The May’s PGRs Research Presentations was held on Wed. 10th July, 2pm, Meeting Room, MC3108 (3rd floor).
This session we had the following presentations:
| Title: “Using anisotropic spatiotemporal smoothing to reduce blurring in images under low light conditions“. | Title: “Automated Robot Control with Behaviour Selection, Object Classification and Background Subtraction” | |
|
By: Gabriel Zahi |
By: Daniel Pashley & Sean Walton |
|
| Abstract:Improving images in low light involves the process of summing neighbouring pixels locally in space and time to improve the reliability of the intensity of each pixel and to reduced the effect of noise acquired when capturing images under low light conditions. The summation process is stronger among neighbourhood pixels closer to the central pixel and reduces as the pixels are further away from the central pixel. Though the summation process is an effective way of improving the reliability of low light images, it has an adverse effect of blurring the image. This presentation introduces the use of the gradient as an anisotropic method to reduce the noise in the images as well as reducing the blurring caused by the summation method. Using the gradient in the three dimension (x,y,t), we can channel the summation process to lean more towards homogenous pixels and less along non-homogenous pixels. This new approach has proven to reduce the blurring while preserving structures and details.. | Abstract: Our research aims to try and tackle some common risks and issues in traditional search and rescue methods by providing an alternative method by which search and rescue can be carried out. We show how automated ground and air based robots can be used to reduce risk and cost but in turn increase speed of searching. To do this, cost differences between current search and rescue methods and commercial robotic platforms that can be used in place of traditional methods are shown. Along with this, we present how search methods can be improved by replacing one or two helicopters with forty ground and air robots laid out in an efficient pattern that allows the fastest search method. We also show how we can utilise these robots through a central system and use all images and data provided by each of the connected robots to perform background removal and face detection, then use the data in a finite state machine based control system to send control commands back to the robots. Finally, we present a series of experiments along with their results to prove the validity of this solution. | |
Additional presentation by Oliver Szymanezyk , titled:
“Brief on the Masters Prizes and his experience of the event”
The Q/A was followed by a brief cath-up meeting.