Computational model of the LGMD neuron for automatic collision detection

Abstract

In many animal species it is essential to recognize approach predators from complex, dynamic visual scenes and timely initiate escape behavior. Such sophisticated behaviours are often achieved with low neuronal complexity, such as in locusts, suggesting that emulating these biological models in artificial systems would enable the generation of similar complex behaviours with low computational overhead. On the other hand, artificial collision detection is a complex task that requires both real time data acquisition and important features extraction from a captured image. In order to accomplish this task, the algorithms used need to be fast to process the captured data and then perform real time decisions. Taking into account the previous considerations, neurorobotic models may provide a foundation for the development of more effective and autonomous devices/robots, based on an improved understanding of the biological basis of adaptive behavior. In this paper, we make a comparative analysis between the new computational model of a locust looming-detecting pathway and the model previously proposed by us. The obtained results proved the improvement provided by the pixel remapping in the model performance.