Research on the Intelligent Carrier Vehicle Obstacle Recognition and Avoidance System Based on the YOLOv8 Model

In response to the challenges of numerous and large-scale equipment, frequent movement, complex construction processes, and variable environments on construction sites, this paper proposes an obstacle detection and avoidance system for intelligent transportation vehicles based on the YOLOv8 algorithm. The aim is to enhance construction efficiency and quality, providing technical support for the automation and intelligentization of construction sites. To meet training requirements, an image dataset containing obstacles such as fences, temporary barriers, protective railings, and safety warning signs was constructed. After image preprocessing, screening, and standardization, a dataset of 1009 images was formed and divided into training and testing sets in a 7:3 ratio. The model training results indicate that in the task of obstacle detection, the classification accuracy reached 89%, and the model achieved an optimal precision value of 1.00 at a confidence level of 0.886, effectively identifying target objects in the scene.