Scholarlink

<p>Video surveillance systems are becoming increasingly important in various applications, including security, traffic monitoring, and human behaviour analysis. However, analysing the vast amounts of video data generated by these systems manually can be a challenging and time-consuming task. This paper proposes a comprehensive methodology for moving object segmentation &amp; classification in video surveillance using Mask R-CNN and a hybrid Deep Learning (DL) algorithm with Owl-Osprey Optimization. The methodology involves collecting raw video data from surveillance cameras, pre-processing the data through various filtering and enhancement techniques, segmenting the moving objects using the optimized Mask R-CNN model, tracking the objects over time, extracting shape, texture, and appearance features of the objects, and classifying them using a hybrid DL-based algorithm. The proposed Owl-Osprey optimization algorithm (OO-OA) is introduced to fine-tune the pre-trained RNN model and improve the accuracy of the classification algorithm. The performance of the segmentation and classification algorithms is evaluated using standard metrics such as precision, recall, and F1 score. The DCSASS dataset is used for this purpose. Experimental results show that he proposed methodology achieves state-of-the-art performance in moving object segmentation and classification in video surveillance. The hybrid DL-based algorithm achieves high accuracy in object classification, and the use of OO-OA further improves the accuracy of the classification algorithm.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Moving Objects Segmentation and Classification in Video Surveillance Using Mask R-CNN and Hybrid Deep Learning Algorithm with Owl-Osprey Optimization

  • 期刊论文

  • 发表日期：

    2025年6月14日

查看全文
  • Dipika Gupta,
  • Manish Kumar

摘要

Video surveillance systems are becoming increasingly important in various applications, including security, traffic monitoring, and human behaviour analysis. However, analysing the vast amounts of video data generated by these systems manually can be a challenging and time-consuming task. This paper proposes a comprehensive methodology for moving object segmentation & classification in video surveillance using Mask R-CNN and a hybrid Deep Learning (DL) algorithm with Owl-Osprey Optimization. The methodology involves collecting raw video data from surveillance cameras, pre-processing the data through various filtering and enhancement techniques, segmenting the moving objects using the optimized Mask R-CNN model, tracking the objects over time, extracting shape, texture, and appearance features of the objects, and classifying them using a hybrid DL-based algorithm. The proposed Owl-Osprey optimization algorithm (OO-OA) is introduced to fine-tune the pre-trained RNN model and improve the accuracy of the classification algorithm. The performance of the segmentation and classification algorithms is evaluated using standard metrics such as precision, recall, and F1 score. The DCSASS dataset is used for this purpose. Experimental results show that he proposed methodology achieves state-of-the-art performance in moving object segmentation and classification in video surveillance. The hybrid DL-based algorithm achieves high accuracy in object classification, and the use of OO-OA further improves the accuracy of the classification algorithm.