Probabilistic neural network based moving object detection method and an apparatus using the same

A moving object detection method based on a Probabilistic Neural Network (PNN) model to distinguish a moving object pixel from a background object pixel in a video image delivered over a bandwidth limited network, and the method comprising:

• receiving a sequence of images of a fixed location;
• generating from the sequence of images a background model based on the PNN model having four layers comprising an input units layer, a pattern units layer, a summation units layer, and an output units layer;
• defining the nodes of the four layers of the PNN model, the weight coefficient for each of the nodes, and the relationships among the nodes;
• receiving a current frame of the sequence of images and dividing the frames into N by N blocks where N is a positive integer greater than 1;
• determining pixel blocks by eliminating from the N by N blocks the blocks which are statistically determined to contain only background pixels;
• and
• determining whether a pixel of the pixel blocks of the sequence of images is a moving object pixel or a background object pixel by comparing the maximum coefficient of the output units layer against an output threshold.