Automated classification of corn is important for corn sorting in intelligent agriculture. This paper presents a reliable corn classification method based on techniques of computer vision and machine learning. To discriminate different damaged types of corns, a line profile segmentation method is firstly used to segment and separate a group of touching corns. Then, twelve color features and five shape features are extracted for each individual corn object. Finally, a maximum likelihood estimator is trained to classify normal and damaged corns. To evaluate the performance of the proposed method, a private dataset consisting of images of normal corn and six kinds of damage corns, including heat-damaged, germ-damaged, cob-rot-damaged, blue eye mold-damaged, insect-damaged, and surface mold-damaged, were collected in this work. The proposed method achieved an accuracy of 96.67% for the classification between normal corns and the first four common damaged corns, and an accuracy of 74.76% was achieved for the classification between normal corns and six kinds of damaged corns. The experimental results demonstrated the effectiveness of the proposed corn classification system.

To better solve the problems associated with optimal pathfinding and dynamic obstacle avoidance in the path planning of mobile robots, a hybrid path planning scheme combining modified gray wolf optimization (MGWO) and situation assessment mechanism is proposed. Firstly, a MGWO algorithm is proposed to plan a global path. Secondly, different situational factors for robots in different regions are extracted from the fusion results of 2D laser measurements and image data, and a Bayesian network model of robot action selection is established. Then, the situational factors of the robot are used as evidence for reasoning. Based on the posterior probability value in the inference result, the grid to be moved is selected and the traveling direction of the robot is adjusted in order to take advantage of both global path planning and local dynamic obstacle avoidance. The simulation results show that the proposed MGWO has better optimization performance. When combined with a situation assessment mechanism, it realizes dynamic obstacle avoidance while keeping the path length as short as possible.

Enhancing cultivated land system resilience is a fundamental prerequisite for improving land use efficiency and thus addressing climate change. Taking the Songnen Plain—a major grain production area in China—as the study region, this study constructs a definition of cultivated land system resilience from three dimensions (“resistance–adaptability–reconstruction”). An index system for resilience evaluation is established, and methods such as three-dimensional Euclidean distance and K-means clustering are employed to investigate the spatiotemporal differentiation characteristics of cultivated land resilience in the Songnen Plain from 2001 to 2021. Based on the findings, zoning is performed and corresponding management strategies are proposed: (1) Overall resilience in the Songnen Plain increased from 0.4450 in 2001 to 0.7469 in 2021; enhanced resistance played the most significant role in promoting this increase. (2) The Songnen Plain exhibited pronounced spatial differentiation in cultivated land resilience, characterized by higher resilience in the eastern region and weaker resilience in the central and western regions. (3) The zoning results reveal significant disparities in resilience levels within the study area; targeted measures are thus required to address key problems in each zone. This study provides theoretical insights and empirical conclusions for formulating differentiated protection policies for cultivated land systems, thereby ensuring the sustainable development of the Songnen Plain’s cultivated land system.

Rapid and efficient evaluation of cultivated land quality in black soil regions at the farm scale using remote sensing techniques is crucial for resource protection. However, current studies face challenges in developing convenient and reliable models that directly leverage raw spectral reflectance. Therefore, this study develops and validates a deep learning framework specifically for this task. The framework first selects remote sensing images from typical periods using a Random Forest model in Google Earth Engine (GEE). Subsequently, the raw spectral reflectance data from these images, without any transformation into vegetation indices, are directly input into an optimized BO-Stacking-TabNet model. This model is enhanced through a two-step Stacking ensemble process and a Bayesian optimization algorithm. A case study at Shuanghe Farm in Northeast China shows that (1) compared to the BO-Stacking-TabNet model using vegetation indices as input, the BO-Stacking-TabNet model based on spectral reflectance as the input indicator achieved an improvement of 10.62% in Accuracy, 1.55% in Precision, 11.05% in Recall, and 10.18% in F1-score. (2) Compared to the original TabNet model, the BO-Stacking-TabNet model optimized by the two-step Stacking process and Bayesian optimization algorithm improved Accuracy by 2.13%, Precision by 12.59%, Recall by 1.83%, and F1-score by 2.19%. These results demonstrate the reliability of the new farm-scale black soil region cultivated land evaluation method we proposed. The method provides significant references for future research on cultivated land quality assessment at the farm scale in terms of remote sensing image data processing and model construction.