원문정보
초록
영어
In traffic monitoring system, moving object monitor is the key part of monitoring system. This paper proposed a non-reference background detection method based on the reference background model for the poor detection effect of Gaussian background modeling. The model utilizes a series of sampling values to estimate the probability model of observed pixels before current pixels; Then binaryzation moving object detection based on the probability model. In terms of moving object identification, this paper proposed several features, and being trained and identified through BP neural network. The experimental result indicated that this background model can detect the foreground moving object effectively, and achieved satisfied effect on pedestrians and vehicles target recognition rate.
목차
1. Introduction
2. Moving Object Detection
3. The Establishment of Non-Reference Model
3.1. Density Estimation
3.2. Variance Estimation
3.3. Restrain Error Detection
4. BP Neural Network Classifier Design
4.1. Feature Extraction
4.2. Network Structure
4.3. Learning Algorithm
5. Experimental Result
6. Conclusion
Acknowledgements
Reference