The rapid development of railway usage is subjecting the rail tracks to more stressful cycles with faster speed that induce defects. With time, these rail defects could trigger train derailment. Non-destructive testing methods (NDTs) are vastly used for rail defect detection. However, traditional NDTs are mostly contact type and have low inspection speed. In this regard, laser ultrasonic technology (LUT) that can ensure a noncontact and efficient inspection becomes popular. In this study, to detect rail surface and internal defects, Rayleigh waves were generated and emitted into rails by laser excitation device and then received the reflected signals by a 3D laser scanner in real-time, making the transduction system fully noncontact. However, the laser-generated ultrasonic signals, such as Rayleigh waves, are in broadband frequency that easily embedded with noise. Hence, weak signals that contain defect information are usually overwhelmed by noise. Duffing chaotic oscillator is a promising signal processing technique that can identify weak ultrasonic signals. It has successfully identified the weak signals reflected by both surface and internal defects. Hence, a novel rail inspection system has been designed by LUT coupled with the oscillator. With the help of this automatic and fully laser-based transduction system, the locations of both surface and subsurface defects occur in the entire structure of rail can be detected. Since the system can perform a line-to-line and long-distance inspection manner, it is superior to the slow and tedious point-to-point inspection systems that employ conventional NDT methods. It is worth to note here that it is the first time to reveal the feasibility of using narrowband Rayleigh waves to inspect rails by using a fully, non-contact and laser-based transduction system. Moreover, the rail inspection process can be automated by an industrial robot for real-time and in field operation.