Path tracking performance and tracking accuracy are often required during the autonomous navigation of the soil sampling platforms. In this study, a path tracking was proposed for the soil sampling platforms using MPC-Stanley. Firstly, a hardware architecture of the navigation system was designed for the soil sampling platform. A control scheme was also formulated for the navigation system. Secondly, a kinematic model was constructed using a bicycle model. Subsequently, the model predictive control (MPC) was selected as the path tracking controller, thus deriving the control processes for both the MPC and Stanley controllers. The Stanley controller was employed to optimize the steering angle of the MPC controller. The high computational complexity of the MPC controller was solved within the specified time constraints. The steering angle was calculated by the Stanley controller, and then served as the input into the MPC controller. The control quantities were modified using an exponential form. The model simplification was fully met the real-time control requirements of the navigation system. Finally, an inertial measurement unit (IMU) and satellite positioning module were employed to obtain the real-time pose information of the platform. The field path tracking experiments were conducted using a soil sampling platform. According to the straight path Tr₁ and curved path Tr₂ as the reference paths, tracking performance tests were conducted at a platform speed of 0.8 m/s for the MPC-Stanley controller, pure pursuit (PP) controller, and proportional integral derivative (PID) controller on both straight and curved trajectories. The results demonstrated that the MPC-Stanley controller was achieved in the best performance of the path tracking. The state variables of the system were predicted using the motion model. The MPC-Stanley controller was effectively resolved the overshoot and oscillation in the PID controller, indicating the superior robustness, compared with the PP controller. Subsequently, the tracking error tests were conducted for the MPC-Stanley controller, PP controller, and PID controller on the straight trajectory Tr₁ and curved trajectory Tr₂ at four speeds: 0.8, 1.6, 2.4, and 3.2 m/s. Test results indicated that the MPC-Stanley controller was achieved in the average absolute deviation, the maximum absolute deviation, and standard deviation of 3.1, 4.7 and 1.2 cm, respectively, during Tr₁ straight-line path tracking at all four speeds, which were reduced by 43.6%, 43.4%, and 14.3%, respectively compared with the PP controller, and 20.5%, 23.0%, and 7.7% respectively, compared with the PID controller. In the MPC-Stanley controller, the dynamic model was fully leveraged to avoid the large system errors that caused by the absence of the angular control in the PP controller. The MPC-Stanley controller demonstrated the more stable performance during path Tr₁ tracking, compared with the PID controller. The MPC-Stanley controller was achieved in the average absolute deviation, maximum absolute deviation, and standard deviation of 3.9, 6.6, and 1.5 cm, respectively, during curved path Tr₂ tracking, which were reduced by 80.2%, 79.8%, and 85.7% over the PP controller, and 93.0%, 89.8%, and 90.5% over the PID controller. The MPC-Stanley controller can be expected to enable the adaptive parameter tuning using system inputs and outputs, delivering the superior control performance. The path-tracking was effectively enhanced the tracking accuracy of the soil sampling platform. Furthermore, the framework can be extended universally applicable to various agricultural platforms with the structures similar to the soil sampling platform. The finding can also provide a technical reference for the high-precision navigation in the field.

