| Abstract |
This paper is presenting the collaborative approach where Siamese networks (a branch of neural networks) and clustering methods work together to improve the visual analytics for enhancing surveillance capabilities of UAVs (Unmanned Aerial Vehicle). The performance of the intelligent UAVs can be improved with the aid of advanced intelligent methods as UAVs are utilized in smart cities for performing various tasks including security surveillance. In this study, a novel clustering method is devised in a combination with Siamese networks to save energy for smarter navigation in complex environments. The UAV network is distributed into polygonal segments in the beginning, and then a few cluster heads are advantageously deployed to shield each polygon area evenly. The deep Siamese networks allow the UAVs to comprehend graphical and visual similarities in the data and extract the relevant features from the visual inputs. These features help in clustering on the basis of similarity in features, to identify the patterns within the data. These patterns further, allow the UAVs to make the decisions based on the trends of the patterns. The UAVs are trained by using intelligent Siamese networks. The results of the proposed study demonstrate that the integration of Siamese networks and clustering techniques effectively reduces the number of clusters and minimizes the handover rate between clusters, and extending the lifespan of the network nodes. The integration of deep learning and clustering techniques holds immense potential for enhancing situational awareness, and advancing the capabilities of UAVs in visual analytics. © 1975-2011 IEEE. |
