Self-adaptive SOM-CNN neural system for dynamic object detection in normal and complex scenarios

作者：

Highlights：

• The paper presents a neuro-inspired method, SOM-CNN, to detect dynamic objects.

• SOM-CNN works with normal and complex scenarios.

• SOM-CNN is a self-adaptive method with two novel neural networks: RESOM and NTCNN.

• SOM-CNN performance in complex scenarios is better than other methods in literature.

• The system can process at 35 fps making it feasible for real time applications.

摘要

•The paper presents a neuro-inspired method, SOM-CNN, to detect dynamic objects.•SOM-CNN works with normal and complex scenarios.•SOM-CNN is a self-adaptive method with two novel neural networks: RESOM and NTCNN.•SOM-CNN performance in complex scenarios is better than other methods in literature.•The system can process at 35 fps making it feasible for real time applications.

论文关键词：SOM,Self-Organizing Map,RESOM,Retinotopic SOM,CNN,Cellular Neural Networks,NTCNN,Neighboorhood Threshold Cellular Neural Network,SOM-CNN,Proposed method,Video analysis,Motion detection,Self-organizing maps,Cellular neural networks

论文评审过程：Received 23 December 2013, Revised 4 September 2014, Accepted 10 September 2014, Available online 21 September 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.09.009