Weeds are among the major factors that could harm crop yield. Site-specific weed management has become an effective tool to control weed and machine vision combined with image processing is an effective approach for weed detection. In this work, an encoder-decoder deep learning network was investigated for pixel-wise semantic segmentation of crop and weed. Different input representations including different color space transformations and color indices were compared to optimize the input of the network. Three image enhancement methods were investigated to improve model robustness against different lighting conditions. The results show that for images without enhancement, color space transformation and vegetation indices without NIR (Near Infrared) information did not improve the segmentation results, while inclusion of NIR information significantly improved the segmentation accuracy, indicating the effectiveness of NIR information for precise segmentation under weak lighting condition. Image enhancement improved the image quality and consequently the robustness of segmentation models against different lighting conditions. The best MIoU value for pixel-wise segmentation was 88.91% and the best mean accuracy of object-wise segmentation was 96.12%. The deep network and image enhancement methods applied in this work provided promising segmentation results for weed detection and did not need large amount of data for model training, which is suitable for site-specific weed management.

The sampling-based Kalman filters have been widely used for GNSS/INS of land vehicle navigation. Although many nonlinear filters are proposed by employing different approximate algorithms, most of the filters update sigma points utilize the moments match and Bayesian filter update framework. In this paper, cubature Kalman filters (CKFs) based on a class of novel sigma-point update framework (NSUF) are developed and compared in terms of accuracy and convergence rate. In order to enhance the reliability of GNSS/INS, an improved CKF, named as ArtCKF is proposed by merging the moments generated by different NSUFs, which not only addresses the problem of observation missing but also improves the efficiency of Bayesian filter update. Numerical simulation based on field test data is employed to verify the superiority of ArtCKF in land vehicle navigation. Our results indicate that the ArtCKF improves the heading of land vehicle by 71.3% compared with CKF, and makes a tradeoff between direct observable and non-direct observable state estimation under GNSS-challenged environment.

A good planting bed is a prerequisite for improving planting quality, while complex ground excitation often leads to machine bouncing and operation vibration, which then affects the operation effect. In order to improve the quality of rotary tillage operations, it is necessary to study the effects of various vibration excitations on the unit during tractor rotary tillage operations and analyze the vibration interaction relationship among the tractor, the three-point suspension mechanism, and the rotary tiller. For this purpose, multiple three-way acceleration sensors were installed at different positions on the rotary tiller unit of a Lexing LS1004 tractor(Lexing Agricultural Equipment Co. Ltd., Qingdao, China) to collect vibration data at different operating speeds and conduct vibration characteristic analysis between different components. The test results showed that when the unit moved forward at 2.1 km/h, 3.6 km/h, and 4.5 km/h, respectively, the vibration acceleration of the tractor, the three-point suspension mechanism, and the rotary tiller increased with the increase in speed, and there was indeed interaction between them. The vertical acceleration change during the test in the three-point suspension mechanism was the most significant (5.914 m/s²) and was related to the increase in the speed of the vehicle and the vibration transfer of the rotary tiller. Meanwhile, the vertical vibration acceleration of the tractor’s symmetrical structure was not similar, suggesting the existence of structural assembly problems. From the perspective of frequency domain analysis, the resonant frequency at the cab of the tractor was reduced in a vertical vibration environment, with relatively low frequencies (0~80 Hz) and small magnitudes, which might be beneficial to the driver’s health. The rotary tillage group resonated around 350 Hz, and this characteristic can be used to appropriately increase the vibration of the rotary tiller to reduce resistance. The tractor cab resonated around 280 Hz, which must be avoided during field operations to ensure driver health and reduce machine wear. The research results can provide a reference for reducing vibration and resistance during tractor rotary tillage operations, as well as optimizing and improving the structure of rotary tillers and tractors.

To satisfy the obstacle avoidance requirements of unmanned agricultural machinery during autonomous operation and address the challenge of rapid obstacle detection in complex field environments, an improved field obstacle detection model based on YOLOv8 was proposed. This model enabled the fast detection and recognition of obstacles such as people, tractors, and electric power pylons in the field. This detection model was built upon the YOLOv8 architecture with three main improvements. First, to adapt to different tasks and complex environments in the field, improve the sensitivity of the detector to various target sizes and positions, and enhance detection accuracy, the CBAM (Convolutional Block Attention Module) was integrated into the backbone layer of the benchmark model. Secondly, a BiFPN (Bi-directional Feature Pyramid Network) architecture took the place of the original PANet to enhance the fusion of features across multiple scales, thereby increasing the model’s capacity to distinguish between the background and obstacles. Third, WIoU v3 (Wise Intersection over Union v3) optimized the target boundary loss function, assigning greater focus to medium-quality anchor boxes and enhancing the detector’s overall performance. A dataset comprising 5963 images of people, electric power pylons, telegraph poles, tractors, and harvesters in a farmland environment was constructed. The training set comprised 4771 images, while the validation and test sets each consisted of 596 images. The results from the experiments indicated that the enhanced model attained precision, recall, and average precision scores of 85.5%, 75.1%, and 82.5%, respectively, on the custom dataset. This reflected increases of 1.3, 1.2, and 1.9 percentage points when compared to the baseline YOLOv8 model. Furthermore, the model reached 52 detection frames per second, thereby significantly enhancing the detection performance for common obstacles in the field. The model enhanced by the previously mentioned techniques guarantees a high level of detection accuracy while meeting the criteria for real-time obstacle identification in unmanned agricultural equipment during fieldwork.