<p>Vehicle tracking in traffic scenarios requires adaptation to varying illumination conditions, while also facing challenges such as similar objects, occlusions, and fast motion. This paper presents a visual tracker that makes a cooperative combination of motion estimation and appearance recognition. To adapt to the target vehicle’s motion and appearance variations, the proposed tracker pre-establishes a Kalman filter-based motion model and a dynamic appearance template using the vehicle’s initial state, and updates both adaptively during tracking. To precisely track the target vehicle under traffic distractions, the motion model first predicts its prior location and then uses it to delineate an adaptive search region in the current frame image. Within this region, candidate vehicles are rapidly detected to undergo location association and appearance matching with the target vehicle’s prior state to calculate the corresponding location and appearance similarity metrics. Further, these metrics are fused via a joint evaluation strategy to judge each candidate vehicle’s overall similarity relative to the target vehicle. Based on the evaluation, the tracker selects the optimal candidate or retains the prior motion estimation as the tracking result, achieving the robust localization of the target vehicle. The proposed tracker’s tracking accuracy (AUC) on the vehicle videos within the UA-DETRAC and UAV123 datasets reaches 81.56% and 78.90%, respectively, outperforming other mainstream trackers and demonstrating superior performance in traffic scenarios.</p>