Accurate recognition of the canopy is a prerequisite for precision orchard yield estimation. This paper proposed an enhanced LA-dpv3+ approach for the recognition of cherry canopies based on UAV image data, with a focus on enhancing feature representation through the implementation of an attention mechanism. The attention mechanism module was introduced to the encoder stage of the DeepLabV3+ architecture, which improved the network’s detection accuracy and robustness. Specifically, we developed a diagonal discrete cosine transform feature strategy within the attention convolution module to extract finer details of canopy information from multiple frequency components. The proposed model was constructed based on a lightweight DeepLabv3+ network architecture that incorporates a MobileNetv2 backbone, effectively reducing computational costs. The results demonstrate that our proposed method achieved a balance between computational cost and the quality of results when compared to competing approaches. Our model’s accuracy exceeded 89% while maintaining a modest model size of only 46.8 MB. The overall performance indicated that with the help of a neural network, segmentation failures were notably reduced, particularly in high-density weed conditions, resulting in significant increases in accuracy (ACC), F1-score, and intersection over union (IOU), which were increased by 5.44, 3.39, and 8.62%, respectively. The method proposed in this paper may be applied to future image-based applications and contribute to automated orchard management.

Early warning of overgrowth in strawberry seedlings is essential to balance vegetative and reproductive growth. However, existing monitoring methods face major challenges, including subtle visual symptoms and limited abnormal samples. To address this, we propose MM-CAPNet, a multimodal fusion framework for early detection of seedling overgrowth. We first developed a representative sample collection of strawberry seedlings through a systematic induction experiment, integrating historical environmental time-series data with contemporaneous plant images. The MM-CAPNet architecture uses a dual-stream design to process these inputs, with a Transformer encoder for environmental sequences and a MobileNetV2 encoder for images. A critical component of the proposed framework lies in the image-guided Cross-Attention mechanism, which uniquely treats the current phenotype as an active query to adaptively retrieve and aggregate the most diagnostically relevant segments of past environmental data. Experiments show MM-CAPNet outperforms baselines, reaching 87.6% accuracy and 0.901 AUC, with strong discriminative ability for early overgrowth categories. Ablation studies confirm its interpretability by linking visual phenotypes to key environmental drivers. This work provides growers with a proof-of-concept framework to regulate fertilization, irrigation, and light management during the nursery stage, thereby reducing the risk of excessive vegetative growth. The proposed framework supports precision cultivation strategies that enhance resource efficiency and crop resilience.