Track: Automation and Control

Abstract

In response to the growing need for autonomous systems that can interpret complex visual information akin to human cognition, this study introduces a probabilistic machine learning approach for visual object detection, demonstrating significant strides in identifying images through a computational model that emulates human cognitive precision. The method begins with capturing images via video cameras, followed by an algorithmic learning phase that employs Blob Analysis to discern diverse object features such as color, shape, and region. These attributes are cataloged in a database, facilitating the matching and recognition of objects against a learned dataset. This technique refines the focus-of-attention mechanism, ensuring high probability that non-target regions do not contain the object of interest. The proposed system shows exceptional promise in optical character recognition and content-based image indexing, with experimental results underscoring its efficiency and accuracy in object detection. This advancement could significantly impact robotic vision systems, enhancing their interpretative capabilities in varied and complex environments.