Ros Learning Note-depth sensor conversion to laser data (Hector_slam)

Prerequisites: 1, ensure that the reader has installed the Kinect or other deep camera driver, if not installed, can be directly on the network disk download: http://pan.baidu.com/s/1hqHB10w Extract Password: wrmn

Create a map with deep camera-like laser data:

Download and install of the dependent packages:

1, install Hector_slam

2, download hector_slam_example:, after compiling, remember to add this file to the Ros package

ROS_PACKAGE_PATH=/home/用户名/catkin_ws/hector_slam_example:$ROS_PACKAGE_PATH　
of course the path is changed to your own path. 3,install The dependency packages:　ROSDEP Install Hector_slam_example4, before booting if you are using Kinect, then please modify the launch file. <include file= "$ (find Openni_launch)/launch/openni.launch" > #由于我们使用的kinect, Openni2 is nearly unsupported. So with Openni and then Roslaunch roslaunch hector_slam_example hector_openni.launch just let the mobile platform to move to build a map, of course, the method used to generate a simulated aurora is Depthimage_to_laserscan, the code is faster but requires the camera to be relatively stable, and as far as possible horizontal placement. Effect: From a single frame, or good, but moved a bit after the discovery of the problem: After a look to know: the 2d slam algorithms commonly used in Ros are mainly gmapping and Hector_slam, in which
Hector_slam is a very good algorithm, but the author in the paper is very clear that hector_slam by the least squares to match the scanning point, and rely on high-precision LIDAR data, but do not need a mileage meter. So the scan angle is small and the noise is large Kinect is not, I tried to match the time will fall into the local point, the map is very chaotic.

Ros Learning Note-depth sensor conversion to laser data (Hector_slam)