Research On Shadow Object Elimination And Shape Extraction Algorithm Of Moving Human Body

Human beings belong to a special group,not only with the universality of the general group,but also with its special characteristics,so human target detection technology is an inevitable trend of social development.In the existing technology,most of the human target detection and morphology extraction algorithms encounter bottlenecks in morphology extraction accuracy,and it is difficult to further improve.From the above discussion,proposes a shadow target elimination and morphology extraction algorithm for moving human body by adding shadow elimination algorithm to the process of morphology extraction of key points of human skeleton,which further improves the accuracy of human body extraction morphology.Aiming at the problem that super-pixel image segmentation is vulnerable to illumination interference and poor segmentation effect,an improved SLIC fusion algorithm is proposed and tested on the laboratory video set taken in the laboratory corridor.The results show that compared with the traditional SLIC algorithm,the average operation efficiency of this algorithm is improved by 9.075%,It can be well applied to super-pixel segmentation under illumination interference.For the traditional ViBe algorithm is often disturbed by environmental changes,producing "ghost images",incomplete detection results and "voids" inside the human body,we propose an improved ViBe algorithm based on multi-frame combined with adaptive thresholding,and through comparison experiments,we verify the effectiveness of the proposed algorithm in processing video frame images,obtaining clearer foreground targets,successfully eliminating ghost images,and improving the accuracy of detecting moving targets in dynamic environments.By analyzing the changes of the internal attributes of HSV color model and LBP texture features in the image of shadow area and background area,a multi feature fusion shadow elimination algorithm based on improved SLIC and ViBe is proposed.Through comparative experiments,it is proved that the proposed algorithm can not only accurately eliminate the shadow of moving human body,but also adapt to different environments and has good robustness.In order to obtain the shape of key points of moving human skeleton more accurately,in the foreground target area without shadow extracted by the improved multi feature fusion shadow elimination algorithm of SLIC and ViBe,the shape of key points of moving human skeleton is extracted by using the OpenPose algorithm framework.Aiming at the problem that there is no sharing mechanism between network parameters in different stages in OpenPose algorithm,and it is easy to miss bone key points,an improved multi person bone key point shape extraction algorithm based on OpenPose is proposed,which improves the sharing between parameters in the model and the recognition degree of human bone key point feature vector.The results of the test on the mixed dataset of Laboratory,Peopleshade,and Bus station show that the algorithm is most prominent in the extraction of human skeletal key points when the similarity determination threshold is set to 0.5,with an average accuracy of 88.08% and a final average accuracy of 79.87% on average.