search

silencht / SG-SLAM

SG-SLAM: A Real-Time RGB-D Visual SLAM toward Dynamic Scenes with Semantic and Geometric Information
GNU General Public License v3.0
168 stars 20 forks source link

Can't get pointcloud map or semantic map via AstraPro camera #22

Closed Cooo1der closed 7 months ago

Cooo1der commented 10 months ago

Thank your for your excellent work! Howerver, when I ran the project code with my AstraPro camera, I was unable to create point cloud map or semantic map. Coud you please give some advice to solve these problems? It bothers me for a long time...

Here are some of my main outputs: 1. roslaunch astra_camera astra_pro.launch SUMMARY

PARAMETERS

NODES /camera/ camera (astra_camera/astra_camera_node)

**2. ./run_astra_pro_camera.sh is map construction consider dynamic: 1 is global pointcloud reconstruction: 1 is octomap semanticObject map construction: 0 camera_valid_depth_Min is 0.5 camera_valid_depth_Max is 5 Sor_Local_MeanK is 50 Sor_Local_StddevMulThresh is 2 Voxel_Local_LeafSize is 0.01 Sor_Global_MeanK is 50 Sor_Global_StddevMulThresh is 3 Voxel_Global_LeafSize is 0.01 Detector3D_Sor_MeanK is 50 Detector3D_Sor_StddevMulThresh is 1 Detector3D_Voxel_LeafSize is 0.01 Detect3D_EuclideanClusterTolerance is 0.02 Detect3D_EuclideanClusterMinSize is 1000 Detect3D_EuclideanClusterMaxSize is 30000 Detect3D_DetectSimilarCompareRatio is 0.1 global_pc_update_kf_threshold is 25 Start PointCloudMapping Viewer...

Camera Parameters:

ORB Extractor Parameters:

Depth Threshold (Close/Far Points): 2.65909

detection_confidence_threshold is 0.99

dynamic_detection_confidence_threshold is 0.05 Detector2D Thread start ... New map created with 794 point Screenshot from 2023-12-26 16-17-53

silencht commented 10 months ago

If the dataset works well, please check:

Cooo1der commented 10 months ago

Thank you for your reply. I tried to recalibrate the internal parameters of the camera, but I found that each time my calibration results were different, even sugnificantly so. Do you know what might be the reason?

silencht commented 10 months ago

I think the correct cx, cy should be around half of the width and height of the image. you can find more informations at these websites:

Cooo1der commented 10 months ago

I think the correct cx, cy should be around half of the width and height of the image. you can find more informations at these websites:

May I ask how the topic /camera/depth_registered/img_raw was generated? The topic of my camera is only /camera/depth/img_raw. Will this affect the construction of the image

silencht commented 10 months ago
  • Subscribe to appropriate depth image topics: I was running OK with the TUM dataset, but once I ran actual (Orbbec) hardware RGBD camera on ROS would fail. I eventually found that the problem was localised to the ROS subscription camera node, the depth image I subscribed to had been /camera/depth_registered/image_raw, and the problem solved after I switched to /camera/depth/image_raw.

This suggestion is only for certain cameras, such as the LeTMC520 camera (actually a cased version of the astra pro) I use. These websites below is when I use LeTMC520 camera reference.