Research Of Target Detection Based On Target Modeling For Video Image

Moving target detection technology of video image is a hot research topic in the field of graphics and image processing,and has certain theoretical research value and practical application value.This paper focuses on video images captured by UAV,mainly studies the moving target detection algorithm based on feature matching and target modeling for dynamic background,and focuses on the target detection algorithm based on target modeling.First of all,in order to improve the detection effect of the m oving target based on feature matching,this paper studies a feature matching moving target detection algorithm with information accumulation.The original image is preprocessed based on Haar wavelet decomposition,the SURF feature points are extracted,the global background motion vector is estimated according to the feature point matching results,and the difference image information is summed up during the target detection process based on the method of information accumulation.Experiments show that the algorithm studied in this paper has a good detection effect of moving targets.Secondly,this paper studies an improved DPM(Deformable Part Model,DPM)target detection algorithm based on feature adaptive weighted fusion.When training the DPM target detection model and performing target detection,fusion characteristics need to be calculated,the principal component analysis of HOG(Histogram of Oriented Gradient,HOG)features and the LBP(Local Binary Pattern,LBP)features are fused with adaptive weighted.Although the algorithm has a large amount of calculation,the target detection model obtained by training has a higher target detection accuracy.Finally,aiming at improving the time-effectiveness of the target detection of the improved DPM-HOG-LBP algorithm,an improved DPM-HOG-LBP target detection algorithm based on selective search is studied.Based on the selective search to detect candidate target regions,the DPM-HOG-LBP algorithm is used to train the target detection model.And according to the target detection model,the candidate regions are classified to obtain th e final target detection result.Experiments show that the improved algorithm has improved real-time perfor-mance and has good detection accuracy in pedestrian and vehicle detection applications.