Adaptive Background-Updating and Target Detection in Motion State

Background Subtraction Based on Gaussian Mixture Model

“…The sequences were manually segmented to generate the ground truth. We compare the improved algorithm with the original algorithm [3] with fixed number of components M = 4. For both algorithms and for different threshold values ( thr c from (2)), we measured the true positives-percentage of the pixels that belong to the intruding objects that are correctly assigned to the foreground and the false positives -percentage of the background pixels that are incorrectly classified as the foreground.…”

“…We assume that the samples are independent and the main problem is how to efficiently estimate the density function and to adapt it to possible changes. Kernel based density estimates were used in [4] and we present here an improvement of the GMM from [3] . There are models in the literature that consider the time aspect of an image sequence and the decision depends also on the previous pixel values from the sequence.…”

“…Gaussian mixture model (GMM) was proposed for background subtraction in [2] . One of the most commonly used approaches for updating GMM is presented in [3] and further elaborated in [10] . These GMMs use a fixed number of components.…”

“…In next section we list some related work. In section 3 the GMM approach from [3] is reviewed. In sections 4 we present how the number of components can be selected on-line and to improve the algorithm.…”