点云算法-点云投影到平面

目录

一、原理推导

二、代码实现

 三、截图

---

一、原理推导

点云是在一个三维空间中由一堆无序的点所组成，那么在三维中理所应该可以在任意的二维平面：Ax + By + Cz + D = 0 中进行投影，获得不同平面中的二维视图。

下面讲给出基础知识与手动推导过程，并使用代码进行验证推导结果。

感谢网络大佬的实现，让我可以偷懒cv

import numpy as np
import mayavi.mlab as mlab

# 参数说明：
# - points：点云数据
# - flat：3d平面的参数，Ax+By+Cz+D=0，参数即为(A,B,C,D)
#         xy平面：(0 0 1 0) | xz平面：(0 1 0 0) | yz平面：(1 0 0 0)
def project_2d(points, flat):

    # 点云投影平面
    A, B, C, D = flat
    distance = A**2 + B**2 + C**2
    t = -(A*points[:, 0] + B*points[:, 1] + C*points[:, 2] + D)/distance
    x = A*t + points[:, 0]
    y = B*t + points[:, 1]
    z = C*t + points[:, 2]
    project_point = np.array([x, y, z]).T
    print(project_point.shape)

    # 投影展现
    mlab.figure(bgcolor=(0, 0, 0), size=(640, 640))
    mlab.points3d(x, y, z, y, mode='point', colormap='spectral')
    mlab.show()


if __name__ == '__main__':
    txt_path = r"E:\airplane_0001.txt"
    points = np.loadtxt(txt_path, delimiter=',')
    project_2d(points, (0, 0, 1, 0))

二、代码实现

"""
代码功能： 将3d点云投影到xyz等不同平面上
"""
import numpy as np
#import mayavi.mlab as mlab
import open3d as o3d
import matplotlib
import matplotlib.pyplot as plt

# 参数说明：
# - points：点云数据
# - flat：3d平面的参数，Ax+By+Cz+D=0，参数即为(A,B,C,D)
#         xy平面：(0 0 1 0) | xz平面：(0 1 0 0) | yz平面：(1 0 0 0)
def project_2d(points, flat):

    # 点云投影平面
    A, B, C, D = (0, 0, 1, 0)
    distance = A**2 + B**2 + C**2
    t = -(A*points[:, 0] + B*points[:, 1] + C*points[:, 2] + D)/distance
    x = A*t + points[:, 0]
    y = B*t + points[:, 1]
    z = C*t + points[:, 2]
    project_point = np.array([x, y, z]).T
    print(project_point.shape)
    #print(project_point)

    A1, B1, C1, D1 = (0, 1, 0, 0)
    distance1 = A1**2 + B1**2 + C1**2
    t1 = -(A1*points[:, 0] + B1*points[:, 1] + C1*points[:, 2] + D1)/distance1
    x1 = A1*t1 + points[:, 0]
    y1 = B1*t1 + points[:, 1]
    z1 = C1*t1 + points[:, 2]
    project_point1 = np.array([x1, y1, z1]).T
    print(project_point1.shape)
    #print(project_point1)

    A2, B2, C2, D2 = (0, 1, 0, 0)
    distance2 = A2**2 + B2**2 + C2**2
    t2 = -(A2*points[:, 0] + B2*points[:, 1] + C2*points[:, 2] + D2)/distance2
    x2 = A2*t2 + points[:, 0]
    y2 = B2*t2 + points[:, 1]
    z2 = C2*t2 + points[:, 2]
    project_point2 = np.array([x2, y2, z2]).T
    print(project_point2.shape)


    c = np.vstack((project_point1,project_point1,project_point2)) 
    print(c.shape)
    print(C)
    # 投影展现
    fig = plt.figure(figsize=(8, 4))
    ax = fig.add_subplot(121,projection='3d')
    ax.scatter(c.T[0], c.T[1],c.T[2], c='r')
    ax = fig.add_subplot(122)
    ax.scatter(c.T[0], c.T[2], c='r')
    #plt.savefig(savepath)
    plt.show()



if __name__ == '__main__':

    pcdpath = "pcd_data_coordinate_transform/3-1-0.pcd"
    points = o3d.io.read_point_cloud(pcdpath)
    print(points)#输出点云的个数
    #print(np.asarray(point.points))#输出点的三维坐标
    pointcloud = np.asarray(points.points)
    #xy 0 0 1
    #xz 0 1 0
    #yz 1 0 0

    project_2d(pointcloud, (0, 0, 1, 0))

    

 三、截图

备注：另外的实现

(17条消息) 3D Vision--将点云投影至平面_诺有缸的高飞鸟的博客-CSDN博客_点云投影到平面