The steel industry is constantly trying to reduce production cost and improve quality by making the steel manufacturing processes continuous and faster. Currently, the rolling process of steel production is largely automated, while the finishing process is not yet appreciably automated. The finishing processes involve many tasks difficult to automate, such as defect inspection and repairing the detected defects. In recent years, however, many automated and labor-saving systems have been developed for use in the finishing processes. The surface defect inspection of steel products is the largest bottleneck in the finishing process. This paper describes an inspection system of steel surface defects for large sections, such as wide flange beams and I-beams. This system is based on applied radiant light and it senses the temperature deviation caused by defects. The wavelength of the detector is optimized to improve the signal-to-noise ratio. An optical attenuator was developed to compensate for the known temperature distribution across the product immediately after rolling. The image processor takes only 50 ms per image frame. Each time frame has the necessary image information to detect defects.