Figure 7.

The template searching process by CEM. The search process by CEM involves sampling, evaluation, and updating. Sampling: samples are drawn from a Gaussian distribution. Evaluation: each sample point is evaluated, and elite samples are selected based on their evaluation results. Updating: the sampling range is updated based on the elite samples. This process is repeated until the sampling range becomes smaller than a predefined threshold. The mean of the optimized sampling range is then used as the identified latent feature to generate the template for matching.