| 摘要 |
<p>The present invention relates to an object extracting and tracking apparatus using a coarse-fine technique, which separates the foreground by a dynamic background model in an image extracted through a surveillance camera, learns about an object such as a person, a vehicle, etc., and recognizes the object such as the person, the vehicle, etc. in the separated foreground in order to enhance accuracy of the object extraction and tracking, and a method thereof. According to the present invention, the apparatus comprises: an input image processing unit (100) for performing calibration for an input image of a fixed camera, and creating a lookup table; and an online processing unit (200) including a coarse object detecting unit (210) using a dynamic background modeling algorithm for an image frame of the input image, a fine object detecting unit (220) for learning and classifying the object to be extracted and tracked with a histogram intersection kernel (HIK) support vector machine (SVM) technique, a camera control unit (230) for controlling movement of the fixed camera according to the object based on the lookup table created in the input image processing unit (100) and information on the learned and classified object in the fine object detecting unit (220), and an object tracking unit (240) for receiving data for the object detected from the camera control unit (230) and tracking the corresponding object along the movement route.</p> |
