Segmentation and classification of streaks in a large-scale particle streak tracking system

In this paper we propose improvements to particle streak tracking. The main focus is an improved segmentation method considering local image statistics combined with classic edge detection and removal of the scene background, which at the same time suppresses the noise. Furthermore, we will describe the streak features used and suggest a greedy algorithm to solve the classification problem of binarized streak segments. Finally, results using the presented algorithm are compared to those using the original particle streak system.