Indoor moving vehicle detection by trajectory matching

【作者】 陈毅鸿； Felix Lütteke； 陈启军； Jrg Franke；

【Author】 CHEN Yi-hong Felix Ltteke,CHEN Qi-jun Jrg Franke (1.School of Electronics and Information,Tongji University,Shanghai 201804,China; 2.The Institute of Manufacturing Automation and Production Systems, Friedrich-Alexander-Universitt Erlangen-Nrnberg,Erlangen 91058,Germany)

【机构】 School of Electronics and Information,Tongji University； The Institute of Manufacturing Automation and Production Systems,Friedrich-Alexander-Universitt Erlangen-Nürnberg,Erlangen 91058,Germany； The Institute of Manufacturing Automation and Production Systems,Friedrich-Alexander-Universitat Erlangen-Nürnberg,Erlangen 91058,Germany；

【摘要】 <正>A solution to determine the positions of transportation vehicles without any optical markers in an area under camera surveillance was introduced.This solution was based on a net of cameras on the ceiling of the vehicle operational areas.A robust background subtraction method was performed to detect the contours of all moving objects in every camera frame.These so called foreground objects contours were fed into a particle filter to estimate the position,orientation and speed of the detected objects over a certain number of frames.Based on the camera data this so-called state estimation was performed for all foreground objects like humans,other moving obstacles and the vehicles.Due to the fact that the vehicles can be totally covered by their payload,they cannot be detected by color or shape.Instead,the vehicles were identified by comparing their movement with the movement of all detected foreground objects.更多还原

【Abstract】 A solution to determine the positions of transportation vehicles without any optical markers in an area under camera surveillance was introduced.This solution was based on a net of cameras on the ceiling of the vehicle operational areas.A robust background subtraction method was performed to detect the contours of all moving objects in every camera frame.These so called foreground objects contours were fed into a particle filter to estimate the position,orientation and speed of the detected objects over a certain number of frames.Based on the camera data this so-called state estimation was performed for all foreground objects like humans,other moving obstacles and the vehicles.Due to the fact that the vehicles can be totally covered by their payload,they cannot be detected by color or shape.Instead,the vehicles were identified by comparing their movement with the movement of all detected foreground objects.更多还原