Hough Transform is a well-known method in the field of computer vision for detecting simple shapes in a photo. This technique is mostly applied to conventional cameras. Recently a new type of imaging devices called Dynamic Vision Sensors (DVSs) have been introduced which only report pixels with intensity change rather than all pixels’ intensity values. The current study is an improvement to a research done previously by
the authors in which Hough Transform was employed in Spiking Neural Network (SNN) for multiple lines detection and tracking. In that study, the events received from DVS were transformed from Cartesian space to the parameter space which was implemented in a spiking neural network. In that study, position of the firing neurons shows the detected lines’ properties. Moreover, lateral inhibitory connections in a rectangular window in the parameter space were used for suppressing neighboring lines. Finally, an event-based clustering algorithm was applied subsequently in the parameter space for tracking detected lines in the video. As an improvement to the work, the current paper deals with detecting small lines at the frame corners which was not considered in the previous study. In addition, the inhibitory window shape is optimized to suppress the lines which are close together in Cartesian space and are not necessarily close together
in parameter space assumed in the previous study. The effectiveness of these improvements is tested and verified by experimental results.
License type:
PublisherCopyrights
Funding Info:
Agency for Science, Technology and Research (A*STAR), Singapore.