In precision agriculture, after vegetable transplanters plant the seedlings, field management during the seedling stage is necessary to optimize the vegetable yield. Accurately identifying and extracting the centerlines of crop rows during the seedling stage is crucial for achieving the autonomous navigation of robots. However, the transplanted ridges often experience missing seedling rows. Additionally, due to the limited computational resources of field agricultural robots, a more lightweight navigation line fitting algorithm is required. To address these issues, this study focuses on mid-to-high ridges planted with double-row vegetables and develops a seedling band-based navigation line extraction model, a Seedling Navigation Convolutional Neural Network (SN-CNN). Firstly, we proposed the C2f_UIB module, which effectively reduces redundant computations by integrating Network Architecture Search (NAS) technologies, thus improving the model’s efficiency. Additionally, the model incorporates the Simplified Attention Mechanism (SimAM) in the neck section, enhancing the focus on hard-to-recognize samples. The experimental results demonstrate that the proposed SN-CNN model outperforms YOLOv5s, YOLOv7-tiny, YOLOv8n, and YOLOv8s in terms of the model parameters and accuracy. The SN-CNN model has a parameter count of only 2.37 M and achieves an mAP@0.5 of 94.6%. Compared to the baseline model, the parameter count is reduced by 28.4%, and the accuracy is improved by 2%. Finally, for practical deployment, the SN-CNN algorithm was implemented on the NVIDIA Jetson AGX Xavier, an embedded computing platform, to evaluate its real-time performance in navigation line fitting. We compared two fitting methods: Random Sample Consensus (RANSAC) and least squares (LS), using 100 images (50 test images and 50 field-collected images) to assess the accuracy and processing speed. The RANSAC method achieved a root mean square error (RMSE) of 5.7 pixels and a processing time of 25 milliseconds per image, demonstrating a superior fitting accuracy, while meeting the real-time requirements for navigation line detection. This performance highlights the potential of the SN-CNN model as an effective solution for autonomous navigation in field cross-ridge walking robots.

The rapid and accurate detection of broccoli seedling planting quality is crucial for the implementation of robotic intelligent field management. However, existing algorithms often face issues of false detections and missed detections when identifying the categories of broccoli planting quality. For instance, the similarity between the features of broccoli root balls and soil, along with the potential for being obscured by leaves, leads to false detections of “exposed seedlings”. Additionally, features left by the end effector resemble the background, making the detection of the “missed hills” category challenging. Moreover, existing algorithms require substantial computational resources and memory. To address these challenges, we developed Seedling-YOLO, a deep-learning model dedicated to the visual detection of broccoli planting quality. Initially, we designed a new module, the Efficient Layer Aggregation Networks-Pconv (ELAN_P), utilizing partial convolution (Pconv). This module serves as the backbone feature extraction network, effectively reducing redundant calculations. Furthermore, the model incorporates the Content-aware ReAssembly of Features (CARAFE) and Coordinate Attention (CA), enhancing its focus on the long-range spatial information of challenging-to-detect samples. Experimental results demonstrate that our Seedling-YOLO model outperforms YOLOv4-tiny, YOLOv5s, YOLOv7-tiny, and YOLOv7 in terms of speed and precision, particularly in detecting ‘exposed seedlings’ and ‘missed hills’-key categories impacting yield, with Average Precision (AP) values of 94.2% and 92.2%, respectively. The model achieved a mean Average Precision of 0.5 (mAP@0.5) of 94.3% and a frame rate of 29.7 frames per second (FPS). In field tests conducted with double-row vegetable ridges at a plant spacing of 0.4 m and robot speed of 0.6 m/s, Seedling-YOLO exhibited optimal efficiency and precision. It achieved an actual detection precision of 93% and a detection efficiency of 180 plants/min, meeting the requirements for real-time and precise detection. This model can be deployed on seedling replenishment robots, providing a visual solution for robots, thereby enhancing vegetable yield.

The operational performance of cereal seeding machinery influences the yield and quality of cereals. In this article, we review the existing literature on intelligent technologies for cereal seeding machinery, encompassing active controllable seeding actuators, intelligent seeding rate control, and intelligent seed position control systems. In this manuscript, (1) the characteristics and innovative structures of existing motor-driven seed-metering devices and ground surface profiling mechanisms are expounded; (2) state-of-the-art detection principles and applications for soil property sensors are described based on different soil properties; (3) optimal seeding rate decision approaches based on soil properties are summarized; (4) the research state of seeding rate measuring and control technologies is expounded in detail; (5) trajectory control methods for seeding machinery and seeding depth control systems are described based on measurement and control principles; and (6) the present state, limitations, and future development directions of intelligent cereal seeding machinery are described. In the future, more advanced multi-algorithm and multi-sensor fusion technologies for soil property detection, optimal seeding rate decisions, seeding rates, and seed position control are likely to evolve. This review not only expounds the latest studies on intelligent actuating, sensing, and control technologies for intelligent cereal seeding machinery, but also discusses the shortcomings of existing intelligent seeding technologies and future developing trends in detail. This review, therefore, offers a reference for future research in the domain of intelligent seeding machinery for cereals.

