Targets Detection Based on the Prejudging and Prediction Mechanism

摘要

The moving target detection is important to video supervision, video content analysis, object identification, and so on. However, some factors such as light, weather, shadow, the falling leaves and objects temporarily stumbled into the video may interrupt the real-time target extraction. In the paper, a new method based on a prejudging and prediction algorithm is proposed to reduce noise, improve the accuracy of segmentation, and decrease the regular computation cost. Six parts are introduced in the paper. In the second part, background subtraction method is simply described for target extraction. In the third part, after comparing two background models, the multi-dimension GMM is chosen and an improved multi-dimension GMM based on the prejudging and prediction algorithm is described in the fourth part. Some experiments are carried out and the experimental results are shown in the fifth part. Experimental results show that the method proposed in the paper could decrease the computation cost, reduce stumbled object noise and improve the accuracy of detection.