This paper presents a robust object tracking algorithm that combines a generative model and a discriminative classifier. The main challenge addressed is the drastic appearance changes in complex and dynamic scenes, such as heavy occlusions, illumination changes, and complex backgrounds. The proposed method uses intensity values for representation, leveraging both holistic templates and local histograms. The discriminative classifier (SD-C) module introduces an effective method to compute confidence values, assigning more weight to foreground than background. The generative model (SGM) module proposes a novel histogram-based method that considers spatial information and handles occlusions. The update scheme integrates both latest observations and original templates, enabling the tracker to effectively deal with appearance changes and reduce drifts. Experiments on various challenging videos demonstrate the proposed tracker's superior performance compared to state-of-the-art algorithms.This paper presents a robust object tracking algorithm that combines a generative model and a discriminative classifier. The main challenge addressed is the drastic appearance changes in complex and dynamic scenes, such as heavy occlusions, illumination changes, and complex backgrounds. The proposed method uses intensity values for representation, leveraging both holistic templates and local histograms. The discriminative classifier (SD-C) module introduces an effective method to compute confidence values, assigning more weight to foreground than background. The generative model (SGM) module proposes a novel histogram-based method that considers spatial information and handles occlusions. The update scheme integrates both latest observations and original templates, enabling the tracker to effectively deal with appearance changes and reduce drifts. Experiments on various challenging videos demonstrate the proposed tracker's superior performance compared to state-of-the-art algorithms.