Corrosion is one of the most universal major defects in structured systems and can cause many safety risks if proper treatment and maintenance is not done. Robotic systems such as drones with integrated machine vision modules for real-time corrosion monitoring have the potential to significantly decrease the inspection time, unreliability of inspection procedures. This study investigates the use of bag of key points method for visual categorization of corroded and non-corroded images. The notion behind bag of key points method is that an image can be represented as a histogram that counts the occurrences of visual words. These visual words are obtained by clustering feature vectors of key points in the training images. We have used SIFT (Scale Invariant Feature Transform) descriptors to represent key points and used the following algorithms for classification after clustering the descriptors and generating features of fixed size: SVM, Decision Tree, Random Forest, KNN, ANN. We present results for corrosion identification problem and the results reveal that the method proposed is robust and produces good performance measures.