Pedestrian detection and tracking are very popular in image processing and computer vision areas. The main purpose of this paper is to compare different pedestrian detection approaches and to build a pedestrian tracking system. This paper trains two different detectors one is based on Histogram of Oriented Gradient feature combined with Support Vector Machine and another one is based on a neural network that imitates the Faster R-CNN then compares their detection results. Both detectors are trained on a dataset that combines part of the INRIA dataset and images taken on the outside of campus. It includes people with different postures, walking in any directions and brightness. And this paper also completes a pedestrian re-identification function using GrabCut segmentation technique in HSV color space to recognize same people. At last, this paper achieves tracking a specific target in a video by using Kernelized Correlation Filters (KCF).