Video anomaly detection has become a popular research area with important security and public safety applications, due to the massive amount of video surveillance data being generated, which humans cannot effectively monitor. Video anomaly detection algorithms are crucial for flagging unusual activity in surveillance video for further review by human operators. While an anomaly can sometimes be due to a simple reason such as a suspicious item (e.g., gun, car in a pedestrian area, etc.) or a suspicious activity (e.g., running, jumping, etc.) under ideal circumstances, it may in general require a more holistic understanding of the scene with multiple objects and be robust to several problems such as weather, wireless connectivity, and natural or artificial noise.
Existing works focus on detecting simple anomalies that involve a single object/actor. In real-world surveillance, in addition to simple anomalies, complex anomalies that involve the interaction of two or more objects/actors, such as person leaving a bag, cyclist running into a car, etc., should also be detected by video anomaly detection algorithms.
Recent advancements in video understanding can enable effective solutions to the real-world surveillance challenges such as interaction anomalies, moving camera, weather conditions (e.g., rain, snow, fog), wireless connectivity problems (e.g., low resolution, glitch, freeze), cyber attacks, and adversarial machine learning attacks.
Research articles contributing new methods/datasets and survey articles reviewing the literature are solicited for this Research Topic.
The topics of interest include, but are not limited to the following:
- Detecting interaction anomalies that involve two or more objects/actors
- Anomaly detection with moving camera such as shaking, turning, or translating camera
- Video anomaly detection for autonomous vehicles
- Robustness to weather conditions such as fog, rain, snow, etc.
- Robustness to noisy video stream such as low resolution, glitch, freeze, etc.
- Cyber attacks and defenses to cyber attacks (e.g., WiFi deauthentication attack) related to video anomaly detection)
- Adversarial machine learning attacks and defenses related to video anomaly detection
- Continual learning for video anomaly detection
- Multimodal video anomaly detection
- Computationally efficient video anomaly detection architectures
- Datasets pertaining to above challenges
Keywords:
video understanding, anomaly detection, video surveillance, computer vision, machine learning
Important Note:
All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
Video anomaly detection has become a popular research area with important security and public safety applications, due to the massive amount of video surveillance data being generated, which humans cannot effectively monitor. Video anomaly detection algorithms are crucial for flagging unusual activity in surveillance video for further review by human operators. While an anomaly can sometimes be due to a simple reason such as a suspicious item (e.g., gun, car in a pedestrian area, etc.) or a suspicious activity (e.g., running, jumping, etc.) under ideal circumstances, it may in general require a more holistic understanding of the scene with multiple objects and be robust to several problems such as weather, wireless connectivity, and natural or artificial noise.
Existing works focus on detecting simple anomalies that involve a single object/actor. In real-world surveillance, in addition to simple anomalies, complex anomalies that involve the interaction of two or more objects/actors, such as person leaving a bag, cyclist running into a car, etc., should also be detected by video anomaly detection algorithms.
Recent advancements in video understanding can enable effective solutions to the real-world surveillance challenges such as interaction anomalies, moving camera, weather conditions (e.g., rain, snow, fog), wireless connectivity problems (e.g., low resolution, glitch, freeze), cyber attacks, and adversarial machine learning attacks.
Research articles contributing new methods/datasets and survey articles reviewing the literature are solicited for this Research Topic.
The topics of interest include, but are not limited to the following:
- Detecting interaction anomalies that involve two or more objects/actors
- Anomaly detection with moving camera such as shaking, turning, or translating camera
- Video anomaly detection for autonomous vehicles
- Robustness to weather conditions such as fog, rain, snow, etc.
- Robustness to noisy video stream such as low resolution, glitch, freeze, etc.
- Cyber attacks and defenses to cyber attacks (e.g., WiFi deauthentication attack) related to video anomaly detection)
- Adversarial machine learning attacks and defenses related to video anomaly detection
- Continual learning for video anomaly detection
- Multimodal video anomaly detection
- Computationally efficient video anomaly detection architectures
- Datasets pertaining to above challenges
Keywords:
video understanding, anomaly detection, video surveillance, computer vision, machine learning
Important Note:
All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.