Visual navigation is developing rapidly and is of great significance to improve agricultural automation. The most important issue involved in visual navigation is extracting a guidance path from agricultural field images. Traditional image segmentation methods may fail to work in paddy field, for the colors of weed, duckweed, and eutrophic water surface are very similar to those of real rice seedings. To deal with these problems, a crop row segmentation and detection algorithm, designed for complex paddy fields, is proposed. Firstly, the original image is transformed to the grayscale image and then the treble-classification Otsu method classifies the pixels in the grayscale image into three clusters according to their gray values. Secondly, the binary image is divided into several horizontal strips, and feature points representing green plants are extracted. Lastly, the proposed double-dimensional adaptive clustering method, which can deal with gaps inside a single crop row and misleading points between real crop rows, is applied to obtain the clusters of real crop rows and the corresponding fitting line. Quantitative validation tests of efficiency and accuracy have proven that the combination of these two methods constitutes a new robust integrated solution, with attitude error and distance error within 0.02° and 10 pixels, respectively. The proposed method achieved better quantitative results than the detection method based on typical Otsu under various conditions.

Near-infrared (NIR) hyperspectroscopy becomes an emerging non-destructive sensing technology for inspection of crop seeds. A large spectral dataset of more than 140,000 wheat kernels in thirty varieties was prepared for classification. Feature selection is a critical segment in large spectral data analysis. A novel convolutional neural network-based feature selector (CNN-FS) was proposed to screen out deeply target-related spectral channels. A convolutional neural network with attention (CNN-ATT) framework was designed for one-dimension data classification. Popular machine learning models including support vector machine and partial least square discrimination analysis were used as the benchmark classifiers. Features selected by conventional feature selection algorithms were considered for comparison. Results showed that the designed CNN-ATT produced a higher performance than the compared classifier. The proposed CNN-FS found a subset of features, which made a better representation of raw dataset than conventional selectors did. The CNN-ATT achieved an accuracy of 93.01% using the full spectra, and keep its high precision (90.20%) by training on the 60-channel features obtained via the CNN-FS method. The proposed methods have great potential for handling the analyzing tasks on other large spectral datasets. The proposed feature selection structure can be extended to design other new model-based selectors. The combination of NIR hyper spectroscopic technology and the proposed models has great potential for automatic non-destructive classification of single wheat kernels.

Smart agriculture development mainly depends on the intelligence and reliability of autonomous agricultural machinery. Automatic navigation systems (ANSs) play a key role in intelligent agricultural machinery design, as they not only reduce farmers’ workloads but also improve their land utilization rates. In this paper, a tractor ANS based on dynamic path search and a fuzzy Stanley model (FSM) was designed, and its capability for whole-field path tracking was tested. First, the tracking performance of the steering control module was validated after the automatic reconstruction of the tractor platform. Then, a navigation decision system was established based on a unified reference waypoint search framework, where the path generation for whole-field coverage was presented. Finally, the gain coefficient of the Stanley model (SM) was adjusted adaptively according to the tracking error by utilizing the fuzzy logic controller. Subsequently, the developed tractor ANS was tested in the field. The experiment’s results indicate that the FSM outperformed the SM in straight path tracking and whole-field path tracking. When the tractor traveled at a speed of 1 m/s, the maximum lateral tracking error for the straight path was 10 cm, and the average lateral tracking error was 5.2 cm, showing improvements of 16.7% and 10.3% compared to the SM. Whole-field autonomous navigation showed that the maximum lateral tracking error was improved from 34 cm for the SM to 27 cm for the FSM, a reduction of approximately 20.6%, illustrating the superiority of the FSM in the application of whole-field path tracking. As the maximum tracking error of whole-field autonomous navigation appears in the turning stage, where tractors often stop working, the designed ANS satisfies the requirements of a self-driving system for unmanned tractors.

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.