This study develops forest road recognition technology using deep learning-based image processing to support the advancement of autonomous driving technology for forestry machinery. Images were collected while driving a tracked forwarder along approximately 1.2 km of forest roads. A total of 633 images were acquired, with 533 used for the training and validation sets, and the remaining 100 for the test set. The YOLOv8 segmentation technique was employed as the deep learning model, leveraging transfer learning to reduce training time and improve model performance. The evaluation demonstrates strong model performance with a precision of 0.966, a recall of 0.917, an F1 score of 0.941, and a mean average precision (mAP) of 0.963. Additionally, an image-based algorithm is developed to extract the center from the forest road areas detected by YOLOv8 segmentation. This algorithm detects the coordinates of the road edges through RGB filtering, grayscale conversion, binarization, and histogram analysis, subsequently calculating the center of the road from these coordinates. This study demonstrates the feasibility of autonomous forestry machines and emphasizes the critical need to develop forest road recognition technology that functions in diverse environments. The results can serve as important foundational data for the future development of image processing-based autonomous forestry machines.

Logging operations comprise a repeated and tedious job in forestry operations because forestry forwarders must keep completing round-trip transportation on forest roads from tree-cutting sites to forest roads where their truck can be accessed. In this study, an autonomous driving system for tracked forwarders was developed using GNSS (Global Navigation Satellite System)/INS (Inertial Navigation System). The mechanical control system of the forwarder was replaced with an electronic control system, and path-planning and -tracking algorithms were implemented. The electronic control system, operated by servo motors to operate the driving levers, exhibited a response that was 150 milliseconds faster in lever control compared to manual operation. To generate an autonomous driving path, a skilled operator drove the forwarder along a forest road, and the recorded path was post-processed using the Novatel Inertial Explorer 8.70 GNSS + INS software to minimize GNSS errors. The autonomous forwarder followed the generated path using the pure pursuit algorithm. Autonomous driving tests conducted along this path achieved a root mean square error (RMSE) within 0.4 m (range: 0.389–0.393). Driving errors were primarily attributed to GNSS positional inaccuracies, especially in environments with dense canopies and landslide prevention structures located higher than the GNSS antenna, obstructing satellite signals. These findings underscore the importance and feasibility of autonomous forwarders in diverse forest environments, providing a critical foundation for advancing autonomous forestry machinery. The proposed technologies are expected to significantly contribute to enhancing the productivity of forestry operations.

Small harvesting compartments and steep terrain in South Korea favor harvesting systems in which a single machine performs multiple processes. Multifunctional systems based on small tracked excavators, equipped with a grapple saw for felling and a clambunk skidder for extraction, have therefore been increasingly adopted to address labor shortages and to improve operational safety. However, because these base machines lack a chassis suspension system and their operator seats provide limited vibration isolation, operators of small-excavator-based systems may be exposed to elevated levels of whole-body vibration (WBV) and shock vibration. This study quantified operator ride-vibration exposure and assessed the associated health risks during small-excavator-based timber harvesting in accordance with ISO 2631-1; the assessment was based on measured vibration exposure rather than on direct medical evaluation. Tri-axial accelerations were measured simultaneously at the operator seat and the cab floor of a 5-ton-class excavator over eight workdays, and one representative workday was analyzed at the work-element level. The harvesting process was divided into timber felling (machine travel, positioning, clearing, and cutting and bunching) and timber extraction (machine travel on forest roads, machine travel on skid trails, and loading and unloading). Exposure was evaluated using the frequency-weighted root-mean-square (RMS) acceleration, the crest factor (CF), and the total vibration dose value (TVDV). The CF exceeded 9 for all work elements, indicating shock-dominated vibration and justifying the TVDV as the primary evaluation metric. Timber extraction generated higher vibration exposure than timber felling, and the seat TVDV differed significantly among the work elements within each operation (p < 0.001). The highest seat TVDV occurred during loading and unloading (123.05 m/s^1.75), exceeding the upper ISO 2631-1 Health Guidance Caution Zone limit for an 8-h exposure by more than eight times, and all major work elements exceeded the ISO 2631-1 health guidance limits. To our knowledge, this study provides the first work-element-level quantification of WBV exposure for small-excavator-based timber harvesting systems in South Korea. The findings underscore the need for improved seat suspension systems, vibration mitigation technologies, and vibration-aware work planning strategies.