ret, img_ref = cv2.threshold(image_reference, 200, 255,0)
# detect contours in image
im, ref_cnts, hierarchy = cv2.findContours(img_ref, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
# store the contour of the reference shape in a variable
ref_cnt = ref_cnts[0]


# load image 
image = cv2.imread("gear.png")
# detect edges in image
edges = cv2.Canny(image,50,50)
# detect contours of edges in image
im, contours, hierarchy = cv2.findContours(edges, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
# for each edge-contour in the image: try to match with te reference.
# if the value is very small, it is a good match. store the result in a variable
found = False
for cnt in contours:
    ret = cv2.matchShapes(cnt, ref_cnt,3,0.0)
    if ret < 0.001:
        cv2.drawContours(image, [cnt], 0, (255), 2)
        found = True
        break

if found:
    print("Gear is in down position")
else:
    print("Gear is in up position")

# show image and reference
cv2.imshow("image", image)
cv2.imshow("image_reference", image_reference)
# release resources
cv2.waitKey(0)
cv2.destroyAllWindows()