In this paper, we propose an efficient hybrid Particle Filter (PF) algorithm for video tracking by employing a genetic algorithm to solve the sample impoverishment problem. In the presented method, the object to be tracked is selected by a rectangular window inside which a few numbers of particles are scattered. The particles' weights are calculated based on the similarity between feature vectors of the scattered particles and that of the central particle. Before the resampling stage of PF algorithm, particles with the highest weights are evolved using a genetic algorithm. The evolved particles' coordinates are transferred to the next frame by a random walk model, and the rectangle involving new particles is specified. Moreover, we utilize the idea of partitioning (selecting parts of target in the first frame with a distinct color/texture) and reducing image size to decrease the number of particles. The partitioning idea also helps our method in resolving the occlusion problem. Simulation results demonstrate the outperformance of the suggested approach comparing with other methods in terms of precision and tracking time when it encounters with the challenges such as full and partial occlusions, illumination and scale variations, fast motions, and color similarity between the object and background.