Robust Pan/Tilt Compensation for Foreground–Background Segmentation

Abstract

In this paper, we describe a robust method for compensating the panning and tilting motion of a camera, applied to foreground–background segmentation. First, the necessary internal camera parameters are determined through feature-point extraction and tracking. From these parameters, two motion models for points in the image plane are established. The first model assumes a fixed tilt angle, whereas the second model allows simultaneous pan and tilt. At runtime, these models are used to compensate for the motion of the camera in the background model. We will show that these methods provide a robust compensation mechanism and improve the foreground masks of an otherwise state-of-the-art unsupervised foreground–background segmentation method. The resulting algorithm is always able to obtain F 1 scores above 80 % on every daytime video in our test set when a minimal number of only eight feature matches are used to determine the background compensation, whereas the standard approaches need significantly more feature matches to produce similar results.

Efficient detection of crossing pedestrians from a moving vehicle with an array of cameras

“…These regions are typically characterized by a larger deviation between the input and the background model or from the detection of “blinky” or isolated foreground pixels. The aforementioned strategies can also be successfully applied to PTZ cameras; the background model can be accurately transformed during camera rotation with a specific form of the homography matrix 12 . However, when the camera also physically moves to another location during the recording time, additional strategies are needed, as is discussed below.…”

“…The aforementioned strategies can also be successfully applied to PTZ cameras; the background model can be accurately transformed during camera rotation with a specific form of the homography matrix. 12 However, when the camera also physically moves to another location during the recording time, additional strategies are needed, as is discussed below.…”

Unsupervised Moving Object Segmentation from Stationary or Moving Camera Based on Multi-frame Homography Constraints

“…When the camera is moving, the pixels corresponding to background do not maintain the same image positions in consecutive frames, which severely make the traditional video object segmentation complicated. In order to segment the object regions, an intuitive idea is to estimate the transformation parameters between consecutive images and creating a difference image [9,10,11]. By now, most of the printed literatures are aware this approach, and this it has become widely adopted as it requires less computational cost and memory storage.…”