Vision-Based Turning Movement Counting at Intersections by Cooperating Zone and Trajectory Comparison Modules

Mohammad Shokrolah shirazi and Brendan Morris

in IEEE 17th international conference on Intelligent Transportation Systems, (ITSC 2014), October, 2014, Qindao, China.

Abstract

This work presents a vision-based vehicle tracking system with an improved method for automatically obtaining turning movement counts at intersections. Improved accuracy is obtained for intersections of varying difficulty through cooperation between complimentary counting modules. When vehicle tracking is robust, a typical zone comparison module quickly accounts for predefined image regions. When vehicle tracking is broken due to occlusion or noise, a trajectory comparison module utilizing the least common subsequence distance is able to count the broken trajectories using the typical scene paths. Experimental evaluations during evening peak hours at two different intersections show an 15% average improvement at two intersections where the trajectory comparison module is in use 22% of time. Finally, counts of five working days are compared for another intersection in 15 minute intervals to highlight the system’s operational utility for turning behavior analysis at high temporal resolution.

Goal

The system provides high accuracy in turning movement count using cooperation of zone and trajectory comparison modules. A zone comparison module examines which image regions are traversed. Unfortunately,

this simple method does not work well in the following situations:

1) When a track is incomplete as might occur when two vehicles move from a stop bar in unison and cannot be distinguished until they get separated.

2) When the centroid of trajectory falls into an undefined region due to various noise sources (e.g. occlusion or poor background subtraction). The trajectory comparison module complements the zone

module by addressing these situations by considering the full trajectory for counting purposes.

Zone comparison module

Predefined regions in the intersection image, called zones, are defined manually to specify the four cardinal directions {north, south, east, west} and the central intersection. The zones are defined based on the set of lines drawn on the stop bars of each intersection direction. The zones are used to define a regular sequence (RS) set, that is, the set of acceptable zone traversals. Examples of the zones for two intersections are displayedbelow.

The zones are used to define a regular sequence (RS) set, that is, the set of acceptable zone traversals. The tracker only keeps record or transitions between zones (when the current tracked zone changes) to build the track zone sequence. If the resulting zone sequence exists in the regular sequence set for the intersection, a counter for the associated TM count is incremented. A zone flag is set to indicate if tracking was successful for cooperation with the trajectory comparison module.

Trajectory comparison module

Although counting by zone comparison is simple, it is unsuccessful when the obtained sequence of zone numbers is not member of the RS. This happens during undesired situations like occlusion. Trajectories and typical paths are compared using like longest common substring (LCSS) because of its robustness to noise and outliers. The typical paths are shown for INT 1.

Results

Experimental evaluations during evening peak hours at two different intersections show an 15% average improvement at two intersections where the trajectory comparison module is in use 22% of time. Finally, counts of five working days are compared for another intersection in 15 minute intervals to highlight the system’s operational utility for turning behavior analysis at high temporal resolution. The turning movement results of five days are shown for turning lefts.

Video