Here's a quick technical post for anyone attempting to harness the capabilities of a Realsense D435 camera on a Jetson TX2. For me, this is about getting usable depth perception on a UAV, but it has proved more problematic than I originally anticipated.
The Intel Realsense D435 depthcam is small enough to mount on a UAV and promises enough range for object detection and avoidance at reasonable velocities. My intention is to couple it with the machine learning capabilities of the Jetson TX2 to improve autonomous flight decision making.
The problem is that the Intel Realsense SDK2 does not apparently support the ARM processor of the Jetson TX2, as I write. This post links to my blog article which aims to provide some simple installation instructions that now work for me, but took a long time to find out!
(Full blog article link is https://mikeisted.wordpress.com/2018/04/09/intel-realsense-d435-on-jetson-tx2/)
Comments
@Mike Isted
I read your blog at
https://mikeisted.wordpress.com/2018/04/22/first-flight-intel-reals...
specification from
https://software.intel.com/en-us/realsense/d400
Intel RealSense Depth Camera D435
Maximum Range Approximately 10 meters
So your IR depth images in your blog are full of artifacts due to limited, maximum range of Intelsense IR technology.
All you need for a flying drone is a single camera blur effect depth technology, giving better results, showing no 10 meters range limit.
Give up Realsense since for a fast flying drone, 10 meters range limit is a serious problem in detecting rigid or soft obstacles.
excellent example of pattern recognition feature
Mike, if you are aiming for autonomous navigation including obstacle avoidance you will need something like an i7 processor, in order to run the entire stack in real time.
That is from perception (Visual SLAM), state estimation, occupancy mapping, reactive path planning to MPC control.
This setup from a recent paper by ETH uses an Intel NUC to run the fulls stack onboard.
https://arxiv.org/abs/1812.03892