forrest-z / mavslam

Visual odometry based on Intel® RealSense™R200

MAVSlam - Visual odometry & SLAM for PX4

Onboard visual odometry for PX4

Integrate Intel® RealSense™ R200 depth camera as an oboard visual input source for PX4 using the UP-Board as companion:

• Estimate position and speed based on stereo depth images up to 60 fps for LPE /EKF2
• Stream video to MAVGCL with overlays
• Feature-based or depth obstacle detection
• Integration with PX4 via MAVLink (MAVComm required)

Prerequisites:

• UP-Board running Ubilinux 3.0 with Java 8 (minimum) stack connected via serial link to PX4 controller

• Intel® RealSense™ R200 (connected with external power supply)

Heading estimation:

XY Estimation:

Obstacle detection:

MicroSLAM:

Desktop odometry demo

A first implementation using Intel® RealSense™ R200 with boofcv on OSX/Linux platforms. Based on https://github.com/IntelRealSense/librealsense and http://boofcv.org/index.php?title=Main_Page.