We propose a target identification algorithm for visual tracking. Target identification is realized by finding out corresponding line segments to the hypothesized model segments of the target. The key idea is the combination of the Mahalanobis distance with the geometrical relationship between model segments and extracted line segments.