Improved traffic detection and increasing computer processing capabilities present a unique opportunity for traffic engineers and researchers to develop real-time predictive models. These models demand new data collection techniques which provide the level of accuracy required for validation and calibration. Automated collection of turning movements at an intersection is difficult and expensive. Accurate observation of the turning movements of a complex intersection is a relatively straightforward process providing a reliable surveyor is equipped with a proprietary data logging device. Simultaneous observation of sets of turning movements throughout a network, however, is not so easy. Many observers collecting data complicate the data entry and collation process. The use of video surveillance and subsequent collection and verification of turning movements collected at a later time has been used to reduce the number of surveyors required and provide a verifying record of the turning movements. This technology has problems of its own including determining a suitable location for video surveillance such that all movements are collected accurately and without being obscured by other vehicles and difficulties in night conditions. This paper is a description of how a network wide turning movement survey technique was tested, improved and eventually implemented both for a manual survey and a video survey. The paper describes how a series of pilot surveys served to define the structure and organization of the final survey, or "Survey Proper". An explanation of how innovative booking sheets evolved for the manual survey is followed by an account of how subsequent surveys modified the technique. The paper concludes with some critical comments that suggest how future surveys may be designed and discusses the benefits and disadvantages of video versus manual turning movement surveys. (A)
Abstract