A performance comparison of automatic detection schemes in wide-area aerial imagery

Abstract

Accurate and efficient detection of vehicles in wide-area aerial imagery is a fundamental task in understanding the automobile traffic patterns in an urban environment so as to help regulate the traffic flow. Vehicles with varying shapes and sizes, background clutter, occlusion, low-resolution and noise in the acquired images make the automatic detection of vehicles a challenging task. We present the performance analysis of six object detection algorithms for moving vehicle detection in low-resolution aerial image sequences. We compare the automatic detection results with manual detection, and evaluate the performance of the six object detection algorithms via several metrics.