Multiple People Detection With Multiple Cameras Based On Foreground Matching

Moving targets detection and tracking is one of the key technologies in computervision and has been widely used in intelligent vision-based surveillance systems,traffic navigation, video searching, medical image analysis and many other fields. Soresearch of moving target detection and tracking methods has great practicalsignificance.This paper mainly aims to detect multiple people using static multiple camerasin an indoor scene. More specifically, we try to detect teacher and standing students ina classroom.We use existing methods to get foreground from each camera, then we match allthe foreground to figure out the position of real people. According to the height wherethe cameras are placed, we use two different methods to accomplish the detectionmission. The methods are as follow:(1)The cameras are placed separately at the different places of the classroom,they are all installed at the height about the height of head when a people sits down.Most area of the classroom can be seen by at least two cameras. To every camera, weestablish background of the region we interest and then get a good foreground. In thispaper we match all the foreground from all cameras to figure out some positionswhere a people maybe be there. We choose several from the positions to make agroup. For each group, we use a method to score it, we choose the group with theminimal score and give the detection results.(2)The cameras are installed at the height near the ceiling, looking down at theclassroom. The people in the classroom can be seen by all cameras. In this situation,we cannot get good background, so we only have approximate foreground. Weemploy a method to match the approximate foreground and figure out a region of theclassroom where the possibly in. Then we find people’s heads by matching the colorof images from the cameras.