Akademik Veri Yönetim Sistemi
EXPERT SYSTEMS WITH APPLICATIONS, cilt.312, 2026 (SCI-Expanded, Scopus)
Traffic anomaly detection (AD) is essential for improving public safety, reducing risks, and enabling quick responses in intelligent surveillance systems. Aerial traffic monitoring, particularly using Unmanned Aerial Vehicles (UAV), has gained attention due to its potential to address challenges like dynamic urban environments, yet it remains underexplored. Detecting anomalies in drone-captured video involves unique obstacles: rare events, small and overlapping objects, multi-scale targets, and complex backgrounds. To address these challenges, we propose the Depthwise Convolutional Variational Autoencoder (DwCVAE), a novel model designed to enhance AD in drone-based traffic surveillance. DwCVAE leverages depthwise convolutions, which allow efficient and detailed feature extraction, improving model sensitivity to subtle and multi-scale anomalies. The proposed DwC-VAE adopts an encoder-latent-decoder VAE architecture, in which stacked depthwise convolutional layers in the encoder emphasize spatially localized feature learning while maintaining channel-wise efficiency, and a compact variational latent space captures the distribution of normal traffic dynamics. Built on variational autoencoder (VAE) architecture, DwCVAE creates compact latent representations that capture normal traffic patterns, enabling reliable detection of deviations. Anomalies are identified through reconstruction-based scoring, where events that deviate from the learned normal representations yield higher reconstruction errors. This depthwise approach marks a key innovation, optimizing both computational efficiency and detection accuracy. We design four additional models: Convolutional Variational Autoencoder (CVAE), Dilated Convolutional VAE (DCVAE), Separable Convolutional VAE (SCVAE), and Convolutional LSTM VAE (CLSTMVAE) to systematically assess the effectiveness of DwCVAE. Additionally, we evaluate DwCVAE against state-of-the-art weakly supervised and unsupervised models on two benchmark datasets, Drone-Anomaly and UIT-Adrone. DwCVAE achieves an AUC of 74.95 with an EER of 0.30 on Drone-Anomaly, and an AUC of 79.77 with an EER of 0.27 on UIT-ADrone, demonstrating its superior performance in complex aerial surveillance tasks.