Object detection in deep learning provides a viable solution for detecting crop-pest-infected regions. However, existing rectangle-based object detection methods are insufficient to accurately detect the shape of pest-infected regions. In addition, the method based on instance segmentation has a weak ability to detect the pest-infected regions at the edge of the leaves, resulting in unsatisfactory detection results. To solve these problems, we constructed a new polygon annotation dataset called PolyCorn, designed specifically for detecting corn leaf pest-infected regions. This was made to address the scarcity of polygon object detection datasets. Building upon this, we proposed a novel object detection model named Poly-YOLOv8, which can accurately and efficiently detect corn leaf pest-infected regions. Furthermore, we designed a loss calculation algorithm that is insensitive to ordering, thereby enhancing the robustness of the model. Simultaneously, we introduced a loss scaling factor based on the perimeter of the polygon, improving the detection ability for small objects. We constructed comparative experiments, and the results demonstrate that Poly-YOLOv8 outperformed other models in detecting irregularly shaped pest-infected regions, achieving 67.26% in mean average precision under $0.5$ threshold ($\mathrm{mA}{\mathrm{P}}_{50}$) and 128.5 in frames per second (FPS).

The Complete Coverage Path Planning (CCPP) is a key technology in the field of agricultural robots, and has great significance for improving the efficiency and quality of tillage, fertilization, harvesting, and other agricultural robot operations, as well as reducing the operation energy consumption. The traditional boustrophedon- or heuristic-search-algorithm-based CCPP methods, when coping with the field with irregular boundaries, obstacles, and other complex environments, still face many problems and challenges, such as large repeated work areas, multiple turns or U-turns, low operation efficiency, and prone to local optimum. In order to solve the above problems, an improved-genetic-algorithm-based CCPP method was proposed in this paper, the proposed method innovatively extends the traditional genetic algorithm’s chromosomes and single-point mutation into chromosome pairs and multi-point mutation, and proposed a multi-objective equilibrium fitness function. The simulation and experimental results on simple regular fields showed that the proposed improved-genetic-algorithm-based CCPP method achieved the comparable performance with the traditional boustrophedon-based CCPP method. However, on the complex irregular fields, the proposed CCPP method reduces 38.54% of repeated operation area and 35.00% of number of U-turns, and can save 7.82% of energy consumption on average. This proved that the proposed CCPP method has a strong adaptive capacity to the environment, and has practical application value in improving the efficiency and quality of agricultural machinery operations, and reducing the energy consumption.

The Complete Coverage Path Planning (CCPP) is a key technology in the field of agricultural robot, and has great significance for improving the efficiency and quality of tillage, fertilization, harvesting and other agricultural robot operations, as well as reducing the operation energy consumption. The traditional boustrophedon or heuristic search algorithm based CCPP methods, when coping with the field with irregular boundaries, obstacles and other complex environments, still face many problems and challenges, such as large repeated work areas, multiple turns or U-turns, low operation efficiency and prone to local optimum. In order to solve the above problems, an improved-genetic-algorithm-based CCPP method was proposed in this paper, which introduces the idea of chromosome pairs and multi-points mutation to improve its global optimization ability in complex environments, reduce the repeated work areas and the number of turns and U-turns, and thereby improve the operation efficiency. The simulation and experimental results on simple regular fields showed that the proposed improved genetic algorithm-based CCPP method achieved the comparable performance with the traditional boustrophedon-based CCPP method. However, on the complex irregular fields, although the proposed CCPP method increased the number of turns by 13.76%, the area of repeated operation and the number of U-turns were decreased by 38.54% and 35.00% respectively, and saves the battery voltage by 7.82% on average compared to the boustrophedon-based CCPP method. This proved that the proposed CCPP method has strong adaptive capacity to the environment, and has practical application value in improving the efficiency and quality of agricultural machinery operations, and reducing the energy consumption.

With the promotion of artificial intelligence in agriculture and the popularization of plug tray seedling-raising technology, seedling raising and transplanting have become the most popular planting modes. Miss-seeding is one of the most serious problems affecting seedling raising and transplanting. It not only affects the germination rate of seeds but also reduces the utilization rate of the plug tray. The experimental analysis of traditional machine vision-based miss-seeding showed that because of uneven lighting, the plug tray was wrongly identified as a seed under bright light, but the seeds in the dark were not easy to identify. When using the seeding area to identify seeds and noise, sweet corn seeds in a small area can be easily screened out. This paper proposes a method using the ResNet network with an attention mechanism to solve the above-mentioned problems. In this paper, the captured image was segmented into the images of a single plug tray, and a residual attention network was built; the detection scheme of miss-seeding was also converted into a dichotomous picture recognition task. This paper demonstrates that the residual attention network can effectively recognize and detect the seed images of sweet corn with very high accuracy. The results of the experiment showed that the average accuracy of this recognition model was 98%. The feature visualization method was used to analyze the features, further proving the effectiveness of the classification method of plug tray seedlings.

Field-road mode classification (FRMC) that identifies “in-field” and “on-road” categories for Global Navigation Satellite System (GNSS) trajectory points of agricultural machinery containing geographic information is essential for effective crop improvement. Most previous studies utilize local trajectory features (i.e., the relationships between a point and its neighboring points), but they ignore global trajectory features (i.e., the relationships between the point and all points of the trajectory), leading to difficulty in improving the overall classification performance. The global trajectory features are useful for FRMC because they contain rich trajectory information (e.g., mode switching and motion tendency). Therefore, a ConvTEBiLSTM network-based method is proposed to improve the overall performance. Firstly, nine statistical features (e.g., speed and direction) are extracted from the original data and fed into the ConvTEBiLSTM network. Then, the ConvTEBiLSTM network combining the Bidirectional Long Short-Term Memory network, 1D Convolution network, and Transformer-Encoder network is used to extract and fuse local and global trajectory features. Finally, a linear classifier is applied to identify the “field” and “road” categories of GNSS points based on the fused features. Experimental results show that compared with the baselines, our method achieves the best accuracy and F1-score of 97.38% and 92.74% on our Harvester dataset, respectively.