Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Pedestrian Detection Design Based on Android System

Author LiZuo
Tutor SuYuTing
School Tianjin University
Course Information and Communication Engineering
Keywords Android pedestrian counting Kalman filtering Java NativeInterface
CLC TP391.41
Type Master's thesis
Year 2012
Downloads 3
Quotes 0
Download Dissertation

With the further development of Intelligent Video Surveillance (IVS) technology,the application of IVS ranges from providing security service to giving guidance formanagement and decision making process. The pedestrian detection system designedin this paper mainly comprises three parts: foreground detection, target tracking andpedestrian counting.In the foreground detection part, foreground image is extracted by usingbackground subtraction based on single Gaussian model which is able to extractforeground image accurately and fast in data processing. The following step is postprocessing of the extracted foreground image. Open operation and closed operationare first conducted, and then the erode operation and dilate operation. The foregroundimage after post processing has smooth edge and a small number of voids, which lay agood foundation for subsequent process. In the target tracking part, this paper adoptsthe target tracing algorithm based on the Kalman filtering model to find out the searchrange of the moving target in the next frame, and then calculate the cost function ofevery target within the search range, and choose the target of the minimal costfunction value as the moving target’s successor. In the pedestrian counting part, thispaper use pedestrian statistical algorithm based on the foreground area as pedestriannumber are linear with the number of foreground pixels. Because the area ofpedestrian’ projection in the camera changes constantly as pedestrian moves close toor walks away from the camera, geometric correction is conducted to the foregroundin order to obtain accurate foreground area for the estimation of the number ofpedestrian.This paper realizes pedestrian detection system on Android. Because of theportability of Android smart phones and tablet PCs, transplanting pedestrian detectionsystem which is developed with OpenCV to Android platform makes the IntelligentVideo Surveillance become more flexible and more mobile, and expands theapplication field of Android platform as well.

Related Dissertations
More Dissertations