Surveyor SRV-1 Trail Video

(1.7 MB) Overhead video of the robot moving along the trail.

(1.3 MB) Standalone video of what the robot sees while moving along the trail for your own testing purposes.

Problems

It is worth pointing out some of the issues and mistakes that we experienced during the creating of this tutorial to help you solve similar issues.

1. Lighting - To illustrate different lighting techniques we purposely made the setup environment non-optimal to run the robot. In reality one would probably attempt as much as possible to ensure that the lighting requirements are better met and no sunlight glare is present. This could also have been better mitigated by using different tiles that reflect less light.

2. Blob-Filter - Removing blobs based on attribute values is somewhat of a black art. As with any image processing applications noise is ever present in the image and can cause some of the squares to become non-square's momentarily. This causes them to disappear from the robots tracking. Better flood filling and blob filtering might be able to reduce this loss of tracking seen as object flicker in the final videos.

3. Motion blur - Capturing images while the robot is moving causes motion blur in the captured image. This motion blur tends to cause the squares to become more irregular in shape and therefore not match the blob filter criteria. To reduce the effects of motion blur the robot is pulsed in-between image captures to ensure that the robot is not moving while an image is being captured. This does slow down the robot movement considerably but is a necessary requirement for stable operation.

4. Frame rate - Despite the SRV-1b being capable of more than 10fps when in 160x128 mode the low/bad lighting further reduces the processed frame rate to less than 4fps. This reduction also requires that the robot move slower in order to capture enough frames to react quickly enough to steer the robot. The pulsing added to reduce motion blur provided enough speed reduction to ensure that the robot could react to the visual scene fast enough.

5. Oscillation - There are still times where the robot has successfully passed a square but during the turn (due to the robot being a tracked vehicle) the robot may back up enough to again see the most recently passed square. This causes the turn to terminate and the robot to reengage tracking. Once passed, the robot returns to turning mode where the process can repeat itself. This issue can be reduced by further increasing the time after which the robot determines it is no longer seeing a square and when to start a turn execution. Addition filters can also be added to eliminate detection of any square in the lower part of the image while turning with the assumption that any square that should now be tracked would appear somewhat distant from the robot.

Your turn ...

Want to try this yourself? Here is the robo-file that you can download and run yourself. Clicking on any of the modules will bring up that interface and allow you to customize it for yourself.

The End

If you have any questions or comments about this tutorial please feel free to contact us.

Please Support Us! Our software is free but our costs are not. Please take the time to visit our sponsor's links on the right of each page. They work hard too!

Link To Us! Please visit our link to us page and add a link to our website to help us get the word out!

Have a nice day!