Detection of Human Bodies in Lying Position based on Aggregate Channel Features

In recent years, detection of human body has drawn a lot of attention from researchers in the field of image recognition, with most work focused on pedestrian detection. The detection of human bodies in lying position also received numerous attention in applications such as elderly fall detection, sleep studies as well as in search and rescue (SAR) operations. Thus, in this paper, feature extraction performed by the Aggregate Channel Features (ACF) algorithm is explored for detection of human bodies in lying positions. ACF makes use of a Boosted Decision Tree (BDT) classifier that has resulted in increase in speed of detection. The classification was carried out using a dataset developed from aerial images of human bodies obtained from the internet. Initial result showed that the accuracy of ACF using the given dataset is 88% and the value of F-measure obtained was 0.9231. This proposed method will be further explored on a more advanced dataset.