Background extraction from the object in python

[Pratiksha2692]

---

Required Info
Camera Model	{ D435 }
Firmware Version	(05.10.06.00)
Operating System & Version	Linux (Ubuntu 16.04)
Platform	PC
SDK Version	{ legacy / 2.. }
Language	{python }
Segment	{3d reconstruction}

Issue Description

I have to extract the object from its background. I had put the realsense camera at the corner of platform. RGB image is attached as below.

I have to do 3d reconstruction of model. I am using opencv_pointcloud_viewer.py code to capture point cloud. How can I extract model from the background in the point cloud?

Kindly advise.

Thanks in advance.

[Pratiksha2692]

I have applied depth clamp min and depth clamp max in opencv_pointcloud_viewer.py.

device = rs.context().query_devices()[0]
advnc_mode = rs.rs400_advanced_mode(device)
depth_table_control_group = advnc_mode.get_depth_table()
depth_table_control_group.disparityShift = 0
depth_table_control_group.depthClampMin = 0
depth_table_control_group.depthClampMax = 1000
advnc_mode.set_depth_table(depth_table_control_group)
print("Depth Table: \n", advnc_mode.get_depth_table())

Above snippet help me to give distance threshold. Now I have to extract the platform from the foot model. Is there any API for background extraction from PLY ? or any color based segmentation from ply file?

Thanks a lot.

[Pratiksha2692]

Any suggestion from Realsense engineering team?

[RealSenseCustomerSupport]

---

Hello @Pratiksha2692,

Sorry for the delay.

Have you taken a look at the documentation of API?

https://intelrealsense.github.io/librealsense/doxygen/pages.html

Also here is some info regarding OpenCV background subtraction.

https://docs.opencv.org/3.4/db/d5c/tutorial_py_bg_subtraction.html

Please take a look at these docs and update ticket with your results.

Thank you!

[Pratiksha2692]

Hi @RealSenseCustomerSupport ,

Thanks a lot.

I had tried OpenCV background subtraction. It worked fine for rgb image.

self.color_image_1 = np.asanyarray(self.color_frame_1.get_data())
which has the shape (480,640,3).
I want to apply same background extraction on the .ply file. But while capturing the .ply file I had done some advance setting regarding depth_table to get particular area of interest.

depth_table_control_group.depthClampMin = 0 depth_table_control_group.depthClampMax = 800
So the x,y,z,r,g,b data points are less than (640*480).
How can I apply this background extraction on .ply file?

Thanks in advance.

[RealSenseCustomerSupport]

---

Hi Pratiksha2692,

Wonder if you have any updates/progress on this one?

Python wrapper newly had threshold filter added in, which is similar to what RS Viewer has in post processing filter section.
You can set min_distance and max_distance to do depth segmentation. Below is an example of using the filter.

a) First claim a filter. which by default uses 0.15~4m): thr_filter = rs.threshold_filter()
b) And then call set_option to adjust min&max distance value (float type):
thr_filter.set_option(rs.option.min_distance, <YOUR_MIN_VALUE>)
thr_filter.set_option(rs.option.max_distance, <YOUR_MAX_VALUE>)

And then from there, you will need to your own way to deal with the data from PLY file.
Hope this can be helpful.

Thanks!

[RealSenseCustomerSupport]

---

Hi Pratiksha2692,

Do you still need help? If nothing else is needed, it will be closed.

Thanks!