Proposing a new feature set for recognition of moving humans in video frames

Human detection and recognition at a distance is recently a matter of great concern among computer vision researchers. This paper introduces a new set of human body features for the recognition of detected human as an object. The feature extraction is performed by an established human model consisting of five parts. These features consist of geometric calculations of detected object and their different ratios. Detection of moving object is achieved using average background model with supportive secondary model and an adaptive threshold selection model based on Gaussian distribution. Recognition of detected object has been carried out using Support vector machine (SVM). Experimental results show better recognition rates as were achieved previously using a relatively smaller feature set.