Human detection algorithm based on edge symmetry

Abstract

One of the most important abilities that personal robots need when interacting with humans is the ability to detecte human timely. Due to the scan of the input image without any disparity in the traditional method for human detection, processing speed can not meet the demand of a real-time system appropriately. Under such a circumstance, an edge symmetry based human detection algorithm is proposed. With mechanism of scan lines, the symmetrical value of each pixel is calculated along scan line and candidate regions are picked out. Then candidate regions are verified by using Histograms of Oriented Gradients (HOG) feature and Support Vector Machine(SVM) classifier. Experiment shows that the algorithm has a good command of keeping the precise of the recognition as well as elevating the speed of calculation.