This paper presents novel application of automated video analysis for the Before/After safety evaluation of a scramble phase treatment. Data availability has been a common challenge to pedestrian studies, especially for proactive safety analysis. The traditional reliance on collision data has many shortcomings in terms of the quality and quantity of collision record. Qualitative and quantitative issues with road collision data are more pronounced in pedestrian safety studies. In addition, little information could be drawn from collision reports regarding the implicated mechanism of action. Traffic conflict techniques have been advocated as a supplement to or an alternative to collision-based safety analysis. Automated conflict analysis has been advocated as a new safety analysis paradigm that empowers the drawbacks of survey-based and observerbased traffic conflict analysis. One of the focus areas of pedestrian safety that could greatly benefit from vision-based road user tracking is before-and-after (BA) evaluation of safety treatments. This paper demonstrates the feasibility of conducting BA analysis using video data collected from a commercial-grade camera in Chinatown, Oakland, California. Video sequences for a period of two hours before and two hours after scramble were automatically analyzed. The

Video-based collection of traffic data is on the rise. Camera calibration is a necessary step in all applications to recover the real-world positions of the road users of interest that appear in the video. Camera calibration can be performed based on feature correspondences between the realworld space and image space as well as appearances of parallel lines in the image space. In urban traffic scenes, the field of view may be too limited to allow reliable calibration based on parallel lines. Calibration can be complicated in the case of incomplete and noisy data. It is common that cameras monitoring traffic scenes are installed before calibration was undertaken. In this case, laboratory calibration, which is taken for granted in many current approaches, is impossible. This work addresses various real world challenging cases, for example when only video recordings are available, with little knowledge on the camera specifications and setting location, when the orthographic image of the intersection is outdated, or when neither an orthographic image nor a detailed map is available. A review of the current methods for camera calibration reveals little attention to these practical challenges that arise when studying urban intersections to support applications in traffic engineering. This study presents the development details of a robust

The video data used in this research was obtained in a previous study conducted at the Institute of Transportation Studies, UC Berkeley Traffic Safety Center. The key contact persons were Jenna Hua and Prof. David Ragland. The authors would like to sincerely thank them. Observation of the traffic conflicts

Amid concerns for the environment and public health, there has been recently a renewed emphasis on active modes of transportation, i.e. walking and cycling. However, these modes have traditionally received research and practice focus secondary to motorized modes. There is consequently a lack of pedestrian data, in particular microscopic data, to meet the analysis and modeling needs. For instance, accurate data on individual stride length is not available in the transportation literature. This paper proposes a simple method to extract automatically pedestrian stride frequency and length from video data collected non-intrusively in outdoor urban environments. Pedestrian walking speed oscillates during each stride, which can be identified through the frequency analysis of the speed signal. The method is validated

In this paper, we introduce MMTrack, a hybrid single pedestrian tracking algorithm that puts together the advantages of descriptive and discriminative approaches for tracking. Specifically, we combine the idea of cluster-based appearance modeling and online tracking and employ a max-margin criterion for jointly learning the relative importance of different cues to the system. We believe that the proposed framework for tracking can be of general interest since one can add or remove components or even use other trackers as features in it which can lead to more robustness against occlusion, drift and appearance change. Finally, we demonstrate the effectiveness of our method quantitatively on a real-world data set. 1