Against to the problem of the low suitability for light, regarding the uncovered background detecting as the foreground region easily and susceptible to environmental interference and other issue of Gaussian mixture model, an improved algorithm of Gauss mixture model was proposed. Through the learning rate updated dynamically to improve the update rate of the background model, making uncovered background area can quickly recover. Replacement method using background model effectively eliminates the external environment interference on moving target detection. Combining the moving target pixel inter-frame difference method with the improved Gaussian mixture model which detecting the moving target pixel into the and calculation, solving the inaccurate problem of moving target detection caused by lighting issues. Since the detected moving target pixel is not focused, the improved algorithm for each pixel movement of the target area was carried out to determine the eight neighborhood threshold, making the extracted moving object more clearly and accurately. Experimental results show that the improved algorithm of this paper has been significantly improved in terms of accuracy, completeness, usefulness and so on. |