Image Detection Using OpenCV in Low-Visibility Environments

Abstract

Object detection is crucial to autonomous vehicles’ ability to navigate environments and avoid collisions. In developing the “Snowbot” autonomous snow plow robot we will employ the tools from Python’s OpenCV package to potentially clean images from the live camera feed so that they can correctly identify objects to avoid. Snowbot is designed to operate in a low visibility environment where snow, clouds, and other vision impairments will obscure objects, people, and animals. A successful snowbot should be able to use and adapt OpenCV object detection models to detect common obstacles, and interface with other programs to determine an updated path and how to avoid collisions. Here we will lay out a workflow for image cleaning, object detection, and object tracking for Snowbot’s navigation system. Reliably identifying and tracking objects is critical to the safety and functionality of Snowbot.