| Summary: | 碩士 === 國立嘉義大學 === 資訊工程學系研究所 === 106 === As the development of monitoring systems towards automation and intelligence, the visual-based monitoring system has also become an important research topic. In an environment where no human intervention is required, automatic analysis is performed by software to detect the position and number of pedestrians in the image. Achieve monitoring, tracking, and identification.
The purpose of this study is to use a single fixed camera for pedestrians identifying and tracking, which are mainly divided into two parts, feature extraction and classifiers.In the feature extraction part, Histogram of Oriented Gradient (HOG) is used to capture pedestrian features, and then Haar-like moment features are used to capture the pedestrian's movement direction. In the classifier part, the pedestrian character cues will train Support Vector Machine (SVM) over for attaining a pedestrian detector, which is used to find out if there is a pedestrian in the image.
In order to verify our proposed approach, we completed pedestrian detection experiments by capturing pedestrian samples from the Internet, and INRIA pedestrian database. Experimental results show that our method can accurately achieve the pedestrian detection and its face orientation in time.
|
|---|
