nvoltex's profile - activity

I'm interested in obtaining an estimation of the linear velocity of the vehicle using optical flow, so which are the best packages for optical flow using the bottom camera of an aerial vehicle? (the camera is a simple webcam).

Or the best approach is to implement optical flow using OpenCV?

Notice that I am using an Odroid X2 (ARM) on the aerial vehicle, so I am searching for something that wont be too demanding for the processor.

Hey there,

Firstly, I'm aware this is not a completely ROS related question, however I have used this web camera before on the exterior with no problems, so I'm wondering if I might be missing something on my setup.

Basically the problem is related with the "saturation" of the image when on the exterior (due to daylight). The image gets either completely white or really "whitish". I'm using the usb_cam ROS package to get the images from the camera.

The launch file I'm using right now is:

<launch>
        <node name="usb_cam" pkg="usb_cam" type="usb_cam_node">
            <param name="video_device" value="/dev/video1"/>
            <param name="image_width"  value="640"   />
            <param name="image_height" value="480"   />
            <param name="pixel_format" value="mjpeg" />
            <param name="auto_focus"   value="true"  /> 
            <param name="framerate"    value="30"    /> 
            <param name="camera_frame_id" value="/camera" />
        </node>
</launch>

Initially I didn't had the auto_focus enabled, but added it on an attempt to solve the problem. The camera is a "Creative Live! Cam Socialize HD 1080". I have used this camera on the exterior on pretty similar conditions with no problems, however I can't seem to find a way to use it now. Am I missing something?

Once again sorry for the question not being totally ROS related, but I'm pretty lost about this. Thanks in advance

You are right, I'm sorry I forgot to add the command. Thanks for warning me.

Hi there,

Whenever I try to use image_view to visualize a topic published by usb_cam with the following command:

rosrun image_view image_view image:=/usb_cam/image_raw

I get this:

init done 
opengl support available 
[ INFO] [1457614339.953957839]: Using transport "compressed"
/opt/ros/indigo/lib/image_view/image_view: symbol lookup error: /opt/ros/indigo/lib/image_view/image_view: undefined symbol: _ZN9cv_bridge18cvtColorForDisplayERKN5boost10shared_ptrIKNS_7CvImageEEERKSsbdd

The same happens even with the "raw" transport:

init done 
opengl support available 
[ INFO] [1457614438.464431134]: Using transport "raw"
/opt/ros/indigo/lib/image_view/image_view: symbol lookup error: /opt/ros/indigo/lib/image_view/image_view: undefined symbol: _ZN9cv_bridge18cvtColorForDisplayERKN5boost10shared_ptrIKNS_7CvImageEEERKSsbdd

Am I doing something wrong? has anyone got similar errors?

Thanks in advance.

I'm trying to use hector_object_tracker, to track some targets from aerial images (captured from a down facing camera on an aerial vehicle).

I already have a node for detection of the targets as well as extracting their position to the world reference frame.

From what I understood of the wiki page, I should publish this information to the "worldmodel/pose_percept" topic, which uses messages of the type: "hector_worldmodel_msgs/PosePercept".

My problem is connected to the PerceptInfo msg: Since I only have one type of target the class information is easy to fill (using class_support=1). However I don't understand how I should handle the object information on that message. Furthermore, from inspection of hector_heat_detection node, that section seems to not be filled.

Thanks in advance

One of my problems is that I can't seem to understand how to setup the packge properly and there are no tutorials or extra information. Do you know of place where I can read more in depth about that package?

What I am searching for is more a solution to the position estimation (noise/outlier filtering and estimation to be precise), since I already detect and convert the dected position to the world referencial frame. However hector_object_tracker, doesn't seem to allow access to the filtering side alone

An aerial vehicle captures images of the ground, using a down facing camera, in order to detect some static ground targets. Using opencv I created a node that detects those targets and converts their position from the image frame to the world frame. Since all targets are on the ground, each sample is a 3D vector (x,y,0). Note: Using robot_localization I also have the attitude and heading of the aerial vehicle.

What's the best approach to estimate the position of the targets from multiple readings?

Is there a package that implements something to this effect? This is basically a problem of noise removal and estimation, so strategies one of the possible strategies is using clustering.

Using opencv I created a node that detects ground targets and converts its position in order to the world reference frame. Each sample is a 3D vector (x,y,0).

Initially, I was only using a really simple strategy where I assumed that if a new sample was within a certain range of another, then it meant it belonged to the same ground target.

However, I would like to test some clustering algorithms to group the samples into detected targets. How should I do it inside the ROS environment? Are there any packages available that do something similar to what I want to achieve? Any kind of direction would be appreciated. Thanks

I'm trying to use the imu_filter_madgwick while feeding it magnetometer data (use_mag=true). However, even though the filter seems to be working properly before adding the magnetometer data, after the addition it stops behaving as expected.

I even calibrated the data for Hard and Soft iron effects using some ellipsoid fitting algorithms, but it still doesn't work. It's as if the information from the magnetometer it's making the orientation to stay "locked" around a certain angle.

I tested the magnetometer by trying to get a heading using the information from the x and y axis, and the heading I obtained was pretty close to the correct one.

Any idea what might be causing this problem? Since I need a global orientation, it's pretty important I'm able to use the magnetometer data.

What i meant when I asked if I should fuse the roll and pitch is that, due to the vehicle being sustained in the air thanks to a ballon, changes on roll and pitch will be small and due to noise from the rotors (or wind). That being said, those changes won't have much effect on the overhall movement

Hey there,

I'm developing an aerial vehicle which is equipped with a GPS, IMU and bottom camera (pointed at the ground for target tracking). I was looking for a way to fuse the data, with two objectives:

• Obtaining a coordinate for the "drone" localization (I could assume it starts at (0,0))

• Extrating a position of the detected targets (on the "local coordinate system" and GPS information)

Which nodes will prove useful for this application?

Thanks in advance

One of the problems I see is that I don't have the heading associated with that velocity and I'm feeding it as the linear.x. However if the wind conditions make the vehicle move sideways, that won't be compensated by the filter.

Oh ok, thanks! Could you give me some insight on the doubts I posted in the question? Another thing, the GPS I'm using is a garmin 18x, which besides the longitude, latitude and altitude also outputs a "speed" parameter. Is it a bad idea to use this as the velocity on the new odometry msg?

The IMU does have a magnetometer and I think I have set both yaw_offset and magnetic_declination_radians properly. (I haven't changed them since my last question)

I updated the question with some new info and corrected some things. I still have the doubts from before, but since I corrected the odometry msg I'm getting a strange behavior from the filter (explained in the question). Thanks for the help!