Using intelligent agricultural machines in paddy fields has received great attention. An obstacle avoidance system is required with the development of agricultural machines. In order to make the machines more intelligent, detecting and tracking obstacles, especially the moving obstacles in paddy fields, is the basis of obstacle avoidance. To achieve this goal, a red, green and blue (RGB) camera and a computer were used to build a machine vision system, mounted on a transplanter. A method that combined the improved You Only Look Once version 3 (Yolov3) and deep Simple Online and Realtime Tracking (deep SORT) was used to detect and track typical moving obstacles, and figure out the center point positions of the obstacles in paddy fields. The improved Yolov3 has 23 residual blocks and upsamples only once, and has new loss calculation functions. Results showed that the improved Yolov3 obtained mean intersection over union (mIoU) score of 0.779 and was 27.3% faster in processing speed than standard Yolov3 on a self-created test dataset of moving obstacles (human and water buffalo) in paddy fields. An acceptable performance for detecting and tracking could be obtained in a real paddy field test with an average processing speed of 5–7 frames per second (FPS), which satisfies actual work demands. In future research, the proposed system could support the intelligent agriculture machines more flexible in autonomous navigation.

Winter jujube is a popular fresh fruit in China for its high vitamin C nutritional value and delicious taste. In terms of winter jujube object detection, in machine learning research, small size jujube fruits could not be detected with a high accuracy. Moreover, in deep learning research, due to the large model size of the network and slow detection speed, deployment in embedded devices is limited. In this study, an improved Yolov5s (You Only Look Once version 5 small model) algorithm was proposed in order to achieve quick and precise detection. In the improved Yolov5s algorithm, we decreased the model size and network parameters by reducing the backbone network size of Yolov5s to improve the detection speed. Yolov5s’s neck was replaced with slim-neck, which uses Ghost-Shuffle Convolution (GSConv) and one-time aggregation cross stage partial network module (VoV-GSCSP) to lessen computational and network complexity while maintaining adequate accuracy. Finally, knowledge distillation was used to optimize the improved Yolov5s model to increase generalization and boost overall performance. Experimental results showed that the accuracy of the optimized Yolov5s model outperformed Yolov5s in terms of occlusion and small target fruit discrimination, as well as overall performance. Compared to Yolov5s, the Precision, Recall, mAP (mean average Precision), and F₁ values of the optimized Yolov5s model were increased by 4.70%, 1.30%, 1.90%, and 2.90%, respectively. The Model size and Parameters were both reduced significantly by 86.09% and 88.77%, respectively. The experiment results prove that the model that was optimized from Yolov5s can provide a real time and high accuracy small winter jujube fruit detection method for robot harvesting.

In modern agriculture, many advanced automated devices are used on farms. To improve the working efficiency of agricultural vehicles, fields are expected to be pre-leveled, because the vehicles work more effectively on a flat field. Leveling a field requires the current field elevation map. Some farmers in Japan have begun to use high-precision real-time kinematic Global Positioning System (RTK-GPS)-based self-steering tractors in the fields. This study uses the RTK-GPS information from a self-steering tractor system to generate a digital elevation model (DEM) especially in mountain regions where the fields are not flat. In addition, all of the information is from the self-steering system with the result that farmers can use the method of this study without additional instruments. However, the GPS receiver sometimes cannot obtain high-quality signals from satellites in mountain regions. Therefore, this study focuses on how to create a high-precision DEM even when a GPS signal is unavailable. It proposes a dynamic interpolation method for generating a DEM. In addition, a test was conducted in a field in a mountain region. The test results show that the dynamic interpolation method can provide an accuracy of less than 0.03 m in the test field for creating a DEM.

Harvesting grapes requires a large amount of manual labor. To reduce the labor force for the harvesting job, in this study, we developed a robot harvester for the vine grapes. In this paper, we proposed an algorithm that using multi-cameras, as well as artificial intelligence (AI) object detection methods, to detect the thin stem and decide the cut point. The camera system was constructed by two cameras that include multi-lenses. One camera is mounted at the base of the robot and named the “base camera”; the other camera is mounted at the robot hand and named the “hand camera” to recognize grapes and estimate the stem position. At the first step, the grapes are detected by using a You Only Look Once (YOLO) method, while the stems of the grapes are detected at the second step using a pixel-level semantic segmentation method. Field experiments were conducted at an outdoor grapes field. The experiment results show that the proposed algorithm and the camera system can successfully detect out the cut point, and the correct detection rate is around 98% and 93% in the indoor and outdoor conditions, respectively. The detection system was integrated to a grape-harvesting robot in the experiment, and the experiment results show the system can successfully harvest the grapes in the outdoor conditions.

In Japan, the aging and decreasing number of agricultural workers is a significant problem. For wine grape harvesting, especially for large farming areas, there is physical strain to farmers. In order to solve this problem, this study focuses on developing an automated harvesting robot for wine grapes. The harvesting robot needs high dust, water, and mud resistance because grapevines are grown in hard conditions. Therefore, a three-axis linear robot was developed using a rack and pinion mechanism in this study, which can be used in outdoor conditions with low cost. Three brushless DC motors were utilized to drive the three-axis linear robot. The motors were controlled using a control area network (CAN) bus to simplify the hardware system. The accuracy of the robot positioning was evaluated at the automated harvesting condition. The experiment results show that the accuracy is approximately 5 mm, 9 mm, and 9 mm in the x-axis (horizontal), y-axis (vertical), and z-axis (depth), respectively. In order to improve the accuracy, we constructed an error model of the robot and conducted a calibration of the robot. The accuracy was improved to around 2 mm of all three axes after calibration. The experimental results show that the accuracy of the robot is high enough for automated harvesting of the wine grapes.

The decline in the number of essential farmers has become a significant issue in Japanese agriculture. In response, there is increasing interest in the electrification and automation of agricultural machinery, particularly in relation to the United Nations Sustainable Development Goals (SDGs). This study focuses on the development of an electric vehicle (EV) crawler-type robot designed for weed cultivation operations, with the aim of reducing herbicide use in organic onion farming. Weed cultivation requires precise, delicate operations over extended periods, making it a physically and mentally demanding task. To alleviate the labor burden associated with weeding, we employed a color camera to capture crop images and used artificial intelligence (AI) to identify crop rows. An automated system was developed in which the EV crawler followed the identified crop rows. The recognition data were transmitted to a control PC, which directed the crawler’s movements via motor drivers equipped with Controller Area Network (CAN) communication. Based on the crop row recognition results, the system adjusted motor speed differentials, enabling the EV crawler to follow the crop rows with a high precision. Field experiments demonstrated the effectiveness of the system, with automated operations maintaining a lateral deviation of ±2.3 cm, compared to a maximum error of ±10 cm in manual operation. These results indicate that the automation system provides a greater accuracy and is suitable for weed cultivation tasks in organic farming.