In facility horticultural production, intelligent disease recognition and precise intervention are vital for crop health and economic efficiency. We construct a multi-source dataset from Bayan Nur, Weifang, and Honghe that integrates handheld camera photos, drone field images, and laboratory-controlled samples. Handheld images capture fine lesion texture for close-up diagnosis common in greenhouses; drone images provide canopy-scale patterns and spatial context suited to open-field management; laboratory images offer controlled illumination and background for stable supervision and cross-crop feature learning. Our objective is robust cross-crop, cross-regional diagnosis and economically rational control. To this end, a model named CCGD-Net is proposed. It is designed as a multi-task framework. The framework incorporates a multi-scale perception module (MSFE) to produce hierarchical representations. It includes a cross-domain alignment module (CDAM) that reduces distribution shifts between greenhouse and open-field environments. The training follows an unsupervised domain adaptation setting that uses unlabeled target-region images. When such images are not available, the model functions in a pure generalization mode. The framework also integrates a regional economic strategy module (RESM) that transforms recognition outputs and local cost information into optimized intervention intensity. Experiments show an accuracy of 91.6%, an F1-score of 89.8%, and an mAP of 88.9%, outperforming Swin Transformer and ConvNeXt; removing RESM reduces F1 to 87.2%. In cross-regional testing (Weifang training → Honghe testing), the model attains an F1 of 88.0% and mAP of 86.5%. These results indicate that integrating complementary imaging modalities with domain alignment and economic optimization provides an effective solution for disease diagnosis across greenhouse and field systems.

A multimodal interactive system, HortiVQA-PP, is proposed for horticultural scenarios, with the aim of achieving precise identification of pests and their natural predators, modeling ecological co-occurrence relationships, and providing intelligent question-answering services tailored to agricultural users. The system integrates three core modules: semantic segmentation, pest–predator co-occurrence detection, and knowledge-enhanced visual question answering. A multimodal dataset comprising 30 pest categories and 10 predator categories has been constructed, encompassing annotated images and corresponding question–answer pairs. In the semantic segmentation task, HortiVQA-PP outperformed existing models across all five evaluation metrics, achieving a precision of 89.6%, recall of 85.2%, F1-score of 87.3%, $mAP@50$ of 82.4%, and IoU of 75.1%, representing an average improvement of approximately 4.1% over the Segment Anything model. For the pest–predator co-occurrence matching task, the model attained a multi-label accuracy of 83.5%, a reduced Hamming Loss of 0.063, and a macro-F1 score of 79.4%, significantly surpassing methods such as ASL and ML-GCN, thereby demonstrating robust structural modeling capability. In the visual question answering task, the incorporation of a horticulture-specific knowledge graph enhanced the model’s reasoning ability. The system achieved 48.7% in BLEU-4, 54.8% in ROUGE-L, 43.3% in METEOR, 36.9% in exact match (EM), and a GPT expert score of 4.5, outperforming mainstream models including BLIP-2, Flamingo, and MiniGPT-4 across all metrics. Experimental results indicate that HortiVQA-PP exhibits strong recognition and interaction capabilities in complex pest scenarios, offering a high-precision, interpretable, and widely applicable artificial intelligence solution for digital horticulture.