Maize (Zea mays L.) is a major crop for global food security at present. Advanced breeding is often required to enhance yield, stress resistance, and adaptability, particularly for high-throughput, non-destructive, and accurate acquisition of plant phenotypic parameters. Three-dimensional (3D) point clouds acquired by LiDAR can provide unprecedented detail of plant architecture, compared with 2D imaging. However, their widespread application has been confined to the accurate instance segmentation of individual plants within dense populations in real-world fields. Furthermore, conventional clustering or geometry algorithms cannot solve the convoluted spatial arrangement, complex plant morphologies, extensive canopy adhesion—where the leaves of adjacent plants are tightly interwoven—and mutual occlusion among plants. The efficient and reliable extraction of phenotypic data has been severely constrained to the fragmented or incorrectly merged plant instances. In this study, an instance segmentation framework, 3D-MaizeNet, was proposed to integrate LiDAR data with deep learning. Individual maize plants were accurately extracted for the high-throughput measurement of key agronomic traits, such as plant height and stem height. Three stages are included. 1) The structural integrity of the individual plant was preserved to avoid the compromise during simplistic preprocessing. An adaptive block segmentation was introduced using crop row detection. The row-planting pattern of farmlands was divided used to divide the large-scale point cloud into plant-centric blocks. This approach was used to effectively minimize the interference from overlapping canopies in adjacent rows. A high-quality, field-derived point cloud dataset was constructed for robust model training. 2) A local spatial encoding module was designed to learn fine-grained geometric features from complex canopy structures (e.g., leaf angles and stem orientations). Concurrently, an attention aggregation down-sampling module was integrated to reduce the loss of key spatial features during feature extraction. Salient information was selectively preserved to distinguish among tightly packed plants. 3) According to the high-fidelity instance segmentation, an pipeline was established for the high-throughput quantification of plant height and stem height—two pivotal phenotypic parameters closely related to yield potential and lodging resistance. Field-scanned data was were collected to validate the efficacy of the framework. Experimental results showed that the 3D-MaizeNet achieved a mean Average Precision (mAP) of 0.959 and an overall accuracy of 0.964 in instance segmentation, indicating the superior performance to identifyin identifying and delineate delineating the individual plant. Furthermore, the key traits were extracted for the strong correlations with manual ground-truth measurements, with coefficients of determination (R²) of 0.91 and 0.89 for plant height and stem height, respectively. The high-throughput and precise phenotyping platform can provide the technical support to advance the maize genomics, Genome-Wide Association Studies (GWAS), and ultimately the molecular breeding for next-generation crops.

Target detection has been widely applied to many scenarios in daily life, including people flow management, items counting, and object searching. Deep learning has also effectively improved the accuracy of target detection in recent years. Therefore, target detection can be expected to improve the efficiency of intelligent management in the livestock industry, such as pig farms. Daily population counting has been one of the most important steps in modern pig farms, such as target tracking and behavior recognition. However, the small targets or occluded pigs are difficult to accurately detect during counting. In this study, an accurate and rapid detection of the pig population was proposed to treat the small targets and occlusion in the dataset using a multi-scale fusion attention mechanism. YOLOpig network was constructed to detect the individual pig target using YOLOv7. Then, the scale network structure was proposed to enhance the detection of small targets. The residual thought of network structure was used to improve the convolution module for the high accuracy of the experiments. A parameter-free attention mechanism was also added to reduce the network weights while accelerating the detection speed. GradCAM was used for the feature visualization to verify the effectiveness of the experimental feature extraction. Finally, target tracking (StrongSORT) was adopted to accurately track the individual IDs of the pigs that were detected by the improved model, providing the identity information required for pig detection tasks. Experiments were conducted to verify the accuracy and real-time performance of the improved model on Large White pigs in the fattening stage. A series of experiments were conducted, including model ablation, model comparison, pig feature information extraction, and tracking. The effectiveness and feasibility of the models were verified to detect the pig groups. The great potential was obtained to solve the difficulties and challenges in pig population counting, providing important support in the agricultural breeding field. The experimental results show that the accuracy, recall, and average accuracy of the counting were 90.4%, 85.5%, and 92.4%, respectively. Furthermore, the average accuracy and the detection speed were improved by 5.1 percentage points and 7.14%, respectively, compared with the basic YOLOv7 model. The average accuracies of the YOLOv5, YOLOv7tiny, and YOLOv8n models were also improved by 12.1, 16.8, and 5.7 percentage points, respectively. Specifically, the pig population counting with a multi-scale fusion attention mechanism can be expected to rapidly and accurately complete the counting task, and then effectively deal with small targets and occlusion. The widespread application of the improved model can greatly contribute to the operational efficiency of pig farms for labor cost-saving, in order to promote the development of intelligent technology in the field of agricultural breeding. The more accurate and efficient counting of pigs can provide strong technical support for the field of agricultural breeding and intelligent farming.

