Vehicle ground-truth database for the vertical-view Ft. Hood imagery

Abstract

This paper reports the work on building the ground-truth databases for the vehicles in the vertical-view Ft. Hood (VVFH) image data set. We briefly describe the protocols followed in manual annotation of the images, how the ground-truth information is inferred from the annotated image data, and the major entities in the ground-truth database. The vehicle detection performance of a few algorithms is evaluated using the data set. The entire data set of images and ground-truth is available on the Internet.

Vehicle detection from high-resolution satellite imagery using morphological shared-weight neural networks

“…The detection rate is defined as the total number of targets correctly detected divided by the total number of targets. For the false alarm rate, we follow the definition in [7] where the false alarm rate is defined by the average number of false alarms per million-pixels in an image. Detection rates of 86% (217 out of 252) and 89% (65 out of 73) were obtained for the training and validation sites, respectively.…”

“…Vehicle detection has applications in military and homeland surveillance, transportation planning and management, and intelligent traffic guidance systems. Most vehicle detection research has been done using aerial imagery with a spatial resolution of 0.35 m or less [3][4][5][6][7][8][9][10]. Vehicle detection research has seldom been done using high-resolution satellite imagery where spatial resolutions of the panchromatic band are presently in the range of 0.6-1.0 m [11,12].…”

“…Vehicle hypothesis was generated by comparing the rectangles centered at the detected centers with the local edge map. Liu and Haralick [7] presented a technique that utilized ground-truth databases for vehicles in a vertical-view image dataset with spatial resolution of 0.3 m. The vehicles were modeled as rectangles and the centers were detected by edge detection and GHT. The performance was improved by a Bayesian classifier algorithm using eight input features.…”