The Algorithm Of Adaptive Target Detection Based On Local Texture Feature

With the rapid development of computer science and technology and intelligent hardware,human beings have entered the intelligent society.Intelligent monitoring is widely used in national defense,transportation and security and so on.As a key technology in intelligent monitoring,moving target detection has become a hot spot in the field of intelligent monitoring.Moving object detection is the basis for the following process,such as image recognition,analysis and tracking of moving targets by using digital image processing,computer vision and pattern recognition technology.Although many scholars have proposed a lot of moving target detection algorithms,but under the complex dynamic scenes,thoses algorithms have defects such as incomplete results,the high false detection rate,sensitive to illumination and other issues.In this thesis,the studies of adaptive background subtraction have important theoretical and practical value.The ViBe(Visual Background Extractor)algorithm,one of the mainstream moving object detection algorithms,has been improved to increase the detection accuracy in complex dynamic scenes as the research focus.A historical pixel value queues has been introduced to implement adaptive segmentation thresholds in dynamic scenes.At the same time,the LBSP(Local Binary Similarity Patterns)operator has been introduced to improve the background model and update mechanism of ViBe algorithm in order to enhance the reliability of the background model.The main contents of this thesis are as follows:1.Combine to the change of pixels in the time domain,a historical pixel value queue has been introduced to describe the change of pixels in the time domain.A self-adaptive segmentation threshold and a self-adaptive update factor have been designed and implemented according to the standard deviation of the historical pixel value queue.According to the experiment result,the improved algorithm increases detection accuracy in complex dynamic background compare to ViBe.2.For the impact of illumination changes on the detection of moving targets,the LBSP texture description and color space has been used to establish the background model.At the same time,in the foreground segmentation,the LBSP texture description operator is used to further determine the pixels.According to experiment result,the improved algorithm effectively improves the accuracy of pixel classification and increases robustness to illumination changes.3.By deeply analyzing the causes of ghost region and combining the principle of spatial neighborhood consistency,a ghost detection and elimination algorithm is proposed and implemented in order to detect and eliminate ghost regions quickly in the result image and exclude the effect of false pixels on the detection results.Finally,the research work is summarized,and the following research direction is pointed out,which has opened up the ideas for further research.