Autonomous agricultural machinery has been emerged as one of the core strategies in recent years. It is ever pressing from the aging population, shrinking agricultural workforce, large-scale rural migration, and agricultural labor costs in China. Autonomous agricultural machinery can be expected to alleviate the labor shortages for the high efficiency and precision. Fully functional unmanned farm can also enhance the agricultural productivity and sustainability. In this study, a comprehensive review was presented on the technical requirements of autonomous agricultural machinery. An emphasis was also put on the key research advancements in China, including the environmental perception, high-precision agricultural mapping, autonomous positioning and navigation, path planning, and tracking control. One of the hallmark features of unmanned farms was attributed to the seamless and fully automated transfer of machinery among storage facilities and fields. Specifically, the autonomous machinery was automatically departed from its designated storage location, thus navigating the roads to perform field operations, and finally returning to park precisely at its original position in the shed after task completion. As such, the autonomous machinery was used in fields, farm roads, and agricultural sheds, indicating the great potential to farming. Nevertheless, the unmanned farm system was remained underdeveloped within the storage shed. Machinery parking, implement attachment, and departure still heavily depended on the manual intervention. Full automation over all scenarios was required the advanced positioning and navigation. Precise perception to shed environment was also necessary for the safety and efficiency during operations. Storage sheds were often densely populated with machinery, rather than the open and relatively predictable environments of fields and farm roads. The maneuvering space was limited to the more rapid and accurate perception and navigation, compared with the open fields. While the significant progress was found in the autonomous machinery under the scenarios of fields and road. The perception and autonomous parking technologies were less developed for shed operations so far. These challenges were also addressed to bridge the automation gap, and then unlock the full potential of unmanned farms. Much effort was focused mainly on the automatic parking of autonomous agricultural machinery, particularly on returning to the shed after field operations. Indoor positioning and precise navigation were also explored with the perception during automatic parking. Some procedures were involved in the indoor navigation from the shed entrance to the parking spot without Global Navigation Satellite System (GNSS) signals. The safe and efficient routes were determined for the final docking after environmental perception and machinery position. The fixed- and non-fixed routes were categorized for the positioning and navigation. Fixed-route navigation offered the simplicity, precision and stability, including the visual navigation (lane tracking and visual marker localization), rail guidance, and magnetic navigation. Safe zones of shed were also delineated to enhance the accuracy of position and parking. In contrast, the non-fixed-route navigation was provided for the superior flexibility, scalability, and adaptability to the complex and dynamic indoor environments, such as the external source-based positioning (e.g., radio frequency identification, WiFi, bluetooth, and ultra-wideband) and simultaneous localization and mapping (SLAM). Finally, the technical and practical challenges were summarized for the autonomous agricultural machinery. Specifically, the diverse technologies were integrated for the dynamic environments, particularly for the robust safety mechanisms. Future directions were outlined to integrate the multiple technologies for all-weather operation in the complex environments. Autonomous agricultural machinery can be expected to serve as the critical technology in unmanned farming. In turn, the operational quality and efficiency can also be enhanced in the modern agriculture.

Accurate extraction of crop row is very important for automation of agricultural production. Crop rows are required for accurate machine guidance in agricultural production such as fertilization, plant protection, weeding and harvesting. In this study, an efficient crop row detection algorithm called Crop-BiSeNet V2 was proposed, which combined BiSeNet V2 with a spatial convolutional neural network. The proposed Crop-BiSeNet V2 detected crop rows in color images without the use of threshold and other pre-information such as number of rows. A data set had 2697 maize crop images was constructed in challenging field trial conditions such as variable light, shadows, presence of weeds, and irregular crop shape. The proposed system was experimentally determined to overcome the interference of different complex scenes. And it can be applied to crop rows of different numbers, straight lines and curves. Different analyses were performed to check the robustness of the algorithm. Comparing this algorithm with the Fully Convolutional Networks (FCN) algorithm, it exhibited superior performance and saved 84.85 ms. The accuracy rate reached 0.9811, and the detection speed reached 65.54 ms/frame. The Crop-BiSeNet V2 algorithm proposed in this study show strong generalization performance for seedling crop row recognition. It provides high-reliability technical support for crop row detection research and assists in the study of intelligent field operation machinery navigation.