The Research Of Pedestrian Detection Method Based On Multi-Features Fusion In Static Image

Pedestrian detection has been one of the hottest research areas in the field of visual monitoring, vehicle intelligent auxiliary driving and human behavior analysis. At the same time, pedestrian detection is also widely used in the emerging areas such as disaster rescue and debris flow warning. Researchers have proposed a variety of human detection methods applied in different environments in recent years by combining classic descriptor including Histograms of Oriented Gradients(HOG), Pyramid Histogram of Oriented Gradients(PHOG)and multi-feature fusion descriptor. These methods are mainly based on machine learning which using extracted image features for training and testing. Such methods are mainly composed of two parts: feature extraction and design of classifier. The purpose of the feature extraction is to extract features that can better describe the characteristics of human information. Faster and more accurate classifier can be constructed by design of classifier.Firstly, we introduce and analysis the research situation of pedestrian detection technology and the research difficulties. Secondly, we summarize some of the main way of feature extraction and classifier. Finally, we introduce the improvement way of feature extraction. The main contributions of the paper are as follows:According to the problem of the inaccurate and unstable pedestrian detection, we present a feature fusion method based on Local Difference Binary Pattern(LDBP) and Local Binary Pattern(LBP). Firstly, we filter the input image by using the 2D discrete haar wavelet transformation, in order to obtain four sub-images with different frequency, including LL, LH,HL and HH. Secondly, we extract the LDBP feature from the low frequency part and the LBP feature from the other three high frequency parts. Then, we reduce the dimension of the LDBP and LBP feature spaces by using Principal Component Analysis(PCA). Finally, we apply the fused LDBP-LBP feature for performing an efficient pedestrian detection.Conducted on INRIA databases by using Support Vector Machine(SVM), the experimental results demonstrate our method can effectively improve the detection accuracy and get a better robustness.According to the real-time deficiency in pedestrian detection, this paper presents a feature fusion method based on Improved Center-Symmetric Local Neighbor Binary Pattern(ICS-LNBP) and Local Binary Pattern(LBP). Firstly, we translate the input image into the grayscale image. Secondly, we extract the ICS-LNBP feature and LBP feature from the grayscale image. Finally, we apply the fused ICS-LNBP-LBP feature for performing an efficient pedestrian detection. The experimental results demonstrate our method can get better detection results in complex background, and can meet the real-time requirements.