Scholarlink

With the proliferation of surveillance systems and the increasing need for public safety, the detection of anomalies in crowded environments has become a critical aspect of video analytics. This research proposes an innovative approach to crowd anomaly detection by integrating Convolutional Neural Networks (CNNs), Radial Basis Function (RBF), and Dynamic Coordinate Search (DCS) called HORD-CNN. The combination of these techniques aims to enhance the accuracy and efficiency of anomaly detection in crowded scenes. The CNN are employed to automatically learn hierarchical features from spatial information within the crowd, enabling the system to capture complex patterns and contextual relationships. RBF is utilized to model the non-linear relationships within the data, allowing for the extraction of subtle anomalies that may be challenging for traditional methods. The proposed methodology is evaluated on a benchmark UMN dataset, demonstrating superior performance compared to existing anomaly detection techniques. The results of the proposed framework revealed an accuracy of 98.6%, which shows a better performance when compared with existing frameworks that used the same dataset.

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

Crowd Anomaly Detection Using Convolution Neural Networks Enhanced with Hyper-Parameter Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Joseph Bamidele Awotunde,
  • Agbotiname Lucky Imoize,
  • Yousef Farhaoui,
  • Abidemi Emmanuel Adeniyi,
  • Shuaib Gbolahan Olatinwo

摘要

With the proliferation of surveillance systems and the increasing need for public safety, the detection of anomalies in crowded environments has become a critical aspect of video analytics. This research proposes an innovative approach to crowd anomaly detection by integrating Convolutional Neural Networks (CNNs), Radial Basis Function (RBF), and Dynamic Coordinate Search (DCS) called HORD-CNN. The combination of these techniques aims to enhance the accuracy and efficiency of anomaly detection in crowded scenes. The CNN are employed to automatically learn hierarchical features from spatial information within the crowd, enabling the system to capture complex patterns and contextual relationships. RBF is utilized to model the non-linear relationships within the data, allowing for the extraction of subtle anomalies that may be challenging for traditional methods. The proposed methodology is evaluated on a benchmark UMN dataset, demonstrating superior performance compared to existing anomaly detection techniques. The results of the proposed framework revealed an accuracy of 98.6%, which shows a better performance when compared with existing frameworks that used the same dataset.