机器人持续学习基准LIBERO系列7——计算并可视化点云

发布时间:2024年01月16日

0.前置

1.前置代码

2.重新获取真实深度信息

  • 之前的由于要显示,进行了整数化处理,所以重新获取一下原始真实深度信息
from robosuite.utils.camera_utils import get_real_depth_map
agentview_depth_real = get_real_depth_map(env.sim, agentview_depth)

3.获取图像尺寸

h,w = env_args['camera_heights'],  env_args['camera_widths']

4.创建像素点序列和颜色序列

i = np.zeros([h*w,2])#(点数,像素点二维坐标)
colors = np.zeros([h*w,3])#(点数,像素点对应的RGB值)
for x in range(h):
    for y in range(w):
        i[x*h+y] = [x,y]
        colors[x*h+y] = agentview_image[x,y]

5.获取相机内外参

from robosuite.utils.camera_utils import get_camera_extrinsic_matrix,get_camera_intrinsic_matrix

camera_intrinsic_matrix_ = np.linalg.inv(get_camera_intrinsic_matrix(env.sim,'agentview', env_args['camera_heights'],  env_args['camera_widths']))
camera_extrinsic_matrix_ = np.linalg.inv(get_camera_extrinsic_matrix(env.sim,'agentview'))

6.计算世界坐标系下三维点坐标

points = np.zeros([i.shape[0],3])
for num,p in enumerate(i):
    p_ = (camera_intrinsic_matrix_@np.array([[p[0],p[1],1]]).T).T
    p_[0,2] = agentview_depth_real[int(p[0]),int(p[1])]
    p_ = (camera_extrinsic_matrix_@np.array([p_[0,0],p_[0,1],p_[0,2],1]).T).T
    points[num] = p_[:-1]
print(points)

7.关闭环境

env.close()
  • 不关闭环境,就是用open3d显示的话,会报错
X Error of failed request:  BadAccess (attempt to access private resource denied)
  Major opcode of failed request:  152 (GLX)
  Minor opcode of failed request:  5 (X_GLXMakeCurrent)
  Serial number of failed request:  183
  Current serial number in output stream:  183

8.open3d显示点云

import open3d as o3d
pcd_show = o3d.geometry.PointCloud()
pcd_show.points = o3d.utility.Vector3dVector(points[:, :3])
pcd_show.colors = o3d.utility.Vector3dVector(colors[:]/255)
o3d.visualization.draw_geometries([pcd_show])

在这里插入图片描述
在这里插入图片描述

文章来源:https://blog.csdn.net/weixin_53610475/article/details/135610636
本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。