In this paper, an infrared thermography technique is used to assess the fiber orientation on the surface of carbon fiber reinforced polymer (CFRP) moulded with randomly-oriented strands (ROS). Due to the randomness of the material, a point by point inspection would be very time consuming. In this paper it is proposed to use a flying laser spot technique to heat a line-region on the surface of the sample instead of a spot. During our experiments, a flying laser spot inspection was performed in 30 seconds while a point by point inspection of the same area would require about 25 minutes. An artificial neural network (ANN) was then used to estimate the fiber orientation over the heated line. The classification rate obtained with the network was 91.2% for the training stage and 71.6% for the testing stage.