This guide is suitable for the Intel® RealSense™ cameras D415/D435/D435I. It describes the steps starting from driver installation/configuration to basic sample programs.
1. Requirements and installation:
The following steps describe the requirements for a standard setup of the 3D camera.
-
The camera can be connected and powered by an USB 3.0 interface.
-
The system requirement is an Ubuntu®16.04 or Microsoft® Windows® 10 operating system on the part of the camera.
-
Download and install the latest version of
.
Select the installer depending on your operating system and platform:32 Bit 64 Bit 32 Bit Windows 64 Bit Windows 32 Bit Linux 64 Bit Linux -
Download and install the transport layer (TL) driver:
-
As a Windows® user download and install the ‘RealSense GenTL producer’.
-
For Linux users an installer (.deb - package) is available via support request.
Note: The terms of service of
apply.
-
1 Like
2. Licensing:
For the RealSense™ cameras to work properly with 
a license is required. See the licensing guide for detailed information about purchase and methods. If is used without a valid license, a watermark (shape of logo) will appear on displayed images. The license terms of 3 party software can be found on the website.
3 Likes
3. Configuration and Discovery:
If a complete overview of the funcionality and especially the driver properties is needed, a first dive into the System Browser (>13.02.000, Windows) or the GenICamBrowser can be helpful.
Fast access (System Browser, Windows only)
- Connect the camera to a USB 3.0 Port:The RealSense™ camera must be connected to an external USB or on-board Host controller card with USB 3.0 ports (usually marked blue). Other recommendations are an USB 3.0 cable with maximum length of 5 meters, and the use of machine vision validated USB cables. See this documentation for more information about the USB 3 vision standard.
- Type “System Browser” in the Start menu
- The device will be available in the list on the left hand side
- A double click on the DEPTH camera in this list will open a depth map live view.
Advanced setup (GenICamBrowser)
The TL implementation allows the standard configuration for GenICam devices.
-
Connect the camera to a USB 3.0 Port:
The RealSense™ camera must be connected to an external USB or on-board Host controller card with USB 3.0 ports (usually marked blue). Other recommendations are an USB 3.0 cable with maximum length of 5 meters, and the use of machine vision validated USB cables. See [this documentation]https://help.commonvisionblox.com/NextGen/14.0/md_theory_of_operation_hardware__gen_i_cam__c_v_b__user_guide.html#gcug_chmtopic45) for more information about the USB 3 vision standard.
-
Open the GenICam Browser from the Start menu (Windows) or from the Dash (linux)
-
Search for available devices. The RealSense™ cameras may appear as two separate devices (RGB and DEPTH) .
For further details refer to the GenICam user guide -
Configure and save the device driver parameters into an .ini - file:
-
Select the cameras (1. and 2.) and add them to the ‘Configured Devices’ section by clicking on the ‘add’ button (3.).
-
In order to select the suitable driver depth image format click on the configured depth device (right hand side), hit ‘Options’ (4.) and select ‘Raw’ as CVB Color Format.
-
Save the configuration selecting ‘Save’ (5.) .
-
Open the device view by double clicks on the desired device in the ‘Configured Devices’ section to see a live capture.
-
By selecting a higher ‘max visibility’ in properties menu (6.) (Beginner, Expert, Guru) advanced features can be displayed.
-
Change RealSense™ specific settings using the NodeMap in the property grid (7.) . Advanced options for sensor settings and processing methods can be adjusted. For feature availability and options refer to camera manuals in the ‘Downloads’ section.
-
Intel releases several documents for methods improving camera performance (e.g. this guide or calibration instructions)
-
The device specific information stored in the property grid (7.) can be saved to a .gcs file via ‘Save device properties’ in the properties menu (6.)
-
To reload these settings programmatically, please refer to the C-API function NMLoadSet().
3 Likes
Other GenTL Consumers
For other software to handle with the TL, an environment variable called GENICAM_GENTL64_PATH or GENICAM_GENTL32_PATH is created. The interface can be found there. Most GenTL consumers such as use this variable to load the TL, hence it should be available automatically.
3 Likes
4. Sample Programs:
These sample programs include a basic acquisition and processing routine in different APIs. A single frame is acquired and transferred from depth map to point cloud by means of a calibrated projection. These snippets make use of the new APIs CVB.Net, CVB++ and CVBpy. A valid installation (version > 13.2.000) is required.
