Anomaly Detection in Wide Area Imagery

This study is about detecting anomalies in wide area imagery collected from an aircraft. The set of anomalies have been identified as anything out of the normal course of action. For this purpose, two different data sets were used and the experiments were carried out on these data sets. For anomaly detection, a convolutional neural network model that tries to generate the next image using past images is designed. The images were pre-processed before being given to the model. Anomaly detection is performed by comparing the estimated image and the true image.