基於超像素追蹤的十字路口車輛軌跡分析

Crossroad Vehicle Trajectory Analysis using Superpixel Tracking

許晉豪(Chin-Hao Hsu)

指導教授 ： 林道通

國立臺北大學/電機資訊學院/資訊工程學系/碩士(2013年)

摘要

物件追蹤在智慧型影像監控中是一種重要技術。本篇論文提出一種新的車輛追蹤演算法，包含了自適性混和背景模型、前景切割，並用卡曼濾波器來預測物件運動行為。此外，我們使用了以超像素為基礎的物件模型進行物件比較的運算。由於使用超像素切割，我們的方法能夠根據物件外觀的空間關係來擷取出局部外觀特徵。此外，為了改善交通，分析車輛交通的資料是必須的。本篇論文提出了一個系統，基於前述車輛追蹤演算法所提取出的車輛軌跡來建立軌跡圖，進而能偵測從直線前進車道違規左轉的車輛。

關鍵字

車輛追蹤 ； 超像素比對 ； 影像監控 ； 軌跡分析 ； 軌跡圖

並列摘要

Vehicle tracking is an important application domain of intelligent video surveillance. This paper presents a novel vehicle-tracking algorithm. The proposed algorithm includes an adaptive background mixture model modeling and segmentation, and Kalman filter to predict object motion. Furthermore, superpixel-based object model is used for object matching. By using superpixel segmentation, the proposed approach is able to capture the local appearance characteristics of objects based on their spatial relationships. For improving transportation, it is necessary to analyze the vehicle traffic data. This paper proposed a system to detect invalid-left-turn from a forward-only-lane using the trajectory map which is constructed based on the trajectory of vehicles extracted by the vehicle-tracking algorithm we proposed.

並列關鍵字

Vehicle tracking ； superpixel matching ； video surveillance ； trajectory analysis ； trajectory map

參考文獻

[4] K. Kalaitzakis V. Kastrinaki, M. Zervakis. A survey of video processing techniques for traffic applications. Image and Vision Computing 21, pages 359–--381, 2003.

[6] Thou-Ho Chen, Yu-Feng Lin, and Tsong-Yi Chen. Intelligent vehicle counting method based on blob analysis in traffic surveillance. In Second International Conference on Innovative Computing, Information and Control, 2007. ICICIC'07., pages 238--238. IEEE, 2007.

[10] Daw-Tung Lin and Kai-Yung Huang. Collaborative pedestrian tracking and data fusion with multiple cameras. IEEE Transactions on Information Forensics and Security., 6(4):1432--1444, 2011.

1252--1261, 1999.

[12] Tsuyoshi Yamane, Yoshiaki Shirai, and Jun Miura. Person tracking by integrating optical flow and uniform brightness regions. In IEEE International Conference on Robotics and Automation., volume 4, pages 3267--3272. IEEE, 1998.