manolovaza's profile - activity

sequoia driver for ROS Do anyone know how to integrate sequoia camera to ROS? A driver?

I used kinect v2 for a series of experiments. I have many bagfiles of the experiment. I want to overlap two georeferenced (total station) point cloud assembly. I have a launch file for assembly and another launch file for filtering (box,ransac, radious, etc). When I assembly the first bagfile everything is fine, but when I play back the second it modifies the tf and complains (Lookup would require extrapolation into the past) and (Detected jump back in time. Clearing TF buffer). I use simulated time. I have tried several solutions but do not solve my problem. I know it is about the time stamp, that is all. Any hint?

I am trying to implement KINFU but it is based on CUDA gpu. I have intel gpu and I havent been able to comopile an emulator called gpuocelot. Now I am wondering if it is possible to use the cpu instead?

Now you can work with the Kinect v2 thanks to the nodelet from Thiemo Wiedemeyer. The kinect v2 is by far better that the kinect v1. Better depth map pixel resolution, an infrared output that allows night vision, and better robustness against sunlight (not perfect, but if the objects of the scene reflect good the near infrared, you can obtain depth information even under sunlight). The tricky part of the Kinect v2 is to handle the high amount of output data (therefore USB 3.0); we have recorded bags at a rate of 2 GB per minute, and sometimes we had problems obtaining the depth data. It is definitely worth changing to the kinect v2.

you can filter out your data, I had some similar problems, here is an example script:

#!/bin/bash
#filter out the bag files for the raw data
source ~/.bashrc

cd /to/your/bag directory/

rosbag filter original.bag filtered.bag 'topic=="/lms1xx_front/scan" or topic=="/lms1xx_front_tilt/sc an" or topic=="/lms1xx_back/scan" or topic=="/fmData/imu_rx" or topic=="/fmInformation/imu" or topic=="/kinect2/depth/camera_info" or topic=="/kinect2/depth/image" '

I am assembling point clouds using the kinect v2. Everything works almost good after adjusting several issues. Now my problem is that the point clouds are tilted with the angle of the kinect and the assembled point clouds don't look like a continuum.

I am applying a voxel filter to a point cloud stream but the topic is not being published and I cannot visualise the added pointclouds. rostopic echo /voxel_grid/kinect_test outputs:

WARNING: no messages received and simulated time is active.
Is /clock being published?

The output of rostopic info /voxel_grid/kinect_test is:

Type: sensor_msgs/PointCloud2

Publishers: 
 * /voxel_grid (http://toshiba:55381/)

Subscribers: 
 * /rqt_gui_cpp_node_2830 (http://toshiba:59223/)
 * /rqt_gui_py_node_2830 (http://toshiba:34321/)

I am unable to open kinect2, neither with rosrun kinect2_bridge kinect2_bridge nor with roslaunch kinect2_bridge.launch, the output is always the same:

[Freenect2Impl] enumerating devices... [Freenect2Impl] 18 usb devices connected [Freenect2Impl] failed to open Kinect v2 @4:5! [Freenect2Impl] found 0 devices Error: no Kinect2 devices found!