A MULTI-ANGLE PEDESTRIAN IMAGE PAIRING METHOD BASED ON PERSPECTIVE TRANSFORMATION

Pedestrian re-identification is to match a person’s identity across different cameras or locations in a video or image sequence. It involves detecting pedestrians from a frame sequence and using features such as human faces and clothes to identify and match. However, the features captured by only one camera are not robust enough with rotation problems. So, two cameras capture multi-angle images as they provide more information than single-angle images. This paper proposes a foreground segmentation method to divide the foreground into pedestrians. Then, a perspective transformation-based pairing method is proposed to match the same pedestrian’s images captured by two cameras at the same place. The results demonstrate that the foreground segmentation method deals with only more than two classes that can be classified as problems. The pairing method can accurately pair the same person’s images from different-angle cameras.