本文提出一種全新的圖像分割方法——連通線多級切割方法，并在此基礎上建立圖像NMI特征的目標識別與跟蹤算法。文章給出了運用連通線多級切割方法實現閾值求取、圖像分割、目標識別與跟蹤的技術技巧。通過實驗驗證了NMI特征的縮放不變性、旋轉不變性和平移不變性以及不同物體的NMI特征差異，表明本文算法具有良好的實時性和精確的目標識別與跟蹤效果。
關鍵詞 不變特征 閾值 OTSU算法 圖像分割 NMI特征 目標識別
Abstract In this paper, a new method of Image Segmentation, namely Consecutive Line Multilevel Segmentation, and the algorithm of Target Recognition and Tracking based on Normalized Moment of Inertia (NMI) feature are introduced. The definition of Centre of Mass and NMI feature is proposed and the principle and methods of Threshold calculation, Image segmentation, Target recognition and tracking are presented. Then, the scaling invariability, rotation invariability, translation invariability of NMI feature and the NMI feature difference of various objects are validated by experiment. The experimental results show that the algorithm discussed in this paper is very practical with good real time and accurate effect of Target recognition and tracking.
Key words Invariant feature/ Threshold/ OTSU algorithm/ Image segmentation/ NMI feature/ Target recognition