Vehicle detection on a pint-sized computer

摘要

Powerful miniature single-board computers have recently gained attention, inviting computer scientists and engineers to develop various kinds of applications on these tiny devices. Having numerous benefits over full-scaled personal computers, their small size enables system mobility and allows operations under limited power resources. Exploring its computing capability, we set up a Raspberry Pi with a high-resolution camera. Its task is to detect vehicles based on image processing techniques. This is generally regarded as a computationally demanding process. Our system implements Haar-like features and a supervised cascade learning model. Empirical results are impressive, achieving good detection rate with an average sustained image rate of 2 Hz.