Threat Posture Detection Based on Weighted HOG of Human-body Infrared Image

摘要

Based on analysis of infrared human image feature,this paper present a betterment method of extracting Histogram of Oriented Gradients (HOG)from infrared human image.Based on weighted hog feature of image.finally using support vector machine (SVM) classify the threatening human from these infrared human image.If making use of the ROIS(Regions of Interest,ROIs).A set of theories of detecting threat from infrared image background were put forward.This system can judge whether there are individual who will attack the object which we defend,and timely alarm.It is useful for security-guard.