Researchers at GW have developed a novel and efficient solution to detect symmetry lines and points within multidimensional spatial data that could be utilized in data processing and robotics. The solution can take the form of one or more artificial neural networks that can be configured to efficiently detect symmetry lines and points within multidimensional spatial data. The solution includes aspects such as (i) amplitude of a tensor space of the distribution of distance; (ii) a specific configuration of a spiking neural network capable of acting on its inputs in a manner identical to a threshold applied to the tensor symmetry space, firing at the points of high geometric symmetry; (iii) a model of a network that could be utilized in both software and in a Field Programmable Gate Array (FPGA). The symmetry-associating behavior of spiking neural networks has immediate applications in image processing.
The disclosed invention can include the following aspects: a simple algorithm for finding a scalar field representing the symmetry of points in a multi-dimensional space; synchronization in input values of spiking neural networks, with the appropriate choice of threshold and spike period, results in the identification of output neurons along points of high spatial symmetry to the network inputs; an implementation of the symmetry selective LIF neural network in common hardware with a high speed, identification of symmetry points in an Manhattan metric space.
Fig. 1 – Aspects of the disclosed Invention
Applications:
Advantages: