Rockfall Detection on Moon and Mars Using Deep Learning Models

The paper contains an in-depth examination of the identification of objects algorithms utilized for recognizing rocks on the surface of the moon. The two most renowned models, MobileNetv2 and YOLO v8, have been evaluated for their effectiveness and precision for detecting pebbles on the moon’s surface. The research capitalizes on of a vast database composed up of images with high resolution obtained from satellites and lunar rovers. In order to produce an accurate representation of lunar surface features, the database encompasses a range of light exposure situations, terrain types, and rock sizes. In order to improve the resilience of the models, an extensive preprocessing approach gets carried out, which includes picture improvement noise mitigation, and data replenishment. Neither the YOLO v8 nor MobileNetv2 models constitute leading edge identification of objects technologies that have gained popularity for their precision and quickness. Using pre-learned parameters via ImageNet, they are trained employing transfer learning approaches on the moon rock database. In order to optimize the algorithms and improve their parameters for the goal of lunar rock detection, thorough evaluations have been carried out. Using pre-learned parameters via ImageNet, they are trained employing transfer learning approaches on the moon rock database. In order to optimize the algorithms and improve their parameters for the goal of lunar rock detection, thorough evaluations have been carried out. All things considered, this paper offers insightful information about deploying recognition of objects models—more especially, MobileNetv2 and YOLO v8—for the identification of lunar rocks. A greater awareness of the moon’s geology and its attendant opportunities for research in science can be facilitated by the discoveries, which can help with the development of automated systems for geological analysis, mapping, and prospective moon-search missions.