In precision agriculture, unmanned aerial vehicles (UAVs) are playing an increasingly important role in farmland information acquisition and fine management. However, discrete obstacles in the farmland environment, such as trees and power lines, pose serious threats to the flight safety of UAVs. Real-time detection of the attributes of obstacles is urgently needed to ensure their flight safety. In the wake of rapid development of deep learning, object detection algorithms based on convolutional neural networks (CNN) and transformer architectures have achieved remarkable results. Detection Transformer (DETR) and Deformable DETR combine CNN and transformer to achieve end-to-end object detection. The goal of this work is to use Deformable DETR for the task of farmland obstacle detection from the perspective of UAVs. However, limited by local receptive fields and local self-attention mechanisms, Deformable DETR lacks the ability to capture long-range dependencies to some extent. Inspired by non-local neural networks, we introduce the global modeling capability to the front-end ResNet to further improve the overall performance of Deformable DETR. We refer to the improved version as Non-local Deformable DETR. We evaluate the performance of Non-local Deformable DETR for farmland obstacle detection through comparative experiments on our proposed dataset. The results show that, compared with the original Deformable DETR network, the mAP value of the Non-local Deformable DETR is increased from 71.3% to 78.0%. Additionally, Non-local Deformable DETR also presents great performance for detecting small and slender objects. We hope this work can provide a solution to the flight safety problems encountered by UAVs in unstructured farmland environments.

An improved YOLOv5 algorithm for the efficient recognition and detection of asparagus with a high accuracy in complex environments was proposed in this study to realize the intelligent machine harvesting of green asparagus. The coordinate attention (CA) mechanism was added to the backbone feature extraction network, which focused more attention on the growth characteristics of asparagus. In the neck part of the algorithm, PANet was replaced with BiFPN, which enhanced the feature propagation and reuse. At the same time, a dataset of asparagus in complex environments under different weather conditions was constructed, and the performance variations of the models with distinct attention mechanisms and feature fusion networks were compared through experiments. Experimental results showed that the mAP_@0.5 of the improved YOLOv5 model increased by 4.22% and reached 98.69%, compared with the YOLOv5 prototype network. Thus, the improved YOLOv5 algorithm can effectively detect asparagus and provide technical support for intelligent machine harvesting of asparagus in different weather conditions and complex environments.

Lightweight detection method for lotus seedpod in natural environment

Efficient scheduling of agricultural machinery is critical for optimizing resource utilization and reducing operational costs in modern farming operations. This study proposes an Adaptive Genetic Algorithm integrated with Ant Colony Optimization (AGA-ACO) to solve the multi-task machinery scheduling problem. The problem is formulated as a Vehicle Routing Problem with Time Windows (VRPTW), considering time constraints, machinery heterogeneity, and task dependencies. The AGA-ACO algorithm employs a two-phase optimization strategy: genetic algorithms for global exploration and ant colony optimization for local refinement through pheromone-guided search. Experimental evaluation using real-world agricultural data from Hangzhou demonstrates that AGA-ACO achieves cost reductions of 5.92–10.87% compared to genetic algorithms, 5.47–7.75% compared to ant colony optimization, and 6.23–9.51% compared to particle swarm optimization, while converging with fewer iterations. The algorithm maintains stable convergence and high robustness across different farmland scales, reducing computational time while preserving solution quality. A scheduling management system integrating IoT sensors, MQTT protocols, and GIS technologies validates the practical applicability of the proposed approach. This research provides a replicable framework for agricultural machinery optimization, contributing to the advancement of sustainable and precision agriculture.

Identifying the maturity of asparagus is a crucial step for machine-assisted harvesting of asparagus in complex environments. This study proposes an innovative method to evaluate the height and diameter of asparagus stems, combining an enhanced YOLOv5 detection algorithm with point cloud data. In this method, first, the YOLOv5 model was improved, enabling efficient recognition and detection of asparagus in complex environments. Subsequently, a RealSense L515 radar camera was deployed to capture both the original RGB images and the point cloud information. The improved YOLOv5 algorithm was then employed to detect asparagus instances within the RGB images, with the pixel positions of the detection frames mapped onto the point cloud dataset to extract comprehensive 3D point cloud details of the asparagus. Finally, noise was reduced through statistical filtering and Euclidean clustering, and asparagus height was determined using the oriented bounding box methodology. Slices, each with a thickness of 10 mm, were extracted at designated measurement points, and the asparagus diameter was calculated by assessing the disparity between the maximum and minimum coordinates perpendicular to the growth direction of the asparagus. Experimental results showed that the mean average precision, precision, and recall of the improved YOLOv5 model increased by 4.85%, 5.09%, and 3.4%, reaching 98.21%, 97.11%, and 95.33%, respectively, which are higher than those of the YOLOv5 prototype network. Therefore, the proposed method could effectively detect asparagus. The algorithm exhibited a mean absolute error of 1.08 cm, a mean absolute percentage error of 4.06%, and a root mean square error of 1.60 cm in its estimation of asparagus height. For asparagus diameter estimation, the algorithm achieved a mean absolute error of 0.86 mm, a mean absolute percentage error of 7.98%, and a root mean square error of 1.23 mm. These results confirm that the proposed method can estimate the height and diameter of asparagus stems accurately, thereby providing invaluable technical support for machine harvesting of asparagus.