Labor shortages in the agricultural sector have become a global concern, especially in countries with aging population. The use of AI technology in agriculture not only addresses the lack of workforce but also increases efficiency by monitoring plant growth and diseases and enabling autonomous transportation. In agriculture, autonomous tractors have been increasingly used, requiring advanced navigation systems to support various agricultural operations. This research focused on designing an autonomous navigation system based on a camera and deep learning for a crawler tractor, allowing it to operate on farm roads without human intervention. The developed system integrated three main components: a vision system, a controlling system, and a Robot Operating System (ROS). The vision system utilized dual cameras, OpenCV, and the YOLOPV algorithm with 3500 trained images to detect drivable areas and traffic lines. The hardware includes a Raspberry Pi, PLC, and a computer dedicated to control tractor actuators. Then ROS facilitates communication between the vision system and control systems, enabling autonomous navigation for tractor. The system demonstrated effective performance with safe and accurate driving after being tested on various road shapes.