Motion target correlation method based on structural similarity in camera array

Aiming to solve the problem that there is limited scope of the motion target correlation method based on geometric constraint in camera array, a novel motion target correlation method based on structural similarity is proposed. Firstly, the target segmentation is addressed on low resolution clear target image in subarray by exploiting the Gaussian mixture model (GMM), then pre-correlation is carried out by utilizing the template matching, finally utilizing the target images' brightness, contrast and structure contrast, the correlation of the same target in different camera images is completed based on structural similarity in order to estimate the continuous motion information of the target. Experiments show that our algorithm can effectively accomplish the correlation of the same moving target between single-target and multi-target scene with high robustness.