Research On Algorithms Of Moving Object Detection In Complex Background

Moving objects detection aimed at the complex dynamic backgrounds is one of the important research topics in the fields of image processing and machine vision.In recent years,moving objects detection has been widely used in remote sensing image,security monitoring,smart cities,military navigation and other fields.The purpose of detecting moving objects is to detect moving targets of interest in video sequences and then provide the information to the higher level of processing modules in the video surveillance system.The results of object detection can directly impact on the accuracy of the following algorithms.But in real living world,there are moving targets by stochastic motion,such as swaying tree branches,flowing current,varied lights and so on.Therefore,the researches of moving object detection aimed at the complex dynamic backgrounds is always a concern of challenging subject.This thesis proposes two moving detection algorithms based on background subtraction and a shadow suppression algorithm for dynamic scenes.The main research contents of this thesis are as follows:1.In view of the problem that the MC-SILTP texture feature can hardly distinguish the colored texture flat area leading to the high false detection rate of the algorithm based on MC-SILTP.This thesis proposes an improved SILTP texture feature which fuses the color information of the HSV color space with the MC-SILTP texture feature so as to own the corresponding characteristics of texture feature and HSV color feature.Therefore,the ability to distinguish the texture features in the texture flat area is greatly improved.At the same time,the codebook model of the original algorithm is improved,and the deadlock prevention strategy and the improved update strategy are added to make the background model fit the background more accurately.The experimental results show that the proposed algorithm compared with the original algorithm can effectively reduce the miss detection rate and greatly improve the integrity and accuracy of the detection results.2.In the FCCV target detection algorithm,fuzzy color coherence vector doesn't make full use of color local spatial information so that it is difficult to distinguish some complex scenes,which makes the precision of detection results lower,so an improved fuzzy color coherence vector target detection algorithm is proposed in this thesis.This algorithm introduces the improved FCCV feature to replace the original feature so as to make full use of coherencearea spatial location information so that the features are more precise and more differentiated.Meanwhile,the image quantization strategy is changed to make the proportion of different color components more reasonable.Thus,the detection accuracy is improved and the superiority of the algorithm is proved by experiments.3.As the improved FCCV algorithm is more sensitive to light changes and can't suppress shadow effectively,so a moving shadow suppression algorithm based on multi-feature and multi-scale is proposed in this thesis.The algorithm firstly extract the brightness,color and multi-scale block MC-SILTP texture feature of the shadow,then fuse the detection results of three features by using decision level fusion strategy.The new algorithm has the ability to suppress the complex background and the shadow meantime.The experimental results show that the algorithm can effectively remove the shadow of the moving object,thus greatly improving the applicability of the algorithm.