To address the conflict between platoon safety and communication efficiency for connected vehicle platoons under false data injection (FDI) attacks and limited communication resources, a neural network-based dynamic event-triggered resilient control strategy integrating explicit attack parameter modeling and platoon error constraints is proposed. A distributed output-feedback framework incorporating time-varying attack characteristics is constructed, and a radial basis function (RBF) neural network state observer is designed to reconstruct unmeasurable vehicle states and achieve online approximation and active compensation of attack signals. A dynamic event-triggered condition driven jointly by attack intensity and neural approximation error is developed to establish an adaptive matching mechanism between attack dynamics and communication strategies, balancing attack suppression and communication optimization. Furthermore, an attack-parameter-dependent Lyapunov functional is constructed, and the semi-global uniform ultimate boundedness (SGUUB) of all signals in the closed-loop system is strictly proven via linear matrix inequalities (LMIs), which reduces the conservatism of stability criteria. Theoretical analysis and simulation results demonstrate that under time-varying FDI attacks, the maximum platoon position error remains below 2 m, and the communication load is reduced by 96.4% compared to time-triggered schemes, verifying the superiority of the proposed strategy in ensuring platoon safety and optimizing communication resources.