Mimicking fly motion tracking and fixation behaviors with a hybrid visual neural network

How do animals like insects perceive meaningful visual motion cues involving directional and locational information of moving objects in visual clutter accurately and efficiently? In this paper, with respect to latest biological research progress made in underlying motion detection circuitry in the fly's preliminary visual system, we conduct a novel hybrid visual neural network, combining the functionality of two bio-plausible, namely the motion and the position pathways, for mimicking motion tracking and fixation behaviors. This modeling study extends a former direction selective neurons model to the higher level of behavior. The motivated algorithms can be used to guide a system that extracts location information of moving objects in a scene regardless of background clutter, using entirely low-level visual processing. We tested it against translational movements in synthetic and real-world scenes. The results demonstrated the following contributions: (1) The proposed computational structure fulfills the characteristics of a putative signal tuning map of the fly's physiology. (2) It also satisfies a biological implication that visual fixation behaviors could be simply tuned via the position pathway; nevertheless, the motion-detecting pathway improves the tracking precision. (3) Contrary to segmentation and registration based computer vision techniques, its computational simplicity benefits the building of neuromorphic visual sensor for robots.

[1]  G. Leitinger,et al.  Two identified looming detectors in the locust: ubiquitous lateral connections among their inputs contribute to selective responses to looming objects , 2016, Scientific Reports.

[2]  N. Strausfeld,et al.  Dissection of the Peripheral Motion Channel in the Visual System of Drosophila melanogaster , 2007, Neuron.

[3]  Damon A. Clark,et al.  Defining the Computational Structure of the Motion Detector in Drosophila , 2011, Neuron.

[4]  Claude Desplan,et al.  The First Steps in Drosophila Motion Detection , 2007, Neuron.

[5]  Qinbing Fu,et al.  Modeling direction selective visual neural network with ON and OFF pathways for extracting motion cues from cluttered background , 2017, 2017 International Joint Conference on Neural Networks (IJCNN).

[6]  Alexander Borst,et al.  Object tracking in motion-blind flies , 2013, Nature Neuroscience.

[7]  Paul F. M. J. Verschure,et al.  Non-Linear Neuronal Responses as an Emergent Property of Afferent Networks: A Case Study of the Locust Lobula Giant Movement Detector , 2010, PLoS Comput. Biol..

[8]  A. Borst,et al.  Common circuit design in fly and mammalian motion vision , 2015, Nature Neuroscience.

[9]  Thomas R. Clandinin,et al.  A Class of Visual Neurons with Wide-Field Properties Is Required for Local Motion Detection , 2015, Current Biology.

[10]  Qinbing Fu,et al.  Collision selective LGMDs neuron models research benefits from a vision-based autonomous micro robot , 2017, 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[11]  Shigang Yue,et al.  Visual motion pattern extraction and fusion for collision detection in complex dynamic scenes , 2006, Comput. Vis. Image Underst..

[12]  Shigang Yue,et al.  A Collision Detection System for a Mobile Robot Inspired by the Locust Visual System , 2005, Proceedings of the 2005 IEEE International Conference on Robotics and Automation.

[13]  Alexander Borst,et al.  ON and OFF pathways in Drosophila motion vision , 2010, Nature.

[14]  Qinbing Fu,et al.  Bio-inspired Collision Detector with Enhanced Selectivity for Ground Robotic Vision System , 2016, BMVC.

[15]  Patrick A. Shoemaker,et al.  A Model for the Detection of Moving Targets in Visual Clutter Inspired by Insect Physiology , 2008, PloS one.

[16]  A. Borst,et al.  Seeing Things in Motion: Models, Circuits, and Mechanisms , 2011, Neuron.

[17]  G. Rubin,et al.  A directional tuning map of Drosophila elementary motion detectors , 2013, Nature.

[18]  Stefan Wernitznig,et al.  Synaptic connections of first‐stage visual neurons in the locust Schistocerca gregaria extend evolution of tetrad synapses back 200 million years , 2015, The Journal of comparative neurology.