Detection of moving objects by background subtraction for foreground detection-a hybrid CNN-Viola-Jones model

• Main Authors: Safaldin, Mukaram, Zaghden, Nizar, Omari, Mahmoud
• Format: Conference Proceeding
• Language: English
• Subjects: Mathematical models; Modelling; Moving object recognition; Subtraction
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0175536

Background subtraction for foreground detection has been commonly applied for varying usages to identify objects in motion within a scene, such as that in video surveillance. In fact, significant publications were noted in the last decade within this area of background modelling. Despite the several surveys noted in the literature, none has offered a comprehensive review in this field. Therefore, this paper elaborates both conventional and recent approaches in light of background modelling. Initially, the approaches listed in the prior works were categorized depending on mathematical models. Next, these models were analyzed based on challenging scenarios that they managed, the challenges and issues are then summarized. After that, an enhanced method is proposed, resulting from hybridizing the weight optimizations from CNN with a set of customized features derived by the Viola-Jones detector to enhance the overall network’s effectiveness. The initial findings show that the proposed method superior the state-of-the-art models with respect to accuracy, recall, precision, and F-measure.