3D scanning of objects max size 2x2x2cm at a distance 5 to 10cm and produce memory accessible pointcloud
Hello,
I would like to scan objects as small as 2x2x2 cm maximum size at a distance of 5-10cm distance. The D405 seems to be suitable. Here are my questions:
1. What is the accuracy in x, y, and z dimensions?
2. Is it possible to take a color picture of the object and superimpose it on the point cloud?
3. How long does it take to obtain the point cloud of the scanned object and download it from the scanner for post-processing by an external program?
4. Is it possible to define ROI of the field of view to reduce download and post-processing time to an external control program?
5. Does the SDK provide means to display the 3D point cloud/heat map on a 3D window to rotate, and zoom in/out the image?
6. Can the scanner be triggered to scan through an external program?
7. Is it possible to control the optical parameters of the scanner and dynamically change them through an external program?
Thank you.
Cem
-
Hi Girit Whilst the default minimum depth sensing range of the D405 camera model is 7 cm, it can be reduced as low as 4 cm using a setting called Disparity Shift.
In regards to your questions:
1. The official accuracy rating of the D405 is less than 2% error at 50 cm distance. D405 provides high quality, high accuracy images at very close range.
2. Yes, it is simple to generate a textured pointcloud where the color data is mapped onto the cloud's depth points.
3. Capture of pointcloud frames should be almost instantaneous. Download of single frames of pointcloud data will also be rapid. If the saving of single frames was automated then it would likely be completed in a second.
This can be tested by using the free RealSense Viewer tool to generate a pointcloud in its 3D pointcloud mode and then using the Viewer's export controls to save a frame as a .ply format file that can be imported into an external program.
4. In terms of RealSense, Region of Interest (ROI) refers to defining a section of the image whose pixels will be maintained at a preset intensity that can be user-defined.
A range of pre-made post-processing filters are provided, such as the Decimation Filter which reduces the complexity of the depth image by downsampling its resolution.
5. Yes, you can rotate a 3D pointcloud with the mouse. The SDK provides example programs for doing so, such as rs-pointcloud
https://github.com/IntelRealSense/librealsense/tree/master/examples/pointcloud
6. The SDK supports hardware sync to initiate capture via an external trigger pulse. However, D405 is not equipped with the hardware sync pins to use this feature.
It is possible to access a RealSense camera with more than one application if that application follows the SDK's Multi-Streaming Model rules.
7. It is possible to change parameters that affect the image during runtime, though resolution and frames per second (FPS) speed cannot be changed once a program is started. It is possible to stop a program momentarily, change the resolution and FPS and restart the program.
Please sign in to leave a comment.
Comments
1 comment