Smart Phone Accuracy of Multi-Camera Pedestrian Tracking in Overlapping Fields of View

In this paper, we analyze the suitability of event mapping via smart phones by evaluating the accuracy of pedestrian tracking. We employ a multiple view tracking and bundle adjustment and recover the scale by an alignment to reference points measured via a total station. The use of multiple perspectives by loosely time synchronized smart phones, enables a least squares assignment of detections while improving the pose estimation of pedestrians and reducing occlusions. Finally, we used a total station to track a pedestrian to evaluate our results and show that the photogrammetric smart phone tracking accuracy of 20 cm is suitable for event mapping in the daily traffic situations.