Research On Video-based Pedestrian Detection Method

Pedestrian detection is an important aspect in the field of computer vision.Researchers have proposed various algorithms for pedestrian detection based on feature extraction and motion information.In the pedestrian detection process,the effect of pedestrian detection is easily disturbed by the external environment(light,shade,background change,etc.).In the light of changing conditions,the phase of the image can identify and locate the features of the image on the gradient-based technique.Therefore,this paper firstly studies the method of pedestrian detection based on phase information.The histogram of Oriented Phase was defined by calculating the histogram of the phase consistency value of the local region in the image and combining the all histograms and the INRIA data set was used to evaluate the performance of HOP descriptors?HOP features can overcome changes in lighting,but it still have limitations for pedestrian detection.In order to be able to get more structural information,and better adapt to changes in light and messy environment,the paper propose method based on feature fusion.The method combines Histogram of Oriented Gradient features(HOG)with HOP to identify and locate more structural information and the INRIA data set was used to assess the performance of the fusion features.In the process of pedestrian detection the paper proposed a method of combining pedestrian characteristics and movement information to detect pedestrians,and a Support Vector Machines(SVM)classifier was used to train pedestrian classifiers for human detection.The paper used the ETHZ and Caltech test sets to evaluate the performance.The results show less error rate and better detection performance than traditional methods in pedestrian detection.