A Bionic Method of Moving Object Detection with Multi-feature Fusion Based On Frog Vision Characteristics

• Published in: Proceedings of the International Conference on Image Processing, Computer Vision, and Pattern Recognition (IPCV) p. 1
• Main Authors: Tang, Xiaogang, Wang, Sun'an, Di, Hongyu, Liu, Litian
• Format: Conference Proceeding
• Language: English
• Published: Athens The Steering Committee of The World Congress in Computer Science, Computer Engineering and Applied Computing (WorldComp) 01.01.2013

In the complex natural background, the image features of moving objects usually change severely. And the kinematics and morphological features of dynamic target are unconspicuous due to the fast movement, unpredictable kinetic law and the accompanied scale transformation. The methods of motion detection based on one single morphological, statistics or kinetic features would not meet the requirements. Inspired by the visual characteristics of frog eye and the physiology characteristics of dynamic response of the retinal neural circuit, a spatial-temporal moving target recognition method based on the frog's vision is provided. This method introduces a biomimetic recognition algorithm of multi-feature fusion for dynamic target detection based on BP neural network. The experimental results show that the method achieves to inhibit the background information effectively and enhance the multidimensional moving target information through the mechanism of spatial and temporal characteristics and multi-feature fusion, which is better than the method based on single feature. The algorithm principle provides a biomimetic approach for motion detection. [PUBLICATION ABSTRACT]