Classroom Occupancy Statistics System Based On Universities’Surveillance Videos

At present, the shortage of studying rooms in every college is a common phenomenon.As the classroom is the main place where students study in, it will take a long time for them tofind the classroom which is not taken and has little students. Therefore, the development ofclassroom occupancy statistics system to help students find a classroom is very valuable andmeaningful. At the same time, all colleges’ classrooms are installed with video surveillancesystems which provide the hardware support for the development. And the growth of thecomputer vision technology also provides a theoretical support. For above reasons, this thesispresents a colleges’ classroom occupancy statistics system based on video surveillance. Ithelps students to find the studying rooms as convenient and improve the utilization rate of theschool teaching resources at the same time.This thesis presents a method of classroom occupancy statistics method which is based onhead detection. First of all, pretreatment the collected video images of classroom, whichinclude removal of interference region, the conversion of HSV color space, binary image,corrosion and expansion. It gets the head region according to hairs information. Then, use themethod of Hough transform detection detecting the heads to obtain the head candidate regionaccording to head shape feature. According to the difference position of the heads size in thepicture, divide the non head region which too large or too small by the method of headdetection based on area. Use the accumulated frame difference method to remove still nonhead region and get the head region. Finally, get the counts of head and the classroomoccupation statistics. And the result of the occupation status is effectively. This paper designthe overall system architecture, functional structure and technical architecture of colleges’classroom occupancy statistics system based on video surveillance and realize the classroomoccupancy statistics module.This thesis mainly analysis three types of classroom’s monitoring video. Theexperimental results show that: in the case of uniform illumination and occlusion are notserious, the three classrooms’ positive detection rates are high. And compare the obtainedclassroom occupancy rate with the actual classroom occupancy rate, the result is basicallysame, which can prove that the method using hair color feature, head shape feature and head area feature is available. The method of classroom occupancy statistics can effective calculatethe occupation status of classroom, and it provides the guarantee of colleges classrooms’occupancy statistics system based on monitoring video for the next step.