An Improved PIC Algorithm of Background Reconstruction for Detecting Moving Object

Abstract

In machine vision, moving object detection and segment pay more attention to the real-time and the accuracy. Generally, familiar case is immovable camera with the fixed focus in moving object detection, however, it is difficult to detect whole and actual object because of the influence of the environment noise and others. This paper makes some improvement in PIC algorithm and presents a new method of detecting moving object. According to normalization the pixels of the chosen images series used to reconstruct the background, quantization statistic, extent the quantized range, reconstruction the background image, the improved PIC algorithm avoids to providing thresholds of the PIC algorithm manually and removes these steps of combining the approximate gray scope, which needs plenty of time and is hard to realize through programming. After acquiring the reconstructed image, the coarse-fine two steps method is suggested to confirm the object position exactly and complete the moving object detection finally. The experiment results show that the method proposed in this paper needs shorter running time of the program and provides more accurate position of the moving object.