Digital intelligence education transformation imposes new requirements on plant landscape teaching. Traditional teaching has obvious limitations in plant cognition, scheme design and assessment. This study adopts VR virtual simulation and intelligent plant databases to build a three-stage blended teaching system including online intelligent autonomous learning, offline on-site project training, and virtual-real iterative evaluation. A full-process multidimensional quantitative assessment mechanism is established to evaluate student competencies objectively. Three-year teaching practice shows that the experimental class achieves 43.8% higher plant identification accuracy and 94.7% design drawing compliance than the control class. The model effectively breaks spatiotemporal constraints in traditional teaching and improves students’ plant recognition, ecological design thinking and digital design abilities.