deletions | additions       diff --git a/section_Literature_Review_subsection_Computer__.tex b/section_Literature_Review_subsection_Computer__.tex  index c23ea58..65147b6 100644  --- a/section_Literature_Review_subsection_Computer__.tex  +++ b/section_Literature_Review_subsection_Computer__.tex 
...
\subsection{Computer Vision in Traffic Applications}  Computer vision is used extensively in traffic applications as an instrument of data collection and monitoring. The two primary branches of computer vision in traffic applications include presence detection (sometimes referred to as virtual loops) and tracking. Presence detection has widespread commercial application due to its relatively high degree of reliability, on par with more common sensors such as inductive loops; its primary application is in providing traffic counts, queue lengths, and basic presence detection. Tracking is a more complex application which aims to extract the road users' trajectories within the camera field of view, from which velocity and acceleration may be derived: it is therefore generally less reliable than presence detection systems. There are three main categories of tracking methods: 1) tracking by detection, which typically relies on background subtraction to detect foreground objects and appearance-based object classification \cite{Zangenehpour_2015}; 2) tracking using flow, also called feature-based tracking \cite{saunier06feature-based}, first introduced in \cite{Coifman_1998}; and 3) tracking with probability based on Bayesian tracking frameworks. The NGSIM project was one of the first large-scale video data collection projects making use of semi-automated vehicle tracking from freeway and urban arterials video data to obtain vehicle trajectories for traffic model calibrations \cite{Kim_2005}. Surrogate safety analysis also makes use of trajectory data, for example with the early SAVEME project \cite{Ervin_2000,Gordon_2012}, and now more recently with extensive open source projects such as Traffic Intelligence \cite{saunier06feature-based,Jackson_2013}. \subsection{Tracking Optimisation}  The work done to optimize parametrization of the various trackers is sparse and usually set manually from experimental results. The instances of automated calibration in \cite{Sidla_2006} and \cite{Ali_2009} used Adaboost training strictly for shape detectors and \cite{P_rez_2006} used evolutionary optimization for the segmentation portion. One of the only cases of systematic improvement of the model as a whole is through evolution algorithms performed on video of pedestrians \cite{Ettehadieh_2014}. There is also a lack of validation of the optimized parameters under different conditions than the data set on which the calibration was done.