Humanoid Detection Based On Support Vector Machines

The pedestrian detection is very important in the pedestrian analysis research direction which pays attention, is foundation which the pedestrian distinguishes. The pedestrian detection in intelligence supervisory system and based on content aspect and so on picture index all has the wide spread application, thus fast studies moreover the accurate human pedestrian detection technology to have the extremely vital significance.This article studies the pedestrian detection question about the gradation picture or the colored picture on the condition of the complex background. From the method based-on sample study a method is presented that is the fast wavelet transformation (DWT) and the pedestrian detection algorithm. This algorithm is composed of the training and the pedestrian detection parts. The training process is collecting the massive "the pedestrian" and "the non pedestrian" samples, then training the sample picture to carry on the wavelet transformation, after the choice transformation the picture low frequency child brings, the high frequency child will bring, a high frequency child will bring the wavelet coefficient will take the characteristic, will train the compound sorter which will be composed by the linear SVM sorter and the non-linear SVM sorter. On the examination stage the compound sorter is used so that it can be determined whether there is the human form in the picture.There is essential difference the method in this algorithm and the based on the movement background technology. Later because must use the movement information, therefore uses in the photograph which monitors confidentially fixed to be motionless, finally causes the application scope reducing. To obtains the picture to carry on the fast wavelet transformation, then uses based on the SVM method the human form which possibly exists is carrying on the examination like the region.