Human detection based on deformable template and SVM
基于可变模板和支持向量机的人体检测

With the development of image processing and pattern recognition technology, human detection system has been widely used in many domains, such as monitor and control system, driver assistance system and image index system. Aiming at the problem of standing human detection within static image, a new method to extract human feature was proposed, and human was detected and positioned within static images by using a deformable template combined with Support Vector Machine (SVM) method. After the image contour was extracted and processed, the image was decomposed into grids, and features of horizontal and vertical grids were grouped into the image feature vector. Then the support vector machine model was trained with feature vectors extracted from example images. Detecting regions in the image were identified with a deformable template. With feature vector of detecting regions as input, the trained SVM model classified the test image and positioned the human in the test image. Experiment results show this method can effectively detect standing humans from static images with various backgrounds, and the correct classifying rate is over ninety-two percent.