A visual neural network for moving object recognition

Noriaki Sato, Masafumi Hagiwara

Research output: Contribution to journalArticlepeer-review


This paper proposes a neural network for moving object recognition based on the visual system. The proposed network has a structure focusing on the parallel hierarchy of visual information processing and can recognize moving objects. The network is composed of a movement detection module, a moving object estimation module, and a moving object recognition module. The movement detection module uses a neural network which is a model of the motion recognition aspect of vision and recognizes motion. The moving object estimation module uses a hierarchical neural network with the backpropagation (BP) algorithm, and infers moving objects based on the features of the motion. The moving object recognition module uses a multiple-structured neural network which is obtained by improvement of the inhibition mechanism and recognizes moving objects on the basis of shape features. The information is transmitted from the movement detection module to the moving object recognition module through the moving object estimation module. With this mechanism, the motion recognition and pattern recognition processes are integrated, and moving objects can be recognized. A computer simulation is performed on the recognition of a walking human, and the effectiveness of the proposed network is demonstrated.

Original languageEnglish
Pages (from-to)46-55
Number of pages10
JournalElectronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)
Issue number9
Publication statusPublished - 2004 Sept


  • Motion recognition
  • Moving object estimation
  • Pattern recognition
  • Visual information processing

ASJC Scopus subject areas

  • Physics and Astronomy(all)
  • Computer Networks and Communications
  • Electrical and Electronic Engineering


Dive into the research topics of 'A visual neural network for moving object recognition'. Together they form a unique fingerprint.

Cite this