Workflow towards autonomous and semi-automized UXO Survey and Detection

This paper presents a workflow for UXO detection based on multibeam data in combination with AUV-based ground truth. An artificial neuronal network (ANN) is trained manually annotated and aims making the generation of target lists faster more objective. Prior to annotation checked according several quality factors ensure that it fits purpose object detection. The accuracy annotations has an influence predicted probabilities ANN, as determined by experts are considered during training.To make this whole even effective terms survey time, check ANN analysis can run automated vessel. While MBES mapping continues, autonomous underwater vehicles (AUVs) be used truth possible targets additional sensors, such geomagnetic under water cameras. precise training data, reliable outcome will be.