Sensors

The agricultural sector remains one of the most hazardous working environments, with viticulture posing particularly high risks due to repetitive manual tasks, pesticide exposure, and machinery operation. This study explores the potential of vision-based Artificial Intelligence (AI) systems to enhance occupational health and safety by evaluating their coherence with human expert assessments. A dataset of 203 annotated images, collected from 50 vineyards in Northern Italy, was analyzed across three domains: manual work activities, workplace environments, and agricultural machinery. Each image was independently assessed by safety professionals and an AI pipeline integrating convolutional neural networks, regulatory contextualization, and risk matrix evaluation. Agreement between AI and experts was quantified using weighted Cohen’s Kappa, achieving values of 0.94–0.96, with overall classification error rates below 14%. Errors were primarily false negatives in machinery images, reflecting visual complexity and operational variability. Statistical analyses, including McNemar and Wilcoxon signed-rank tests, revealed no significant differences between AI and expert classifications. These findings suggest that AI can provide reliable, standardized risk detection while highlighting limitations such as reduced sensitivity in complex scenarios and the need for explainable models. Overall, integrating AI with complementary sensors and regulatory frameworks offers a credible path toward proactive, transparent, and preventive safety management in viticulture and potentially other high-risk agricultural sectors. Furthermore, vision-based AI systems inherently act as optical sensors capable of capturing and interpreting occupational risk conditions. Their integration with complementary sensor technologies—such as inertial, environmental, and proximity sensors—can enhance the precision and contextual awareness of automated safety assessments in viticulture.

Pavement surface defects such as potholes pose significant safety risks and accelerate infrastructure deterioration. Accurate and automated detection of such defects requires both advanced sensing technologies and robust deep learning models. In this study, we propose PoFormer, a Transformer–CNN hybrid framework designed for precise segmentation of pavement potholes from heterogeneous image datasets. The architecture leverages the global feature extraction ability of Transformers and the fine-grained localization capability of CNNs, achieving superior segmentation accuracy compared to state-of-the-art models. To construct a representative dataset, we combined open source images with high-resolution field data acquired using a multi-sensor pavement inspection vehicle equipped with a line-scan camera and infrared/laser-assisted lighting. This sensing system provides millimeter-level resolution and continuous 3D surface imaging under diverse environmental conditions, ensuring robust training inputs for deep learning. Experimental results demonstrate that PoFormer achieves a mean IoU of 77.23% and a mean pixel accuracy of 84.48%, outperforming existing CNN-based models. By integrating multi-sensor data acquisition with advanced hybrid neural networks, this work highlights the potential of 3D imaging and sensing technologies for intelligent pavement condition monitoring and automated infrastructure maintenance.

Eye movement is an important tool used to investigate cognition. It also serves as input in human–computer interfaces for assistive technology. It can be measured with camera-based eye tracking and electro-oculography (EOG). EOG does not rely on eye visibility and can be measured even when the eyes are closed. We investigated the feasibility of detecting the gaze direction using EOG while having the eyes closed. A total of 15 participants performed a proprioceptive calibration task with open and closed eyes, while their eye movement was recorded with a camera-based eye tracker and with EOG. The calibration was guided by the participants’ hand motions following a pattern of felt dots on cardboard. Our cross-correlation analysis revealed reliable temporal synchronization between gaze-related signals and the instructed trajectory across all conditions. Statistical comparison tests and equivalence tests demonstrated that EOG tracking was statistically equivalent to the camera-based eye tracker gaze direction during the eyes-open condition. The camera-based eye-tracking glasses do not support tracking with closed eyes. Therefore, we evaluated the EOG-based gaze estimates during the eyes-closed trials by comparing them to the instructed trajectory. The results showed that EOG signals, guided by proprioceptive cues, followed the instructed path and achieved a significantly greater accuracy than shuffled control data, which represented a chance-level performance. This demonstrates the advantage of EOG when camera-based eye tracking is infeasible, and it paves the way for the development of eye-movement input interfaces for blind people, research on eye movement direction when the eyes are closed, and the early detection of diseases.