Please note: The 3D calibration method (pinhole camera model) required for the transfer between depth image and point cloud for RealSense™ cameras isn’t available yet for CVB++/CVBpy, but will included as soon as possible. Interested developers are asked to use the .Net Wrappers or to send a support request. The said function also can be found in the C-Api and can be included via a handle.
CVB.Net
using System;
using System.IO;
using System.Linq;
using System.Text.RegularExpressions;
using Stemmer.Cvb;
using Stemmer.Cvb.Driver;
using Stemmer.Cvb.Utilities;
using Stemmer.Cvb.GenApi;
namespace GrabPointCloud
{
class Program
{
static void Main()
{
// open device and get information
var genPath = Path.Combine(SystemInfo.InstallPath, "Drivers", "GenICam.vin");
var discovery = DeviceFactory.Discover();
var depthDiscover = discovery.First(i => Regex.IsMatch(i[DiscoveryProperties.DeviceModel], @".*Intel RealSense.*_DEPTH"));
using (var device = DeviceFactory.Open(depthDiscover))
{
NodeMap nm = device.NodeMaps[NodeMapNames.Device];
// get The camera intrinsics and create a Calibrator
var fXNode = nm["IntrinsicsFocalLengthX"] as FloatNode;
var fYNode = nm["IntrinsicsFocalLengthY"] as FloatNode;
var pXNode = nm["IntrinsicsPrincipalX"] as FloatNode;
var pYNode = nm["IntrinsicsPrincipalY"] as FloatNode;
var sZ = nm["DepthUnits"] as IntegerNode;
Point2Dd principalPoint = new Point2Dd(pXNode.Value, pYNode.Value);
var calibrator = Calibrator3D.FromPinholeCameraModel(fXNode.Value, fYNode.Value, principalPoint, sZ.Value);
// start acquisition
Stemmer.Cvb.Driver.Stream stream = device.Stream;
stream.Start();
try
{
using (StreamImage image = stream.Wait())
{
image.Save("rangemap.tif");
var PointCloud = calibrator.CreatePointCloud<Point3Df>(image);
PointCloud.Save("pointcloud.tif");
}
}
catch (System.IO.IOException e)
{
Console.WriteLine("Error :{1}", e.Message);
}
finally
{
stream.Stop();
}
}
}
}
}
CVBpy
import cvb
import os
import re
flags = cvb.DiscoverFlags.IgnoreGevSD | cvb.DiscoverFlags.IgnoreVins
discover = cvb.DeviceFactory.discover_from_root(flags)
info = next(filter(lambda x: re.match(".*Intel RealSense.*_DEPTH",x[cvb.DiscoveryProperties.DeviceModel]),discover),None)
with cvb.DeviceFactory.open(info.access_token) as device:
# get device intrinsic information
nmp = device.node_maps["Device"]
zDivisor = nmp.DepthUnits.value
focalX = nmp.IntrinsicsFocalLengthX.value
focalY = nmp.IntrinsicsFocalLengthY.value
principalX = nmp.IntrinsicsPrincipalX.value
principalY = nmp.IntrinsicsPrincipalY.value
principalPoint = cvb.Point2D(principalX,principalY)
calibration = cvb.Calibrator3D(focalX,focalY,zDivisor,principalPoint)
# start acquisition
stream = device.stream
stream.start()
rec_image, status = stream.wait()
image = rec_image.clone()
rec_image.save("rangeimage.tif")
# create and save point cloud
flags = cvb.PointCloudFlags.NoExtrinsic | cvb.PointCloudFlags.Double
pc = cvb.PointCloudFactory.create(image.planes[0], calibration, flags)
pc.save("pointcloud.tif")
CVB++
#include <iostream>
#include <cvb/string.hpp>
#include <cvb/image.hpp>
#include <cvb/image_plane.hpp>
#include <cvb/point_cloud.hpp>
#include <cvb/point_cloud_factory.hpp>
#include <cvb/calibrator_3d.hpp>
#include <cvb/device_factory.hpp>
#include <cvb/utilities/system_info.hpp>
#include <cvb/driver/stream.hpp>
int main(int /*argc*/, char* /*argv[]*/)
{
try
{
// open device and grab a depth image
auto path = Cvb::InstallPath();
path += CVB_LIT("Drivers/GenICam.vin");
auto device = Cvb::DeviceFactory::Open(path);
auto nodeMap = device->NodeMap(CVB_LIT("Device"));
auto stream = device->Stream();
stream->Start();
auto waitResult = stream->WaitFor(std::chrono::seconds(10));
if (waitResult.Status == Cvb::WaitStatus::Timeout)
throw std::runtime_error("acquisition timeout");
auto image = waitResult.Image;
stream->Stop();
// get the camera's intrinsics
double fx = nodeMap->Node<Cvb::FloatNode>(CVB_LIT("IntrinsicsFocalLengthX"))->Value();
double fy = nodeMap->Node<Cvb::FloatNode>(CVB_LIT("IntrinsicsFocalLengthY"))->Value();
