The purpose of the SLAM system is very specific, for detecting cones in an image and triangulate their position to create a map.
The data input would be the camera data, odometry and the LIDAR data.
I have been going through SLAM algorithms on openSLAM.org and through other implementations of SLAM systems.
I would like to know if there are a set of SLAM algorithms specific for the problem I have and what are the most efficient and least time consuming SLAM algorithms available. Any leads would be helpful.


You should probably also add information regarding your sensors, for instance what LIDAR are you to use 2D or 3D etc. The nature of SLAM algorithm shall also depend upon what kind of system do you need to use it on specifically what are the rates that you need, do you need the SLAM to be online or offline etc. Further the machine you run your algorithms may also be the reason which decides which algorithm you finally use.

Regardless I think these are some SLAM algorithms that may help you :

  1. Orb SLAM : http://webdiis.unizar.es/~raulmur/orbslam/ uses only camera information but performs very well at least the RGB-D version.
  2. Google Cartographer : https://github.com/googlecartographer/cartographer needs no introduction.
  3. You should also checkout gmapping - slam-gmapping

  4. Definitely checkout the LOAM algorithm, it uses only LIDAR data but produces very good quality maps, you can use EKF on the odometry that it provides with the one you have, that should give you a good positional estimate.


I did a little search and I found some existing research for cone detections in Youtube. One of them is: OpenCV Traffic Cone Detection Tutorial in Visual Basic https://youtu.be/Y7SkyY58aUw

I saw in the title you mentioned LIDAR and Stereo camera, however, in the first sentence you just mentioned image rather than 3D point cloud. So I wonder if you mean to get a 3D map first and then detect cones from the 3D map.

As far as I know it is a really challenge to detect objects in 3D maps. It just like an extension of the image processing problem in RGB images.

Maybe you want to learn something like this. It seems to be a cutting-edge technology.

Real-Time Semi-Automatic Object Detection and Measurement with Intel RealSense https://www.youtube.com/watch?v=nH-bKOZjOec

If you don't mind the short detect range, the RGBD camera like Intel RealSense should be a good option for you.

  • $\begingroup$ Thanks! But what SLAM method do I use to generate a map using the cones? My task is to build the SLAM system. The inputs are the mentioned data. I have been going through algorithms that have been tested on the KITTI dataset. I was wondering if there is a SLAM system built for the dataset I have to work with. $\endgroup$ Jan 17 '17 at 12:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.