I am trying to create a 3D point cloud with color information based on captures from multiple azure kinect cameras from different angles. For this I used some of the code from the green screen example, right now I capture color and depth images from 2 cameras. I transform both the depth image from camera 2 and the the depth image from camera 1 to the color "domain" of camera 1. For this I use the "depth_image_to_color_camera" function.
The output of my program right now is: color image camera 1 in color domain camera 1, depth image camera 1 in color domain of camera 1, depth image camera 2 in color domain camera 1, color image camera 2 still in color domain of camera 2.
I use Open3D in python to create 3D point clouds from these files, (one point cloud for each camera), and then overlap them to create a more complete 3D point cloud. However this does not work so well, because the color image of camera 2 is not transformed to the perspective of camera 1. The reason for this is because the SDK does not have a function for this.
jhkjames1110
commented
6 months ago
I am trying to create a 3D point cloud with color information based on captures from multiple azure kinect cameras from different angles. For this I used some of the code from the green screen example, right now I capture color and depth images from 2 cameras. I transform both the depth image from camera 2 and the the depth image from camera 1 to the color "domain" of camera 1. For this I use the "depth_image_to_color_camera" function.
The output of my program right now is: color image camera 1 in color domain camera 1, depth image camera 1 in color domain of camera 1, depth image camera 2 in color domain camera 1, color image camera 2 still in color domain of camera 2.
I use Open3D in python to create 3D point clouds from these files, (one point cloud for each camera), and then overlap them to create a more complete 3D point cloud. However this does not work so well, because the color image of camera 2 is not transformed to the perspective of camera 1. The reason for this is because the SDK does not have a function for this.
Is anyone aware of a solution to this problem?