Universal Background Subtraction based on Arithmetic Distribution Neural Network

TL;DR

This is the first method to propose network layers based on arithmetic distribution operations for learning distributions during background subtraction and is able to utilize the probability information of the histogram and achieve promising results with a very simple architecture compared to traditional convolutional neural networks.

Abstract

We propose a universal background subtraction framework based on the Arithmetic Distribution Neural Network (ADNN) for learning the distributions of temporal pixels. In our ADNN model, the arithmetic distribution operations are utilized to introduce the arithmetic distribution layers, including the product distribution layer and the sum distribution layer. Furthermore, in order to improve the accuracy of the proposed approach, an improved Bayesian refinement model based on neighboring information, with a GPU implementation, is incorporated. In the forward pass and backpropagation of the proposed arithmetic distribution layers, histograms are considered as probability density functions rather than matrices. Thus, the proposed approach is able to utilize the probability information of the histogram and achieve promising results with a very simple architecture compared to traditional convolutional neural networks. Evaluations using standard benchmarks demonstrate the superiority of the proposed approach compared to state-of-the-art traditional and deep learning methods. To the best of our knowledge, this is the first method to propose network layers based on arithmetic distribution operations for learning distributions during background subtraction.

Figures (3)

• Figure 1: Illustration of the arithmetic distribution neural network for background subtraction. Histograms of subtractions between the current observations and their historical counterparts in pixels are input into the arithmetic distribution layers, containing the product and sum distribution layers for distribution learning. In particular, the learning kernels of arithmetic distributions layers are also distributions described by histograms. A classification architecture is then attached to label the pixels according to the output of these layers.
• Figure 2: An illustration of the arithmetic distribution neural network for background subtraction. Histograms of subtractions between a pixel's current observation and its historical counterparts are used as the input to arithmetic distribution layers for learning distributions. The output histograms of arithmetic distribution layers are combined by a convolution and input into a classification architecture containing a convolution layer, a rectified linear unit (Relu) layer, and a fully connected layer for classification.
• Figure 3: An illustration of the Gaussian approximation function which approximates using a piecewise function controlled by the parameters of the Gaussian function.