Changeset 690

Timestamp:

02/21/10 20:54:14 (2 years ago)

Author:

mosttw

Message:

Report checkpoint

Location:

branches/2010-image-rec/docs

Files:

• report.pdf (modified) (previous)
• report.tex (modified) (2 diffs)

branches/2010-image-rec/docs/report.tex

@@ -23 +23 @@
 
 \section{Introduction}
-The ability of a autonomous mobile robot to navigate in an unknown environment opens it 
-to many applications. Robots placed in unfamiliar environments can navigate them based on 
-many sensor inputs. One sensor in this capacity is a color camera, returning a high density 
-of information about an environment, at a low component cost. This information must be 
-interpreted into information useful for navigation and function. 
+The ability of an autonomous mobile robot to navigate in an unknown environment
+opens it to many applications.  A variety of sensors are available for robotics
+use, but the color camera stands out for the high density of information it
+provides for a low component cost, particularly compared to techniques like
+laser rangefinding.  However, the raw sensor data provided by a camera is not
+in a form suitable for direct use in robot navigation --- processing must be
+done to identify significant features in the image and determine their location
+relative to the camera's view.  Doing this processing is a non-trivial task,
+particularly given varying lighting or obstacles.
 
-Our algorithm is designed to be used on Rose-Hulman's entry, Moxom's Master, to the Intelligent Ground 
-Vehicle Competition. The IGVC is a robotics competition for collegiate teams of engineers,
-with the goal in the competition to design and build a robot for autonomous navigation of 
-an obstacle course in the shortest time. The obstacle course is defined as white lines layed
-out on grass or pavement, between 10 and 15 ft apart. Obstacles are placed between these lines
-randomly, and this course is changed between runs. There are complex arrangements of these 
-including swithcbacks and center islands. 
 
-The robot has a camera mounted at the top of a mast, angled to return an image of its forward 
-path. These images must be 
+The Intelligent Ground Vehicle Competition (IGVC) is an intercollegiate
+robotics competition where vehicles must autonomously navigate an obstacle
+course delimited by lines painted on the ground and strewn with obstacles.
+Obstacles, which are generally colorful, plastic construction barrels, may be
+arranged in configurations such as switchbacks and dead ends, requiring that
+vehicles must maintain a high level of spatial awareness to avoid crossing
+lines or touching obstacles, both of which are forbidden.  Additionally, as the
+competition takes place over an entire day, lighting conditions change.  This,
+combined with the rainy weather in Rochester, MI, means that identifying these
+features is difficult.
+
+
+\begin{figure}[htb]
+        \centering
+                \includegraphics[width=0.25\textwidth]{figures/robot.jpg}
+        \caption{Moxom's Master, the Rose-Hulman Robotics Team's IGVC entry.}
+        \label{fig:moxoms-master}
+\end{figure}
+
+Our algorithm is designed to be used by the Rose-Hulman Robotics Teams' entry
+to the Intelligent Ground Vehicle Competition, a mobile robot named Moxom's
+Master.  The robot has a high-resolution camera mounted at the top of a mast
+which images its forward path.  Our approach to reliably classifying obstacles,
+lines, and ground is to take advantage of the camera's color and high
+resolution by using histograms based on a ``safe region'' at the bottom of the
+image to classify the ground, then use texture to differentiate the smooth
+plastic of obstacles from the roughness of lines painted on the grass.
+Obstacles are then identified by searching upward from the bottom of the image,
+and those pixel co\"ordinates transformed into physical locations relative to
+the camera.  We achieve workable results, reliably differentiating obstacles from grass and accurately classifying nearby lines.
 
 \section{Literature review}
 
-
+Our approach is primarily based on that of Ulrich and Nourbakhsh \cite{ulrich2000appearance}.  They describe a color-histogram--based approach to classifying 
 
 \section{Process}
@@ -67 +92 @@
 
 
+\bibliographystyle{plain}
+\bibliography{references.bib}
+
 \end{document}