Deep learning methods applied to anomaly detection in vehicle manufacturing and operations
As one of the most common modes of transportation, vehicles are very closely related to our lives. As a result, safety is an important issue in both vehicle production process and vehicle operations. Recently, unmanned vehicles have received much attention from both companies and academia. The first thing we need to consider for unmanned vehicles is safety. With the adoption of deep learning (DL) methods, DL-based defect detection and fault detection technology has evolved into a powerful tool with increased accuracy and autonomy compared with traditional detection technology. This thesis presents novel deep learning methods which can help detect defects in X-ray images of vehicle engines during the manufacturing process and various faults in the operation of autonomous vehicles. This thesis is focused on applying deep neural networks in image-based classification operations. The results show that the algorithms can successfully detect the anomalies with satisfactory accuracy.