Radial Basis Function Based Neural Network for Motion Detection in Dynamic Scenes.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Motion detection is crucial for video surveillance, but challenging in dynamic scenes with elements like swaying trees or rippling water.
• Existing methods struggle with complete and accurate motion detection in complex, dynamic environments.

Purpose of the Study:

• To develop a novel motion detection approach using radial basis function artificial neural networks.
• To accurately detect moving objects in both dynamic and static scenes.

Main Methods:

• A multibackground generation module creates a flexible probabilistic model for static or dynamic backgrounds via unsupervised learning.
• A moving object detection module uses a block alarm and object extraction procedure to process only relevant blocks.

Main Results:

• The proposed method demonstrated superior performance compared to state-of-the-art techniques.
• Achieved Similarity accuracy of 69.37% and F1 accuracy of 65.50% on diverse natural video sequences.

Conclusions:

• The novel radial basis function artificial neural network approach offers robust and accurate motion detection.
• This method significantly enhances motion detection capabilities, especially in challenging dynamic scenes.