Neural network feature recognition system

A method comprising:

• obtaining image data representing a rotor blade that is imaged at different times;
• determining changes in a candidate feature on the rotor blade over time based on image analysis of the image data, the changes in the candidate feature representing one or more of a changing size or a changing position of the candidate feature relative to the rotor blade over time;
• and
• automatically identifying the candidate feature on the rotor blade as a defect feature responsive to the changes in the candidate feature over time being one or more of the same or similar to a predicted progression of the defect feature over time according to an action-guidance function that is associated with the defect feature, the action-guidance function generated by an artificial neural network via a machine learning algorithm.

A method comprising:

• detecting a candidate feature on an object in a series of images acquired at different time and determining changes in the candidate feature relative to the object over time;
• generating, using an artificial neural network, an action-guidance function that is used to predict a progression of the defect feature relative to the object over time, the action-guidance function generated via reinforcement learning using a classifier network trained to detect the defect features in images as a supervisor to provide feedback to the artificial neural network;
• and
• automatically identifying the candidate feature on the rotor blade as the defect feature responsive to the changes in the candidate feature over time being one or more of the same or similar to the predicted progression of the defect feature according to the action-guidance function.