抄録
Precision seeding and harvesting are increasingly evolving from equipment-based automation toward data-driven, closed-loop field operations, yet the integration of sensing, intelligent decision-making, and adaptive control across the two critical stages remains fragmented. This review establishes an integrated framework of multi-source perception–intelligent decision-making–adaptive control–real-time feedback to systematically synthesize recent advances in precision seeding and harvesting. For precision seeding, we examine multi-source sensing, variable-rate seeding, and machine vision- and deep learning-based monitoring of seeding quality, with emphasis on how heterogeneous field information is transformed into real-time seeding decisions and control actions. For precision harvesting, we evaluate crop maturity detection, yield mapping, and adaptive regulation of harvesting parameters, highlighting the role of multi-sensor fusion and edge computing in enabling responsive harvesting operations. Across both processes, we identify data heterogeneity, interoperability among machinery and sensing systems, model generalizability, and the balance between technological complexity and economic benefits as major barriers to large-scale deployment. Based on these findings, we propose a transition pathway from isolated precision operations toward agronomy–machinery integration, digital-twin-enabled process optimization, and autonomous field systems. By unifying precision seeding and harvesting within a common perception–decision–control framework, this review clarifies the technological logic and key bottlenecks underlying intelligent field operations and provides a systems-level basis for advancing precision agriculture from component-level automation toward autonomous, adaptive production.
