This paper presents a real-time vision-based navigation system designed for the efficient detection of environmental stimuli in mobile robots. The system integrates a vision module implemented using OpenCV and multithreading with a finite state machine to autonomously guide the robot toward a light source even under resource-constrained conditions. The incorporation of RAOI behavior rules, which include repulsion, attraction, orientation, and influence, facilitates precise target localization and enables effective obstacle avoidance. Validation experiments conducted with the robot starting from various initial orientations demonstrate that combining a vision system with simple behavior rules significantly enhances navigation performance. The results highlight the trade-offs between maneuver complexity, energy consumption, and navigation accuracy, providing a robust foundation for future developments in swarm robotics for collaborative tasks.