This paper presents a method for multi-camera people tracking using a probabilistic occupancy map combined with dynamic programming. The approach effectively tracks up to six individuals across thousands of frames despite significant occlusions and lighting changes. The key contributions are: (1) demonstrating that a generative model can handle occlusions independently, even with poor input data, and (2) showing that multi-person tracking can be reliably achieved by processing individual trajectories separately over long sequences with a reasonable heuristic to avoid confusion. The method uses a generative model to estimate the probabilities of occupancy of the ground plane at each time step, combined with dynamic programming to track people over time. This results in a fully automated system that can track up to six people in a room for several minutes using only four cameras without producing false positives or negatives. The system provides location estimates accurate to within tens of centimeters and maintains performance even when up to 20% of images are lost. The approach involves two main algorithmic steps: (1) estimating the probabilities of occupancy of the ground plane using background subtraction and a generative model, and (2) combining these probabilities with color and motion models using a Viterbi algorithm to track individuals across thousands of frames. The method avoids combinatorial explosion by processing trajectories individually over long sequences and using a greedy approach to rank individuals and avoid confusion. The system is tested on two indoor and four outdoor video sequences, showing excellent performance in tracking individuals despite occlusions and varying lighting conditions. The probabilistic occupancy map is computed separately at each time step and used as input to the dynamic programming approach. The results demonstrate the system's robustness and accuracy in tracking individuals across multiple frames.This paper presents a method for multi-camera people tracking using a probabilistic occupancy map combined with dynamic programming. The approach effectively tracks up to six individuals across thousands of frames despite significant occlusions and lighting changes. The key contributions are: (1) demonstrating that a generative model can handle occlusions independently, even with poor input data, and (2) showing that multi-person tracking can be reliably achieved by processing individual trajectories separately over long sequences with a reasonable heuristic to avoid confusion. The method uses a generative model to estimate the probabilities of occupancy of the ground plane at each time step, combined with dynamic programming to track people over time. This results in a fully automated system that can track up to six people in a room for several minutes using only four cameras without producing false positives or negatives. The system provides location estimates accurate to within tens of centimeters and maintains performance even when up to 20% of images are lost. The approach involves two main algorithmic steps: (1) estimating the probabilities of occupancy of the ground plane using background subtraction and a generative model, and (2) combining these probabilities with color and motion models using a Viterbi algorithm to track individuals across thousands of frames. The method avoids combinatorial explosion by processing trajectories individually over long sequences and using a greedy approach to rank individuals and avoid confusion. The system is tested on two indoor and four outdoor video sequences, showing excellent performance in tracking individuals despite occlusions and varying lighting conditions. The probabilistic occupancy map is computed separately at each time step and used as input to the dynamic programming approach. The results demonstrate the system's robustness and accuracy in tracking individuals across multiple frames.