python - 获取numpy矩阵中正方形的坐标

Question

给定以下 numpy 矩阵

import numpy as np
np_matrix = np.array([[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,2,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,2,0,0,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,1,0,0,0]
                    ,[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,0,0,0]
                    ,[0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
                    ,[0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1]
                    ,[0,0,0,1,0,0,0,2,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,2,0]
                    ,[1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
                    ,[2,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]])

print(np_matrix)
print(np_matrix.shape)

可以表示为图像。

这里有一些关于输入矩阵的假设:

• 矩阵形状可以变化。它不是固定尺寸。
• 方 block 的数量因矩阵而异。但是矩阵中总是至少有一个正方形。
• 每个绿点周围都有一个正方形，正方形内部有一个点。正方形内的点用数字 2 标记
• 正方形在 X 轴上是连续的，正方形的边界永远不会与其他正方形的边界重叠。正方形边框用数字 1 标记
• 空白用数字0标记

我有两个问题。以最有效的方式如何获得包含所有绿点坐标的数组？我怎样才能得到一个包含绿点周围正方形所有角的数组？

这是我为这个例子寻找的结果:

green_dots_coordinates = [
    [0,0],  # dot 1 with coordinates x1, y1 inside square 1
    [7,12], # dot 2 with coordinates x2, y2 inside square 2
    [16,27],# dot 3 with coordinates x3, y3 inside square 3
    [29,21],# dot 4 with coordinates x4, y4 inside square 4
    [34,7], # dot 5 with coordinates x5, y5 inside square 5
]



sqaures_corners_coordinates = [
    #square nr 1
    [
        [0,6], # x1, y1
        [2,6], # x2, y2
        [0,0], # x3, y3
        [2,0], # x4, y4
    ],

    #square nr 2
    [
        [3,14], # x1, y1
        [9,14], # x2, y2
        [3,7], # x3, y3
        [9,7], # x4, y4
    ],

    #square nr 3
    [
        [12,31], # x1, y1
        [23,31], # x2, y2
        [12,24], # x3, y3
        [23,24], # x4, y4
    ],

    #square nr 4
    [
        [25,22], # x1, y1
        [32,22], # x2, y2
        [25,15], # x3, y3
        [32,15], # x4, y4
    ]

    ,#square nr 5
    [
        [33,13], # x1, y1
        [35,13], # x2, y2
        [33,0], # x3, y3
        [35,0], # x4, y4
    ],

]

Answer 1

坦率地说，您的输出格式非常奇怪，需要大量颠倒索引的顺序，(并且使输出对于索引原始数组毫无用处)但这有效:

def find_boxes(np_matrix):
    np_mat = np_matrix[::-1, :]  # reversed in expected output
    def find_extent(arr, val):
        xn = arr.size
        x0 = np.flatnonzero(arr == 1)
        xi = np.searchsorted(x0, val, side = 'right')
        if xi == x0.size:
            x1 = x0[xi-1]
            x2 = xn - 1
        elif xi == 0:
            x1 = 0
            x2 = x0[xi]
        else:
            x1 = x0[xi-1]
            x2 = x0[xi]
        return np.array([x1, x2])

    green = np.where(np_mat == 2)
    green = tuple(g[np.argsort(green[-1])] for g in green)
    coords = np.empty((green[0].size, 2, 4))

    for i, (x, y) in enumerate(zip(*green)):
        coords[i, 0] =   np.tile(find_extent(np_mat[x, :], y),       2)
        coords[i, 1] = np.repeat(find_extent(np_mat[:, y], x)[::-1], 2)  # reversed again
    return np.stack(green)[::-1].T, coords.swapaxes(1,2).astype(int)
    # reversed again and transposed

测试:

find_boxes(np_matrix)
Out: 
(array([[ 0,  0],
        [ 7, 12],
        [16, 27],
        [29, 21],
        [34,  7]], dtype=int32), 
 array([[[ 0,  6],
         [ 2,  6],
         [ 0,  0],
         [ 2,  0]],

        [[ 3, 14],
         [ 9, 14],
         [ 3,  7],
         [ 9,  7]],

        [[12, 31],
         [23, 31],
         [12, 24],
         [23, 24]],

        [[25, 22],
         [32, 22],
         [25, 15],
         [32, 15]],

        [[33, 13],
         [35, 13],
         [33,  0],
         [35,  0]]]))