Unlike other metaheuristics, differential Evolution (DE) employs a crossover operation filtering variables to be mutated, which contributes to its successful applications in a variety of complicated optimization problems. However, the underlying working principles of the crossover operation is not yet fully understood. In this paper, we try to reveal the influence of the binomial crossover by performing a theoretical comparison between the $(1+1)EA$ and its variants, the $(1+1)EA_{C}$ and the $(1+1)EA_{CM}$. Generally, the introduction of the binomial crossover contributes to the enhancement of the exploration ability as well as degradation of the exploitation ability, and under some conditions, leads to the dominance of the transition matrix for binary optimization problems. As a result, both the $(1+1)EA_{C}$ and the $(1+1)EA_{CM}$ outperform the $(1+1)EA$ on the unimodal OneMax problem, but do not always dominate it on the Deceptive problem. Finally, we perform exploration analysis by investigating probabilities to transfer from non-optimal statuses to the optimal status of the Deceptive problem, inspired by which adaptive strategies are proposed to improve the ability of global exploration. It suggests that incorporation of the binomial crossover could be a feasible strategy to improve the performances of randomized search heuristics.