Teaching large-scale image processing over worldwide network cameras

摘要

This paper presents a software system for large-scale image processing. Through this system, students may choose to analyze the images from several thousand network cameras deployed worldwide. This system allows both real-time analysis of live data or storing the data for off-line analysis. This system currently supports image processing using OpenCV-Python. The system allocates cloud instances as the computational engine and, as a result, allows users to analyze the images from many cameras simultaneously. The system demonstrates the ability to process 5,000 images from 500 cameras for lane detection in less than 2 minutes.