Foreground Object Detection Based on Multi-model Background Maintenance

Abstract

This paper addresses the problem of background maintenance for foreground object detection. A Multimodel Background Maintenance (MBM) framework that contains two principal features is proposed. Under this framework, a pure time-varying background image is maintained and learned using the statistical information of the multi-model Gaussian distribution with principle features. The principal features consist of static and dynamic pixels to represent the characteristic of background. Experiments are conducted on ten image sequences containing targets of interest in a variety of environments.Quantitative evaluation and comparison with the existing method show that the proposed method provides much improved results.

Design of vision‐based indoor positioning based on embedded system

“…The second step is background subtraction, and the next step is to apply a region filter to remove the noise. We modified the algorithm of [21] to perform the background modelling with more rapid convergence rate. The proposed algorithm utilises Gaussian mixture models (GMM) and adaptive mixture learning.…”

“…Since the updated formula in [21] converges very slowly, the model of the background may be very inaccurate during the convergence completion. To solve this problem, we improved the updated formula.…”

Design for an intelligent surveillance system based on system-on-a-programmable-chip platform

“…The foreground detection algorithm is based on a part of our previous work [6], namely multiple background maintenance (MBM) algorithm. We utilize the look-up table method to reduce the computational complexity of MBM, and the ping-pong buffer is used to solve the data dependencies problem.…”

Design and Integration for Background Subtraction and Foreground Tracking Algorithm

“…is a Gaussian probability density function, Gaussian distribution relates to an estimated weight, , i t ω , which indicates that the portion of the data is counted for the i th Gaussian at time t. In general, a new pixel value will be presented by one of the major components of the multi-model and used to update the distribution. To handle (1) on-line, the incremental update procedure, which contains dynamic and static background pixels update loop are used [8]. Then, multiple background maintenance maintains and learns a time adaptive background image using the statistical information of the multiple Gaussian distribution.…”

“…A hardware-oriental background subtraction algorithm, which is modified from our prior algorithm [8], is proposed. Besides, a sequential foreground labeling algorithm [9] and Kalman filter based tracking method [10] are integrated.…”