Anomaly Detection in Pedestrian Walkways using Imperialist Competitive Algorithm with Deep Learning on Remote Sensing Images

Remote sensing images (RSI) describe images captured by sensors aboard satellites, drones, aircraft, or alternative platforms from a large distance above the Earth's surface. These images are useful for numerous applications, comprising agriculture, ecological monitoring, urban development, disaster management, and land use planning. Anomaly Detection (AD) in pedestrian walkways is a significant research domain, generally employed to increase the security of pedestrians. The conventional manual analysis of labeling abnormal actions is a tedious task owing to the extensive application of video surveillance systems and the increased quantity of captured videos. Currently, the expansion of deep learning (DL) methods are gained great interest in various methods of CV such as object classification and object detection, and such applications are dependent upon supervised learning that needs labels. This manuscript devises an Anomaly Detection in Pedestrian Walkways using Imperialist Competitive Algorithm with Deep Learning (ADPW-ICADL) technique on Remote Sensing Images. The ADPW-ICADL technique starts with Wiener filter (WF) scheme to develop the quality of the images by decreasing the noises. To derive features, the ADPW-ICADL technique applies a DenseNet100 architecture for capturing complicated patterns in pedestrian walkways. Also, the Imperialist Competitive Algorithm (ICA) is utilized for the hyperparameter tuning process to assure the optimal parameter selection of the DenseNet model. At last, Gradient Boosting Machine (GBM) is applied for proper identification of the anomalies in the walkway patterns. By wide-ranging experimental results on RSIs, the ADPW-ICADL technique demonstrates higher AD performance, emphasizing its impending for improving the dependability and productivity of pedestrian walkway investigation in several geographic settings.