Multi-Object Tracking and Segmentation (MOTS) is a critical task in autonomous driving, robotic perception, and video analysis. The challenge lies in accurately identifying and associating objects within video sequences, especially when the number of objects is uncertain, motion patterns vary, and frequent object overlaps occur. In this paper, we propose a method with pre-matching and selective association (PS-Track) to adaptively combine motion and appearance cues to cope with continuously changing scenes. Unlike solely relying on the single cue or predetermined combination schemes, our method facilitates data association by discerning similarities in appearance among tracks across different scenarios through the dynamic selection of suitable schemes. Through experimentation, we also found that our method exhibits advantages in tracking efficiency compared to complex models. Our method achieved outstanding results on the MOTS dataset, with scores of 61.0 for sMOTA, 56.4 for IDF1, 76.0 for MOTSA, and 76.5 for FPS.