Study Of Human Tracing Based On Binocular Vision And3D-reconstruction

In recent years, human tracking has been taken care of the field of computer vision,intelligent monitoring at the scene, virtual reality, pose estimation and recognition, motionanalysis and game animation and other fields have been widely used and have good prospects.With the continuous development of computer image processing technology, the traditionaltracking algorithm based on monocular video sequences has some limits on the light coverand it can not adapt the complex environment of the human body tracking. Trackingalgorithm based on binocular vision came into being, many experts and scholars have studiedfor binocular vision of human tracking algorithm. There are three steps. First, human targetdetection is extracting detected the body from the video sequence. Second, match the humanbody contour and3d human body model and tracking the detected human. Third, reproducecolumnar model or skeleton model based.This paper presents a particle filter tracking method based on binocular visionmulti-feature fusion. Body contour characteristics and three-dimensional human to determinethe three-dimensional position of the human skeleton model. Improved multi-feature particlefilter algorithm to track the position of joints and restore the body’s three-dimensional motionskeleton sequence. The main contents are as follows:First, introduce the theory of binocular stereo vision, camera calibration, featureextraction, stereo matching and3D reconstruction. Select the model of binocular vision ofparallel obtain the three-dimensional coordinates of the joint point.Next, choose a three-dimensional human body joint model, It has20degrees of freedom.It can be reflected on the body posture and calculate simple.Once again, particle filter tracking algorithm of a multi-feature fusion is proposed. Itovercomes the effects of illumination, complex background. Particle degradation problem hasimproved and there is a strong tracking accuracy and stability.Finally, the binocular particle filter algorithm is applied to the human body tracking. Wepropose a multi-feature binocular vision particle filter tracking algorithm and restored the three-dimensional human skeleton sequence. It shows that this is an effective identificationmethod by experiment.