Bridge Detection using Satellite Images

Abstract

The detection of bridges played a significant role in providing construction status. In general, satellite images contain information about geographical capabilities such as bridges, which are extremely useful to both military and civilian personnel. The detection of bridges in major infrastructure projects is critical for providing data about the fame of those structures and guiding feasible decision-making processes. There are traditional methods for inspecting and identifying bridges that use IOT sensors and lasers, but these can only be identified if the object is within a medium range of distance. Convolutional neural networks and Deep learning techniques can be used to perform this identification. In addition, the Geographic Information System aids in the analysis, collection, capture, and management of geographical features. For tracking bridge health, GIS is used to control and combine disparate assets of spatial and characteristic records. The proposed method makes use of YOLOv5's advanced features, such as improved architecture and training methods, to achieve greater accuracy in detecting bridges. On the bridge dataset, transfer learning is used to fine-tune the pre-trained models of YOLOv5 and YOLOv3. The experiments are carried out on a large dataset of satellite images containing a variety of bridge types. In terms of accuracy and mean average precision (mAP) of loss, the results show that YOLOv5 outperforms YOLOv3. YOLOv5 has a mean average precision of 0.92, while YOLOv3 has a mean average precision of 0.54. This approach can be applied to a variety of infrastructure detection tasks and can help to improve the efficiency and accuracy of bridge inspections.