This research investigates the automatic detection of a dismounted human from a single image as a function of range.
The histogram of oriented gradients (HOG) method provides the feature vector and a support vector machine performs
the classification. This work presents, for the first time, an understanding of how HOG for human detection holds up as
range increases. The results indicate that HOG remains effective even at long distances; for example, the average miss
rate and false alarm rate were both kept to 5% for humans only 12 pixels high and 4-5 pixels wide. The impact of the
amount and type of training data needed to achieve this long-range performance is examined.