Unattended object intelligent analyzer for consumer video surveillance

Detection of stationary foreground objects: A survey

“…Whereas most approaches use a unique BG model, some strategies propose constructing two models [80], which are statistically updated at different speeds [86] or apply two different selective BG updates [107]. These models are commonly called short-term model and long-term model.…”

“…Additionally, some of them are able to deal with other specific challenges. For example, the strategy in [20] includes mechanisms to be robust to illumination changes, the strategy in [12] avoids false detections due to camera jitter, and the strategy in [80] includes a post-processing stage to remove shadows from the detections and provide robustness against lighting changes.…”

“…In the strategy proposed by Porikli, and some other later works [69,71,98], the models are constructed using multiple Gaussians. However, other modeling choices can also be found in the literature: non-statistical models (basic) in [80], [86], [87] and [107], SGMs in [61], MMs in [63], or CMs in [44].…”

Moving object detection using unstable camera for video surveillance systems

“…[1] , [2] , [3] The detection of a moving object can be considered as a low-or intermediatelevel processing step for various object recognition and analysis applications.…”

Multiple moving object segmentation using motion orientation histogram in adaptively partitioned blocks for high-resolution video surveillance systems

“…To address this problem, Kim employed RGB color-based background modeling [12], and Zin used a combination of periodic background models for object segmentation [13]. But, these methods are not suitable for real-time high-resolution video surveillance systems because they have high computational complexity for the background modeling.…”