Mobile robots in automation construction have been designed for applications of craning, conveying, excavating, and floor polishing. These robots nowadays are equipped with various sensors to detect environmental information and finish tasks autonomously. When robot's navigating path needs to be rescheduled, the supervisor of robot can duly interrupt system and then redefines a new route for robot. In addition to robot remote control by radio signals, using digital camera to receive instructions from supervisor's gestures is also effective and can avoid the drawback of networked data routing. In this paper, we propose a gesture tracking system by simulating a traffic light baton to guide a differential drive robot in construction site. Here a real-time moving object detection first tracks supervisor's waves (gestures) with digital camera. Next the system determines guiding direction and steering angles based on fuzzy logic. All of our designs are implemented in single FPGA chip for operating under rigor environments. The experimental results demonstrate that proposed gesture tracking system is accurate and promising for chip-based gesture guidance on construction robots in the future.