为什么 HoughLinesP 输出的是 3D 数组，而不是 2D 数组？

Question

我在图像上使用函数 (Python、Numpy、OpenCV3)，这是我的示例输出 -

[[[539 340 897 538]]

 [[533 340 877 538]]

 [[280 460 346 410]]

 [[292 462 353 411]]

 [[540 343 798 492]]]

它的大小是 (5,1,4)

我想了解函数输出类似 (5,2,4) 或 (5,3,4) 的情况。但是我想不出我现在处理过的所有图像，它是一个列数为 1 的 3D 数组。

难道仅仅一个二维数组就足够了吗？也许效率更高？

Answer 1

如果你不想要额外的维度，因为它是 1，只需使用挤压

>>> a = np.arange(5*4).reshape(5,1,4)
>>> a
array([[[ 0,  1,  2,  3]],

       [[ 4,  5,  6,  7]],

       [[ 8,  9, 10, 11]],

       [[12, 13, 14, 15]],

       [[16, 17, 18, 19]]])
>>> a.squeeze()
array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8,  9, 10, 11],
       [12, 13, 14, 15],
       [16, 17, 18, 19]])

有时额外的轴会派上用场

a.swapaxes(1,2)
array([[[ 0],
        [ 1],
        [ 2],
        [ 3]],
     ... Snip
       [[16],
        [17],
        [18],
        [19]]])

鉴于更新并假设需要第一个 return 相同的结果假设不需要额外的维度。

>>> a[0].squeeze()
array([0, 1, 2, 3])
>>> a.squeeze()[0]
array([0, 1, 2, 3])

Answer 2

我在 OpenCV Q&A 上询问并得到了以下回复 -

opencv is a c++ library, and the python wrappers are auto-generated from some scripts, so in c++ we have:

vector lines; to hold the hough results.

now unfortunately , Vec4i is a descendant of Matx , which is actually a 2d thing, so in python you get:

[ #one for the vector

[ #one for the 1st dim of Vec4i (1, pretty useless, admittedly :)

[ #one for the 2nd dim of Vec4i (4 elements)

again, i think, you'll just have to live with it.

为什么 HoughLinesP 输出的是 3D 数组，而不是 2D 数组？

Why does the HoughLinesP output a 3D array, instead of a 2D array?

opencv

numpy

hough-transform

python-3.x

opencv3.1