A Profiling Based Approach to Safety Surrogate Data Collection

This study identifies the challenges and requirements for a surrogate safety data collection system and provides a robust methodology for surrogate safety data collection. The current study develops a comprehensive methodology for obtaining objective and detailed vehicle profile data from video. This data is used to generate surrogate measures that may be employed for the evaluation of potential safety benefits of improvement strategies at an intersection. Methods for the collection of three surrogate safety measures have been developed and evaluated in this research: Acceleration-Deceleration profiles, Post Encroachment Time (PET), and intersection approach speed. A custom semi-manual video processing software is developed for the purpose of efficiently reducing the video to a useable format ready for analysis. This semi-manual approach allows for the use of lower camera angles with larger perspective views thereby limiting equipment needs, a limitation of most of the automatic video detection equipment based approaches. Video recording is at 30 hz resulting in the generation of discrete position-time trajectory data. This process of generating discrete data and the error inherent in manually identifying the key frames required in the data reduction method causes noise in the data. Hence, low pass filters are developed to smooth the noise in the speed and acceleration-deceleration data of sampled vehicles. To demonstrate the developed methodologies the results for a high speed rural intersection in North Georgia are presented. The video reduction methodology and smoothing algorithms are validated with vehicle traces from global positioning system (GPS) instrumented vehicles. Sampling guidelines are also presented to minimize the required manual effort to reduced the video and determine representative surrogate measures.