Three-dimensional (3D) reconstruction for crops is often limited to a prolonged duration and suboptimal quality. In this study, an improved framework was proposed as the Oriented FAST and Rotated BRIEF-SLAM2 (ORB-SLAM2) system, in order to specifically reconstruct the seedling-stage maize plants. Several advanced computer vision techniques were integrated to achieve a robust and efficient reconstruction. Initially, the Fast Library for Approximate Nearest Neighbors (FLANN) algorithm was employed to combine with the Random Sample Consensus (RANSAC) algorithm, in order to perform high-accuracy feature matching on the multi-view images of the maize seedlings. This hybrid matching strategy was effectively identified to align the keypoints over the different images. While the outliers and incorrect matches were filtered out for the subsequent 3D processing. Feature matching and sparse point cloud generation were then realized using the modified ORB-SLAM2 front-end. A Multi-View Stereo (MVS) algorithm was seamlessly integrated into the pipeline. This MVS module was utilized for the pose estimations using a precise camera. The feature was performed on dense matching and depth estimation. Ultimately, a dense and accurate 3D point cloud model was generated to represent the morphology and spatial structure of the target maize plants. After that, the key architectural parameters were extracted from two maize cultivars, such as the plant height, leaf length, and stem diameter. A comparison was then made between the conventional and manual measurements in order to evaluate the practical utility and accuracy of the reconstructed models. Experimental results demonstrate that the superior performance of the integrated approach was achieved in the accuracy, reliability, and overall effectiveness of the non-contact phenotype using imaging. Specifically, the FLANN+RANSAC combination achieved a high correctness rate of 89.00% in the feature matching. The dense reconstruction produced the point clouds with an average density of 7.13×10⁵ points per model. Remarkably, the entire reconstruction, from the image input to dense point cloud output, was required for an average time of only 15.32 min, indicating its significant efficiency. All architectural parameters were extracted from the 3D models for the seedling maize plants. The errors were consistently maintained within a 10% threshold. There was a significant correlation between the extraction and the manual measurements, indicating the strong robustness. In conclusion, the high speed and precision were realized in plant phenotyping. The computational time was substantially reduced without compromising the reconstruction quality, thus offering a rapid, accurate, and non-destructive solution. Therefore, this work can provide a solid theoretical foundation and a practical, high-performance technical framework for the automatic acquisition of the 3D architectural traits in the maize plants during the critical seedling stage. The considerable potential was provided to advance the high-throughput plant phenotyping. The efficient morphological analysis can also accelerate crop genetics and breeding in precision agriculture.

Accurate, real-time, and in-situ acquisition of soil electrical conductivity can provide effective data support for the precise management of agricultural production. The current-voltage four-terminal approach as an invasive technology has considerable performance in the in-situ measurement of soil electrical conductivity on a large scale. This study aims to improve the accuracy of soil electrical conductivity measured by the traditional current-voltage four-terminal approach. A systematic analysis was made to determine the constant current source and electrode spacing in the three measurement arrays. The soil bin test was carried out to explore the influence of the main factors (soil moisture content, electrode embedded depth, soil compaction, and soil texture) on the measurement accuracy of three measurement arrays at different levels. The results showed that two measurement arrays of Wenner and Schlumberger were better applied to different soil environmental conditions. The measured values of soil electrical conductivity were further used as the inputs into the model. The regression model of soil electrical conductivity was constructed using the BP neural network. The R² of the model fit was 0.99762 in the training set, and the RMSE of the model between the calculated and standard value was 0.12 μS/mm in the testing set, indicating the smaller than that of individual measurement. All RMSE values were smaller than those in the individual array measurements. The measurement device of dual-array fusion soil electrical conductivity was designed using a regression model. The components of the device included the touchable LCD display, electrode sockets, switches, differential amplifier module, constant current source module, power supply, STM32 microcontroller data acquisition module, JESTON nano, and sensor. The soil electrical conductivity was then optimized using the measured values. The working stability test showed that the standard deviation of measured data was less than 0.43 μS/mm under different soil electrical conductivity gradient conditions. The comparative field-site performance test showed that the absolute, relative error range, and RMSE of measured soil electrical conductivity were -2.1-1.8 μS/mm, -8.0%-5.8%, 0.18 μS/mm respectively. The RMSE of 0.18 μS/mm was smaller than that of the traditional individual measurement array and the commonly used soil conductivity meters in the market. The measurement device can be expected to rapidly and accurately detect the soil's electrical conductivity, indicating better working stability and higher accuracy. The finding can provide high-precision detection and technical means for the real-time in situ collection of soil information in the field.

Efficient and non-destructive detection of rubber tree diseases is of great significance for optimizing disease control measures for pesticide application and fertilization. In this study, the feasibility of rubber yellow-leaf disease monitoring based on a low-altitude unmanned aerial vehicle (UAV) remote sensing platform was explored, and a low-cost method for detecting yellow-leaf disease based on visible light sensors was proposed. We compared the difference between the spectral response of each band of the visible light sensor in the diseased area and the healthy area, and then decorrelated and stretched the image in the RGB color space, thereby enhancing the color separation between highly correlated channels and enhancing the color difference of the image. Then we converted the image to the HSV color space, comparing the detection effect of different morphological parameters on yellow-leaf diseases and optimizing the extraction of the diseased area. The experimental results showed that this study provides the distribution information of yellow-leaf disease of rubber trees, and the R² of the regression model of rubber trees was greater than 0.8. This study holds significance for optimizing disease control and sustainable development of the rubber industry.