Abstract:Objective: Many computer vision and robotic applications require the implementation of robust and efficient target-tracking algorithms on a moving platform. However, deployment of a real-time system is challenging, even with the computational power of modern hardware. Lightweight and low-powered flying insects, such as dragonflies, track prey or conspecifics within cluttered natural environments, illustrating an efficient biological solution to the target-tracking problem. Approach: We used our recent recordings from 'small target motion detector' neurons in the dragonfly brain to inspire the development of a closed-loop target detection and tracking algorithm. This model exploits facilitation, a slow build-up of response to targets which move along long, continuous trajectories, as seen in our electrophysiological data. To test performance in real-world conditions, we implemented this model on a robotic platform that uses active pursuit strategies based on insect behaviour. Main results: Our robot performs robustly in closed-loop pursuit of targets, despite a range of challenging conditions used in our experiments; low contrast targets, heavily cluttered environments and the presence of distracters. We show that the facilitation stage boosts responses to targets moving along continuous trajectories, improving contrast sensitivity and detection of small moving targets against textured backgrounds. Moreover, the temporal properties of facilitation play a useful role in handling vibration of the robotic platform. We also show that the adoption of feed-forward models which predict the sensory consequences of self-movement can significantly improve target detection during saccadic movements. Significance: Our results provide insight into the neuronal mechanisms that underlie biological target detection and selection (from a moving platform), as well as highlight the effectiveness of our bio-inspired algorithm in an artificial visual system.

Moving object detection based on bioinspired background subtraction

Insect Vision Inspired Object Detection in the Video with a Moving Cluttered Background.

Modeling bio-inspired visual neural for detecting visual features of small- and wide-field moving targets synchronously from complex dynamic environments

Looming Detection in Complex Dynamic Visual Scenes by Interneuronal Coordination of Motion and Feature Pathways

Drosophila-Inspired 3D Moving Object Detection Based on Point Clouds

Modeling Drosophila vision neural pathways to detect weak moving targets from cluttered backgrounds

Vision-Based Moving Objects Detection with Background Modeling

An autonomous robot inspired by insect neurophysiology pursues moving features in natural environments

Panoramic Motion Perception Inspired Fly Visual Brain Joint Neural Network on Omnidirectional Collision Detection

Modelling Drosophila motion vision pathways for decoding the direction of translating objects against cluttered moving backgrounds

A fly inspired solution to looming detection for collision avoidance

Shape Prior Based Moving Object Detection

Flying Bird Object Detection Algorithm in Surveillance Video Based on Motion Information

Attention and Prediction Guided Motion Detection for Low-Contrast Small Moving Targets

Moving Object Detection in Aerial Video.

Fly visual system inspired artificial neural network for collision detection.

A Robust Visual System for Small Target Motion Detection Against Cluttered Moving Backgrounds

A bioinspired retinal neural network for accurately extracting small-target motion information in cluttered backgrounds

A Fuzzy Background Model For Moving Object Detection

Moving Object Detection with a Freely Moving Camera Via Background Motion Subtraction.

FlyTracker: Motion Tracking and Obstacle Detection for Drones Using Event Cameras.