Feature extraction and tracking for haptic systems integration with Dynamic Active-pixel Vision Sensor (DAVIS)

Feature detection and tracking objects in computer vision have been widely used in industry applications to control and enhance robots’ performance. However, standard cameras are limited to frames that lead the system to have considerable blind time between frames. Also, frame-based algorithms are computationally expensive. In contrast, Dynamic Visions Sensors (DVS) provides stream of events that include only intensity changes in the pixels to eliminate redundant information. In this research proposal, artificial intelligence neural networks algorithms will be developed based on the hybrid vision sensors to extract and track object features that assist high-speed robots to control adaptively their gripping force in real time. Further, this approach could provide more accurate and faster solutions for industrial control systems with minimum power consumption. The research proposal is interdisciplinary in nature includes artificial intelligence neural networks, image processing, robotics, and control systems to achieve desirable solution that can be embedded in Baxter robot platform.