To achieve efficient and accurate detection of blueberry fruit ripeness, this study proposes a lightweight yet high-performance object detection model—CES-YOLO. Designed for real-world blueberry harvesting scenarios, the model addresses key challenges such as significant visual differences across ripeness stages, complex occlusions, and small object sizes. CES-YOLO introduces three core components: the C3K2-Ghost module for efficient feature extraction and model compression, the SEAM attention mechanism to enhance the focus on critical fruit regions, and the EMA Head for improved detection of small and densely packed targets. Experiments on a blueberry ripeness dataset demonstrated that CES-YOLO achieved 91.22% mAP50, 69.18% mAP95, 89.21% precision, and 85.23% recall, while maintaining a lightweight structure with only 2.1 M parameters and 5.0 GFLOPs, significantly outperforming mainstream lightweight detection models. Extensive ablation and comparative studies confirmed the effectiveness of each component in improving detection accuracy and reducing false positives and missed detections. This research offers an efficient and practical solution for automated recognition of fruit and vegetable maturity, supporting broader applications in smart agriculture, and provides theoretical and engineering insights for the future design of agricultural vision models. To further demonstrate its practical deployment capability, CES-YOLO was successfully deployed on the NVIDIA Jetson Orin Nano platform, where it maintained real-time detection performance, with low power consumption and high inference efficiency, validating its suitability for embedded edge computing scenarios in intelligent agriculture.

To develop a high-performance detection system for automated harvesting on resource-limited edge devices, we introduce FSA-DETR-P, a lightweight detection framework that addresses challenges such as illumination inconsistency, occlusion, and scale variation in complex orchard environments. Unlike traditional computationally intensive architectures, this model optimizes real-time detection transformers by integrating an efficient backbone for fast feature extraction, a simplified aggregation structure to minimize complexity, and an adaptive mechanism for multi-scale feature fusion. The optimized backbone improves early-stage texture extraction while reducing computational demands. The streamlined aggregation design enhances multi-level interactions without losing spatial detail, and the adaptive fusion module strengthens the detection of small, partially occluded, or ambiguous fruits. We created a domain-specific pomegranate dataset, expanded to 13,840 images with a rigorous 8:1:1 split for training, validation, and testing. The results show that the pruned and optimized model achieves a Mean Average Precision (mAP50) of 0.928 and mAP50–95 of 0.632 with reduced parameters (13.73 M) and lower computational costs (34.6 GFLOPs). It operates at 24.6 FPS on an NVIDIA Jetson Orin Nano, indicating a strong balance between accuracy and deployability, making it well-suited for orchard monitoring and robotic harvesting in real-world applications.

Papaya ripeness identification is a key task in orchard management. To achieve efficient deployment of this task on edge computing devices, this paper proposes a lightweight detection model, ABD-YOLO-ting, based on YOLOv8. First, the width factor of YOLOv8n is adjusted to construct a lightweight backbone network, YOLO-Ting. Second, a low-computation ADown module is introduced to replace the standard downsampling structure, aiming to enhance feature extraction efficiency. Third, an enhanced BiFPN is integrated into the neck structure to achieve efficient multi-scale feature fusion. Finally, to strengthen the model’s capability in identifying small objects, the dynamic detection head DyHead is introduced to improve ripeness recognition accuracy. On a self-constructed Japanese quince orchard dataset, ABD-YOLO-ting achieves a mAP50 of 94.7% and a mAP50–95 of 77.4%, with only 1.47 M parameters and 5.4 G FLOPs, significantly outperforming mainstream models such as YOLOv5, YOLOv8, and YOLOv11. On edge devices, the model achieves a well-balanced trade-off between detection speed and accuracy, demonstrating strong potential for practical applications in intelligent harvesting and orchard management.

To address persistent problems such as clogging, high digging resistance, incomplete soil removal, and severe pod loss during the operation of shovel-chain peanut harvesters, a hybrid excavation approach was developed based on an in-depth analysis of the mechanical interaction between the peanut plant–soil complex (hereafter referred to as the “complex”) and the harvesting mechanism. The proposed approach integrates vertical and horizontal excavation directions to enhance soil fragmentation and reduce operational resistance. A progressive soil disintegration process was introduced, in which the complex undergoes lateral and longitudinal compression-bending deformation during movement. A driven soil–plant separation scheme was implemented through coordinated operation of upper conveying and lower combing–lifting mechanisms, promoting efficient and continuous material flow. A resistance-reducing digging device consisting of opposing round plow blades and horizontally sliding digging shovels was designed to minimize excavation resistance and soil adhesion. Meanwhile, an anti-clogging separation mechanism, integrating squeezing and feeding rollers and harrow-chain, was developed to improve soil removal and pod separation. Key structural and operational parameters—such as the chain-to-machine speed ratio, tooth-to-chain rotation speed ratio, harrow-tooth spacing ratio, and pushing-tooth transmission ratio—were optimized through theoretical analysis and prototyping. The final design also refined the number of pushing-tooth rows, squeezing and feeding roller geometry, conveying-tooth radius, and the configuration and distribution of rake and stick-tooth shafts. Field experiments were conducted using the developed prototype under sandy loam conditions (11–15% moisture content) with Yu Hua 22 peanut plants (35–40 cm height, 70 cm ridge spacing, 30 cm narrow-row spacing) at a working speed of 1.5–1.6 km·h⁻¹. Results demonstrated that the prototype achieved average ground pod loss, buried pod, and soil carryover rates of 1.13%, 0.95%, and 7.87%, respectively. The entire operation proceeded smoothly without clogging, and continuous conveying of peanut plants was maintained. These findings confirm that the proposed combined excavation and separation system meets and in some respects exceeds the performance requirements for efficient peanut harvesting under typical field conditions.