With the rapid development of spatio-temporal crowdsourcing, its application scenarios are becoming increasingly complex, often involving multiple subtasks with sequential dependencies. This paper defines such workloads as multi-stage spatio-temporal crowdsourcing tasks with sequential dependencies. Traditional centralized crowdsourcing platforms face significant issues, including single points of failure, privacy risks, insufficient fairness, and high intermediary fees, making them unsuitable for complex tasks. To address these challenges, we propose a blockchain-based decentralized system architecture for multi-stage spatio-temporal crowdsourcing as an alternative to conventional platforms. In the resource-constrained blockchain environment, we design an efficient worker selection algorithm that verifies subtask dependencies in advance to prevent invalid assignments and minimize task costs. Additionally, we introduce a truthful incentive mechanism based on the VCG reverse auction to ensure workers report their costs honestly, balancing task requester interests with economic efficiency. Experimental results demonstrate the approach’s advantages in performance, cost-effectiveness, and incentive compatibility, confirming its feasibility and practical value.