المجلة الدولية للعلوم والتقنية

Maintaining the high accuracy of landmines recognition is an essential objective for deminres safety. Those who are involved in mine actions need to be informed about the true type, and the current condition, of the unexploded ordinance they will be dealing with. Therefore, timely and accurate definition of such lethal items is a crucial task in explosives hazard areas. Deep Learning models have been experimented in landmines recognition, and have shown impressive result. However, the lack of field realistic data sets disrupts adequate model training efficiency. The deterioration and damage of landmine structures have caused huge safety concerns. Corrosion has been identified as a big reason for the deterioration and damage, which causes demining members confused with other materials. Hence, this paper proposes a landmine data sets enrichment method based on Generative Artificial Intelligence tools that are used to produce more realistic images for an augmented landmines data set. The proposed process is based on OpenAI Dall-E image generator as the landmine data augmentation tool, and MobileNetV2 deep learning model as the landmine recognition tool. The goal is to effectively predict the corrosion behavior of steel structured landmines. The system can simulate corrosion based on the dataset collected from different internet resources. Based on comparative experiments, this system demonstrates outstanding performance and outperforms the baseline model. Thus, the method proposed in this study can be a promising tool to assist in building DL agents deployed in mobile applications useful for landmines and UXOs identification.