Recognition of geological legends on a geological profile via an improved deep learning method with augmented data using transfer learning strategies

作者：Tian, Miao; Ma, Kai; Liu, Zhihao; Qiu, Qinjun^*; Tan, Yongjian; Xie, Zhong

来源：Ore Geology Reviews, 2023, 153: 105270.

DOI：10.1016/j.oregeorev.2022.105270

• Geological profile
• Transfer learning
• DT-SE-ResNet50
• Legend recognition
• Data augmentation

Accurately and efficiently legend recognition for geological maps aims to not only helps users quickly obtain the name and meaning of symbols, but also understand the geological profile and establishing a connection between the profile and the contextual text. Detection and recognition entail high complexity and uncertainty because of the diversity of geological legend types and the randomness of legend distribution; thus, automatically recognizing dynamic legends for geological profiles is challenging. Deep learning (DL) algorithms have a strong ability to recognize high-level features in mineral exploration data and have been widely employed for the recognition of multiple image data. In this study, we developed a geological legends identification model, DT-SE-ResNet50, optimized by SENet and pre-trained by ImageNet to form a new model, SE-ResNet50. The SE-ResNet50 model was trained with the constructed database to obtain a new weight. The SE-ResNet50 model with the new weight was then trained by transfer learning with a self-built database to create the DT-SE-ResNet50 model for the identification of geological legends. We also propose an automatic workflow for generating datasets by data enhancement for training geological legend models to deal with the difficulty of lacking of large amount of samples. We present the results of a wide-ranging evaluation of the performance of the proposed method using the constructed geological profile legend dataset. These results show that the proposed method can significantly outperform previous studies, as well as previous methods based on supervised deep learning.