OpenCV python findContours 错误的结果

Question

我正在尝试使用 python 和 OpenCV 查找图像中白色区域的坐标。这应该是一个简单的任务，使用 erode => threshold => findContours.

这是我的代码：

th_er = cv2.erode(th, np.ones((15, 15), np.uint8))
th_er = cv2.bitwise_not(th_er)

contours, _ = cv2.findContours(th_er, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
for cntr in contours:
    x, y, w, h = cv2.boundingRect(cntr)
    cv2.rectangle(th_er, (x, y), (x + w, y + h), (100, 100, 100), 5)

cv2.imshow('il', th_er)
cv2.waitKey()

我的问题是“findContours”返回奇怪的结果，如图中所示 here。

所以，有人遇到过这种行为或知道任何可能的解决方法吗？

here为原图

Answer 1

img = cv2.imread('try.jpg', 0) # (200, 1427)
img2 = cv2.imread('try.jpg', -1) # (200, 1427, 4)

# gray = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY) # <-- you can use this to convert into grayscale image and then feed it to cv2.erode(img2, .....)
th_er = cv2.erode(img, np.ones((15, 15), np.uint8))
th_er = cv2.bitwise_not(th_er)

contours, _ = cv2.findContours(th_er, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
for cntr in contours:
    x, y, w, h = cv2.boundingRect(cntr)
    cv2.rectangle(img2, (x, y), (x + w, y + h), (200, 100, 100), 5)
plt.figure(figsize=(15,20))

plt.imshow(img2)
plt.show()

编辑：

现在可以正常使用了。

th_er1 = 255-cv2.bitwise_not(th_er) 正如我所说，物体应该是白色的，背景应该是黑色的。反之亦然。通过减去 255，它现在是正确的格式。

# img = cv2.imread('try.png', 0) # (200, 1427)
img = cv2.imread('try.png') # (200, 1427, 4)

gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # <-- you can use this to convert into grayscale image and then feed it to cv2.erode(img2, .....)
_, th = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY+ cv2.THRESH_OTSU) 
th_er = cv2.erode(th, np.ones((15, 15), np.uint8))
th_er1 = 255-cv2.bitwise_not(th_er) # <----- here

contours, _ = cv2.findContours(th_er1, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
for cntr in contours:
    x, y, w, h = cv2.boundingRect(cntr)
    cv2.rectangle(img, (x, y), (x + w, y + h), (200, 100, 100), 5)
plt.figure(figsize=(15,20))

plt.imshow(img)
plt.show()

OpenCV python findContours 错误的结果

OpenCV python findContours wrong results

python

opencv-python