double px = nodeMap->Node<Cvb::FloatNode>(CVB_LIT("IntrinsicsPrincipalX"))->Value();
double py = nodeMap->Node<Cvb::FloatNode>(CVB_LIT("IntrinsicsPrincipalY"))->Value();
Cvb::Point2D<double> principal(px, py);
double Sz = nodeMap->Node<Cvb::IntegerNode>(CVB_LIT("DepthUnits"))->Value();
// transform to point cloud
auto calib = Cvb::Calibrator3D::FromPinholeModelParams(fx, fy, Sz, principal);
auto flags = Cvb::PointCloudFlags::Double | Cvb::PointCloudFlags::NoExtrinsic;
auto pc = Cvb::PointCloudFactory::Create(image->Plane(0), *calib.get(), flags);
pc->Save(CVB_LIT("pointcloud.tif"));
image->Save(CVB_LIT("rangemap.tif"));
}
catch (const std::exception& error)
{
std::cout << error.what() << std::endl;
}
return 0;
}
Result:
The sample scene shows a cup, a RealSense box and a garbage can. With the code above, a depth map and the corresponding point cloud are created.
The acquired depth map and the generated point cloud are obviously stored as .tif, respectively. As a result they can be visualized by using Viewer and Player3D.
Range map:
Point cloud:
3 Likes
5. FAQ
The Depth module provides data (by default) as a 16 bit Image.
The value 0 means that there is no distance value available.
Values 1 to 65535 represent depth data (calibrated, value depending on “depth unit” parameter of the device)
The TL or the device do not appear
Avast and Kaspersky might block the TL. Typically, the C-API call fails when opening the camera device. An antivirus exception for the calling software might be needed.
3 Likes
Hallow,I am honored to see your case post. I used a D435I camera, but I could not find the camera with Browser.In websearch also could not find the camera . Do you know how to solve it? Thanks.
Glad to receive a reply。Using the browser or license manager could not find the camera.In the WEB search,all camera ID in the camera box could not search the code
Hi @352206277, sorry, I forgot, that Realsense are not provided with a camsuite license by default. In order to use the Realsense with , you are kindly asked to purchase a different license method here
Thanks for your reply.Could you give me a more accurate connection or more detailed graphic tutorials?
Your connection is the picture show.I do not know where to buy the lincense.thanks.
This really gets off topic now. Here is the english version. Please select either foundation or image manager package. There you can find the different licenses.
If you have further questions, please contact our sales support. This thread is for realsense questions.
Thank you very much for your reply and patient guidance. I’ll try it, and I’ll ask you if i encounter any problems later.Thanks.
If you have further questions, please contact our sales support. This thread is for realsense questions.
OK. Thank you very much.
Thank you for your post. How can I get the color image on display window?
From your Advancedsetup, GenICam Browser can show color depth image.
Hi Charlene,
do you want a point cloud or do you want the the depth image in color (like in the image below)?
If you just want the coloured depth image, just change the “Pixelformat” to RGB8. (Warning this is not persistant).
Is there ar define, that turns off any reference to realsense in CVB library? We are having an issue with using librealsense with depth cameras and at the same time CVB with line scan cameras. including
#include <librealsense2/rs.hpp>
#include <cvb/device_factory.hpp>
together causes basic realsense code to crash with a message
free(): invalid pointer on pipe.start()
This does not happen, if CVB is not included.
Hi @KKulikovskis ,
There are no references to realsense in the CVB headers.
We do have a transportlayer, which you have to manually install (which is librealsense wrapped via the GenTL interface, so no includes there either).
Do you use the Transportlayer? If no, then there should be no interation between CVB and librealsense.
If yes, then there is no reason to include anything from librealsense.
Hope this helps.