Research On Moving Human Detection And Tracking Algorithms In Complex Scenes

Moving human detection and tracking technology as a research hotspot in the field of computer vision.it is widely used in intelligent monitoring,motion analysis,traffic control,advanced human-computer interaction and other fields.It is also a necessary research content in security robot's visual navigation technology high research value.this text studies the detection and tracking of moving human bodies with occlusion in complex scenes.the main work is as follows:Firstly,a detection algorithm for the deformable part model of a moving human body is introduced in the text.the trained deformable part model is used to detect the human body region on the image to be detected.experiments show that the method can be used to perform the detection and positioning of moving human body in different states in a complex environment.In addition,a discriminative scale space tracking algorithm is introduced,a position filter and a scale filter are used to locate and scale the moving human body area.analysis shows that the tracking method is effective in complex situations such as illumination changes and scale transformation.Then,the DSST algorithm is studied when the target is completely blocked for a long time,the tracking model is contaminated and the target is lost in the follow-up tracking.a anti-occlusion tracking algorithm for moving human body based on DSST is proposed.on the basis of DSST tracking,the historical mean calculation strategy based on Fmax and APCE is used to detect occlusion;when it is determined that there is occlusion,the DPM detection algorithm is used to reposition the moving human body and update the target position,and then update the DSST filter model.experimental analysis shows that this article the anti-occlusion tracking algorithm is used to improve the distance accuracy and success rate compared to DSST by 4.3% and 6%,respectively,in the case of occlusion,deformation,motion blur,scale change,and out-of-plane rotation,and is more accurate than DSST.Finally,in order to solve the problem of tracking errors in complex situations such as the posture change of the moving human body or the temporary obstruction by obstacles,in this paper a anti-occlusion tracking algorithm combining Kalman filtering and DSST was designed.the occlusion is determined by using multi-peak detection and image similarity.if occlusion,the Kalman filter is combined to predict the position of the current frame and the model is reinitialized according to the correct position area of the previous tracking.otherwise,the optimal response is calculated according to the response value of the position filter and the similarity of the two frames of images,and the target is located according to the optimal response under the four-neighborhood search.experiments show that the tracking algorithm has a distance accuracy and success rate higher than DSST by 26.3% and 28.4% respectively in complex situations such as occlusion,deformation,scale transformation,and out-of-plane rotation,and is more accurate than DSST.