The rapid advancement of Unmanned Aerial Vehicle (UAV) technology has led to increasingly widespread applications of Multi-UAV Systems (MUS) in both military and civilian domains. Efficient task allocation strategies are crucial for MUS to successfully accomplish complex missions. Traditional centralized task allocation methods are ill-suited to large-scale, dynamic, and distributed environments. To address this issue, a distributed task allocation algorithm based on negotiation and consensus, termed NCDTA (Negotiation and Consensus-based Distributed Task Allocation Algorithm), is proposed. This algorithm adopts a two-stage task allocation framework: first, leveraging the Contract Net Protocol (CNP) for rapid initial task assignment; and subsequently, utilizing the Consensus-Based Bundle Algorithm (CBBA) for task adjustment and optimization, incorporating an obstacle avoidance strategy based on minimizing heading angle cost to achieve adaptivity to dynamic environments, task requirements, and the safe operation of UAVs. Simulation results demonstrate that, compared to standalone CNP and CBBA algorithms, NCDTA enhances task completion rates, reduces average task execution times and average flight distances, while simultaneously ensuring safe separation distances between UAVs, thereby effectively addressing the complexities of multi-UAV task allocation in dynamic and intricate environments.