Pedestrian Distance Measurement Based on Stereo Vision

  • 葉 保宏

Student thesis: Master's Thesis


With the development of technology computer vision is now widely applied in daily life in applications such as face recognition and obstacle avoidance In this thesis we use the principles of stereo vision to measure distance Firstly we calibrate two cameras to obtain the intrinsic and extrinsic parameters and then we use the two cameras to capture images of the same scene However the two cameras are not necessarily on the same horizontal level and thus we would use intrinsic and extrinsic parameters to rectify the images so that the corresponding points of two images are on the same horizontal level After that we use the Oriented FAST and Rotated BRIEF (ORB) feature algorithm to obtain the corresponding points and calculate disparity and distance This thesis proposes two distance measurement systems One manually select objects from two images The other system combines the concept of a Histogram of Oriented Gradient (HOG) and Support Vector Machine (SVM) to train a pedestrian classifier We use the classifier to detect pedestrians in two images automatically and then we use the ORB features to calculate the disparity and distance in the pedestrian region In addition we revise the original distance formula to compare with the existing distance formula The results of this research can be used to avoid collisions with pedestrians
Date of Award2015 Aug 5
Original languageEnglish
SupervisorTa-Chung Wang (Supervisor)

Cite this