Developing Agent-Oriented Video Surveillance System through Agent-Oriented Methodology (AOM) Agent-oriented methodology (AOM) is a comprehensive and unified agent methodology for agent-oriented software development. Although AOM is claimed to be able to cope with a complex system development, it is still not yet determined up to what extent this may be true. Therefore, it is vital to conduct an investigation to validate this methodology. This paper presents the adoption of AOM in developing an agent-oriented video surveillance system (VSS). An intruder handling scenario is designed and implemented through AOM. AOM provides an alternative method to engineer a distributed security system in a systematic manner. It presents the security system at a holistic view; provides a better conceptualization of agent-oriented security system and supports rapid prototyping as well as simulation of video surveillance system.
Road traffic accidents are a leading cause of injuries and fatalities globally, prompting extensive research into deep learning-based accident recognition models for their superior performance in computer vision tasks. However, most studies focus on non-mixed traffic environments, where detection is simpler due to predictable traffic patterns and uniform vehicle types. In contrast, mixed-traffic scenarios present greater challenges as diverse vehicles, motorcyclists, and pedestrians move unpredictably. Models relying on a single type of perception are effective in structured traffic but struggle to handle the complexities of mixed-traffic environments. This study proposes a novel multi-stream deep learning model called Accident Recognition in Mixed-Traffic Scene (ARMS), which integrates three distinct streams: the first stream analyzes the overall accident scene, the second focuses on mixed-traffic accident features, and the third examines vehicle motion abnormalities through object detection and tracking, aimed at improving road accident recognition accuracy in mixed-traffic environments at intersections. This model is trained and evaluated using datasets from CADP, UA-DETRAC, and supplementary online sources. The results demonstrate that the ARMS model achieves an accuracy of 93.3%, with performance improving significantly through the fusion of the individual streams. Additionally, the ARMS model was evaluated using two publicly available standard datasets, which further highlights its improved performance in recognizing mixed-traffic accidents compared to existing studies.
23 December 2024 in Journal of Technology in Behavioral Science
Human-following capability has been widely implemented in mobile robotic systems, yet its application to closeproximity assistive devices such as smart walkers remains limited. Contemporary front-following robots often rely on lowdimensional kinematic variables or reactive models that respond to user motion but neglect richer semantic, perceptual, and cognitive cues, resulting in limited adaptability and anticipation. This paper presents a review of state-of-the-art smart walker and mobile robot designs and highlights the need for intention-aware navigation in close-proximity contexts. Besides, to address this gap, a conceptual framework based on Model-Based Reinforcement Learning (MBRL) is proposed to utilize an intention-aware predictive model that encodes the relationship between gait features and the user's intended motions. This proposed framework aims to observe and learn the user's spatial and temporal gait features, identify richer motion cues, predict the motion intention, and then execute the optimal walker action that maximizes the expected reward of aligning its movement with the user's anticipated motion. Future work will focus on the gait dataset preparation, implementation of the proposed framework, experimental validation, and performance evaluation under realworld walking scenarios.
Road accidents remain a major global concern, particularly at urban intersections where cars, motorcycles, and pedestrians interact in unpredictable and often chaotic ways. Traditional accident detection systems struggle in these mixed-traffic environments, especially in areas with minimal lane discipline and diverse road user behavior. To address this challenge, we introduce accident recognition in mixed-traffic scene (ARMS), an artificial intelligence (AI)-powered, multistream deep learning model that combines scene-level context, object-specific features, and motion abnormality tracking in a unified framework. Designed to handle the visual and behavioral complexities of real-world traffic, ARMS offers a promising step toward more accurate and timely accident detection in intelligent transportation systems.