KRK. Radhika

K. Radhika

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


K. Radhika has a total of 10 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Intelligent Multi-Group Marine Predator Algorithm With Deep Learning Assisted Anomaly Detection in Pedestrian Walkways

    by S. Rama Sree, E. Laxmi Lydia, Mohammed Altaf Ahmed, K. Radhika, Mohamad Khairi Ishak, Khalid Ammar, Mohamed Fauzi Packeer Mohamed

    Anomaly Detection (AD) in Pedestrian Walkways (PWs) is critical to urban security and safety systems. It is widely used to detect abnormal or unusual behaviours, situations, or events in areas dedicated to pedestrian traffic, namely crosswalks, sidewalks, or pedestrian bridges. The main objective is to improve efficiency, safety, and security in the urban environment by identifying deviations and monitoring pedestrian activities from established norms. This kind of AD typically includes surveillance cameras, sensors, and advanced software algorithms. Using advanced machine learning (ML) and computer vision (CV) approaches, this technique continuously monitors the pedestrian area to detect potential threats and irregularities. Deep Learning Assisted AD in Pedestrian Walkways presents a novel and very efficient method to enhance security and safety in urban environments. Therefore, this study designs an Intelligent Multi-Group Marine Predator Algorithm with Deep Learning Assisted Anomaly Detection (MMPADL-AD) in Pedestrian Walkways. The MMPADL-AD system aims to ensure security in PWs via the AD process. The MMPADL-AD technique incorporates a NASNet feature extractor that proficiently extracts high-level features from surveillance data, allowing a deep understanding of pedestrian behaviours. Besides, the MMPADL-AD technique applies convolutional long short-term memory (ConvLSTM), inheriting the benefits of convolutional neural networks) and LSTM for the AD process. Finally, the MMPA has been used for the hyperparameter tuning mechanism, which optimizes the model’s performance, assuring accuracy and adaptability. Benchmark data accompanied an extensive set of experiments to ensure the higher effectiveness of the MMPADL-AD approach. The experimental values highlighted the supremacy of the MMPADL-AD approach over other DL methods.


  • Open Access

    Rider Optimization With Deep Learning Based Image Encryption for Secure Drone Communication

    by N. Kannaiya Raja, E. Laxmi Lydia, Thumpala Archana Acharya, K. Radhika, Eunmok Yang, Okyeon Yi

    In recent years, drones or Unmanned Aerial Vehicles (UAVs) got significant attention among researchers because of their extensive application in commercial applications, border surveillance, etc. As the conventional terrestrial communication system does not work effectively on heavy calamities namely floods, landslides, cyclones, earthquakes, etc., UAVs can offer a potential solution for inexpensive, rapid, and wireless communication. Despite the drones’ benefits in emergency monitoring, security is been a main factor because of the existence of wireless connections for transmission. Therefore, this article introduces optimal deep learning with image encryption-based secure drone communication (ODLIE-SDC) technique. The major intention of the ODLIE-SDC technique lies in the effectual secure communication and classification process in emergency monitoring scenarios. To accomplish this, the presented ODLIE-SDC technique designs a hyperchaotic map-based image encryption technique and its optimal keys are produced by the use of a rider optimization algorithm (ROA). The image classification process is performed encompassing EfficientNet-B4-CBAM feature extraction and enhanced stacked autoencoder (ESAE) classification. Finally, the hyperparameter tuning of the EfficientNet-B4-CBAM technique takes place using the Bayesian optimization (BO) algorithm. The experimental validation of the ODLIE-SDC technique is tested on the AIDER dataset. The comprehensive comparative analysis reported the enhanced performance of the ODLIE-SDC technique over other existing approaches.