3D reconstruction

[yzhan72]

Hello，Is there anyone konw how to use inter sr300 to reconstruct objects three-dimensionally,I have got the color image and depth image,I don't know what should i do next ?

sincerely appreciate,Please!

[MartyG]

Hi there! Are you wanting to make 3D model scans of objects? If so, the '2016 R2' and '2016 R3' Windows versions of the RealSense SDK come with a sample program called 3DScan that enables you to scan objects and save them as an .OBJ format model without needing any programming knowledge.

You can find 3DScan by going to a folder called 'Intel RealSense SDK Gold' that should have been placed on your computer desktop when you installed R2 or R3. Inside that folder is an application called the Sample Browser, where you can browse a range of RealSense sample apps and run them from that interface. You can easily find 3DScan by typing '3dscan' into the Sample Browser's search box.

You should place a tick in 3DScan's "Landmarks" tick-box before clicking the 'Start Camera'' button, or else the program will not work. You can use a combination of tick-box selections, but Landmarks should always be one of them.

[yzhan72]

okay,i have found it.Thank you very much.In addition,i have read some article about 3D reconstruction based on kinect,i don't have kinect but inter SR300.So i wonder that whether i can take some measure on depth image to achieve 3D reconsruction based on inter SR300 just like the 3D reconstruction based on kinect.

[MartyG]

Kinect 1 and the SR300 both use Structured Light for their scanning. So they will share some basic principles, even if Kinect and RealSense are not compatible with each other, and a Kinect programming script will not work with a RealSense camera.

Kinect V2 changed to a different scanning method called Time Of Flight.

RealSense cameras are depth sensing cameras just like Kinect cameras are. So they will already be using depth to create 3D reconstructions.

[yzhan72]

Thank you for your reply. Should i understand that if i want to get 3D model of the cup,it just need through the 3DScan in the inter SR300,instead of get the depth image of the cup ,and then take some measures like Gaussian filter,Image alignment,create pointcloud???

[MartyG]

In most circumstances, you can get a very good 3D model in .OBJ format by using the free 3DSCan program. That OBJ file can then be imported into other 3D model programs such as Blender for further editing.

If you want more advanced object scanning programs with extra features that will work with the SR300 camera then there a couple of good options, such as ItSeez3D and Easy 3D Scan.

https://realsenseapp.intel.com/apps/itseez3d/ itSeez3D - Intel RealSense

You would only likely need to consider using a point cloud for depth scanning if you were getting a bad quality image using the usual scanning methods.

[yzhan72]

okay,i got it. Thank you once again.Otherwise, I am doing some research about robot vision and computer vision,i don't konw how to apply the inter SR300 in this study,could you give some advanced suggestion?

[MartyG]

For moving robots such as Turtlebots, RealSense camera models such as the R200 and ZR300 are more commonly used than the SR300. For robots that do not move from their position, there have been cases where an SR300 has been used with a robotic arm. Here are some example YouTube videos.

https://www.youtube.com/watch?v=Ze6dpPPydKs Object sorting with Intel SR300 and Turtlebot 2i - YouTube

https://www.youtube.com/watch?v=HRMNDO-1K4E Intel SR300 depth cam on my 6DOF toy robot arm - YouTube

If you plan to develop a mobile robot in Windows with an SR300, I would recommend looking at the new RealSense SDK 2.0 which is SR300 compatible, as SDK 2.0 can be used with the Robot Operating System (ROS).

https://github.com/IntelRealSense/librealsense/releases/ Releases · IntelRealSense/librealsense · GitHub

[yzhan72]

All right, Thank you.