add research

1 files changed, 65 insertions, 2 deletions

diff --git a/doc/dui.md b/doc/dui.md
index 6260fb5..e5a425f 100644
--- a/doc/dui.md
+++ b/doc/dui.md
@@ -34,8 +34,8 @@ description in section \ref{problem-statement}.
   \begin{enumerate}
     \item Driving when not on a road is not allowed
     \item The vehicle can follow a road by steering itself accordingly
-    \item Driving off the road is only allowed when necessary for the camera to
-          keep seeing the road
+    \item \label{req:offroad} Driving off the road is only allowed when
+    necessary for the camera to keep seeing the road
   \end{enumerate}
   \item \label{req:traffic-lights}
   The vehicle handles traffic lights in the following way
@@ -237,6 +237,38 @@ For more information about Radon transform algorithms check the below links:
 - \citetitle{matlab:radon}
 - \citetitle{opencv:radon}
 
+#### Blob detection for roads
+
+As mentioned in section \ref{specifications} requirement \ref{req:offroad}, the
+car is allowed to drive off-road to keep the road visible. This means that a
+naive approach where the car drives towards where 'the most road' is could
+suffice for our road detection needs.
+
+A simple prototype for this approach was made using Matlab, shown in figure
+\ref{fig:matlab-roaddetect}. The top part of the figure shows the raw camera
+image (flipped), with a gray line down the middle, and a red arrow showing the
+steering value. The red arrow is the only 'output' of this algorithm.
+
+The bottom part of the figure shows the detected blobs (green bounding boxes)
+on a copy of the original top image with the following transforms:
+
+1. Reverse perspective-transform
+2. Gaussian blur (3x3 kernel) to smooth out any noise caused by the floor
+   texture
+3. Threshold to match most of the light image parts (road)
+
+The steering value (red arrow) is calculated by averaging the horizontal screen
+position (normalized to between -1 and 1) using a weight factor calculated by
+using each blobs bounding box area. The weight factor has a minimum 'base'
+value that is added, and has a maximum value so large blobs don't 'overpower'
+smaller blobs. This is so the inside road edge of a turn doesn't get lost
+because the outer edge has a larger bounding box.
+
+![Road detection prototype in Matlab](../assets/blob_invpers.pdf){#fig:matlab-roaddetect}
+
+The implementation of this road detection algorithm is provided in the source
+code tree.
+
 ### Which algorithm is suitable for our project?
 
 We have identified four different types of line detection algorithms that could potentially be used for our project. To decide on the best algorithm, we need to consider various factors such as accuracy, efficiency, and ease of use. While processing time is not a critical factor in our case, we need to ensure that the algorithm we choose meets our requirements and is accessible through the platform we are using, which is currently openMV but may change to openCV in the future. Therefore, our priority is to select an algorithm that is easy to implement, provides accurate results, and is compatible with our platform.
@@ -516,6 +548,37 @@ solution (this requires testing).
 }
 \signRecognitionConclusion
 
+## Traffic light detection using blobs
+
+A simple traffic light detection algorithm using blob detection can take
+advantage of the following properties of traffic lights (as seen by the Nicla
+module):
+
+- Traffic lights are mostly dark
+- Traffic lights are generally rectangular
+- Traffic lights have fixed spots where the saturation and hue value ranges are
+  known if the light is on
+
+The algorithm has the following steps:
+
+1. Apply a threshold to keep only dark areas
+2. Prune any blobs with too little surface area
+3. Prune any blobs that aren't a vertical rectangle with approximately the same
+   aspect ratio as a traffic light
+4. Poke three points down the center of each blob, at 20%, 50%, and 80% of the
+   blob's height
+5. Check if any of the three points has a matching hue and saturation range of
+   a lit up light
+6. The first point that matches is deemed to be the traffic light's color, if
+   no points match, it's probably not a traffic light
+
+Figure \ref{fig:matlab-trafficlight} shows this algorithm on an example image:
+
+![Traffic light detection prototype in Matlab](../assets/blob_traffic_lights.pdf){#fig:matlab-trafficlight}
+
+The implementation of this road detection algorithm is provided in the source
+code tree.
+
 # Conclusion
 
 \communicationConclusion