To facilitate rehabilitation training for patients, we proposed the implementation of an anthropomorphic exoskeleton structure that incorporates a variable instantaneous center of rotation (ICR). This design considers the variability in knee ICR among individuals, resulting from the irregular form of the human knee joint, and leverages a double-degrees-of-freedom (2DOF) five-bar mechanism to adapt to these differences. The walking gait of the human lower limb and the corresponding knee ICR were measured and calculated using an optical 3D motion capture system. The optimal dimension parameters of the five-bar mechanism were then obtained through the optimization of human movement position inputs and rod length constraints to minimize the error in knee ICR, gait angle, and ankle trajectory between the human and the exoskeleton. Finally, we established an exoskeleton prototype to conduct relevant experimental tests. The experiment results showed that the average errors of knee ICR trajectory, hip angle, knee angle, and ankle trajectory were 5.52 × 10⁻⁴ m, 0.010 rad, 0.014 rad, and 1.57 × 10⁻³ m, respectively. The experimental results demonstrated that the exoskeleton’s movement trajectory was close to the human’s, reducing the human–mechanism interaction force and improving patient comfort during rehabilitation training.

The selection of effective and representative spectral bands is extremely important in eliminating redundant information and reducing the computational burden for the potential real-time applications of hyperspectral imaging. However, current band selection methods act as a separate procedure before model training and are implemented merely based on extracted average spectra without incorporating spatial information. In this paper, an end-to-end trainable network framework that combines band selection, feature extraction, and model training was proposed based on a 3D CNN (convolutional neural network, CNN) with the attention mechanism embedded in its first layer. The learned band attention vector was adopted as the basis of a band importance indicator to select effective bands. The proposed network was evaluated by two datasets, a regression dataset for predicting the relative chlorophyll content (soil and plant analyzer development, SPAD) of basil leaves and a classification dataset for detecting the drought stress of pepper leaves. A number of calibration models, including SVM, 1D-CNN, 2B-CNN (two-branch CNN), 3D ResNet and the developed network were established for performance comparison. Results showed that the effective bands selected by the proposed attention-based model achieved higher regression R² values and classification accuracies not only than the full-spectrum data, but also than the comparative band selection methods, including traditional SPA (successive projections algorithm) and GA (genetic algorithm) methods and the latest 2B-CNN algorithm. In addition, different from the traditional methods, the proposed band selection algorithm can effectively select bands while carrying out model training and can simultaneously take advantage of the original spectral–spatial information. The results confirmed the usefulness of the proposed attention mechanism-based convolutional network for selecting the most effective band combination of hyperspectral images.

Aflatoxin contamination of maize is a major food safety issue worldwide. The problem is of special significance in African countries because maize is a staple food. This manuscript describes a low-cost, portable, non-invasive device for detecting and sorting aflatoxin-contaminated maize kernels. We developed a prototype employing a modified, normalized difference fluorescence index (NDFI) detection method to identify potentially aflatoxin-contaminated maize kernels. Once identified, these contaminated kernels can be manually removed by the user. The device consists of a fluorescence excitation light source, a tablet for image acquisition, and detection/visualization software. Two experiments using maize kernels artificially infected with toxigenic Aspergillus flavus were implemented to evaluate the performance and efficiency of the device. The first experiment utilized highly contaminated kernels (71.18 ppb), while mildly contaminated kernels (1.22 ppb) were used for the second experiment. Evidently, the combined approach of detection and sorting was effective in reducing aflatoxin levels in maize kernels. With a maize rejection rate of 1.02% and 1.34% in the two experiments, aflatoxin reduction was achieved at 99.3% and 40.7%, respectively. This study demonstrated the potential of using this low-cost and non-invasive fluorescence detection technology, followed by manual sorting, to significantly reduce aflatoxin levels in maize samples. This technology would be beneficial to village farmers and consumers in developing countries by enabling safer foods that are free of potentially lethal levels of aflatoxins.

Aflatoxin contamination in peanut products has been an important and long-standing problem around the world. Produced mainly by Aspergillus flavus and Aspergillus parasiticus, aflatoxins are the most toxic and carcinogenic compounds among toxins. This study investigated the application of fluorescence visible near-infrared (VNIR) hyperspectral images to assess the spectral difference between peanut kernels inoculated with toxigenic and atoxigenic inocula of A. flavus and healthy kernels. Peanut kernels were inoculated with NRRL3357, a toxigenic strain of A. flavus, and AF36, an atoxigenic strain of A. flavus, respectively. Fluorescence hyperspectral images under ultraviolet (UV) excitation were recorded on peanut kernels with and without skin. Contaminated kernels exhibited different fluorescence features compared with healthy kernels. For the kernels without skin, the inoculated kernels had a fluorescence peaks shifted to longer wavelengths with lower intensity than healthy kernels. In addition, the fluorescence intensity of peanuts without skin was higher than that of peanuts with skin (10 times). The fluorescence spectra of kernels with skin are significantly different from that of the control group (p<0.001). Furthermore, the fluorescence intensity of the toxigenic, AF3357 peanuts with skin was lower than that of the atoxigenic AF36 group. Discriminate analysis showed that the inoculation group can be separated from the controls with 100% accuracy. However, the two inoculation groups (AF3357 vis AF36) can be separated with only ～80% accuracy. This study demonstrated the potential of fluorescence hyperspectral imaging techniques for screening of peanut kernels contaminated with A. flavus, which could potentially lead to the production of rapid and non-destructive scanning-based detection technology for the peanut industry.

Autonomous navigation is key to improving efficiency and addressing labor shortages in the fruit industry. Semi-structured orchards, with straight tree rows, dense weeds, thick canopies, and varying light conditions, pose challenges for tree identification and navigation line extraction. Traditional 3D lidars suffer from a narrow vertical FoV, sparse point clouds, and high costs. Furthermore, most lidar-based tree-row-detection algorithms struggle to extract high-quality navigation lines in scenarios with thin trunks and dense foliage occlusion. To address these challenges, we developed a 3D perception system using a servo motor to control the rolling motion of a 2D lidar, constructing 3D point clouds with a wide vertical FoV and high resolution. In addition, a method for trunk feature point extraction and tree row line detection for autonomous navigation has been proposed, based on trunk geometric features and RANSAC. Outdoor tests demonstrate the system’s effectiveness. At speeds of 0.2 m/s and 0.5 m/s, the average distance errors are 0.023 m and 0.016 m, respectively, while the average angular errors are 0.272° and 0.146°. This low-cost solution overcomes traditional lidar-based navigation method limitations, making it promising for autonomous navigation in semi-structured orchards.