Visual Tracking Through Dual-Camera Based On Sparse And Low-Rank

With the rapid development of artificial intelligence these year,the requirements of the visual field are getting higher,the visual field in the development of artificial intelligence is also more significant,target tracking is an important aspect of computer vision.In the classical low rank sparse model,there is a problem of repeated calculation,and speed is one of the problems in the field of target tracking.Therefore,this paper improves the traditional tracking model and improves the tracking speed.However,in the small range or large range of occlusion conditions,the usual target tracking algorithm will drift and lead to target tracking failure.Based on this,this paper studies and proposes a binocular vision tracking algorithm based on low rank sparseness,and uses the distance information of target and equipment to track the target to realize the feasibility of target tracking after occlusion.This paper mainly includes the following aspects:1.Introduce the current mainstream target tracking model,dictionary update algorithm,low rank sparse algorithm principle and binocular stereo vision to obtain distance information;2.A single camera tracking model is enhanced.Based on the sparse theory modeling method,the sparse matrix is constructed by using the low rank sparsity of the image on the basis of the particle filter model.The background modeling problem is reduced to the sparse linear combination problem of atoms in the positive and negative sample dictionaries under the L21 norm.Aiming at the problem of target updating,a dictionary updating strategy is proposed to ensure that the background model can adapt well to the environment;3.The single-camera tracking algorithm is based on the combination of qualitative and quantitative methods in the common data set under different environmental conditions.Compared with the classical,mainstream and sparse theory based on the moving target tracking method,this paper validates the The L21 norm has the advantage.4.Set up binocular camera device to obtain a binocular video with light changes,front and rear movement,left and right movement,shape change,small area occlusion,large area occlusion,and calibration.The binocular video depth map is obtained and applied to the single camera target tracking model constructed under the L21 norm.