China's socialized service industry of agricultural machinery emerged as the times required, which resolved the institutional contradiction between small farm households and large-scale agriculture. However, as cross-operation harvesters are still monitored based on phone calls, the monitoring efficiency and real-time performance are relatively poor. To improve the dynamic monitoring accuracy of harvester working areas, we proposed a working progress monitoring method based on harvester traveling trajectory and header status identification. The position and gap bridge angle information of harvesters was synchronously acquired in real-time. The height of the header above ground was calculated to determine whether the harvester was harvesting on a valid trajectory. Finally, the valid trajectory of the harvester was selected for area calculation. Field experiments showed that the header heights at the working and non-working tracing points collected by the system were discretely distributed at intervals of [0m, 0.5m] and (0.5m, 2m), respectively. The working progress monitoring algorithm proposed in this paper effectively improved the monitoring accuracy of harvester working areas. The mean error between the working areas calculated with identifying the header status and the actual cultivated land area in the five test plots was 0.09 hm², with a mean error rate of 3.10%, 8.59% lower than that without identifying the harvester header status.

Rice production is highly seasonal, and the timing for conducting harvest operations has a relatively significant impact on yields, especially the appropriate timing of the harvesting operations. Using scientific methods to monitor the appropriate harvest period can effectively reduce harvesting losses. Most of the existing studies in this field focus on yield changes that occur during the harvest period of rice and also use multispectral remote sensing for inversion of the crop canopy. However, there has been no research on combining the yields and spectral characteristics of the harvest period. In this article, a monitoring method for the appropriate harvest period of rice based on multispectral remote sensing was established. Remote sensing data of the rice canopy were acquired using UAV equipped with multispectral (550 nm, 660 nm, 735 nm, and 790 nm) cameras, and physiological characteristic data of rice moisture content, thousand‐grain weight, and tassel ratios were determined simultaneously. Single‐band and combined‐band spectral reflectance were introduced as model input variables to compare the BP neural network, SVM, and decision tree; a quantitative inversion model of the physiological characteristics of rice was established. The accuracy of the inversion model with different input variables and model methods was evaluated, and the optimal inversion model for rice moisture content and thousand‐grain weight was selected, which could be used to determine whether the crop was harvested during the appropriate period. The continuous monitoring of two rice varieties (South Japonica 46 and South Japonica 5055) indicated that the moisture content of South Japonica 46 decreased linearly during the 20‐day trial period, while its thousand‐grain weight showed a trend of increasing first and then decreasing. The moisture content of South Japonica 5055 showed a fluctuating downward trend, and its thousand‐grain weight also showed a trend of increasing first and then decreasing. The spectral reflectance at 735 nm increased gradually, and that of both rice varieties increased from about 35% at the beginning of the trial to about 75% in the later stage. The spectral reflectance at the other three bands did not show significant change with the harvest date. The inverse model of spectral reflectance with rice thousand‐grain weight and moisture content was established using the BP neural network, SVM, and decision tree. In single‐band inversion models, the regression effect at 735 nm was relatively good in the BP neural network model, with the highest determination coefficients of spectral reflectance and thousand‐grain weight of both rich varieties and the smallest RMSE of prediction results. In combined‐band inversion models, the regression effects at 550 + 660+735 nm and 660 + 735+790 nm were the best in the SVM and thousand‐grain weight inversion models.

To improve the seedling-picking efficiency of the vegetable transplanter and reduce the damage rate of the seedling pot, a reciprocating seedling-picking device driven by full air pressure was designed. In this paper, the structure and working principle of the pneumatic seedling-picking device are introduced. Through the mechanical analysis between the seedling-picking claw and the seedling pot, working parameters such as the stroke and driving force of the pneumatic seedling-picking claw clamping cylinder were determined. According to the action sequence of the seedling-picking mechanism, which is horizontally dispersed and longitudinally conveyed, the pneumatic control scheme of the seedling-picking and -dropping system was formulated. The simulation model for the control loop of the longitudinal cylinder was created with AMESim simulation software, and the simulation analysis was carried out. The Box–Behnken response surface design optimization method was used to determine the best operating parameters of the cylinder. The optimized peak value of shock vibration at the end of the cylinder was optimized from −65.64 mm·s⁻² to 35.41 mm·s⁻², proving that the optimization of pneumatic working parameters has a positive effect on the success rate of seedling picking. The bench test of the seedling-picking mechanism was conducted on 72-hole plug seedlings with two picking frequencies of 120 plants·min⁻¹ and 144 plants·min⁻¹, respectively, and the average seedling leakage rate, seedling damage rate, and seedling pot damage rate at different picking frequencies were counted. The experimental results show that under the two seedling-picking frequencies, the average success rate of seedling picking and throwing after optimization is increased from 96.4% and 92.4% to 97.9% and 95.3%, respectively. This is in line with the requirements of high-speed seedling picking and confirms the rationality of the seedling-picking mechanism design.