search

qh-huang / QMapping

Integrate openslam GMapping and OpenCV within Visual Studio for testing on Windows platform
Apache License 2.0
1 stars 2 forks source link

The evalution of the result map #1

Open wtc1993 opened 8 years ago

wtc1993 commented 8 years ago

I use the windows version to build the map, and it is worser than the map built by ros gmapping. I use the same dataset. Why? It seems that the algorithm and the code is same.

qh-huang commented 8 years ago

Any screenshot? And which dataset did you use?

wtc1993 commented 8 years ago

Dataset: Intel Research Lab and MIT Killian Court (2D) download from http://kaspar.informatik.uni-freiburg.de/~slamEvaluation/datasets.php

Below is my result map. intel_my

mit

Below is the result in ros. I only test the intel research lab dataset in ros. intel_ros

I don't know why the result is different?

qh-huang commented 8 years ago

Obviously there's a transformation between these 2 result. Assuming left-up point is origin point (0,0) then the transformation is x' = y y' = x

On the other hand, I changed some parameters such as skipping frames, patch size, laser angle resolutions..., which would effect results.

wtc1993 commented 8 years ago

I use the default parameters same as the ros gmapping. Though the two results have a transformation,the whole performace should be same. But the map is worser.

qh-huang commented 8 years ago

Interesting...how did you play with .clf file on ros?

wtc1993 commented 8 years ago

Sorry,I was sleeping last night. I used the .clf file to make a rosbag by python.The python code is someone else gave to me.Then I just paly the rosbag. In ros, the laserdata is published every 50ms simulated. In windows, just like the code, I read every frame laserdata and then processed it. I just feel map in ros is better than in windows.