Motion Detection

Motion detection is to detect moving target objects in an input image sequence. Conventional approaches for motion detection methods use background subtraction, temporal differencing, and optical flow. With the growing usage of active cameras in recent visual surveillance environments, motion detection algorithms using active cameras are becoming an important component. Motion detection using active cameras has been developed by making background mosaics, modified background subtraction methods which compensate for camera motion using optical flow, feature matching, and geometrical camera models to detect moving objects by registering the current frame to the background image.

Figure 1. Result of the motion detection

Object Tracking

Tracking is the problem of generating an inference about the motion of an object given a sequence of images. In a visual surveillance system, the goal of object tracking is to follow a moving object detected in motion detection stage from one frame to another in an image sequence. The performance of the high-level image interpretation module such as behavior understanding depends highly on the object tracking result. For example, different kinds of trucks should move in different, fixed patterns in an airport; if they do not, then something is going wrong. Similarly, there are combinations of places and patterns of motions that should never occur(e.g., no truck should ever stop on an active runway). These examples need high-performance tracking. Difficulties in tracking an object can arise from abrupt object motion, changing the appearance of object and scene, self-occlusion, occlusion by a structure. Thus, these difficulties should be solved to track the target object accurately.

Figure 2. Result of the object tracking system


Surveillance Platform: PIL-EYE 

PIL-eye is an intelligent visual surveillance software package with pan/tilt cameras. With PIL-eye, detecting and tracking the intruders or understanding behaviors of objects are possible in selected area. PIL-eye contains several sub-modules such as image enhancement, motion detection, object tracking, behavior understanding and the module integrating all sub-modules. Algorithms developed and implemented exclusively in our lab can be selected in each sub-modules, and each sub-modules itself also can be turned on/off.