Abstract
INTRODUCTION: Accurate orchard yield data are essential for economic assessment and management optimization, but traditional manual estimation is labor-intensive and often inaccurate for modern precision orchard management. METHODS: Following the core principles of the PRISMA guidelines, this systematic review summarizes recent domestic and international studies on orchard yield estimation and compares machine vision, remote sensing, and multi-source data fusion methods from both methodological and application perspectives. RESULTS: Existing studies show that machine vision and remote sensing can effectively support automated orchard yield estimation, while multi-source heterogeneous data fusion generally improves robustness and estimation accuracy by integrating fruit detection information with canopy physiological and structural traits. DISCUSSION: Despite rapid progress, major challenges remain, including fruit occlusion and missed detection, data heterogeneity, and limited automation and adaptability across orchard scenarios. Future work should strengthen advanced algorithm integration, promote multi-modal data fusion, and develop intelligent, automated yield-estimation platforms for diverse orchard environments.